PREVALENCE AND PREDICTORS OF IMMUNOLOGIC FAILURE AMONG HIV

PATIENTS ON HAART IN SOUTHERN ETHIOPIA

by

KESETEBIRHAN DELELE YIRDAW

submitted in accordance with the requirements

for the degree of

MASTER OF PUBLIC HEALTH

at the

UNIVERSITY OF SOUTH AFRICA

SUPERVISOR: PROFESSOR SP HATTINGH

November 2014
 Student number: 4836-405-3

DECLARATION

I declare that PREVALENCE AND PREDICTORS OF IMMUNOLOGIC FAILURE

AMONG HIV PATIENTS ON HAART IN SOUTHERN ETHIOPIA is my own work and
that all the sources that I have used or quoted have been indicated and acknowledged
by means of complete references and that this work has not been submitted before for
any other degree at any other institution.

Kesetebirhan Delele November, 2014

Full name and signature Date
 ACKNOWLEDGEMENTS

My heartfelt gratitude goes to my parents who laid down the foundations for my life that
have led to its being the way it is now, and my lovely wife, Fitih Zelalem, who has been
considerate, patient and supportive throughout my education even at times when it was
necessary for me to be absent from the family so that I could succeed, and our baby
girl, Gelila, whom I am grateful to have and dedicate this dissertation to.

In addition, I would like to thank the following people who contributed to the completion
of this current research:

 Special thanks to my supervisor, Professor SP Hattingh, for her insightful

guidance and unrelenting support throughout.
 My gratitude also goes to the dedicated team of health care workers in Butajira
and Hosanna Hospitals who provide care to HIV patients who made this study
possible.
 Heartfelt thanks to Kabtamu Tolossie who guided me with statistical aspects of
the study, Mrs. Rina Coetzer for professionally formatting the dissertation,
Emebet Haile who did the editing.

Last but not least, I would like to thank Dr. Zewdu Gashu, Ato Kemal Woliyu, and Ato
Tekalign Mekuria, all UNISA students, who provided me with technical information that
made life at UNISA easy for me.
 PREVALENCE AND PREDICTORS OF IMMUNOLOGIC FAILURE AMONG HIV
PATIENTS ON HAART IN SOUTHERN ETHIOPIA

STUDENT NUMBER: 4836-405-3

STUDENT: DR KESETEBIRHAN DELELE YIRDAW
DEGREE: MASTER OF PUBLIC HEALTH
DEPARTMENT: HEALTH STUDIES, UNIVERSITY OF SOUTH AFRICA
SUPERVISOR: PROFFESOR SP HATTINGH

ABSTRACT

Immunologic monitoring is part of the standard care for patients on antiretroviral

treatment. Yet, little is known about the routine implementation of immunologic
monitoring in Ethiopia. This study assessed the pattern of immunologic monitoring,
immunologic response, level of immunologic treatment failure and factors related to it
among patients on antiretroviral therapy in selected hospitals in southern Ethiopia. A
retrospective longitudinal analytic study was conducted using documents of patients
started on antiretroviral therapy.

A total of 1,321 documents of patients reviewed revealed timely immunologic monitoring

were inadequate. Despite overall adequate immunologic response, the prevalence of
immunologic failure was 11.5% (n=147). Having WHO Stage III/IV of the disease and a
higher CD4 (cluster differentiation 4) cell count at baseline were identified as risks for
immunologic failure.

These findings highlight the magnitude of the problem of immunologic failure.

Prioritizing monitoring for high risk patients may help in effective utilisation of meager
resources.

Key concepts: Human Immunodeficiency Virus; CD4 cell; highly active antiretroviral
therapy (HAART); adherence; immunologic treatment failure; antiretroviral treatment
failure; prevalence; patient; predict.
 i

TABLE OF CONTENTS

LIST OF TABLES ..................................................................................................................... viii

LIST OF FIGURES .................................................................................................................... xi
LIST OF ANNEXURES.............................................................................................................. xii
ABBREVATIONS ..................................................................................................................... xiii
CHAPTER 1 ............................................................................................................................ 1
ORIENTATION TO THE STUDY ................................................................................................ 1
1.1 INTRODUCTION ................................................................................................... 1
1.2 BACKGROUND TO THE RESEARCH PROBLEM ................................................ 2
1.2.1 HIV infection, disease causation, and HAART ....................................................... 2
1.2.2 HIV and AIDS globally ........................................................................................... 3
1.2.3 HIV and AIDS in sub-Saharan Africa...................................................................... 3
1.2.4 HIV and AIDS in Ethiopia ....................................................................................... 4
1.2.5 The goal of HAART and adherence ....................................................................... 4
1.2.6 Monitoring HAART treatment response.................................................................. 6
1.3 STATEMENT OF THE RESEARCH PROBLEM .................................................... 7
1.4 DEFINITIONS OF KEY CONCEPTS...................................................................... 8
1.4.1 Human Immunodeficiency virus ............................................................................. 9
1.4.2 Describing CD4 cell ............................................................................................... 9
1.4.3 Highly active antiretroviral therapy (HAART) .......................................................... 9
1.4.4 Adherence ........................................................................................................... 10
1.4.5 Treatment failure .................................................................................................. 10
1.4.6 Antiretroviral treatment failure .............................................................................. 10
1.4.7 Prevalence ........................................................................................................... 11
1.4.8 Patient or client .................................................................................................... 11
1.4.9 Predict ................................................................................................................. 11
1.5 OPERATIONAL DEFINITIONS ............................................................................ 12
1.5.1 Adherence to medication refill .............................................................................. 12
1.5.2 Immunological failure ........................................................................................... 12
1.6 RESEARCH AIM/PURPOSE ............................................................................... 13
1.7 RESEARCH OBJECTIVES .................................................................................. 13
1.8 RESEARCH QUESTIONS/HYPOTHESES .......................................................... 13
1.9 PARADIGM.......................................................................................................... 13
1.10 RESEARCH DESIGN .......................................................................................... 14
1.10.1 Correlational elements ......................................................................................... 14
1.10.2 Longitudinal elements .......................................................................................... 15
1.11 SAMPLING .......................................................................................................... 15
 ii

1.11.1 Site sampling ....................................................................................................... 16

1.11.2 Data source sampling .......................................................................................... 16
1.11.2.1 Eligibility criteria ................................................................................................... 16
1.11.2.2 Respondent/participant sampling technique ......................................................... 17
1.11.2.3 Respondent/participant sample size .................................................................... 17
1.12 DATA COLLECTION ........................................................................................... 18
1.12.1 The data collection instrument ............................................................................. 18
1.12.2 Administering the data collection instrument ........................................................ 18
1.12.3 Pretesting the data collection tool ........................................................................ 18
1.12.4 Data collection procedure .................................................................................... 19
1.13 DATA MANAGEMENT AND ANALYSIS .............................................................. 19
1.14 DATA AND DESIGN QUALITY ............................................................................ 21
1.14.1 Quality of research design ................................................................................... 21
1.14.1.1 Internal validity ..................................................................................................... 21
1.14.1.2 External validity .................................................................................................... 21
1.14.2 Quality of the data gathering instrument............................................................... 22
1.14.2.1 Reliability and validity ........................................................................................... 22
1.15 ETHICAL CONSIDERATIONS ............................................................................. 22
1.15.1 The institution/site ................................................................................................ 22
1.15.2 Scientific integrity of the research ........................................................................ 23
1.15.3 Ethical principles .................................................................................................. 23
1.15.3.1 Beneficence ......................................................................................................... 23
1.15.3.2 Respect for human dignity ................................................................................... 23
1.15.3.3 Justice ................................................................................................................. 24
1.16 SIGNIFICANCE OF THE STUDY ........................................................................ 24
1.17 SCOPE AND LIMITATIONS ................................................................................ 25
1.18 STRUCTURE OF THE DISSERTATION .............................................................. 25
1.19 CONCLUSION ..................................................................................................... 25
CHAPTER 2 .......................................................................................................................... 27
LITERATURE REVIEW ............................................................................................................ 27
2.1 INTRODUCTION ................................................................................................. 27
2.2 PURPOSE OF THE LITERATURE REVIEW ....................................................... 27
2.3 CLASSIFICATION AND HISTORY OF THE HIV VIRUS ...................................... 28
2.4 EPIDEMIOLOGY OF HIV/AIDS ........................................................................... 28
2.4.1 Transmission and risk factors .............................................................................. 29
2.4.2 Prevalence and geographic distribution ............................................................... 29
2.4.3 The state of the epidemic in Ethiopia ................................................................... 30
 iii

2.4.3.1 The state of the epidemic in Southern Nations Nationalities and People‘s Region of
Ethiopia................................................................................................................ 31
2.5 PATHOGENESIS OF HIV/AIDS........................................................................... 31
2.5.1 HIV infection and replication ................................................................................ 31
2.5.2 Immune activation and CD4 T-cell depletion ........................................................ 33
2.5.3 Reservoirs of HIV infection................................................................................... 33
2.6 IMMUNE RESPONSES TO HIV INFECTION ...................................................... 34
2.7 LABORATORY DIAGNOSIS OF HIV INFECTION ............................................... 34
2.8 THE NATURAL HISTORY OF HIV INFECTION .................................................. 37
2.8.1 Acute phase ......................................................................................................... 37
2.8.2 Asymptomatic phase............................................................................................ 38
2.8.3 Symptomatic phase ............................................................................................. 38
2.9 ANTIRETROVIRAL THERAPY ............................................................................ 38
2.9.1 Historical perspective ........................................................................................... 39
2.9.2 Eligibility criteria ................................................................................................... 39
2.9.3 Access to antiretroviral therapy ............................................................................ 40
2.9.4 Antiretroviral drug classes .................................................................................... 41
2.9.5 Goal of therapy .................................................................................................... 41
2.10 ADHERENCE PREPARATION BEFORE INITIATING ANTIRETROVIRAL
THERAPY ............................................................................................................ 41
2.11 ADHERENCE ASSESSMENT DURING ANTIRETROVIRAL THERAPY INTAKE 42
2.11.1 Level of adherence to HAART in developed countries ......................................... 42
2.11.2 Level of adherence to HAART in sub-Saharan Africa........................................... 43
2.11.3 Level of adherence to HAART in Ethiopia ............................................................ 43
2.12 MONITORING OF ANTIRETROVIRAL THERAPY .............................................. 44
2.12.1 The pattern of immunologic monitoring in developed countries ............................ 45
2.12.2 The pattern of immunologic monitoring in sub-Saharan Africa ............................. 45
2.12.3 The pattern of immunologic monitoring in Ethiopia ............................................... 46
2.13 IMMUNOLOGIC RESPONSE AFTER TAKING ANTIRETROVIRAL THERAPY .. 47
2.13.1 The pattern of immunologic response in developed countries .............................. 47
2.13.2 The pattern of immunologic response in sub-Saharan Africa ............................... 48
2.13.3 The pattern of immunologic response in Ethiopia ................................................. 49
2.14 IMMUNOLOGICAL TREATMENT FAILURE ........................................................ 50
2.14.1 Immunological treatment failure criteria................................................................ 50
2.14.1.1 The prevalence of immunologic treatment failure ................................................. 50
2.14.1.1.1 The prevalence of immunologic treatment failure in sub-Saharan Africa .............. 51
2.14.1.1.2 The prevalence of immunologic treatment failure in Ethiopia ............................... 51
2.15 DETERMINANTS OF IMMUNOLOGIC TREATMENT FAILURE .......................... 51
 iv

2.16 INTERVENTION FOR THOSE WITH TREATMENT FAILURE ............................ 52

2.16.1 Managing poor treatment adherence ................................................................... 53
2.16.1.1 Factors related to the health system .................................................................... 53
2.16.1.2 Factors related to the individual ........................................................................... 53
2.16.2 Managing treatment failure .................................................................................. 54
2.17 CONCLUSION ..................................................................................................... 55
CHAPTER 3 .......................................................................................................................... 56
RESEARCH METHODOLOGY ................................................................................................ 56
3.1 INTRODUCTION ................................................................................................. 56
3.2 RESEARCH DESIGN .......................................................................................... 57
3.2.1 Paradigm ............................................................................................................. 57
3.2.2 Research design .................................................................................................. 57
3.2.2.1 Quantitative elements .......................................................................................... 58
3.2.2.2 Correlational elements ......................................................................................... 58
3.2.2.3 Longitudinal elements .......................................................................................... 58
3.3 RESEARCH METHODS ...................................................................................... 59
3.3.1 Population ............................................................................................................ 59
3.3.2 Sampling .............................................................................................................. 60
3.3.2.1 Eligibility criteria ................................................................................................... 60
3.3.2.2 Sample size ......................................................................................................... 61
3.3.2.3 Sampling process ................................................................................................ 61
3.3.3 Data collection ..................................................................................................... 62
3.3.3.1 Data collection instrument .................................................................................... 62
3.3.3.2 Administering the data collection instrument ........................................................ 62
3.3.3.3 Pre-testing the data collection tool ....................................................................... 63
3.3.3.4 Data collection procedure .................................................................................... 63
3.3.4 Data management and analysis ........................................................................... 63
3.3.4.1 Descriptive statistics ............................................................................................ 64
3.3.4.2 Inferential statistics .............................................................................................. 65
3.3.4.2.1 The Chi square test.............................................................................................. 65
3.3.4.2.2 The log-rank test .................................................................................................. 65
3.3.4.2.3 The Cox proportional hazards regression ............................................................ 66
3.3.4.2.3.1 Extensions of the proportional hazards regression model .................................... 66
3.3.4.2.3.2 Variable selection for Cox proportional hazards regression modelling.................. 67
3.3.4.2.3.3 Purposeful selection of covariates ........................................................................ 67
3.3.4.2.3.3.1 Step 1 ................................................................................................................ 67
3.3.4.2.3.3.2 Step 2 ................................................................................................................ 67
3.3.4.2.3.3.3 Step 3 ................................................................................................................ 68
 v

3.3.4.2.3.3.4 Step 4 ................................................................................................................ 68

3.3.4.2.3.3.5 Step 5 ................................................................................................................ 68
3.3.4.2.3.3.6 Step 6 ................................................................................................................ 68
3.3.4.2.3.3.7 Step 7 ................................................................................................................ 69
3.3.4.2.3.3.7.1 Collinearity between covariates .................................................................... 69
3.3.4.2.3.3.7.2 Proportional hazards assumption ................................................................. 69
3.3.4.2.3.3.7.3 Checking for influential records .................................................................... 69
3.4 RESEARCH DESIGN AND DATA INSTRUMENT QUALITY ............................... 70
3.4.1 Quality of research design ................................................................................... 70
3.4.1.1 Internal validity ..................................................................................................... 70
3.4.1.2 External validity .................................................................................................... 71
3.4.2 Quality of the data gathering instrument............................................................... 71
3.4.2.1 Validity and reliability ........................................................................................... 71
3.5 ETHICAL CONSIDERATIONS ............................................................................ 72
3.5.1 External Review ................................................................................................... 73
3.6 SCOPE AND LIMITATIONS ................................................................................ 74
3.7 CONCLUSION ..................................................................................................... 74
CHAPTER 4 .......................................................................................................................... 75
PRESENTATION AND DESCRIPTION OF THE RESEARCH FINDINGS................................ 75
4.1 INTRODUCTION ................................................................................................. 75
4.2 DATA MANAGEMENT AND ANALYSIS .............................................................. 75
4.2.1 Data management ............................................................................................... 75
4.3 RESEARCH RESULTS ....................................................................................... 76
4.3.1 Sample characteristics ......................................................................................... 76
4.3.1.1 Demographic characteristics ................................................................................ 76
4.3.1.1.1 Age ...................................................................................................................... 76
4.3.1.1.2 Gender................................................................................................................. 77
4.3.1.2 Clinical information............................................................................................... 77
4.3.1.2.1 Baseline WHO Stage ........................................................................................... 78
4.3.1.2.2 Baseline CD4 cell count ....................................................................................... 78
4.3.1.2.3 Diagnosis of Tuberculosis during follow-up .......................................................... 79
4.3.1.2.4 Adherence to medication refill .............................................................................. 79
4.3.1.2.5 Final follow-up status ........................................................................................... 80
4.3.1.2.5.1 Final follow-up status by age group...................................................................... 80
4.3.1.2.5.2 Final follow-up status by gender .......................................................................... 81
4.3.1.2.5.3 Final follow-up status by baseline WHO Stage..................................................... 81
4.3.1.2.5.4 Final follow-up status by baseline CD4 cell count ................................................ 81
4.3.1.2.6 Follow-up time ..................................................................................................... 81
 vi

4.3.2 Objective 1: assess the pattern of immunologic (CD4 cell count) monitoring in
selected hospitals in southern Ethiopia ................................................................ 82
4.3.2.1 Pattern of CD4 cell count monitoring by age ........................................................ 84
4.3.2.2 Pattern of CD4 cell count monitoring by gender ................................................... 84
4.3.2.3 Pattern of CD4 cell count monitoring by baseline WHO Stage ............................. 85
4.3.2.4 Pattern of CD4 cell count monitoring by baseline CD4 cell count ......................... 85
4.3.3 Objective 2: assess the pattern of immunologic (CD4 cell count) response for
patients on HAART in selected hospitals in southern Ethiopia ............................. 85
4.3.3.1 Pattern of CD4 cell count response by age .......................................................... 86
4.3.3.2 Pattern of CD4 cell count response by gender ..................................................... 87
4.3.3.3 Pattern of CD4 cell count response by baseline WHO Stage ............................... 87
4.3.3.4 Pattern of CD4 cell count response by baseline CD4 cell count ........................... 88
4.3.4 Objective 3: describe immunologic treatment failure and identify its predictors
among patients on HAART in selected hospitals in southern Ethiopia.................. 89
4.3.4.1 Prevalence of immunologic failure ....................................................................... 89
4.3.4.2 Immunologic failure by criteria of diagnosis .......................................................... 89
4.3.4.3 Immunologic failure at the final immunologic evaluation ....................................... 90
4.3.4.4 Patients on second line HAART among those with confirmed immunologic failure
.......................................................................................................................... 91
4.3.4.5 Predictors of immunologic failure ......................................................................... 91
4.3.4.5.1 Event free survival ............................................................................................... 91
4.3.4.5.2 Univariable analysis ............................................................................................. 93
4.3.4.5.2.1 Age and immunologic failure ................................................................................ 94
4.3.4.5.2.2 Gender and immunologic failure .......................................................................... 94
4.3.4.5.2.3 Baseline WHO Stage and immunologic failure ..................................................... 95
4.3.4.5.2.4 Baseline CD4 cell count and immunologic failure ................................................. 96
4.3.4.5.2.5 Tuberculosis disease and immunologic failure ..................................................... 97
4.3.4.5.2.6 Adherence to medication refill and immunologic failure ........................................ 98
4.3.4.5.3 Multivariable analysis with Cox proportional hazards regression modelling .......... 99
4.3.4.5.3.1 Step 1 .................................................................................................................. 99
4.3.4.5.3.2 Step 2 ................................................................................................................ 100
4.3.4.5.3.3 Step 3 ................................................................................................................ 100
4.3.4.5.3.4 Step 4 ................................................................................................................ 100
4.3.4.5.3.5 Step 5 ................................................................................................................ 101
4.3.4.5.3.6 Step 6 ................................................................................................................ 103
4.3.4.5.3.7 Step 7 ................................................................................................................ 103
4.3.4.5.3.7.1 Collinearity between covariates ....................................................................... 104
4.3.4.5.3.7.2 Proportional hazards assumption..................................................................... 104
 vii

4.3.4.5.3.7.3 Checking influential records ............................................................................. 105

4.3.4.5.3.7.4 Goodness-of-fit test ......................................................................................... 107
4.3.4.5.3.8 Summary of regression analysis to predict immunologic failure ......................... 107
4.4 OVERVIEW OF RESEARCH FINDINGS ........................................................... 109
4.5 CONCLUSION ................................................................................................... 111
CHAPTER 5 ........................................................................................................................ 112
CONCLUSIONS AND RECOMMENDATIONS ...................................................................... 112
5.1 INTRODUCTION ............................................................................................... 112
5.2 RESEARCH DESIGN AND METHOD................................................................ 112
5.3 CONCLUSIONS ................................................................................................ 113
5.3.1 Objective 1: assess the pattern of immunologic (CD4 cell count) monitoring in
selected hospitals in southern Ethiopia .............................................................. 113
5.3.1.1 Pattern of immunologic monitoring ..................................................................... 113
5.3.2 Objective 2: assess the pattern of immunologic (CD4 cell count) response for
patients on HAART in selected hospitals in southern Ethiopia ........................... 113
5.3.2.1 Pattern of immunologic response ....................................................................... 113
5.3.3 Objective 3: describe immunologic treatment failure and identify its predictors
among patients on HAART in selected hospitals in southern Ethiopia................ 114
5.3.3.1 Prevalence and predictors of immunologic failure .............................................. 114
5.4 RECOMMENDATIONS ...................................................................................... 115
5.4.1 Treatment plan to identify and manage patients with immunologic failure .......... 115
5.4.2 Recommendations to health care providers ....................................................... 115
5.4.3 Recommendations to program managers .......................................................... 117
5.4.4 Recommendations for further research .............................................................. 118
5.5 CONTRIBUTIONS OF THE STUDY .................................................................. 118
5.6 LIMITATIONS OF THE STUDY ......................................................................... 119
5.7 CONCLUDING REMARKS ................................................................................ 119
BIBLIOGRAPHY..................................................................................................................... 120
 viii

LIST OF TABLES

TABLE 2.1a SUMMARY OF PREFERRED SECOND-LINE ART REGIMENS FOR ADULTS

AND ADOLESCENTS .......................................................................................... 54
TABLE 2.1b SUMMARY OF RECOMMENDED FIRST- AND SECOND-LINE ART REGIMENS
FOR CHILDREN AND ADOLESCENTS .............................................................. 54
TABLE 4.1 DESCRIPTIVE STATISTICS FOR CONTINOUS VARIABLES OF PATIENTS ON
ANTI-RETROVIRAL THERAPY (N=1,321) .......................................................... 77
TABLE 4.2 CHARCTERISTICS OF PATIENTS ON ANTI-RETROVIRAL THERAPY (N=1,321)
............................................................................................................................ 77
TABLE 4.3 FINAL FOLLOW-UP STATUS BY DEMOGRAPHIC AND BASELINE
CHARACTERISTICS OF STUDY PATIENTS ON ANTI-RETROVIRAL THERAPY
(N=1,321)............................................................................................................. 80
TABLE 4.4 DETERMINATION OF TIMELY FOLLOW-UP CD4 CELL COUNT TESTING BY
FOLLOW-UP PERIOD IN MONTHS AMONG PATIENTS ON ANTI-RETROVIRAL
THERAPY (N=1,321) ........................................................................................... 83
TABLE 4.5 DETERMINATION OF TIMELY FOLLOW-UP CD4 CELL COUNT TESTING BY
DEMOGRAPHIC AND BASELINE TREATMENT VARIABLES AMONG PATIENTS
ON ANTI-RETROVIRAL THERAPY (N=1,321) .................................................... 84
TABLE 4.6 MAGNITUDE OF FOLLOW-UP CD4 CELL COUNT BY DEMOGRAPHIC AND
BASELINE TREATMENT VARIABLES AMONG PATIENTS ON ANTI-
RETROVIRAL THERAPY (N=1,321 PATIENTS WITH 5241 FOLLOW-UP CD4
CELL COUNT TESTS) ........................................................................................ 87
TABLE 4.7 IMMUNOLOGIC TREATMENT FAILURE AMONG PATIENTS ON ANTI-
RETROVIRAL THERAPY (N=1,321) ................................................................... 89
TABLE 4.8 DIAGNOSTIC CRITERIA FOR FIRST INSTANCE IMMUNOLOGIC FAILURE
AMONG STUDY PATIENTS ON ANTI-RETROVIRAL THERAPY (n=226) .......... 90
TABLE 4.9 FOLLOW-UP IMMUNOLOGIC EVALUATION AND FINAL IMMUNOLOGIC
STATUS OF PATIENTS WHO EVER HAD IMMUNOLOGIC FAILURE (n=226) .. 90
TABLE 4.10 FREQUENCY DISTIBUTION OF PATIENTS ON SECOND LINE TREATMENT
AMONG THOSE WITH CONFIRMED IMMUNOLOGIC FAILURE (n=93) ............ 91
TABLE 4.11 DESCRIPTION OF FOLLOW-UP TIME IN YEARS FOR STUDY PATIENTS ON
ANTI-RETROVIRAL THERAPY (n=1282) ............................................................ 92
TABLE 4.12 LIFE TABLE FOR IMMUNOLOGIC FAILURE AMONG PATIENTS ON ANTI-
RETROVIRAL THERAPY (n=1282) ..................................................................... 92
TABLE 4.13 STRATIFIED LOG-RANK TEST FOR EQUALITY OF FAILURE FUNCTIONS BY
COVARIATES AMONG PATIENTS ON ANTI-RETROVIRAL THERAPY (n=1,282)
............................................................................................................................ 93
 ix

TABLE 4.14 INDIVIDUAL COVARIATE COX PROPORTIONAL HAZARDS REGRESSION

ANALYSIS FOR PREDICTORS OF IMMUNOLOGIC FAILURE AMONG
PATIENTS ON ANTI-RETROVIRAL THERAPY (n=1,282) .................................. 99
TABLE 4.15 INITIAL MULTIVARIABLE MODEL FOR PREDICTORS OF IMMUNOLOGIC
FAILURE AMONG PATIENTS ON ANTI-RETROVIRAL THERAPY (n=1,282) .... 99
TABLE 4.16 PARTIAL LIKELIHOOD RATIO TEST FOR THE EFFECT OF REMOVAL OF
COVARIATES ONE AT A TIME FROM THE INITIAL MULTIVARIABLE MODEL
(n=1,282) ........................................................................................................... 100
TABLE 4.17 PARTIAL LIKELIHOOD RATIO TEST FOR THE EFFECT OF THE ADDITION OF
COVARIATES NOT INCLUDED IN STEP 1, ONE AT A TIME, TO THE INITIAL
MULTIVARIABLE MODEL (n=1,282) ................................................................ 101
TABLE 4.18 CHANGES IN THE COEFFICIENT OF WHO STAGE DURING ADDITION OF
COVARIATES EXCLUDED DURING STEP 1 OF MODEL BUILDING (n=1,282)
.......................................................................................................................... 101
TABLE 4.19 CHANGES IN THE COEFFICIENT OF BASELINE CD4 CELL COUNT DURING
ADDITION OF COVARIATES EXCLUDED DURING STEP 1 OF MODEL
BUILDING (n=1,282).......................................................................................... 101
TABLE 4.20 SUMMARY OF A FRACTIONAL POLYNOMIALS ANALYSIS OF THE BASELINE
CD4 CELL COUNT (n=1,282) ............................................................................ 102
TABLE 4.21 PARTIAL LIKELIHOOD RATIO TEST FOR THE ADDITION OF INTERACTION
TERM BETWEEN THE BASELINE CD4 CELL COUNT AND THE WHO STAGE
TO THE PRELIMINARY MULTIVARIABLE MODEL (n=1,282) .......................... 103
TABLE 4.22 COLLINEARITY BETWEEN COVARIATES IN THE FINAL MODEL (n=1,282).. 104
TABLE 4.23 TEST OF THE PROPORTIONAL HAZARDS ASSUMPTION OF THE
PRELIMINARY MULTIVARIABLE MODEL (n=1,282) ........................................ 104
TABLE 4.24 EXAMINING EXTREME VALUES FOR POTENTIALLY INFLUENTIAL
VARIABLES (n=1,282)....................................................................................... 106
TABLE 4.25a HIGHEST DIFFERENCES IN PARAMETER ESTIMATES OF VARIABLES
INCLUDED IN THE PRELIMINARY MODEL, APPROXIMATE DELTA-BETAS
FOR THE BASELINE CD4 CELL COUNT (n=1,282) ......................................... 106
TABLE 4.25b HIGHEST DIFFERENCES IN PARAMETER ESTIMATES OF VARIABLES
INCLUDED IN THE PRELIMINARY MODEL, APPROXIMATE DELTA-BETAS
FOR THE WHO STAGE (n=1,282) .................................................................... 107
TABLE 4.26a COEFFICIENTS OF VARIABLES IN FINAL MULTIVARIABLE MODEL FOR
PREDICTORS OF IMMUNOLOGIC FAILURE AMONG PATIENTS ON ANTI-
RETROVIRAL THERAPY (n=1,282) .................................................................. 108
 x

TABLE 4.26b HAZARD RATIO OF VARIABLES IN FINAL MULTIVARIABLE MODEL FOR

PREDICTORS OF IMMUNOLOGIC FAILURE AMONG PATIENTS ON ANTI-
RETROVIRAL THERAPY (n=1,282) .................................................................. 108
TABLE 5.1 IMMUNOLOGIC FAILURE RISK ASSESSMENT AND MANAGEMENT TOOL..116
 xi

LIST OF FIGURES

Figure 2.1 Geographic distribution of HIV: estimated HIV prevalence by country .................. 29
Figure 2.2 Geographic distribution of HIV-1 subtypes and recombinant forms ...................... 30
Figure 2.3 Structure of HIV-1, including envelope structures (gp120 envelope, gp41), genetic
material (RNA), enzyme, inner membrane (matrix), and core protein (capsid) ..... 32
Figure 2.4 Life cycle of HIV ................................................................................................... 32
Figure 2.5 Antibody responses to HIV infection .................................................................... 34
Figure 2.6 HIV testing algorithm in Ethiopia .......................................................................... 36
Figure 2.7 Natural history of HIV infection in untreated individual: virological, immunological,
and clinical progression of typical progressors ..................................................... 37
Figure 2.8 Eligibility for antiretroviral therapy versus coverage, low- and middle-income
countries, by region, 2011 .................................................................................... 40
Figure 2.9 Box plot for changes in magnitude of CD4 cell count among the cohort of patients
on anti-retroviral treatment across months of retrospective follow-up over five years
............................................................................................................................ 49
Figure 4.1 Magnitude of baseline CD4 cell count distribution for the study sample (N=1321) 78
Figure 4.2 Pattern of CD4 cell count monitoring by follow-up time(N=1321) ......................... 82
Figure 4.3 Box plot for magnitude of baseline and follow-up CD4 cell count among cohort of
patients on anti-retroviral therapy by follow-up month (N=1,321) ......................... 86
Figure 4.4 Kaplan-Meier survival functions for survival free of immunologic treatment
failure(n=1,282).................................................................................................... 93
Figure 4.5 Kaplan-Meier survival functions by age group(n=1,282)....................................... 94
Figure 4.6 Kaplan-Meier survival functions by gender(n=1,282)............................................ 95
Figure 4.7 Kaplan-Meier survival functions by WHO Stage(n=1,282) .................................... 96
Figure 4.8 Kaplan-Meier survival functions by baseline CD4 cell count(n=1,282) ................. 97
Figure 4.9 Kaplan-Meier survival functions by diagnosis of tuberculosis disease(n=1,282) ... 98
Figure 4.10 Kaplan-Meier survival functions by adherence(n=1,282) ...................................... 98
Figure 4.11 Martingale residual vs baseline CD4 cell count(n=1,282) ................................... 102
Figure 4.12 Plot of the Grambsch, Therneau and Fleming smooth and fitted two-term (2, 3)
fractional polynomials model(n=1,282) ............................................................... 103
Figure 4.13 Graphs of the scaled Schoenfeld residuals and their Lowess smooth curves for the
WHO Stage(n=1,282) ........................................................................................ 104
Figure 4.14 Graphs of the scaled Schoenfeld residuals and their Lowess smooth curves for the
baseline CD4 cell count(n=1,282) ...................................................................... 105
Figure 4.15 Graph of the score residuals for baseline CD4 cell count(n=1,282) .................... 106
Figure 4.16 Cumulative hazard plot of the Cox-Snell residuals(n=1,282) .............................. 107
 xii

LIST OF ANNEXURES

Annexure A Data abstraction form .......................................................................................... 135

Annexure B UNISA ethics clearance letter ............................................................................. 138
Annexure C Application to conduct study sent to Ethiopian National Research Ethics Review
Committee ......................................................................................................... 139
Annexure D Ethics clearance letter from the Ethiopian National Research Ethics Review
Committee ......................................................................................................... 141
Annexure E Southern Nations Nationalities and Peoples‘ Regional State Health Bureau support
letter .................................................................................................................. 142
Annexure F Nigist Eleni Mohamod Memrial (Hosanna) hospital support letter ........................ 143
Annexure G Butajira hospital support letter ............................................................................ 144
 xiii

ABBREVATIONS

ABC Abacavir
AIDS Acquired Immunodeficiency Syndrome
ART Antiretroviral Therapy
ARV Antiretroviral
AZT Zidovudine
CD4 Cluster of differentiation 4
D4T Stavudine
DNA Deoxyribonucleic acid
EFV Efavirenz
FACS Fluorescent Activated Cell Sorter
FHAPCO Federal HIV/AIDS Prevention and Control Office
FMOH Federal Ministry of Health
FTC Emitricitabine
HAART Highly Active Antiretroviral Therapy
HIV Human Immunodeficiency Virus
INRUD-IAA International Network for the Rational Use of Drugs Initiative on
Adherence to Antiretrovirals
KHB Shanghai Kehua Bio-Engineering Co. Ltd
MDG Millennium Development Goal
MSF Médecins Sans Frontières
NGO Non-Governmental Organisation
NVP Nevirapine
LPV/r Ritonavir boosted lopinavir
PEPFAR President‘s Emergency Plan for AIDS Relief
RNA Ribonucleic acid
SNNPR Southern Nations Nationalities and Peoples‘ Region
TB Tuberculosis
TDF Tenofovir
UNAIDS Joint United Nations Programme on HIV/AIDS
UNISA University of South Africa
USA United States of America
WHO World Health Organization
3TC Lamivudine
 1

CHAPTER 1

ORIENTATION TO THE STUDY

1.1 INTRODUCTION

The Human Immunodeficiency Virus (HIV) is a retrovirus that infects people across all
ages and from all walks of life (Zuckerman, Banatvala, Griffiths, Schoub & Mortimer
2009:899). The most common means of transmission among adults is unprotected
sexual intercourse, while among children mother to child transmission accounts for
much of it (Zuckerman et al 2009:899).

The Joint United Nations Programme on HIV/AIDS (Acquired Immunodeficiency

Deficiency Syndrome) (UNAIDS) (2014:A35) further states that, in 2013, 35 million
people had already been infected with the HIV virus worldwide. The impact of the virus
is being mitigated by the use of highly active antiretroviral therapy also referred to as
HAART. Access to this therapy has been increasing and this is reflected in the number
of people taking antiretroviral medication. By 2013, the number of people on HAART
had already reached 12.9 million globally while it increased to 8.4 million in sub-
Saharan Africa (UNAIDS 2014:A67, A69).

In Ethiopia, access to life saving antiretroviral medication, also referred to by its

acronym, ARVs started on a massive scale in 2005, and by 2012 it had reached some
270,000 people according to Federal Ministry of Health (FMOH) and Federal HIV AIDS
Prevention and Control Office (FHAPCO & FMOH) (2012:33) and UNAIDS (2013b:A85,
A91). Treatment has been successful, as demonstrated by the improvement in the
survival rate in many settings in Ethiopia (Alemu & San Sebastian 2010:online; Tsegaye
&Worku2012:105). This has changed the course of the epidemic and made HIV a
chronic manageable disease.

As important as it is to get people who need the treatment to be on HAART, it is even

more important to ensure that patients adhere to the treatment so that they benefit from
it in the long run as well. This is because poor adherence to treatment is associated with
increased risk of drug resistance and thus increased risk of treatment failure (El-Khatib,
 2

Katzenstein, Marrone, Laher & Mohapi 2011:online; Jaka, Mshana, Liwa, Peck &
Kalluvya 2009:7). This statement is supported by Beyene, Gedif, Gebre-Mariam and
Engidawork (2009:1010) who further state that achieving adherence to treatment of
≥95% is not always possible for all people on HAART because of, for example, lack of
family support, unemployment, alcohol intake, poverty and the presence of other
conditions and diseases that could limit insight and knowledge. In this regard, Keiser,
Tweya, Braitstein, Dabis, MacPhail and Boulle (2010:254) report that the longer a
person is on a failing HAART regimen, the higher the mortality risk. For this reason,
monitoring HAART treatment response, especially in the immunological and virological,
aspects which are challenging issues in many resource limited chronic HIV care
settings, becomes very important. In addition, detection and management of treatment
failure could be compromised if the monitoring is not up to standard (Azzoni, Foulkes,
Liu, Li, Johnson, Smith, Kamarulzaman, Montaner, Mounzer, Saag, Cahn, Cesar,
Krolewiecki, Sanne & Montaner 2012:online; Bélec & Bonn 2011:1255)

This study, therefore, intends to investigate the level of immunologic monitoring, the
level of treatment response to antiretroviral medication as well as the prevalence of
treatment failure and its determinants in selected areas in Ethiopia with a focus on
immunologic criteria.

1.2 BACKGROUND TO THE RESEARCH PROBLEM

In this section, a brief background of HIV and HAART in the world with specific
reference to Ethiopia is given.

1.2.1 HIV infection, disease causation, and HAART

The HIV is a viral infection from the family Retroviridae and subfamily of Lentivirinae.
Among adults, it is predominantly spread by unprotected sexual intercourse among
heterosexuals. Other important means of transmission include unprotected sexual
intercourse between men, the sharing of sharp materials such as needles, and
exposure of mucous membranes or broken skin to bodily fluids for example blood and
pus. Among children, mother to child transmission (during pregnancy, delivery or breast
feeding) accounts for most of the transmission (Zuckerman et al 2009:899).
 3

The path to HIV disease progression is viral replication in great numbers. This is
inevitably accompanied by destruction of immune cells (especially cluster of
differentiation 4 cells, also known as CD4 cells) that weakens the body‘s defence
mechanisms and gradually leads to an escalation of susceptibility to various life
threatening opportunistic infections (Longo, Fauci, Kasper, Hauser, Jameson &
Loscalzo 2012:1510). The stage of the disease whereby an HIV infected person
experiences serious opportunistic infections is called Acquired Immunodeficiency
Syndrome (AIDS). It takes an average of 10-15 years for the disease to reach this stage
after infection. Once in this stage, even the short term mortality is extremely high (Longo
et al 2012:1500; Zuckerman et al 2009:910).

To mitigate the impact of this disease, the use of HAART is warranted (Longo et al
2012:1580). Antiretroviral therapy works by suppressing viral replication which leads to
halting and reversing of the destruction of immune blood cells. With this therapy, the
immune system starts to regain its ground and the frequency of opportunistic infections
declines (Longo et al 2012:1568). With further restoration of the immune system, it is
possible for affected individuals to regain their health status depending on other factors
like the timing of treatment initiation, presence of co-morbid medical conditions,
nutrition, psychosocial support, life style, and the quality of care. Achieving meaningful
control of the pandemic requires access to life saving HAART for all infected people
who need the treatment (Department of Health and Human Services [USA] 2013:K-1;
Longo et al 2012:1568).

1.2.2 HIV and AIDS globally

The number of people affected globally had reached 35.3 million in 2012. All nations are
affected but the most affected are the least developed and developing nations from sub-
Saharan Africa and South Eastern Asia. By 2012, 9.7 million people had begun
receiving antiretroviral therapy in sub-Saharan Africa alone (UNAIDS 2012:19, 24). The
global AIDS related mortality reached a peak in 2005 at 2.3 million yearly deaths. With
widespread access to HAART, this number had decreased to 1.5 million by 2013
(UNAIDS 2014:A95).

1.2.3 HIV and AIDS in sub-Saharan Africa

 4

Sub-Saharan Africa is the worst affected by the HIV/AIDS pandemic, accounting for two
thirds of all HIV-infected individuals in the world. From the latest information, it is
indicated that 70% of new HIV infection in the world for 2012 also came from the same
region (UNAIDS 2013b:A24, A27). Access to HAART has gradually increased, and by
2012, 68% of all HIV infected people in the region who needed the treatment urgently
had access to it. For this reason, mortality had also been decreasing, having declined to
1.1 million by 2013 from a peak of 1.8 million recorded in 2005 (UNAIDS 2013b:A85;
UNAIDS 2014:A56).

1.2.4 HIV and AIDS in Ethiopia

The prevalence of HIV in Ethiopia among adults age 15-49 is 1.2% (1.1-1.4%)
according to the latest estimates (UNAIDS 2014:A9). The estimated number of all
people with HIV was around 790,000 (720,000-890,000) for 2013. In the same year,
new HIV infections were estimated to be 21,000 (15,000-30,000). This number declined
from 51,000 in 2005 (UNAIDS 2014:A38). Access to HAART had then been ensured
through the President‘s Emergency Plan for AIDS Relief (PEPFAR) and Global Fund
programmes utilized throughout the country that started in 2005. This resulted in the
fact that 68% of the 400,000 people who need the treatment commenced with HAART
(UNAIDS 2013b:A85). Various studies in Ethiopia showed this roll out of treatment
improved survival of people with HIV/AIDS (Alemu & San Sebastian 2010:online;
Mulissa, Jerene & Lindtjørn 2010:online; Tsegaye & Worku2012:105; UNAIDS
2013b:A80).

1.2.5 The goal of HAART and adherence

Achieving the goal of HAART, that is sustained viral suppression to an undetectable

level, requires a high degree of adherence to antiretroviral medications (≥95%). Unless
this level of adherence is maintained throughout the therapy, treatment will not be
effective. In order to make sure that patients are adherent to treatment, it should be
monitored by responsible professionals and caregivers (FHAPCO & FMOH 2008a:82).

There are various types of adherence to HAART medication that need monitoring under
ideal conditions. These include adherence to medication refill, adherence to timing in
taking the medication, adherence to the prescribed dosing of medication, and so on
 5

(WHO 2003:118). There are various tools that can be used to monitor adherence: which
may include self-report, pill count, review of treatment or drug dispensary records,
electronic devices, and/or biochemical assays that measure the blood levels of ARV
drugs. In Ethiopia though, only the first two are available in routine clinical care settings
(FHAPCO & FMOH 2008a:130; Turner 2002:S143).

A number of studies conducted in Ethiopia have demonstrated variable levels of

adherence to HAART in various clinical settings. Azmeraw and Wasie (2012:120) report
that 80% of patients showed an optimal adherence (at ≥95%) among 340 children in
two referral hospitals in Northwest Ethiopia. Among adult patients, Giday and Shiferaw
(2010:191) report on the level of adherence of ≥95% as being 88% for patients
investigated for one month. Kebede and Wabe (2012:67) report a 95% adherence for
HAART among 24 tuberculosis (TB)/HIV co-infected patients in a hospital in Southwest
Ethiopia. In addition, in a study conducted in a university hospital in Southwest Ethiopia
which evaluated the timing of ARV intake in addition to adherence to doses, 303 of the
patients (95%) were adherent based on dose adherence during the week before the
actual interview. The same study also indicated the adherence including timing and food
intake was at 72.4% (Tiyou, Belachew, Alemseged & Biadgilign 2010:online). All the
above studies used self-report as the measure of adherence. A different set of studies
used other means to evaluate adherence besides self-report. One such study is a multi-
country survey of adherence in East Africa, in which Ethiopia was one of the countries.
This study revealed that the percentages of patients with optimal adherence to
antiretroviral medications based on self-reporting adherence, adherence to appointment
date, and pill count were 94%, 86%, and 76% respectively (Chalker, Andualem, Gitau,
Ntaganira, Obua, Tadeg, Waako, Ross-Degnan & International Network for the Rational
Use of Drugs Initiative on Adherence to Antiretrovirals (INRUD-IAA) 2010:online).

The above reported discrepancy between different measures of adherence is noted in

other studies as well. Beyene et al (2009:1010) report optimal adherence among 93% of
the study participants using a 15-day self-report, while it was 88% when using an
unannounced pill count. This shows that self-report as a measure of adherence can be
unreliable for some patients as they tend to overestimate adherence due to social
desirability or recall bias. A similar problem exists when using pill count as a measure of
adherence: conscious of being monitored, patients may dispose of pills before showing
 6

up for the clinic visit in order to make it look like they took it (Department of Health and
Human Services 2013:K-2; Turner 2002:S144; WHO 2003:118).

Authors in nearly all the referred to research studies mentioned above (Beyene et al
2009:1010; Department of Health and Human Services 2013:K-2; Turner 2002:S144;
WHO 2003:118) uniformly agree that the means or tools used in routine practice to
assess adherence in Ethiopia (self-report and pill count) are not sensitive indicators for
all types of adherence to ARV treatment. In practical terms, this means that a lot more
patients on HAART have poor or suboptimal adherence than reported in published
literature and that identifying all these patients is going to be a persistent challenge.

1.2.6 Monitoring HAART treatment response

If left unattended, the consequence of poor adherence to HAART is drug resistance and
treatment failure (El-Khatib et al 2011:online). Therefore, the best thing to do besides
assessing and promoting adherence to therapy is to monitor response to HAART and
identify patients who have treatment failure as soon as possible. This may also involve
the changing of the current antiretroviral treatment regimen to second line regimen.
Early intervention will preserve the effectiveness of second line ARVs and prevents
transmission of drug resistant strains to others (Jaka et al 2009:6).

The WHO (2010a:48) refers to three means used to monitor response to HAART,
namely clinically, immunologically, and virologically. The WHO (2010a:49) is of the
opinion that, successful treatment means suppressed viral replication, which is
measured by viral load monitoring. Suppressed viral replication eventually leads to
increase in CD4 cell count; and this is the basis for immunological monitoring. Tsegaye
and Worku (2012:105) state that after a period of time on HAART, the amount of CD4
cells gained is somewhat predictable: an increase by roughly 50 to 150 cells is expected
after six months on treatment, after which the rate of increase decreases slightly. The
increase in CD4 cell count eventually leads to improvement of the immune status and
this is reflected by decrease in the occurrence of opportunistic infections. This forms the
basis for clinical monitoring: the better the immunity, the lower the probability of a
person experiencing opportunistic infections for example tuberculosis, oral candidiasis,
toxoplasmosis, etc. The gold standard to monitor HAART treatment response is viral
load measurement (FHAPCO & FMOH 2008a:159). But since this is not available in
 7

routine care in Ethiopia, clinical and immunologic criteria are mostly used. Clinical
monitoring is recommended at every visit, while immunologic monitoring is
recommended at baseline and every six months for the first year and every year after
that, plus whenever the clinicians deem it necessary (FHAPCO & FMOH 2008a:159).
The same tools are used to identify treatment failure and effect the decision to change
the current regimen the patient is on.

In resource limited settings where viral load is not available for routine clinical practice,
the WHO (2010a:48) recommends using clinical and immunologic criteria. However,
immunologic criteria have poor sensitivity and positive predictive value in picking
virological failure. However, they have good negative predictive value and thus can be
used to rule out treatment failure (Keiser, MacPhail, Boulle, Wood, Schechter, Dabis,
Sprinz &Egger 2009:1221; Rawizza, Chaplin, Meloni, Eisen, Rao, Sankale´, Dieng-Sarr,
Agbaji, Onwujekwe, Gashau, Nkado, Ekong, Okonkwo, Murphy & Kanki 2011:1285;
Reynolds, Nakigozi, Newell, Ndyanabo, Galiwongo, Boaz, Quinn, Gray, Wawer &
Serwadda 2009:697). Clinical criteria are the least effective in terms of accuracy as well
as in terms of detecting treatment failure in time to remedy it. For this reason, the WHO
(2010a:48) recommends the utilisation of immunologic criteria supplemented by
targeted viral load testing when available in these settings to lessen any unnecessary
changing of the regimen of treatment.

1.3 STATEMENT OF THE RESEARCH PROBLEM

A number of studies in sub-Saharan Africa have assessed the prevalence of

immunologic failure and have come up with variable results. El-Khatib et al
(2011:online) report the prevalence of immunologic failure to be 19% among 456
patients followed for a median period of 15 months in a South African clinic. In Uganda,
after a median follow-up time of 22 months, immunologic failure was 11% among 1,133
patients (Reynolds et al 2009:697). In Nigeria, Rawizza et al (2011:1286) reported that
the prevalence of immunologic treatment failure after a median follow-up of 33.2 months
was 32.2% among 9,690 patients. In another study in Nigeria, though, the prevalence of
first line HAART failure was reported to be very low, 1.19% among 9,449 patients during
the study period (Eze & Uyagu 2009:46). Jaka et al (2009:6) report that, in one clinic in
Tanzania, the prevalence of immunologic treatment failure was 17.1% among 362
 8

patients followed for a mean duration of 29 months. In short, these studies show that
different settings have different levels of immunologic failures.

For Ethiopia, data on immunologic failure among patients on HAART is available only
for paediatric patients. Bacha, Tilahun and Worku (2012:online) report immunologic
failure of 8.2% in selected hospitals in Addis Ababa among 1,186 children with a mean
age of six years and mean follow-up period of 37 months. In another study in Jimma
university in Ethiopia, among a smaller cohort of paediatric patients (mean age six
years, mean follow-up duration 13.7 months), the percentage of immunologic failure
was 11.5% among 96 patients (Workneh, Girma & Woldie 2009:78). Not much similar
data could be found for adult patients despite an extensive search. However, one study
was reported that the prevalence of patients on second line HAART in selected health
facilities across Ethiopia, among 7,451 clients actively on follow-up at 24 months, was
2.13% (Assefa, Kiflie, Tesfaye, Haile-Mariam, Kloos, Edwin, Laga & Van Damme
2011:online). This is comparable to what is reported by WHO (2009:53), but much
higher than that in studies reported from 48 Médecins Sans Frontières (MSF)
supported programmes in resource limited countries (Pujades-Rodriguez, O‘Brien,
Humblet & Calmy 2008:1307).That study showed change to second-line regimens at a
rate of 4.1 per 1,000 person-years after 20 months of follow-up among 48,000 patients.

Because of difficulties in monitoring treatment response in many settings and difficulty in

the utilisation of available resources by health care workers, the actual prevalence of
treatment failure is believed to be higher (Azzoni et al 2012:online; Bélec & Bonn
2011:1251). Only one publication could be found that has assessed adherence to
immunologic monitoring of children with HIV in Ethiopia (Berhan 2011b:205) despite a
wide literature search for adult patients in Ethiopia indicating the level of adherence to
immunologic monitoring guidelines. Furthermore, the researcher could not find any
study indicating the prevalence and determinants of patients with immunological failure
that may benefit targeted viral load monitoring in the Ethiopian care setting. Information
obtained from this study is vital for clinicians that manage patients at facility level and for
decision makers at health system level for the scaling up and the proper utilisation of
immunological and viral load tests.

1.4 DEFINITIONS OF KEY CONCEPTS

 9

In this section the researcher outlines the definitions, descriptions and application of the
terms related to this study.

1.4.1 Human Immunodeficiency virus

According to the Macmillan Dictionary Online (2013, Sv ―HIV‖), the HIV is defined as a
virus that causes AIDS. Longo et al (2012:1506) define HIV more technically as a
retrovirus which has two major types, HIV-1 and HIV-2. It infects cells of the immune
system which leads to progressive damage to the immune system, which leads to
diminished defence mechanisms and increased susceptibility to various infections that
lead to ill health and death if left unattended (Zuckerman et al 2009:897, 904). For the
purpose of this study, HIV infection refers to that by HIV-1 because, in the Ethiopian
context, HIV-1 is the predominant type of the virus known to infect people (Longo et al
2012:1506).

1.4.2 Describing CD4 cell

The CD4 is a special type of immune cell referred to as a lymphocyte with a marker on
its surface called cluster of differentiation or CD4, that which is responsible for
coordinating the defence mechanisms of the body (FHAPCO & FMOH 2008a:36). The
term ‗CD4 cell count‘ refers to the number of CD4 cells within a micro litre of blood. A
machine called FACS (Fluorescent Activated Cell Sorter) Count is used to determine
CD4 cell count. In the Ethiopian context, the normal value in HIV uninfected people
ranges between 600-1,300 per micro litre, with a mean value closer to the lower end
(FHAPCO & FMOH 2008a:38). For HIV infected patients, this number decreases
significantly as the disease progresses to a more advanced stage (Longo et al
2012:1519).

1.4.3 Highly active antiretroviral therapy (HAART)

The Oxford Dictionary Online (2013, Sv ‖HAART‖) defines the term HAART as a drug
therapy (or regimen) for HIV infection containing at least three antiretroviral drugs. The
WHO (2013:15) refers to HAART as a combination of three or more antiretroviral drugs
given for life in order to effect the sustained suppression of the HIV virus. The
combination of antiretroviral medications used in Ethiopia as first line therapy includes
 10

one of tenofovir, zidovudine, or stavudine, plus lamivudine, plus either nevirapine or

efavirenz (FHAPCO & FMOH 2008a:122).

1.4.4 Adherence

According to the Macmillan Dictionary Online (2013, Sv ―adherence‖), the term

‗adherence‘ means ―accepting and following a prescribed programme‖. A patient or
client is said to have good adherence for antiretroviral therapy if one follows the
treatment regimen as prescribed in terms of dosage, timing of medication intake, and
timing of medication pick up from the treatment facility (Department of Health and
Human Services 2013:122). In this study, adherence to antiretroviral therapy was
assessed by the timing of medication pick up from the treating facility (FHAPCO &
FMOH 2008a:100; FHAPCO & FMOH 2008b:66).

1.4.5 Treatment failure

The Macmillan Dictionary Online (2013, Sv ―treatment‖) defines the term ―treatment as
medical care provided to improve a situation‖. In this study, antiretroviral therapy is the
treatment or medical care provided to improve the health of those with the disease as a
result of HIV infection. ‗Failure‘ is defined as a lack of success in accomplishing an
intended objective (Oxford Dictionaries Online 2013, Sv ―failure‖). The WHO (2010a:50)
defines antiretroviral treatment failure as ―the state of failure for HAART to suppress HIV
viral replication while it is being taken‖. The same definition was used in this study to
refer to treatment failure. This is further elaborated in the next section.

1.4.6 Antiretroviral treatment failure

The WHO (2010a:50, 51) recommends three criteria to define antiretroviral treatment
failure namely clinical, immunologic and virologic. Clinical failure for adults and
adolescents is when a patient develops a new or recurrent WHO HIV disease stage four
condition or WHO HIV stage three conditions for example pulmonary tuberculosis and a
severe bacterial infection. For children, this includes WHO HIV disease stage three or
four conditions as per the WHO classification, except TB. Immunologic failure is
diagnosed when there is a decline in the CD4 cell count by 50% from the peak level
achieved during treatment or when it falls to or below the baseline CD4 cell count level,
 11

or when it is persistently below 100 cells/micro litres. Virologic failure is when the
plasma viral level is more than 5,000 copies. All three definitions are used in the
Ethiopian context, but for the purpose of this study, only immunologic criteria were used
(FHAPCO & FMOH 2008a:155).

1.4.7 Prevalence

The term ‗prevalence‘ refers to the proportion of individuals having a characteristic

within a given population (Oxford Dictionary Online 2013, Sv ―prevalence‖). Webb and
Bain (2011:34) define prevalence as the ―proportion of behaviour or disease among a
particular population at a given time‖. In this study, the prevalence of immunologic
treatment failure among HIV patients on HAART was determined as a proportion
expressed as a per cent.

1.4.8 Patient or client

A patient is someone who is receiving medical care because of illness while a client is a
person who seeks professional help from for instance a lawyer or a doctor (Macmillan
Dictionary Online 2013, Sv ―patient‖). Glynn and Drake (2012:3) describe a ‗patient‘ as
a term derived from the Latin term meaning ‗sufferance‘. The term ‗patient‘ in this study
is regarded as being synonymous with the term ‗client‘, both designating a person who
has a confirmed diagnosis of HIV through laboratory tests and who has been prescribed
antiretroviral medication at the selected setting in Ethiopia used as a source of data in
this research.

1.4.9 Predict

The Macmillan Dictionary Online (2013 Sv “predict”) defines the term ‗prediction‘ as a
―statement about what will happen in the future or the process of making such a
statement‖. Polit and Beck (2012:738) provide a more technical definition of prediction,
stating that ―prediction is the use of objective data in forecasting a relationship between
variables of interest‖. Therefore, a predictor variable is a variable that has a measurable
influence on a different variable of interest. In this study, independent variables
including age, gender, baseline CD4 cell count and others were used to predict or
 12

forecast the occurrence of immunologic treatment failure in order to develop a care plan
for patients at risk of immunologic treatment failure.

1.5 OPERATIONAL DEFINITIONS

Polit and Beck (2012:736) define an operational definition as a selected meaning when
there are various meanings for a term.

In this research, clarification for the following terms relative to their specific use in the
study must be given. The terms are adherence to medication refill and immunological
treatment failure.

1.5.1 Adherence to medication refill

Antiretroviral therapy involves a lifelong treatment. This means that medications must
be picked up at regular time intervals from treatment facilities in order for treatment to
continue. A patient who has run out of medication but fails to collect more from a
treatment centre for one month is labelled as being ‗lost to follow-up‘. A patient who has
not collected medication for more than three consecutive months is labelled as a ‗drop
out‘ (FHAPCO & FMOH 2008a:130).

1.5.2 Immunological failure

Although the WHO (2010a:50, 51) recommends the use of three types of criteria with
which to define antiretroviral treatment failure, namely clinical, immunologic and
virologic, the focus of this study is on immunologic criteria which the WHO defines as
follows:

 A falling of follow-up CD4 cell count to baseline (or below), or

 A 50% fall from on-treatment peak value, or
 CD4 cell count levels persisting below 100 cells/mm3.

These same criteria were used in this study to identify patients with immunologic failure.
 13

1.6 RESEARCH AIM/PURPOSE

The aim of the study is to describe the prevalence of immunologic failure and identify its
determinants among people living with HIV receiving treatment at selected hospitals.

1.7 RESEARCH OBJECTIVES

The objectives of this study are to:

 Assess the pattern of immunologic (CD4 cell count) monitoring in selected

hospitals in southern Ethiopia
 Assess the pattern of immunologic (CD4 cell count) response for patients on
HAART in selected hospitals in southern Ethiopia
 Describe immunologic treatment failure and identify its predictors among patients
on HAART in selected hospitals in southern Ethiopia

1.8 RESEARCH QUESTIONS/HYPOTHESES

The research questions posed in this study are as follows:

 What is the pattern of immunological monitoring in the selected hospitals in

southern Ethiopia?
 What is the level of CD4 cell count response of patients on HAART in the
selected hospitals in southern Ethiopia?
 What do the results reveal with respect to the extent of and factors characterising
immunologic treatment failure among patients on HAART treatment in selected
hospitals in southern Ethiopia?

1.9 PARADIGM

In the quantitative research design, reality exists independent of the observer. A number
of measures are taken to make sure the observation of phenomena is objective and free
of bias, unaffected by the beliefs of the observer. Values are assigned to variables by
counting or measuring with an instrument (as for example, a checklist, questionnaire, or
structured interview schedule) in an objective manner (Polit & Beck 2012:13). This
 14

design is appropriate for this study because the aspect of the dependant variable being
studied; that is, immunologic treatment failure is determined by examining and
interpreting the patterns of quantitative measures of CD4 lymphocyte counts (WHO
2010a:50, 51).

1.10 RESEARCH DESIGN

This study has a quantitative, correlational and longitudinal design. Since this is a
quantitative, non-experimental study, the researcher simply observes what is going on
without interference or manipulation. In some cases, manipulation would not be
possible, as in the case of gender. In other cases, manipulation is possible but wouldn‘t
be ethical, as in the case of providing a beneficial supplement for one group and
withholding it from another just to study the additional effects it may have (Polit & Beck
2012:227). Exposure and outcome status is determined when the condition of interest is
experienced by the study subjects. Nor is there randomization of study subjects into
exposure and control group. That is exactly what is happening in this study. Patients‘
information gathered in their natural environment during routine clinic visits is used to
evaluate prevalence and determinants of immunologic failure. It would be unethical to
manipulate who gets what sort of treatment by allocating different intervention
categories to various patients just to satisfy one‘s curiosity about who is likely to
experience treatment failure. So this design is appropriate for this study (LoBiondo-
Wood & Haber 2010:310). The next two items (see 1.10.1 and 1.10.2) focus on the
elements of this design.

1.10.1 Correlational elements

Polit and Beck (2012:226) state that in the correlational study design, the association
between variables which the researcher cannot manipulate is determined. In addition to
assessing the mere association between variables, this type of design addresses the
issue of whether one variable (a dependant variable) is predicted by another (an
independent variable). In this study, the major aim was to assess the prevalence of
immunologic treatment failure and factors that predict the development of immunologic
treatment failure. With respect to the second aim, independent variables (for example
age, gender, and poor adherence history) that may be correlated to immunologic
treatment failure are studied, while in the first, features of patients with immunologic
 15

failure are described and hence, this study has a descriptive component as well. In a
descriptive research design, a health event of interest is simply described (LoBiondo-
Wood & Haber 2010:200,201).

1.10.2 Longitudinal elements

In a longitudinal study design, a group of subjects with common characteristic of interest

are followed for some length of time. The outcome of interest is measured repeatedly as
the follow-up continues (Webb & Bain 2011:100). A longitudinal study design can have
two forms: prospective and retrospective. In the prospective longitudinal design,
subjects are followed for a time, and observations are made for the occurrence of
outcomes of interest, while in the retrospective longitudinal design, both the exposure
and outcome of interest have already occurred in the past, and the exposure precedes
outcome (Kestenbaum 2009:37).

In this study, a group of patients on HAART were followed and the outcome of interest,
i.e. immunologic treatment failure, was evaluated at various times. Clearly, patients take
the treatment for a while before they experience treatment failure. So, there is a timed
sequence of events. All these features make a longitudinal design appropriate for this
study. Both exposure and outcome status were determined at the start of the study and
hence that makes it a retrospective study (Chernick 2011:12). This design is particularly
advantageous for this study. This is because, since the specific objective includes
assessment of the pattern of immunological monitoring and whether immunologic failure
was dealt with appropriately and in a timely fashion, health care workers may modify
their behaviour and exercise greater vigilance than they normally do in response to their
awareness of a prospective study being done addressing these issues (Polit & Beck
2012:225).

1.11 SAMPLING

Sampling is the process of selecting part of the accessible population with the intention
of studying the selected cases and making inferences about the population from which it
is drawn. Taking a sample and studying it has a number of advantages. This includes
saving time and minimizing the cost associated with collecting data from a large
population (LoBiondo-Wood & Haber 2010:224).
 16

1.11.1 Site sampling

In this study, a total of 22 hospitals made up the site population. Twenty were under
government control and two administered by non-governmental organizations
(NGO).There are also 110 antiretroviral treatment providing health centres in the
Southern Nations Nationalities and Peoples‘ Region (SNNPR).All 22 of the hospitals
formed the site target population and hence the site sampling frame. Two of them have
been accessible and convenient for the investigator to draw data from. In this study,
these hospitals will be referred to as Hospital A and Hospital B. Convenience sampling
is chosen as these facilities have been accessible to the investigator. These facilities
are accessible to the investigator since they are found in the place where the
investigator lives (Hospital A) and works (Hospital B).

1.11.2 Data source sampling

The population of clients‘ on HAART in the two hospitals at the time of sampling was as
follows: 881 in hospital A and 1,266 in hospital B. Documents of all patients enrolled for
HAART in the selected facilities formed the site population.

1.11.2.1 Eligibility criteria

The eligibility criteria are the ones that specify which subgroup of the population is to be
studied (Polit & Beck 2012:274). For this study, the eligibility criteria were as follows
listed as inclusion and exclusion criteria.

Inclusion criteria

 The documents of people living with HIV on antiretroviral therapy enrolled and
started on ART between September 1, 2005 until December 31, 2012
 Age ≥5 years old (because for patients of an age lower than this immunologic
monitoring is rare as a result of the difficulty in taking blood samples)
 Patients who had two or more CD4 cell count tests

Exclusion criteria
 17

 Having taken HAART for less than six months

 Being a case transferred from a facility other than the study facilities for
continuation of treatment
 Being a case that does not fall within the time frame of the study

All patients‘ documents fulfilling the eligibility criteria in the study in the two hospitals
were accessible; and hence, formed the accessible population. The sample was
selected through a sampling process. It was important for the researcher to select a
representative sample from the population of each hospital‘s ART clinic, a sample which
accurately and directly represented the characteristics as reflected by the population
(Polit & Beck 2012:283).For the purposes of this study, the sample was selected from
the total population of clients‘ from hospitals A and B whose documents complied with
the eligibility criteria for this study.

1.11.2.2 Respondent/participant sampling technique

Simple random sampling of all accessible patients fulfilling the eligibility criteria was
used to determine the required study sample.

1.11.2.3 Respondent/participant sample size

The total sample size required was estimated for the combined group of accessible
patients on HAART treatment in the two hospitals. Sample size calculations were
conducted assuming random sampling for a finite population based on the estimation of
a 50% proportion, at a 5% significance level and 2.5% precision as described by Daniel
(2009:192). The formula used to estimate the sample size for the study was:

n= [z2p(1-p)N]/[(N-1)d2+z2p(1-p)]

Where
n = calculated sample size
N = population size
z = critical value at the chosen significance level
p = proportion to be estimated
 18

d = precision

Thus, for this study taking N=2,147, z=1.96, p=0.50 (50%) and d=0.02 (2 %), sample
size was thus 1,133.703 which is rounded off to 1,134. Assuming the proportion of
missing records to be 15%, the sample size was inflated to 1,303. This means after
applying the eligibility criteria, of the remaining patients‘ files, 1,303 patients‘ files were
randomly selected. In other words, proportionate sampling was used (Daniel 2009:13).

Permission was obtained from the management of hospitals under study, and the ethics
committee of the Regional Health Bureau of The Southern Nations Nationalities and
Peoples‘ Region and the National Research Ethics Review Committee of Ethiopia (see
Annexure B to F) to access the list of all clients that were started on. A list was compiled
indicating the patients‘ file number. The statistician then applied the eligibility criteria to
assess which patient records on the list fulfilled the eligibility criteria and deleted those
who did not fulfil it.

1.12 DATA COLLECTION

This study utilised a self-designed document analysis checklist. Personal and social
information was not collected from human data sources.

1.12.1 The data collection instrument

The data collection instrument used in this study is provided in Annexure A.

1.12.2 Administering the data collection instrument

The existing medical records were used as a data source. The advantage of using
existing records was to save time and reduce cost than if data was collected
prospectively. In addition, since the researcher was familiar with the records and the
data collection procedures, appropriate precautions were taken where the records and
the data entered had limitations (LoBiondo-Wood & Haber 2010:270, 279).

1.12.3 Pretesting the data collection tool

 19

The data collection document checklist and data cleaning program codes were
pretested prior to data collection and cleaning. This was done by a statistician among
transferred-in patients found in one of the study facilities. The purpose of the pre-test
was to assess the adequacy of the program codes, the length of time required to do the
data collection, and cleaning, and to make sure the clarity of the procedures was
adequate (LoBiondo-Wood & Haber 2010:280). A full description of this process is given
in chapter 3.

1.12.4 Data collection procedure

Trained data clerks at each facility provided the statistician with electronic version of
patients‘ records onsite, where he applied eligibility criteria and identified charts that
were eligible for inclusion in the study. This list was passed to data collectors who
abstracted patients‘ records then after.

The data collection instrument utilised to collect the information was a self-designed
document checklist which consisted of the following four sections:

Section A included the age, and gender. Such demographic variables were used to
characterise the study population.

Section B included clinical characteristics in the form of baseline CD4 cell count,
baseline WHO stage, and other variables that might have affected treatment failure,
such as, occurrence of tuberculosis, a history of treatment interruption, and weight
status. All these variables were used to characterise immunologic monitoring,
immunologic response, and immunologic treatment failure.

Section C included immunologic information like timing of immunologic (CD4 cell count)
monitoring, magnitude of CD4 cell count, and occurrence of immunologic failure.

1.13 DATA MANAGEMENT AND ANALYSIS

In this study, Epi InfoTM 3.5.1 statistical software (United States Centers for Disease
Control and Prevention 2007) and Stata/IC 12.0 (StataCorp LP) were used to undertake
both descriptive and inferential statistics.
 20

The purpose of descriptive statistics is to summarize and organise raw data so as to

understand the information contained in it in a simple way (Daniel 2009:20). A number
of descriptive statistical measures were used in this study to summarise data. These
included frequency distribution depicted in the form of tables or graphs used for
categorical variables. Data from section C was used to address the following objectives:

 Describe the expected pattern of immunological monitoring as well as the

observed immunological (CD4 cell count) monitoring (section C.1). This was
shown graphically.
 Describe the pattern of CD4 cell count response after HAART initiation (section
C.2).
 The prevalence of immunologic treatment failure was calculated as proportion
with 95% confidence interval (section C.3). Mean and standard deviation were
used as summary statistics whenever the variable was normally distributed or
median and inter quartile range when it was not. These are the so called
univariable statistics that are used to describe a single variable.

Inferential statistics must go one step further from just describing and summarising
variables and compares two or more variables in order to assess similarity or difference
amongst them in order to reach a conclusion (LoBiondo-Wood & Haber 2010:310,318).
In this study, variables from sections A and B of the checklist were used to characterise
pattern of immunologic monitoring, immunologic response, and to characterise patients
with immunologic treatment failure as compared to those without immunologic treatment
failure. A Chi- square test was used to compare variables in the nominal or ordinal
scale, while Student‘s t-test was used to compare the mean of continuous variables if
they were normally distributed; otherwise non-parametric tests were used.

Time to event, that is time from HAART initiation to immunologic treatment failure, was
determined using survival analysis as described in Chernick (2011:164). The use of this
technique was justified for two reasons. Firstly, immunologic treatment failure is a time
dependent variable. This means the risk of immunologic treatment failure increased with
time as the exposure to antiretroviral medications increased. Secondly, it was expected
that patients would have different length of follow-up duration. In addition, the variation
in follow-up duration was accounted for in this technique. Factors thought to affect the
 21

probability of developing immunologic treatment failure were studied by applying a

multivariable regression model using the Cox proportional hazards model as explained
in Daniel (2009:648,659).

1.14 DATA AND DESIGN QUALITY

This section deals with data and design quality as it applies to this study.

1.14.1 Quality of research design

In this section the quality of the research design is discussed, in terms of internal and
external validity.

1.14.1.1 Internal validity

Internal validity refers to the extent to which it is possible to draw a conclusion from the
study for the fact that there is a true causal relationship between an independent and
dependant variables, independent of the influence of other variables (Webb & Bain
2011:230).

Threats to the internal validity include the absence of temporality, the presence of
historical events that might have influenced one group only, maturation or fatigue,
attrition, and selection bias. In this study, since it was a longitudinal study, the
investigator evaluated whether predictor variables occurred prior to the occurrence of
the outcome of interest (Polit & Beck 2012:495).

The survival analysis carried out involved the analysis of choice and therefore, by its
being conducted, the researcher took into consideration statistically the effects of
variable duration of follow-up. Selection bias is a common threat to internal validity but
in this study, since cases were randomly selected from the selected facilities and the
Cox regression analysis was used, the investigator was able to provide a control for the
difference in baseline characteristics (Polit & Beck 2012:228,450).

1.14.1.2 External validity

 22

External validity refers to the generalizability of the findings of the study concerning the
studied population to similar groups of people outside the study. Generalisation to the
accessible population in this case would be straightforward as there was simple random
sampling of all eligible cases such that the selection took place without any bias. But,
generalization to the target population should be made carefully as convenience
sampling was used to select study sites (Polit & Beck 2012:250).

1.14.2 Quality of the data gathering instrument

In this section, the quality of the data gathering instrument is described in terms of its
reliability and validity.

1.14.2.1 Reliability and validity

Polit and Beck (2012:332) state that ―reliability refers to the accuracy and consistency of
information obtained in a study.‖ Since this study was retrospective and based on
existing medical records, a number of potential measurement biases were essentially
avoided. These include interviewer bias, recall bias, and the Hawthorn effect. One
particular variable that needed to be measured and might have had some influence on
the outcome variable, that is immunologic treatment failure, was adherence. Self-
reported adherence could have been unreliable because of a social desirability bias. For
this reason, the investigator measured patients‘ adherence to their appointment for a
drug refill. This measure of adherence also prevented misclassification errors with
respect to adherence that could have occurred if one relied on self-report to assess
adherence. (LoBiondo-Wood & Haber 2010:279)

1.15 ETHICAL CONSIDERATIONS

This section addresses ethical considerations as applied to this study.

1.15.1 The institution/site

After getting approval for the proposal from the Department of Higher Degrees of the
Department of Health Studies at The University of South Africa (UNISA), ethical
approval was sought from the National Research Ethics Review Committee, and the
 23

Southern Nations Nationalities and People‘s Regional Health Bureau of Ethiopia.

Permission was also obtained from the directors of each of the hospitals from whose
HIV clinics data was going to be used for the purpose of this study. (See Annexures D,
E, F and G)

1.15.2 Scientific integrity of the research

This is a quantitative study and such studies are planned from beginning to end. This
helped in ensuring that the standards used at various levels were visible and accounted
for in the research process. Preparing a study is only half of the job, though, and making
sure that everything is implemented as planned is equally important. Guidelines at every
level were followed from start to finish and the investigator faithfully upheld the desired
standards at various levels. Whenever that was not possible for operational reasons,
the researcher provided alternatives to compensate for the deficit, with the intention of
openly discussing such during the reporting of results about the limitations of the study.

1.15.3 Ethical principles

In this section, the three basic ethical principles and how they relate to this study is
discussed.

1.15.3.1 Beneficence

According to LoBiondo-Wood and Haber (2010:250), research and the research

process must maximise benefit (beneficence) and minimise harm (non-malfeasance) at
all times. In this study, there was no interaction of the researcher with study subjects.
Therefore, the possibility of harm being inflicted during the study was remote.

1.15.3.2 Respect for human dignity

LoBiondo-Wood and Haber (2010:250) state that this principle entails ensuring that the
researcher will respect the confidence of research subjects, and recognize that they
have the right to decide for themselves. This is ensured through the implementation of
informed consent whenever it applied. In this study, in fact, since there was no
 24

interaction between study subjects and investigator or data collector, there was no need
for informed consent.

1.15.3.3 Justice

LoBiondo-Wood and Haber (2010:251) state that this principle deals with the right to fair
treatment and the right to privacy. In this study the privacy of patients was respected in
that information about them was kept in confidence. Access to electronic records and
computer sat study sites was limited to the trained data clerks already working in the
study facilities. The computer used to update the electronic records was password
locked and medical records were kept in a medical records room which is secured
under normal circumstances and this security was maintained throughout the study.
Since random selection of all eligible patients was to be used, it was ensured that all
eligible would have an equal chance to be involved in the study, and thus, equality was
taken into consideration as well.

1.16 SIGNIFICANCE OF THE STUDY

This study may help decision makers in Ethiopia to identify gaps in the quality of care
pertaining especially to immunologic monitoring. In addition it may provide valuable
information in providing an estimate of patients who experience immunologic failure and
need targeted viral load testing. This will facilitate the estimating of the potential number
of people who may need second-line therapy in years to come.

For health care providers at the point of care, the study will prove a help in addressing
operational issues relevant to monitoring immunologic response. The analysis of
specific risk factors for immunologic failure will also help them to identify and pay
attention to those patients who are in greater risk of immunologic failure and addressing
the barriers to their adherence.

For people living with HIV in the study area, and beyond, the information obtained from
this study could be beneficial to improve the quality of care and ultimately contribute to
the betterment of their health status.
 25

1.17 SCOPE AND LIMITATIONS

The scope of this study is limited in many ways. Firstly, the study focused on patients in
antiretroviral care in the Southern Nations Nationalities and People‘s Region and no
other regions in Ethiopia. Secondly, the study was conducted among patients in
selected hospitals and not health centres. Thirdly, existing client records were used as
the data source with all the limitations such may have, including human error in
reporting, the omitting of patient information, and the lack of other pertinent information.

1.18 STRUCTURE OF THE DISSERTATION

This dissertation is divided into the following chapters:

Chapter 1: Orientation to the study

Chapter 2: Literature review
Chapter 3: Research methodology
Chapter 4: Presentation and discussion of the results of the research
Chapter 5: Conclusions, limitations and recommandations.

1.19 CONCLUSION

This chapter is an orientation to the study. Separate sections have described the
background to the study, the research problem, key concepts, and objectives of the
study. This was followed by the research methodology which specified the research
design, sampling, data collection, and data analysis plan to be used in order to achieve
these objectives. The final section identified possible ethical issues and limitations that
might arise in the course of the study and outlined the means employed to handle them.

All in all, it is the researcher‘s strong belief that the findings of this study might fill in a
knowledge gap which now exists with respect to the management of HIV patients on
antiretroviral therapy. The health care system might benefit specifically from the
estimate provided of the prevalence of immunologic treatment failure among adults and
children and factors associated with it.
 26

The next chapter will deal with the review of the relevant literature which formed the
basis for this study.
 27

CHAPTER 2

LITERATURE REVIEW

2.1 INTRODUCTION

The previous chapter briefly addressed several issues related to this study. Initially it
addressed the background to the study, the origin of the problem, and the problem
statement. Next it addressed the specific objectives of the study. This was later followed
by the methodology in brief to be used so as to answer the study question in ethically
and scientifically sound ways. The major components of this section included the
sampling plan, the data collection tool, the data cleaning plan, the data analysis plan,
and ethical issues relevant to the study.

This chapter commences by outlining the purpose of literature review. This will be
followed by the actual review.

2.2 PURPOSE OF THE LITERATURE REVIEW

According to Polit and Beck (2012:57), the purpose of a literature review is to familiarize
oneself with existing knowledge. That will serve two main purposes: the identification of
existing knowledge gaps or problems and the identification of existing solutions to the
problem. A literature review can further be used to outline research directions in terms
of: refining research questions, selecting an appropriate study design to answer the
research question, developing or using an existing data collection and analysis plan,
creating an outline to serve as the basis for interpretation of the research findings, and
the exploration of improved means to finding a solution to a problem.

For those who will ultimately use the research, a literature review can also be used to
enhance evidence based practice in designing a policy or choosing an appropriate
policy (Polit & Beck 2012:58). In this study, the researcher conducted an extensive
review of the literature and this will delineate the existing knowledge in relation to
HIV/AIDS and ART globally with special reference to Ethiopia with respect to
immunologic treatment failure, and factors associated with it. Also the pattern of
 28

immunologic response among patients on treatment and the practices used in

immunologic monitoring in different settings were reviewed. In order to achieve a
comprehensive review of the most recent and relevant literature, various research
engines were used to obtain research articles; at the same time, international and local
HIV treatment guidelines were reviewed in order to get an idea of existing standards
and practices.

2.3 CLASSIFICATION AND HISTORY OF THE HIV VIRUS

Zuckerman et al (2009:869) state that HIV is a virus from the family Retroviridae and
subfamily Lentivirinae. They further elaborate that this family of viruses is unique
because, in the various type of it, genetic material is encoded with ribonucleic acid
(RNA) instead of deoxyribonucleic acid (DNA). The HIV is further classified into two
subtypes, HIV-1 and HIV-2. HIV-1 has well defined subgroups designated M, N, O, and
P while HIV-2 has subgroups designated A to H (Longo et al 2012:1506).

Historically, HIV/AIDS was first reported in the United States of America among
homosexual men who exhibited Kaposi‘s Sarcoma and Pneumocystis Carinii
Pneumonia (Weiss 2008:201). Later, it was observed among injection drug users, sex
workers, and blood recipients, which led to the thinking that this disease was probably
transmitted by an infectious agent. This was followed by the isolation of HIV-1 and HIV-
2 in 1983 and 1986 (Barré-Sinoussi, Chermann, Rey, Nugeyre, Chamaret, Gruest,
Dauguet, Axler-Blin, Vézinet-Brun, Rouzioux, Rozenbaum & Montagnier 1983:869;
Clavel, Mansinho, Chamaret, Guetard, Favier, Nina, Santos-Ferreira, Champalimaud &
Montagnier 1987:1180). Even though HIV was first isolated from human samples,
Heeney, Dalgleish, and Weiss (2006:463) state that HIV has a zoonotic origin with HIV-
1 coming from chimpanzees while HIV-2 developed in Sooty Mangabey monkeys in
West Africa.

2.4 EPIDEMIOLOGY OF HIV/AIDS

This section addresses the epidemiology of HIV with particular attention to the risk
factors and its distribution.
 29

2.4.1 Transmission and risk factors

Zuckeman et al (2009:899) state that unprotected sexual intercourse, vertical

transmission, and injection drug use continue to constitute the major risk factors for the
transmission of HIV. Even though almost everyone is at risk, the WHO (2013:14)
identifies men who have sex with men, sex workers, and injection drug users as those
most at risk for HIV infection. Next in the tier of vulnerability to infection are adolescents,
especially girls, orphans, street children, prisoners, people with disabilities and migrant
and mobile workers. In general, the higher the level of the virus in the blood or body
fluids of a source case, the higher the risk of transmission (FHAPCO & FMOH
2008a:39).

2.4.2 Prevalence and geographic distribution

(UNAIDS 2013a:online)

Figure 2.1 Geographic distribution of HIV: estimated HIV prevalence by country

 30

(Longo et al 2012:1513)
Figure 2.2 Geographic distribution of HIV-1 subtypes and recombinant forms

As can be seen from figures 1 and 2, the HIV pandemic affects sub-Saharan Africa the
most even though no region of the world is spared. The geographic distribution of the
various subtypes of HIV-1 also indicate different predilections in various regions, as for
example, subtype B for western countries, while it is subtype C for eastern Africa
(Longo et al 2012:1513). The UNAIDS (2013a:online) estimated the global adult HIV
prevalence in 2012 to be 0.8% (range: 0.7 to 0.9%) from which it can be projected there
were then 35.3 million people with HIV. It further estimated the number of new infections
and deaths in 2012 among adults and children to be 2.3 and 1.6 million respectively.
The global trend for both new infections and mortality is a decreasing one.

2.4.3 The state of the epidemic in Ethiopia

 31

The prevalence of HIV in Ethiopia among adults age 15-49 is 1.2% (1.1-1.4%)
according to the latest estimates (UNAIDS 2014:A9). The estimated number of all
people with HIV was around 790,000 (720,000-890,000) for 2013. In the same year,
new HIV infections were estimated to be 21,000 (15,000-30,000). This number declined
from 51,000 in 2005 (UNAIDS 2014:A38). The annual HIV mortality rate is declining as
evidenced by the fact that there were 45,000 deaths in 2013 as opposed to 120,000 in
2005 (UNAIDS 2014:A57). As can be seen from figure 2.2, the subtype C of HIV-1 is
responsible for the epidemic in Ethiopia.

2.4.3.1 The state of the epidemic in Southern Nations Nationalities and People’s
Region of Ethiopia

The Ethiopian Health and Nutrition Research Institute (EHNRI) and FMOH (2012:36)
estimate that in the Southern Nations Nationalities and People‘s Region, the prevalence
of HIV from the latest estimates for adults was 0.7%. This puts the estimated number of
people with HIV at 97,236. Thus, prevalence in the study area is well below the national
prevalence of 1.2% (EHNRI & FMOH 2012:6).

2.5 PATHOGENESIS OF HIV/AIDS

This section discusses how HIV causes AIDS, and what makes it difficult to treat and
cure.

2.5.1 HIV infection and replication

To gain entry into cells the virus attaches its surface structure, gp120 (see figure 2.3), to
CD4 receptors found mainly on CD4 T lymphocytes. Once attachment is established, it
fuses its membrane with that of the cell and releases its RNA and enzymes to the inside
of the cell (see figure 2.4). This is followed by the enzyme reverse transcriptase making
a DNA copy of the viral RNA. The new viral DNA is integrated within the host DNA of
the infected CD4 cell. Then, as the cell carries out its routine chores, new viral
components are produced. Once produced, the different components are assembled
together and new viruses are released (FHAPCO & FMOH 2008a:36).
 32

(Longo et al 2012:1513)

Figure 2.3 Structure of HIV-1, including envelope structures (gp120 envelope, gp41),
genetic material (RNA), enzyme, inner membrane (matrix), and core
protein (capsid)

(FHAPCO & FMOH 2008a:36)

Figure 2.4 Life cycle of HIV
 33

This process repeats itself time and time again, producing hundreds of millions of
viruses per day, each cycle producing a genetically different copy as a result of
mutations and the recombination of viral genetic material. This continuous high
replenishment of viral particles creates the opportunity for adaptation to selective
pressure from the immune system or antiretroviral drugs and drug resistance
(Zuckerman et al 2009:904).

2.5.2 Immune activation and CD4 T-cell depletion

The availability of the viral antigen in the body activates the immune system, which
leads to the creation of activated CD4 cells (Grossman, Meier-Schellersheim, Paul &
Picker 2006:290; Sodora & Silvestri 2008:440). Activated CD4 cells are further infected
and lead to more viral particles being released in the body. This positive feedback
process leads to chronic activation of the immune system, which in turn results in the
depletion of CD4 cells and the dysfunction of the immune system. The CD4 cells are
depleted because of the destructive effect of the virus directly or indirectly through the
cell death of infected cells programmed by the immune system (Levy 2009:151). Since
CD4 T cells play a central role in the defence mechanism of the body, their depletion
coupled with the dysfunction of the immune system leads to enhanced susceptibility of
affected individuals to opportunistic infections and malignancies which ultimately lead to
ill health and to death in those who are not treated (FHAPCO & FMOH 2008a:40).

2.5.3 Reservoirs of HIV infection

A small portion of infected CD4 cells survive and those cells are converted into resting
or memory cells that lie dormant for a long time, up to ten years (Zuckerman et al
2009:904). Even in patients on antiretroviral medication, HIV in such cells is not affected
as long as the cells remain dormant (Chun & Fauci 2012:1262). However, when such
cells get activated and start to function, full viral replication capability resumes. It is this
type of immune cells that serves as a reservoir of HIV infection. This forms the Achilles
heel for any attempt to cure HIV (Chun & Fauci 2012:1262; Levy 2009:150; Katlama,
Deeks, Autran, Martinez-Picado, Van Lunzen, Rouzioux, Miller, Vella, Schmitz, Ahlers,
Richman & Sekaly 2013:2110). What is even more important is the fact that the genetic
material of any dominant virus is continuously archived in these cells including drug
resistant strains, which updates and maintains HIV latency.
 34

2.6 IMMUNE RESPONSES TO HIV INFECTION

As with any infection, the immune system reacts when it detects a foreign antigen in the
body (see figure 2.5). In the case of HIV, this results in a cell-mediated and humoral
(antibody) mediated response (Levy 2009:153). The antibody response is rather
effective, especially with the neutralizing antibodies because they target the more stable
surface proteins, but it comes months after the infection occurs, by which time, usually
new variants of the virus are present, as a result of which, the virus manages to
continue to reproduce, circumventing the body‘s defence mechanism (Longo et al
2012:1534).

(Longo et al 2012:1534)
Figure 2.5 Antibody responses to HIV infection

2.7 LABORATORY DIAGNOSIS OF HIV INFECTION

The FMOH (2010b:95) states that the diagnosis of HIV is made through detection of
HIV specific antibodies in the body for anyone older than 18 months. Rapid antibody
tests are employed to serve this purpose. According to the WHO (2012b:35), the use of
 35

rapid tests instead of laboratory based tests (enzyme immune assays) makes it possible
for HIV counselling and testing services to be accessible at point of care in health
facilities. A serial testing algorithm is used in order to make sure that a person‘s HIV
status is determined accurately (WHO 2012a:36). There are various combinations of
tests in the world, but the testing algorithm used in Ethiopia at present is displayed in
figure 2.6 (FMOH 2010b:98).

The HIV (1/2) Antibody Colloidal Gold test from Shanghai Kehua Bio-Engineering (also
known as KHB test), which has high sensitivity, is used initially to screen patients. If the
result is negative, then the person is HIV negative. If positive, a confirmatory test is
carried out using HIV 1/2 Stat-Pak™ Dipstick. If the result is positive, then the individual
is HIV positive, while if negative, a tie breaker test is done using the Uni-Gold™ HIV test
(FMOH 2010b:100). The limitation of this test algorithm is that it cannot detect patients
in the window period (in the first three months after infection); that is, patients with
recent HIV infection are at a stage where anti-HIV antibodies used for the detection
purpose are not produced by the immune system yet. For this reason, a negative result
of HIV with a such testing algorithm only means the person in question was not infected
with HIV three months prior to the test (FHAPCO & FMOH 2008a:39). For this reason,
repeat testing is recommended if there was exposure within three months prior to
testing.

Earlier detection is possible these days with the development of test technology that
detects the viral genetic material directly. This is used in diagnosing infants exposed to
HIV (FMOH 2010b:96; WHO 2010b:4).

The WHO (2012a:35) states that HIV testing is not a simple series of lab tests. It must
be accompanied by pre-test counselling and post-test counselling sessions in order to
make sure that the person being tested understands what the test is for and why the
testing is necessary and what should be done to reduce future risks. For HIV positive
patients, the counselling provides an opportunity to educate them about the care they
need and where they may find it in addition to giving them advice on positive living
(FMOH 2010b:44).
 36

(FMOH 2010b:100)
Figure 2.6 HIV testing algorithm in Ethiopia
 37

2.8 THE NATURAL HISTORY OF HIV INFECTION

The natural history of HIV follows three general patterns: rapid progressors, typical
progressors, and long term non-progressors. Typical progressors account for 90% of
patients and it takes eight to ten years before they develop AIDS symptoms. Rapid
progressors account for five per cent of individuals infected, and it takes them a much
shorter time to develop AIDS, between three to five years. Finally, long term non-
progressors are patients who remain clinically asymptomatic for more than 15-20 years
(FHAPCO & FMOH 2008a:39; Levy 2009:153; Sabin & Lundgren 2013:314). In any of
these types of cases, untreated people go through three phases which are discussed
next.

(Longo et al 2012:1520)
Figure 2.7 Natural history of HIV infection in untreated individual: virological,
immunological, and clinical progression of typical progressors

2.8.1 Acute phase

 38

This phase follows HIV infection and lasts around six to nine months (Zuckerman et al
2009:509). Initially there is uncontrolled viral replication and rapid destruction of CD4 T
lymphocytes (see figure 2.7). As the body mounts its defence, CD4 cells show some
recovery and viral replication drops rapidly but fails to clear and is maintained around a
certain level called the viral set point. This marks the end of this phase. The level of viral
copies at this stage could range between 1,000 and 100,000. Zuckerman et al
(2009:511) explain that in most individuals, this phase is accompanied by the acute HIV
syndrome, which is characterised by the acute onset of a constellation of flu like
symptoms including fever, rashes, inflammation of lymph nodes, and pharyngitis. This
condition is self-limiting and resolves by itself.

2.8.2 Asymptomatic phase

This phase begins with the establishment of a viral set point (Longo et al 2012:1520).
There are no clinical symptoms during this stage, but viral replication increases
gradually over the years, accompanied by a gradual decrease in CD4 T lymphocytes
(see figure 2.7). This phase could last three to 20 years depending on the viral set point.
The lower the viral set point, the longer the duration of this phase and vice versa (Longo
et al 2012:1520).

2.8.3 Symptomatic phase

Longo et al (2012:1540) state that this phase is marked by the onset of clinical
symptoms. In untreated patients, nonspecific, mild infections appear in the form of skin
fungal infections, herpes zoster, ear, and sinus infections. As the number of CD4 cells
declines to below 200 per micro litre of blood, the more serious AIDS defining infections
follow. The FHAPCO and FMOH (2007:2) identify the common AIDS defining conditions
as oral candidiasis, severe bacterial infections, chronic diarrheal diseases,
Pneumocystis Carinii Pneumonia, tuberculosis, and Kaposi‘s sarcoma. The authors
further state that, as time goes on, these opportunistic infections become more frequent
and eventually lead to death.

2.9 ANTIRETROVIRAL THERAPY

 39

By definition this is therapy directed at the HIV virus (Longo et al 2012:1563). The
coming sections further discuss this therapy from different perspectives.

2.9.1 Historical perspective

Zidovudine (AZT) was the first antiretroviral drug ever used for treating HIV. It was used
as mono therapy and marked the beginning of a revolution towards managing this
disease (Fischl, Richman, Grieco, Gottlieb, Volberding, Laskin, Leedom, Groopman,
Mildvan, Schooley, Jackson, Durack, King and the AZT Collaborative Working Group
1987:187). The short term benefit was marked but a different study demonstrated its
limitation in regard to the long term benefit, since there was no difference in survival,
disease progression, or death comparing the effects of AZT and a placebo (Seligmann,
Warrell, Aboulker, Carbon, Darbyshire, Dormont, Eschwege, Girling, James, Levy, Peto,
Schwarz, Stone, Weller, Withnall, Gelmon, Lafon, Swart, Aber, Babiker, Lhoro, Nunn &
Vray 1994:873). This was the result of drug resistance occurring with mono therapy.
This finding resulted in the launching of combination therapies starting with dual drug
use in the form of zidovudine with didanosine or zalcitabine (Aber, Aboulker, Babiker,
Bragman, Breckenridge, Carbon, Charreau, Chene, Collis, Cooper, Darbyshire,
Dormont, Fiddian, Flepp, Gazzard, Goebel, Hooker, Lange, Luthy, Peto, Reiss,
Seligmann, Stone, Thomis, Vella, Walckenaer, Warrell, Weller, Wilber, Yeni, Yeo &
Withnall 1996:286). Such studies showed that dual therapy is superior to mono therapy
and paved the way for subsequent studies using three drugs. Three drugs combinations
are the mainstay of therapy at present. Zuckerman et al (2009:917) state that studies
done with early initiation of HAART identified challenges with adherence, and toxicity,
for which reason the need for eligibility criteria became evident in order to balance those
factors with respect to long term benefits.

2.9.2 Eligibility criteria

Eligibility criteria for the initiation of antiretroviral therapy are based on CD4 cell count
and the staging of the disease. Currently, the WHO (2013:92) recommends initiation for
all adult and adolescent (>14 years) patients with a CD4 cell count below 500, with
those below 350 or those whose disease stage is advanced being given top priority.
The same criteria apply for children above five years of age. For children below five
years of age, all are eligible for initiation but priority is being given to those whose
 40

disease is at advanced stage. The CD4 cell count cut-off for adults and adolescents
was 350 during the 2010 revision of eligibility criteria according to the WHO (2010a:25).

2.9.3 Access to antiretroviral therapy

Access to antiretroviral therapy has been progressively increasing all over the world,
and especially in low and middle income countries, where it reached 9.7 million people
in 2012 (UNAIDS 2013b:6). This represents great progress in considering all eligible for
the treatment. For sub Saharan Africa, the most affected region by the HIV/AIDS
pandemic, nearly 56% of eligible people has been put on HAART (see figure 2.8)
(UNAIDS 2012:23). For Ethiopia, 68% of the 400,000 people who need the treatment
have started HAART (UNAIDS 2013b:A85). Treatment eligibility cut-off for these
assessments was at CD4 cell count of ≤350/mm3 according to the WHO (2010a:25)
treatment guideline.

(UNAIDS 2012:23)
Figure 2.8 Eligibility for antiretroviral therapy versus coverage, low- and middle-
income countries, by region, 2011

UNAIDS (2013b:A85) plans to put 15 million people on antiretroviral therapy by 2015 as

part of fulfilment of HIV component of Millennium Development Goal (MDG) 6 (United
Nations 2012:38).
 41

2.9.4 Antiretroviral drug classes

Various classes of drugs have been invented over the years. There are five classes of
drugs in clinical practice globally. These include nucleoside reverse transcriptase
inhibitors (as for example, zidovudine, tenofovir, lamivudine, didanosine, abacavir), non-
nucleoside reverse transcriptase inhibitors (example nevirapine, efavirenz), protease
inhibitors (as for example ritonavir boosted lopinavir, ritonavir and boosted atazanavir),
fusion inhibitors (example maraviroc), and integrase inhibitors (like raltegavir)
(Department of Health and Human Services 2013:P-1; Longo et al 2012:1563). Three
combinations of drugs from at least two different classes must be used in order to
ensure effective treatment. All drugs listed in the examples are available in Ethiopia
except for the fusion inhibitors and integrase inhibitors (FHAPCO & FMOH 2008a:122).

2.9.5 Goal of therapy

The WHO (2013:176) states that the goal of antiretroviral therapy is to achieve viral
suppression to an undetectable level. It further clarifies that achieving this goal means
very limited viral replication, meaning improved immunologic and clinical outcomes. In
order to achieve this, adherence to treatment must be more than 95% (FHAPCO &
FMOH 2008a:82).

2.10 ADHERENCE PREPARATION BEFORE INITIATING ANTIRETROVIRAL

THERAPY

Adherence preparation for ART is an important step before initiating treatment. This is
because poor adherence to treatment can lead to drug resistance and ineffectiveness of
the medication being taken (WHO 2012b:10). It is said that the first regimen taken is the
most effective. Even if the treatment regimen is changed once the first regimen fails,
drug resistance may continue to pose a problem as there is cross resistance among
drugs within the same class (FHAPCO & FMOH 2008a:103).

The WHO (2013:89) states that once eligibility is determined, adherence preparation
should start as soon as possible by assessing patient‘s willingness, readiness for, and
knowledge about HIV/AIDS and ART. This should be followed by addressing the
 42

knowledge gap of the client in a coordinated way so as not to overwhelm the client with
too much information. So, such counselling sessions must be planned ahead of time,
when possible, to address specific issues in multiple visits. A systemic review done by
Bärnighausen, Chaiyachati, Chimbindi, Peoples, Haberer and Newell (2011:946) show
the added value in having a structured education program rather than informal sessions
in providing adherence education. Two to three well-planned visits are usually adequate
to address the most important aspects of treatment, namely the effectiveness of the
therapy, the medication dosage, the drug side effects, opportunistic infections,
recommended nutritional intake, personal hygiene, and the risk of transmission to
others. Once a common understanding is established, all potential barriers to
adherence in the individual (for example, the lack of a care giver, mental illness,
substance abuse, forgetfulness) or the health system (for example distance to facility)
must be discussed including the means to overcome them (WHO 2013:176). The WHO
(2013:178) states that client specific arrangements need to be made to address any
major barrier. Such assessment should be made continuously even after treatment is
commenced in order to identify any potential problem and manage it. Treatment should
be initiated once the client and provider agree on all important aspects of the treatment.

2.11 ADHERENCE ASSESSMENT DURING ANTIRETROVIRAL THERAPYINTAKE

The WHO (2013:181) states that in routine care settings, adherence can be assessed
by viral load measurement, pill count, pharmacy refill records, or self-report. Additional
complex adherence monitoring tools are also available according to Turner (2002:S143)
which include electronic devices, and biochemical assays that measure blood level of
antiretroviral drugs. The level of adherence in different care settings is reported below.

2.11.1 Level of adherence to HAART in developed countries

Cohen, Meyers and Davis (2013:online) calculate that the mean medication possession
ratio, which is a proxy measure of adherence to picking up a prescription refill, among
8,721 patients on ART in the United States of America (USA) insured with Medicaid was
81% (with a standard deviation of 15%). The level of adherence was higher for patients
who took single tablet regimens once a day. This study was representative as it was
conducted in 11 different states. In another study with a wider geographic coverage in
North America that spanned Canada, the USA and Puerto Rico, the mean level of self-
 43

reported adherence over the 30 days prior to the data collection day was 85.7% (with a
standard deviation of 21.8) among 1,875 patients receiving ART (Phillips, Webel, Rose,
Corless, Sullivan, Voss, Wantland, Nokes, Brion, Chen, Lipinge, Eller, Tyer-Viola,
Rivero-Méndez, Nicholas, Johnson, Maryland, Kemppainen, Portillo, Chaiphibalsarisdi,
Kirksey, Sefcik, Reid, Cuca, Huang & Holzemer 2013:online). The level of adherence in
North America was, in general, comparable to that found in a study in Italy where in it
was reported that adherence among 2,763 patients who received ART in March and
May 2010 was 100% and 79% respectively in the week prior to the data collection. No
association was found between self-reported adherence and the frequency and number
of pills taken (Gianotti, Galli, Bocchiola, Cahua, Panzini, Zandonà, Salpietro, Maillard,
Danise, Pazzi, Lazzarin & Castagna 2010:156). In a meta-analysis that included 9,931
HIV patients in 23 observational studies done in Spain, Ortegoa, Huedo-Medinab, Vejoa
and Llorcac (2011:286), it was reported that the level of adherence more than 90% was
54% (95%CI: 49-59%). The findings of this study showed levels of adherence much
lower than those reported from other studies. The authors explained that this was
because 52% of studies used more than one tool to assess adherence, like for
example, pharmacy dispensary records, plasma drug concentration, left over pill count,
viral load and electronic devices in addition to self-report, and emphasised the need to
use multiple strategies in order to unmask overestimation of adherence as obtained
from self-reports.

2.11.2 Level of adherence to HAART in sub-Saharan Africa

Studies from Kenya, Uganda, and Cameroon show that the level of adherence to ART
ranges between 61% and 98%. The actual level varies depending on the type of
adherence measure (in the form of self-report, record monitoring, or plasma drug levels)
employed and the time patients have been on treatment. (Boruett, Kagai, Njogo,
Nguhiu, Awuor, Gitau, Chalker, Ross-Degnan, Wahlström and Tomson 2013:online;
Mghamba, Minzi, Massawe & Sasi 2013:online; Roux, Kouanfack, Cohen, Marcellin,
Boyer, Delaporte, Carrieri, Laurent & Spire 2011:S42).

2.11.3 Level of adherence to HAART in Ethiopia

Most of the studies done on the level of adherence in Ethiopia are based on the self-
report of adherence. Azmeraw and Wasie (2012:120) reported that in their study, 80%
 44

of patients had optimal adherence (at 95%) among 340 children in two referral hospitals
in Northwest Ethiopia. Among adult patients, Giday and Shiferaw (2010:191) have
reported adherence of ≥95% among 88% of patients followed for one month. Kebede
and Wabe (2012:67) report a 95% adherence for HAART among 24 TB/HIV co-infected
patients in a hospital in Southwest Ethiopia.
A study that assessed adherence for the month prior to the study found adherence to be
80.2% among 349 patients in central Ethiopia (Dagnew 2009:89). Another study in a
university hospital in Southwest Ethiopia evaluated the timing of pill intake in addition to
adherence to doses. The finding of this study was that 303 (95%) of the study subjects
were adherent based on dose adherence during the week before the actual interviews
while adherence including timing and food intake was found to be 72.4% (Tiyou et al
2010:39).

Chalker et al (2010: online) report that adherence to appointment date was 86% while
the adherence by to the regimen measured by self-reporting and by pill count was 94%,
and 76% respectively. A tendency of overestimate adherence was observed in another
study by Beyene et al (2009:1010); in that study, optimal adherence among 93% study
participants was reported when15-days of self-report were used, while it was 88% when
the measure was on unannounced pill count. While adherence has been suboptimal in
most studies reviewed, a degree of overestimation found in data based on self-report
leads one to assume the true level of adherence to be lower in most study-settings than
reported.

2.12 MONITORING OF ANTIRETROVIRAL THERAPY

The WHO (2013:131) identifies three options for monitoring treatment response. These
are virological, immunological and clinical options. With virological monitoring, the level
of plasma RNA is measured as an indicator of the viral load. It is recommended that
viral load measurements be taken at baseline followed by another measurement taken
at six months after antiretroviral treatment is initiated. Hereafter, it should be measured
yearly. Viral load measurement is the gold standard for assessing treatment response
because a decrease in the level of viral load is highly associated with improved clinical
outcome (Murray, Elashoff, Iacono-Connors, Cvetkovich & Struble 1999:799).
 45

With immunological monitoring, the number of CD4 T lymphocytes in the blood is

measured. The WHO (2013:131) recommends that the CD4 cell count be determined at
baseline and every six months during treatment. The CD4 cell count is a measure of the
strength of the immune system and is a good predictor of patient survival (Egger, May,
Chêne, Phillips, Ledergerber, Dabis, Costagliola, D'Arminio, De Wolf, Reiss, Lundgren,
Justice, Staszewski, Leport, Hogg, Sabin, Gill, Salzberger & Sterne 2002:125). Clinical
monitoring is a documentation of the clinical stage of the patient and must be done at
every visit. In resource poor settings, viral load monitoring, which is the gold standard, is
too expensive to use, for which reason the WHO (2013:133) recommends
immunological and clinical monitoring.

2.12.1 The pattern of immunologic monitoring in developed countries

One multi-national study by Podlekareva, Reekie, Mocroft, Losso, Rakhmanova,

Bakowska, Karpov, Lazarus, Gatell, Lundgren and Kirk (2012:online) evaluated the
frequency of immunologic monitoring among patients started on ART as part of an
evaluation of the provision of care for patients in chronic HIV care. It included 5,859
patients from 34 countries in Europe and Argentina. The study found that the median
number of CD4 cell count tests done on a patient varied considerably among the
different regions of Europe. West-Central European countries had the highest number
of tests per patient per year at 3.5, followed by northern and southern European
countries having three annually and Eastern-Central European countries having an
average of 2.5 per year. This indicates that, countries from the aforementioned regions
have more tests than the WHO recommends for CD4 cell count testing, and implies that
a more stringent monitoring strategy exists in those countries (WHO 2013:132). In
eastern European countries, the level of monitoring is two tests per patient per year, as
per the recommendation, while in Argentina, it is slightly lower, 1.5. In general, viral load
testing followed the same pattern and from this one can conclude that the system of
monitoring treatment response among those in the study population Europe met the set
targets.

2.12.2 The pattern of immunologic monitoring in sub-Saharan Africa

In general, there is limited published literature addressing the issue of laboratory

monitoring for patients on ART in sub-Saharan Africa including immunologic monitoring.
 46

Fawibe, Olafimihan, Salami, Desalu and Odeigah (2010:290) report that in a teaching
hospital in Nigeria providing chronic HIV care, evaluation of the baseline and follow-up
CD4 cell count was documented for 56% and 28% of patients respectively among those
who had had follow-up for at least one year (N=440) over a period of 12 months. A
prospective study in Kenya that evaluated the effect of electronic reminders for ordering
CD4 cell count tests reveals that overdue baseline CD4 cell count tests were 29%
(N=1482) at baseline. The percentage of requests for these overdue tests was 42% and
36% respectively for patients in intervention and control clinics (Were, Shen, Tierney,
Mamlin, Biondich, Li, Kimaiyo & Mamlin 2011:152). These studies point to a much lower
frequency of lab monitoring for ART patients in the study settings.

In resource limited settings such as existing in most sub-Saharan African countries,

there are a number of challenges that hamper the effective implementation of
immunologic monitoring. Zachariah, Reid, Chaillet, Massaquoi, Schouten and Harries
(2011:39, 40) state that maintaining the traditional laboratory based or non-point of care
CD4 cell count tests is problematic because of technology, logistic, and patient factors.
The authors further explain that the problem with the technology is that the machines
are expensive and prone to breakdown and require regular maintenance. Even when
they are present and functional, they require skilled workers to run them, which there is
a scarcity of. Logistics issues include problem with sustained availability of reagents,
test tubes, and other supplies. In Malawi, such factors contributed to the progressive
decline of functional CD4 cell count machines from 52 in 2009 to 41 in 2010. Taking this
into consideration, it would not be surprising to get a much lesser frequency of
monitoring in resource poor settings.

2.12.3 The pattern of immunologic monitoring in Ethiopia

Literature is scanty on this issue in Ethiopia, but the available data reveals a similar
picture of suboptimal CD4 cell count monitoring in chronic HIV care settings in Ethiopia
similar to that in settings in other sub-Saharan African countries. Berhan (2011b:205)
reports that in a multicentre study that included seven public hospitals in Ethiopia, 6th
and 24th month follow-up CD4 cell count tests were done for 68% and 37% of children
respectively (N=1,163). Another study in northwest Ethiopia showed that only 21% of
patients had had follow-up CD4 test in the previous six months (Alemayehu, Bushen &
Muluneh 2009:359).
 47

2.13 IMMUNOLOGIC RESPONSE AFTER TAKING ANTIRETROVIRAL THERAPY

According to the Department of Health and Human Services (2013:C-7) a minimum of a

threefold decrease in viral load measurement in the first three to six months of therapy
is an indication of adequate treatment response, while sustained suppression is said to
occur when there is undetectable viral replication during follow-up. The same guideline
further clarifies that it takes a year or two to achieve complete suppression in patients
with proper follow-up.

In patients with treatment judged to be successful, the CD4 cell count increases by 50-
150 in the first year of treatment. A further yearly increase of 50 to 100 cells per micro
litre of blood is expected per year in patients with suppressed viral replication
(Department of Health and Human Services 2013:C-5). The rate of the CD4 cell
increment is lower for patients with a low starting CD4 cell count and for the elderly in
spite of viral suppression (Althoff, Justice, Gange, Deeks, Saag, Silverberg, Gill, Lau,
Napravnik, Tedaldi, Klein & Gebo 2010:2473).

2.13.1 The pattern of immunologic response in developed countries

In a study conducted in San Diego primary infection cohort, the rate of CD4 cell
recovery and the peak CD4 cell count achieved after 48 months of HAART was lower
for patients who initiated treatment later as well as for patients with a lower CD4 cell
counts at baseline (Le, Wright, Smith, He, Catano, Okulicz, Young, Clark, Richman,
Little & Ahuja 2013:222). In a different study done among members of the military, the
United States Military HIV Natural History Study, the first six months showed the
greatest increase of CD4 recorded, at 93-151 CD4 cells. Then after, the yearly CD4 cell
increment was 22-36 CD4 cells per year. Furthermore, it was shown that patients who
started treatment with a lower CD4 cell count (≤200) were seen to follow a trajectory of
CD4 cell recovery lower than those that started when there was higher CD4 cell count,
for which reason their mean CD4 cell count after 12 years on HAART failed to reach
400, while those with a higher baseline CD4 cell count managed to have a CD4 cell
count above 500 (Lifson, Krantz, Eberly, Dolan, Marconi, Weintrob, Crum-Cianflone,
Ganesan, Grambsch & Agan 2011:online).
 48

Touloumi, Pantazis, Chaix, Bucher, Zangerle, Kran, Thiebaut, Masquelier, Kucherer,

Monforte and Meyer (2013:online) report that immunologic response was similar among
B (predominant subtype in most developed nations) and non-B subtypes except for
subtype A of the HIV virus, which showed a lower increase of CD4 cells in the first three
months followed by a higher rate of long term CD4 cell recovery after controlling for
gender, sexual orientation, injecting drug use, age at initiation of HAART, prior
antiretroviral drug exposure, time from sero-conversion, initial HAART regimen, and
viral load at baseline. Apart from that, they reported a similar pattern of CD4 cell
recovery, which was an initial rapid CD4 cell count increase in the first three months and
a slower one thereafter.

2.13.2 The pattern of immunologic response in sub-Saharan Africa

Kanters, Nachega, Funk, Mukasa, Montaner, Ford, Bucher and Mills (2013:online)
explain that in their prospective cohort study involving 5,271 patients on HAART in
Uganda, the largest increase in CD4 cell count was recorded in the first year of therapy.
After five years of therapy, they reported that CD4 cell recovery had reached a plateau.
In their study, the baseline CD4 cell count was the strongest predictor of CD4 cell
recovery, and adolescents experienced the greatest CD4 increase. A different study
comparing clinical outcomes between patients with and without Kaposi‘s sarcoma in
South Africa found that the pattern of CD4 cell recovery still held a similar pattern of a
greater initial step increment followed by a protracted and gradual increase later in the
course of treatment. Patients with Kaposis Sarcoma were also found to be less likely to
show an increase of 50 in their CD4 cell count in the first six months as compared to
those without Kaposis Sarcoma (Maskew, Fox, van Cutsem, Chu, MacPhail, Boulle &
Egger 2013:online).

Nakanjako, Ssewanyana, Nabatanzi, Kiragga, Kamya, Cao and Mayanja-Kizza

(2013:online) demonstrate that patients in Uganda with lowest CD4 cell count increase
having been on HAART for four years despite suppressed viral replication (also called
suboptimal immune responders) tend to have lower baseline CD4 cell count as
compared to optimal immune responders. They further substantiate this by
demonstrating that cells from these patients have impaired T-cell proliferation (to which
CD4 cells are part of) in vitro studies in reaction to stimulation with antigens from
 49

common pathogens infecting HIV patients hinting at depletion of reserves. This same
phenomenon was observed in vitro study among patients in the USA (Lederman,
Calabrese, Funderburg, Clagett, Medvik, Bonilla, Gripshover, Salata, Taege, Lisgaris,
McComsey, Kirchner, Baum, Shive, Asaad, Kalayjian, Sieg & Rodriguez 2011:1222).

2.13.3 The pattern of immunologic response in Ethiopia

In general, in Ethiopian patients on HAART, CD4 cell increment was found to be related
with duration of HAART, the greatest increase observed in the first six to twelve months.
A pattern of slight decrease in mean CD4 cell count in later years of follow-up was also
noted (see figure 2.1). Age, functional status, and baseline WHO stage were all not
found to be associated with CD4 cell increment (Berhan 2011a:294; Tsegaye & Worku
2012:107; Kassa, Gebremichael, Alemayehu, Wolday, Messele & Van Baarle
2013:online; Reda, Biadgilign, Deribew, Gebre & Deribe 2013:online).

(Reda et al 2013:online)

Figure 2.9 Box plot for changes in magnitude of CD4 cell count among the cohort of
patients on anti-retroviral treatment across months of retrospective
follow-up over five years
 50

2.14 IMMUNOLOGICAL TREATMENT FAILURE

The consequence of poor treatment adherence is treatment failure, which is then

managed by switching treatment to another effective regimen than the patient was
taking when the treatment failed (Department of Health and Human Services 2013:K-2;
WHO 2013:176). The criteria used to measure adequate treatment response cannot be
used to decide which ones require treatment switching especially for those with a poor
immunological response, as they may actually have adequate viral suppression. Even
for viral load measurement, a temporary rise of plasma RNA could be wrongly assessed
for treatment failure (Department of Health and Human Services 2013:C-5, C-7). For
this reason, in order to identify those with immunological failure, the WHO criteria must
be used to declare a patient as having immunologic treatment failure. In keeping with
the objective of this study, only immunological criteria are further discussed.

2.14.1 Immunological treatment failure criteria

Immunologic failure is when there is a decline of CD4 cells by 50% from the peak level
achieved during treatment or when it goes to or below the baseline CD4 cell count level,
or when it is persistently below 100 (WHO 2010b:50). These criteria should be applied
only after a minimum of six months of therapy with HAART. Interpretation of
immunologic criteria should be made carefully since patients with suppressed viral
replication may have immunological failure. This is because these criteria have low
sensitivity and positive predictive value for virological failure (WHO 2010b:48). But,
since they have very high negative predictive value, they can be used to identify those
who may need a further work up with viral load measurement. This strategy will prevent
unnecessary switching to second line therapy (WHO 2013:135).

2.14.1.1 The prevalence of immunologic treatment failure

Since developed nations use virological parameters, the so-called gold standard, to
define treatment failure (WHO 2013:133), data is in general lacking on prevalence and
determinants of immunologic treatment failure among patients on HAART for those
nations. For this reason, discussion for this and the next section is limited to sub-
Saharan Africa.
 51

2.14.1.1.1 The prevalence of immunologic treatment failure in sub-Saharan Africa

A number of studies in sub-Saharan Africa have assessed the prevalence of

immunologic failure and have come up with variable results, but in general, most studies
point to a conclusion that, in most settings, the prevalence of immunologic treatment
failure has been more than 10% (range 10-32%) (El-Khatib et al 2011:online; Jaka et al
2009:6; Rawizza et al 2011:1286; Reynolds et al 2009:697). In these studies, the follow-
up duration was variable, ranging from 15 to 33 months. From the reviewed literature,
one study in Nigeria has a contradictory finding. In this study the prevalence of first line
HAART failure was reported to be very low, at 1.19% among 9,449 patients during the
study period, but the follow-up duration was not reported (Eze & Uyagu 2009:46).

2.14.1.1.2 The prevalence of immunologic treatment failure in Ethiopia

For Ethiopia, data is available only for paediatric patients. Bacha et al (2012:online)
report immunologic failure to exist for 9.2% of 1,186 children with a mean age of six
years and a mean follow-up of 37 months. In another study among a smaller cohort of
96 paediatric patients (with a mean age six years, mean follow-up duration of 13.7
months) immunologic failure was at 11.5% (Workneh et al 2009:78). Despite an
extensive search, similar data could not be found for adult patients. Assefa et al
(2011:online) report that the prevalence of patients on second line HAART among 7,451
clients actively on follow-up at 24 months was 2.13%. But because of difficulties in
monitoring treatment response in many settings and difficulties concerning the
utilisation of available resources by health care workers, the actual prevalence of
treatment failure is expected to have been higher (Azzoni et al 2012:online; Bélec &
Bonn 2011:1255). The level of patients on second line HAART was comparable to what
the WHO (2009:79) reports but much higher than that reported by 48 MSF supported
programs in resource limited countries (Pujades-Rodriguez et al 2008:1307), which was
a switch to a second-line regimen of 4.1 per 1,000 person-years after 20 months of
follow-up among 48,000 patients.

2.15 DETERMINANTS OF IMMUNOLOGIC TREATMENT FAILURE

Few studies have been conducted to assess for predictors of immunologic treatment
failure among patients on HAART. Kassa et al (2013:online) demonstrated that at six
 52

months post HAART initiation, among patients in Addis Ababa, Ethiopia, the odds of
immunologic failure were 5.6 times higher for patients with a baseline CD4 cell counts of
less than 100 as compared to those with a higher CD4 cell counts (95% CI 1.6-20.1; p
value=0.008). The same study found that patients in an advanced disease stage at
treatment initiation were also found to be more likely to experience immunologic failure
than those who were asymptomatic (an odds ratio of 4.3, 95% CI 1.4-13.4; p
value=0.01). The problem with this assessment is that the authors used a different set
of criteria to define immunologic treatment failure than what the WHO (2010a:50) has
recommended. They used the criteria ‗an increase of CD4 cell count by fewer than 50
cells/μl at month six,‘ and the criteria ‗fewer than 100 cells/μl at months 18 and 24 of
HAART‘. These criteria are rather similar to what other authors have defined as a sub-
optimal immunologic response, which is completely different from immunologic
treatment failure (Lederman et al 2011:1222; Maskew et al 2013:online; Nakanjako et al
2013:online). For this reason, any comparison of their findings with those in other
studies is difficult.

A similar problem exists with the interpretation of a study conducted in Nigeria that
assessed immunologic failure. The criteria the authors used were: a fall of the CD4 cell
count to baseline or below; or an increase of CD4 cell count by less than 50cells/mm3 at
one year after HAART initiation, which is still different than the WHO recommendation
(2010b:50). The findings of the study were that patients with immunologic failure were
more likely to be of the male sex and of an age less than 30 years (OR 1.46; 95% CI
1.04-2.45; p value=0.04) (Anude, Eze, Onyegbutulem, Charurat, Etiebet, Ajayi, Dakum,
Akinwande, Beyrer, Abimiku & Blattner 2013:online). Overall, there is limited literature
on predictors of immunologic treatment failure as defined by the WHO (2010b:50) for
use in resource limited settings.

2.16 INTERVENTION FOR THOSE WITH TREATMENT FAILURE

Since one of the most common causes of treatment failure is poor treatment adherence,
any attempt to manage treatment failure by simply switching the HAART drug regimens
are futile, as the same problems that caused the treatment to fail in the first place may
result in treatment failure for the new regimen as well (Department of Health and
Human Services 2013:H-2, H-3).
 53

2.16.1 Managing poor treatment adherence

This should be done by addressing barriers to adherence at the various levels which are
discussed below.

2.16.1.1 Factors related to the health system

The WHO (2013:183) states that such factors include the high cost of travel or
treatment, use of regimens with a high pill burden or a high level of severe side effects,
the frequency of drugs being ―out of stock‖, and the lack of adequate and trained health
care workers to provide high quality care. The WHO further states that these factors
could be addressed if the health care system were designed based on the
recommended continuum of care and followed the public health approach. Such system
would provide care free of charge, and provide fixed dose combination drugs with
minimal side effects and pill burden. In addition, there would need be some task shifting
used to address the overburdened health care providers through deployment of
adherence supporters to assist in providing appropriate counselling, adherence
preparation, and patient tracing (WHO 2008:14). Decentralization and integration of
services would help in addressing accessibility issues.

2.16.1.2 Factors related to the individual

The WHO (2013:183) states that such personal factors include forgetfulness, not
appearing for clinic visits or not taking medications at home for fear of being seen taking
HAART medication, the lack of a care providers for guidance and co-morbid conditions
such as poor mental health and substance abuse. The most important intervention for
removing such barriers would be adequate adherence preparation and on-going
personalised follow-up during therapy in the form of education and counselling
concerning the issues that prove troublesome, both for patients and their care givers.
This could be done by adherence supporters or community health workers and could
help those providing care identify needs and get the proper care for patients, with
consultation to health care providers when required. A review by Bärnighausen et al
(2011:944) has demonstrated the effectiveness of such interventions in multiple
countries. In addition to this, the WHO (2013:183) recommends the use of text
messaging to manage forgetfulness when it is applicable.
 54

2.16.2 Managing treatment failure

Patients with treatment failure should be managed by switching treatment to an

appropriate regimen. In developed countries, doing this can be facilitated by the
determination of resistance patterns, which is very useful in selecting patient specific
regimens that will be able to effectively suppress viral replication in the face of drug
resistance (Department of Health and Human Services 2013: H-1). In resource poor
settings, though, this cannot be done because these tests are too expensive. For this
reason, the WHO (2013:147,150) has come up with standardised second line regimens
to use when failure is diagnosed with known first line regimens based on the most likely
resistance pattern expected to occur (see tables 2.1 A and B). These regimens are
used in Ethiopia except emitricitabine (FTC).

TABLE 2.1a SUMMARY OF PREFERRED SECOND-LINE ART REGIMENS FOR ADULTS

AND ADOLESCENTS

(WHO 2013:147)

TABLE 2.1b SUMMARY OF RECOMMENDED FIRST- AND SECOND-LINE ART

REGIMENS FOR CHILDREN AND ADOLESCENTS

(WHO 2013:150)
AZT = zidovudine; d4T = stavudine; TDF = tenofovir; 3TC = lamivudine; FTC = emitricitabine;
LPV/r = ritonavir boosted lopinavir; ABC = abacavir; NVP = nevirapine; EFV = efavirenze
 55

2.17 CONCLUSION

This chapter dealt with a number of issues related to HIV and pertinent to this study.
Initially, it dealt with the pathogenesis of the disease. This was followed by its
epidemiology at various levels: global, African, and Ethiopian. The review also
addressed treatment objectives of antiretroviral therapy and how treatment can be
monitored with particular attention to immunologic criteria. Finally, determinants of
treatment failure were briefly discussed. Chapter 3 addresses issues related to research
methodology as used in this study.
 56

CHAPTER 3

RESEARCH METHODOLOGY

3.1 INTRODUCTION

This chapter deals with the research methodology. It has four components: the research
design, the research method, validity and reliability, and ethical considerations.

The research design pertains to the general approach the researcher chooses to
answer the research questions. Research designs depend on the paradigm chosen. For
quantitative studies, the design is further specified by aspects relating to the presence
of interventions, the presence of comparison, and the timing of measurement (Polit &
Beck 2012:58,59). The research methods outline the specifications with respect to the
study population, aspects of sampling, data collection, and data analysis. Validity and
reliability will be discussed further with respect to both study design and data collection
instrument.

The purpose of the research methodology is to systematically address the research

problem. The aim of the study was to generate knowledge by describing the practice of
immunological monitoring at the facility level, and identifying determinants of
immunologic failure among people living with HIV in selected hospitals. The specific
objectives of this study were to:

 assess the pattern of immunologic monitoring in selected hospitals in southern

Ethiopia
 assess the pattern of CD4 cell count response of patients on HAART in selected
hospitals in southern Ethiopia
 describe immunologic treatment failure and identify its predictors among patients
on HAART in selected hospitals in southern Ethiopia

The following sections of this chapter will address the details of research design,
research methodology, validity and reliability, and ethics.
 57

3.2 RESEARCH DESIGN

3.2.1 Paradigm

Polit and Beck (2012:736) defines paradigm as ―a way of looking at natural phenomena
that encompasses a set of philosophical assumptions and that guide one‘s approach to
inquiry‖. It is like a model that is followed to answer research questions. There are two
paradigms, positivist and naturalistic, whose distinction is made by answers to the
following philosophical questions: what is the nature of the reality being studied
(ontologic assumption); how is the researcher related to that which is being studied
(epistemologic assumption); what is the role of researcher‘s values and views (axiologic
assumption); and how is data acquired (methodologic assumption). (Polit & Beck
2012:13)

Quantitative research design is related to the positivist research tradition which dictates
that, reality exists independent of the observer. A number of measures are taken to
make sure the observation of phenomenon is objective and free of bias, unaffected by
the beliefs of the observer including measurement of observations by an objective tool.
A fixed and predetermined study design generates numerical data which is collected by
counting or measuring a phenomenon of interest. (Polit & Beck 2012:13). In contrast to
this, qualitative research design follows the positivist tradition where subjectivity and
values of the investigator are deemed inevitable and demanded. The investigator‘s
interaction with what is studied generates the information. A flexible study design, that
which evolves as the study progresses, is followed. Narrative type of data is collected
and analysed. (Burns & Grove 2009:22; Polit & Beck 2012:13)

3.2.2 Research design

Research design is defined as the overall plan to answer a research question (Polit and
Beck 2012:741). In this study, a quantitative, correlational (or non-experimental) study,
the researcher simply observes what is going on without interference or manipulation.
Manipulation was not required nor was it possible in this study since all events occurred
prior to data collection (Polit & Beck 2012:227). Since there was no intervention
randomization of study subjects into exposure and control groups was not required.
Patients‘ information gathered in their natural environment during routine clinic visits
 58

was used to evaluate prevalence and determinants of immunologic failure. So this

design was appropriate for this study. (LoBiondo-Wood & Haber 2010:310).

3.2.2.1 Quantitative elements

This design is appropriate for this study because the aspect of the dependant variable
being studied; that is, pattern of immunologic monitoring, response and treatment failure
were determined by examining and interpreting the pattern of quantitative measure of
CD4 lymphocyte counts (WHO 2010a:50,51). The number of CD4 tests and magnitude
of CD4 cell count and related data was measured by a pre-designed document checklist
by data collectors making the data measurement independent of the observer. Also
controlling mechanisms were used to analyse the data.

3.2.2.2 Correlational elements

This study is a correlational or non-experimental study as there was no intervention. In

such correlational or non-experimental quantitative study design, the association
between variables as they occur naturally is measured (Bruce, Pope & Stanistreet
2008:423). In addition to assessing the mere association between variables, this type of
design can address the issue of whether one variable (a dependant variable) is
predicted by another (an independent variable). In this study, the major aim was to
assess the prevalence of immunologic treatment failure and factors that predict the
development of immunologic treatment failure. With respect to the second aim,
independent variables that correlated to immunologic treatment failure as well the
degree of correlation were studied, while in the first, it intended to describe the point
prevalence of immunologic failure, and describe other features of patients with
immunologic failure and hence, this study had a descriptive component as well as a
predictive analytic component (Burns & Grove 2009:13). In a descriptive research
design, a health event of interest is simply described. (LoBiondo-Wood & Haber
2010:200,201; Polit & Beck 2012:226).

3.2.2.3 Longitudinal elements

In a longitudinal study design, subjects are followed for some length of time. Exposure
and the outcome of interest are measured repeatedly as the follow-up continues. This is
 59

in contrast to cross-sectional study where observation is made at one point in time (Polit
& Beck 2012:181; Shaughnessy, Zechmeister & Zechmeister 2012:154,158).In this
study, there were no distinctive cohorts or exposure groups that are followed. Two types
of study designs exist based on the timing of data collection: prospective and
retrospective. In the prospective design, a group of subjects is followed forward in time,
and observations are made on the occurrence of outcomes of interest. In the
retrospective design, both the exposure and outcome of interest has already occurred in
the past and exposure precedes outcome (Polit & Beck 2012:187).

In this study, records of patients that were on HAART were followed and CD4 cell count
monitoring, magnitude of CD4 cell count as well as the main outcome of interest, i.e.
immunologic treatment failure, were evaluated at various times. Patients took treatment
for some time before experiencing treatment failure. So there needed to be time
sequence of events. All these features made the longitudinal design appropriate for this
study. Both exposure and outcome status were already determined at the start of the
study and hence that makes it a retrospective longitudinal study (Chernick 2011:12).
This design was particularly advantageous for this study. This is because since the
specific objective include assessment of the patterns of immunological monitoring and
whether immunologic failure was being dealt with appropriately and in a timely fashion,
health care workers might have modify their behaviour and been unnaturally more
vigilant in reaction to their awareness of a prospective study being done addressing
these issues (Polit & Beck 2012:225).

3.3 RESEARCH METHODS

‗Research methods‘ refers to the specific techniques followed to structure a study

including the specific means to collect and analyse data. It is specified by choices made
in identifying study population, sampling technique, data collection approach, and
analysis. (Polit & Beck 2012:12)

3.3.1 Population

‗Study population‘ refers to the collection of all subjects being considered for a study
that have some common characteristics (Polit & Beck 2012:273; Shaughnessy et al
2012:141). The study population in this study was medical records of HIV patients aged
 60

five years and older who took HAART for at least six months in selected hospitals
(Butajira and Nigist Eleni Mohamed Memorial, also called Hosanna) in SNNP regional
state. Of these records, all were accessible for the investigator hence forming the
‗accessible population‘. At the time of sample size determination, there were a total of
881 patients in hospital A (Butajira hospital) and 1,266 in hospital B (Nigist Eleni
Mohamed Memorial hospital).

3.3.2 Sampling

LoBiondo-Wood and Haber (2010:224) state that sampling is the process of selecting
part of the accessible population with the intension to study the selected cases and
make inferences about the target population. They further highlight advantages of
sampling which includes saving time and cost associated with collecting data from large
population. The ‗target population‘ is the collection of all patients about which the
researcher would like to make inferences about including patients beyond the
accessible population (Polit & Beck 2012:273).

3.3.2.1 Eligibility criteria

These criteria are the ones that specify which subgroup of the study population is to be
studied (Polit &Beck 2012:274). For this study, the eligibility criteria were as follows
listed as inclusion and exclusion criteria:

Inclusion criteria

 People living with HIV on antiretroviral therapy enrolled and started on

HAART ever since the program was started in the facility until 31 December
2012.
 Age ≥5 years

Exclusion criteria
 Patients who had been initiated on HAART and took it for less than six
months,
 Patients who had fewer than two CD4 cell count tests, and
 61

 Transferred in cases. These were cases that started ART in other facility than
the study facility and transferred to the study facilities for continuation of
treatment.

3.3.2.2 Sample size

The total sample size required was estimated for the combined accessible patients who
were on HAART treatment in the two hospitals. Sample size calculations were done
assuming random sampling for a finite population based on the estimation of a 50%
proportion, at a 5% significance level and 2.5% precision (Daniel 2009:192). The
formula used to estimate the sample size for the study was:

n= [z2p(1-p)N]/[(N-1)d2+z2p(1-p)]
Where
n = calculated sample size
N = population size
z = critical value at the chosen significance level
p = proportion to be estimated
d = precision

Thus, for this study taking N=2,147, z=1.96, p=0.50 (50%) and d=0.02 (2%), the sample
size was thus 1,133.703 which was rounded up to 1,134. Assuming the number of
missing records to be 15%, the sample size was inflated to 1,303. This meant that after
applying the eligibility criteria, of the remaining patients, 1,303 patients would be
randomly selected. In other words, proportionate sampling was used (Daniel 2009:13).

3.3.2.3 Sampling process

The sample was selected through a sampling process. Any sampling plan should
deliver a sample that accurately represents characteristics of the study population (Polit
& Beck 2012:283). For the purposes of this study, sample was selected from the total
population of clients that fulfilled the eligibility criteria. Of the two types of sampling
designs, that is probability and non-probability sampling, probability sampling was
chosen for this study (Shaughnessy et al 2012:144). In probability sampling, subjects
are chosen randomly and have equal chance of being selected (Polit & Beck 2012:280;
 62

Shaughnessy et al 2012:144). For this study, simple random sampling of all accessible
patients fulfilling the eligibility criteria was planned.

Permission was obtained from the management of the hospitals under study and ethics
committee of Regional Health Bureau of Southern Nations Nationalities and Peoples‘
region and National Ethics Review Committee of Ethiopia to access the list of all clients
started on HAART. A list was compiled indicating the patient file number. The
statistician then applied the eligibility criteria by going through the list and deleting those
who didn‘t fulfil the eligibility criteria.

3.3.3 Data collection

This study utilized a self-designed document checklist. Personal and social details were
not collected from human data sources.

3.3.3.1 Data collection instrument

Data collection was carried out by a self-designed document checklist (see annexure A)
which consisted of the following three sections:

Section A included patients‘ age, and gender. These demographic variables were used
to characterise the study population.

Section B This section included clinical characteristics in the form of baseline CD4 cell
count, baseline WHO stage, and other variables that might predict treatment failure
such as occurrence of tuberculosis, history of treatment interruption, and weight status.
All these variables were used to characterise immunologic monitoring, immunologic
response, and immunologic treatment failure.

Section C In this section immunologic information like timing of immunologic (CD4 cell
count) monitoring (section C.1), timing and magnitude of CD4 cell count (section C.2),
and timing and occurrence of immunologic failure (section C.3) were recorded.

3.3.3.2 Administering the data collection instrument

 63

Data source were existing medical records. The advantage of using existing records
was to save time and reduce cost than if data was collected prospectively. In addition,
since the researcher was familiar with the records and the data collection procedures,
appropriate precautions were taken where the records and the data entered had
limitations (LoBiondo-Wood & Haber 2010:270, 279).

3.3.3.3 Pre-testing the data collection tool

The data collection tool and cleaning program codes were pre-tested prior to data
collection and cleaning. The statistician did this among transfer-in patients. These were
patients who started HAART treatment in other facilities than study hospitals and sent
for follow-up to the study sites. Such patients were excluded from this study but had the
same set of data as required for this study and hence provided ideal ground for pre-
testing the data collection instrument. Twenty five such charts were randomly selected
from one of the study facilities and the pre-test was carried out. The purpose of the pre-
test was to assess the adequacy of the tool, the length of time required to do the data
collection, and to make sure of the clarity of the procedures. (LoBiondo-Wood & Haber
2010:280; Shaughnessy et al 2012:168)

3.3.3.4 Data collection procedure

Data clerks at each facility provided electronic records to the statistician in the study
hospitals. The database had a built in program to de-identify or remove patient
identifiers like name and address. That program was run before the data was provided
to the statistician. So total anonymity was ensured and even the researcher had no
access to the identity or any personal information of the clients under study. After the
statistician gave the data collectors the selected sample of cases, data collectors
abstracted the data from medical records.

3.3.4 Data management and analysis

In this study, Epi InfoTM 3.5.1 statistical software (United States Centers for Disease
Control and Prevention 2007) and Stata/IC 12.0 (StataCorp LP) were used to undertake
both descriptive and inferential statistics. P value was set at 0.05.
 64

3.3.4.1 Descriptive statistics

The purpose of doing descriptive statistics was to summarize and organize raw data so
as to understand the information contained in it in a simple way (Daniel 2009:20;
Shaughnessy et al 2012:348). Data from section C of the checklist was used to address
the following objectives:

 Assess the pattern of immunologic monitoring in selected hospitals in southern

Ethiopia.
 Assess the pattern of CD4 cell count response for patients on HAART in selected
hospitals in southern Ethiopia.
 The prevalence of immunologic treatment failure was calculated as a proportion
with a 95% confidence interval.

A number of descriptive statistical measures were used in this study to summarise data.
These include frequency distribution depicted in the form of tables for categorical and
ordinal variables(Polit & Beck 2012:379; Shaughnessy et al 2012:122). The distribution
of continuous variables (age and CD4 cell count) was displayed graphically in addition
to summary of their mean, standard deviation, median, minimum and maximum
values(Polit & Beck 2012:385; Shaughnessy et al 2012:355). Pattern of immunologic
monitoring and response were examined graphically.

Pattern of immunologic monitoring was assessed using the national monitoring and
evaluation standard for determination of follow-up CD4 cell count for patients on
antiretroviral therapy. The timing for evaluation of patients was every six months the first
year and yearly then after (FMOH 2010a:107). A CD4 cell count test was said to have
been done timely if it was done within one month of that particular follow-up month the
test was supposed to be done. The expected number of follow-up CD4 cell count tests
was calculated for those patients who were still on follow-up in that facility. A bar chart
was used to show and contrast expected and actual CD4 cell count tests that were
determined timely. The use of bar chart is justified as it is used for categorical data with
natural ordering (Chernick 2011:41). In this study, data are ordered by month of follow-
up. Pattern of immunologic response was evaluated at cohort level every six months. All
CD4 cell counts within two months of follow-up month of interest were used to
 65

determine CD4 cell count response at that month. A box-plot was used to view this
graphically. A box-plot is a graph that shows median, inter-quartile range, as well as
outliers (Chernick 2011:41,42).

Descriptive survival analysis was used to assess the cumulative probability of

immunologic failure. Survival analysis is used when analysing data with time dependent
outcome that may or may not have complete follow-up information. Follow-up may be
incomplete either because subjects don‘t experience event of interest or they are lost to
follow-up before end of study (Daniel 2009:648). Result of analysis is shown in a life
table as well as graphically by Kaplan-Meier curves for this study. Life table shows
survival probability estimates for specific time intervals whereas Kaplan-Meier survival
curves show continuous probability over time (Chernick 2011:159,160). In addition,
Kaplan-Meier curves were drawn to graphically evaluate similarities and differences with
respect to values of covariates on event free survival.

3.3.4.2 Inferential statistics

The Chi Square test, the log-rank test and multivariable analysis using Cox proportional
hazards regression were used to analyse data. The P value was set at 0.05.

3.3.4.2.1 The Chi square test

The Chi square test is a non-parametric test that is used to compare two categorical or
ordinal variables (Daniel 2009:598; Jacobsen 2011:207). It was used to identify
variables that have effect on pattern of immunological monitoring and immunological
response after HAART initiation.

3.3.4.2.2 The log-rank test

Daniel (2009:656) describes the log-rank test as a non-parametric test that is used to
compare the survival experience of two or more groups. This test was used to
supplement the graphical description and comparison of survival experiences between
groups.
 66

3.3.4.2.3 The Cox proportional hazards regression

Regression is a technique used to determine the effect of covariates on an outcome

variable. In more practical terms, it is used to predict the value of a dependent variable
given the value of independent variables (Daniel 2009:410). Cox proportional hazards
regression techniques are those techniques used to model dependent variables that
measures time to event data with or without the censoring of observations (Daniel
2009:659). For this study, the event was the first record of immunologic treatment
failure. The term ‗censoring‘ refers to the fact that observations may be incomplete in
certain individuals because they discontinued follow-up in the selected facilities as a
result of death, transfer to another facility or loss to follow-up. The hazard rate ratio is
calculated during the Cox proportional hazards regression as compared to the odds
ratio during the logistic regression or rate ratio during the Poisson regression (Daniel
2009:659). The term ‗hazard rate‘ refers to the instantaneous risk of an event happening
after a time ‗t‘ given that the subject survived up to time ‗t‘ (Hosmer, Lemeshow & May
2008:62).

3.3.4.2.3.1 Extensions of the proportional hazards regression model

In longitudinal studies, the values of some factors may vary with time. Two such
variables in this study were adherence and tuberculosis status. The variable
‗adherence‘ was assigned a value of ‗0‘ as long as patients were adherent to treatment
with respect to their appointment to HAART medication prescription refill. If they failed to
show up, this variable was assigned a value of ‗1‘ for the time of follow-up after the time
of interruption. This enabled the investigator to compare if the risk of immunologic
treatment failure varied between times after treatment interruption as compared to that
before treatment interruption. In practical terms, this required splitting the observation
time of patients into that up to the point at which treatment was discontinued, where the
value to adherence was ‗0‘, and the time period after that, where the value given to
adherence was ‗1‘. So, the time an individual spent being adherent to appointments was
taken into account. This is justified for this study since patients who failed to show up for
their medication refill in effect discontinued treatment for at least more than two weeks
and thus fell beneath the 95% adherence limit, thus putting themselves at risk of
experiencing treatment failure (WHO 2012b:10). Tuberculosis was treated in the same
way to assess if occurrence of tuberculosis was a predictor of immunologic failure. Such
 67

modelling of time-varying covariates requires the use of extensions to the Cox

proportional hazards regression modelling which was employed in this study (Collett
2003:251; Hosmer et al 2008:213;Machin, Cheung& Parmar 2006:176). In addition,
tuberculosis cases after six months of therapy were considered during regression
analysis. This was important in order to use it as proxy measure of follow-up WHO
staging during treatment (treatment or T-staging).Finally, since samples were taken
from two different sites, accounting for within site correlation was made by stratification
of hazard ratios by site. The same was done for calculating the log-rank test (Hosmer et
al 2008:208; Machin, Cheung & Parmar 2006:181).

3.3.4.2.3.2 Variable selection for Cox proportional hazards regression modelling

There are a number of techniques that can be used to develop a regression model.
These include purposeful, stepwise and best subsets selection of covariates. In the first
method variable selection is under the control of the data analyst, while in the latter two
methods, covariates are selected by statistical procedures (Hosmer et al 2008:132).
The former method of variable selection procedure was used in this study as
recommended by Hosmer et al (2008:132).

3.3.4.2.3.3 Purposeful selection of covariates

In this section, the details of purposeful selection of covariates as used in this study are
discussed as described by Hosmer et al (2008:132).

3.3.4.2.3.3.1 Step 1

In this step, a multivariable model containing all variables significant in the individual
regression analysis at the 20-25 per cent level was generated.

3.3.4.2.3.3.2 Step 2

In this step, p-values from the Wald tests of the individual coefficients were used to
identify covariates that might be deleted from the model. The deletion was made one
variable at a time. P-value of the partial likelihood ratio test was used to determine
whether the deleted covariate was significant or not.
 68

3.3.4.2.3.3.3 Step 3

In this step, the effect of the deletion of covariates on coefficients of those variables still
remaining in the model during Step 2 was evaluated. A change in coefficient was
considered significant if its coefficient changed by 20 per cent or more. If a variable that
had this much effect was deleted in Step 2, then it was added back to the model.

3.3.4.2.3.3.4 Step 4

In this step, variables left out in the initial multivariable model were added one at a time
and checked to see if that made any significant change in the model or in the
coefficients in the model. Then, partial likelihood ratio test was used to check whether
there were significant changes to the model. Any variable with such an effect was kept
in the model, thus giving a preliminary main effects model.

3.3.4.2.3.3.5 Step 5

In this step, the scale of continuous covariates found in the model was checked. This
was graphically checked by plotting Martingale residuals from a model excluding the
variable of interest and plotting that against the variable. A Lowess curve was then fitted
to aid in estimating the functional form of the covariate in the log hazard. If this line was
non-linear, then fractional polynomials analysis was used to decide if transformation
was needed. If it was linear, the variable was unchanged in the final model, but if the
transformation fit the data better, then the new term(s) was kept in the model to serve
as the main effects model.

3.3.4.2.3.3.6 Step 6

In this step, there was testing done to assess whether or not there was any interaction
of variables to be used in the final model. The interaction terms significant at the p-value
of 0.05 were used to select interaction terms. A partial likelihood ratio test was used to
decide if inclusion of such terms had statistically significant addition to the model. The
model at the end of this step was called the preliminary final model.
 69

3.3.4.2.3.3.7 Step 7

In this step, the preliminary final model was checked for key assumptions in regression
analysis in order to come up with the final model. These included checking for
collinearity between covariates, the fulfilment of proportional hazards assumptions, the
existence of influential records, and the goodness-of-fit test.

 Collinearity between covariates

Extreme collinearity between covariates may produce clinically implausible results

(Hosmer et al 2008:167). For this reason, whether there was any collinearity between
covariates in the preliminary model was checked.

 Proportional hazards assumption

This assumption entails that the hazard function for values of a covariate remain
equidistant over time (Machin et al 2006:121). Since the Cox proportional hazards
regression modelling is based on this assumption, it was necessary to test for the
fulfilment of this assumption after modelling. Hosmer et al (2008:167) suggest a two-
step procedure to test for this assumption: evaluate score test based on the scaled
Schoenfeld residuals and examine plots of the scaled and smoothed scaled Schoenfeld
residuals obtained from the model. P values for score tests of covariates and the global
test are>0.05 in cases in which this assumption is fulfilled. This evidence should be
supported by a smoothed scaled Schoenfeld residuals plotting that shows no pattern of
distribution, or must be randomly scattered about zero. In addition, the smoothed curve
should have a slope close to zero.

 Checking for influential records

In this step, regression diagnostic statistics were examined to assess the presence of
outliers or extreme values of a covariate, their influence on estimates of parameters,
and their influence on the fit of the overall model. In order to assess this, values of score
tests were plotted against the covariate of interest in a scatter plot for continuous
variables and examined for outliers. Records of outliers were further examined to see if
 70

those values were meaningful. The effect of the deletion of those values was checked to
see if it significantly changed parameter estimates. The effect on parameter estimates
was measured by calculating delta-betas. A record was said to have an influence if the
change in parameter estimate, as measured by delta-beta (Δβ), following its deletion
form the data set was more than one standard error of the parameter in the preliminary
model. In cases in which the deletion of a record had a big effect, the change in
magnitude of the hazard ratio was examined as well. Those with a very high effect
needed to be deleted before the final model was generated. (Collett 2003:242; Hosmer
et al 2008:191)

 The goodness-of-fit test

The fit of the model was evaluated by using Cox-Snell residuals. If the model fit the data
well, then the true cumulative hazard function conditional on the covariate vector was
considered to have an exponential distribution with a hazard rate of one (Collett
2003:238).

3.4 RESEARCH DESIGN AND DATA INSTRUMENT QUALITY

Research design and data instrument quality must be kept to the highest possible
standard in order to capture the true essence of reality. For this to happen, one must
address these issues properly at design or analysis stage.

3.4.1 Quality of research design

Quality of research design is reflected in internal and external validity, which are
discussed further below.

3.4.1.1 Internal validity

This refers to the extent to which it is possible to draw a conclusion from the study
because there is a true causal relationship between independent and dependant
variables (Polit & Beck 2012:244). But prior to checking internal validity, it is important to
first establish that there is a relationship between the two variables to begin with. This is
 71

termed statistical conclusion validity. The power of a study must be high to detect this
which in turn is reflected in the sample size (Polit & Beck 2012:241).

Threats to internal validity include absence of temporality, attrition, and selection bias
(Polit & Beck 2012:295; Shaughnessy et al 2012:187). In this study, temporality of
independent variables was maintained during analysis by implementing statistical
control of time-varying covariates. This means value of time-varying covariate occurring
prior to diagnosis of immunologic failure was included in analysis while that occurring
after the outcome was ignored (Collett 2003:251; Hosmer et al 2008:213). As for
attrition bias, the researcher had little control over the prevalence of attrition but
assessment was made to see if there was such bias by comparing if patients lost to
follow-up differed by baseline characteristics than those remaining in care.

Selection bias is a common threat to internal validity but in this study since cases were
randomly selected from the selected facilities and Cox regression analysis was used,
the investigator was able to control for difference in baseline characteristics (Polit &
Beck 2012:228,450; Shaughnessy et al 2012:453).

3.4.1.2 External validity

This refers to the generalizability of the findings of the study to different settings, and
populations (Shaughnessy et al 2012:94). Generalization to the accessible population in
this case would be straight forward as simple random sampling of all eligible cases was
taken (Polit & Beck 2012:250). But generalization beyond that to the target population of
HIV patients on HAART in SNNPR might not be made straight forward especially for
prevalence estimate of immunologic failure since patients from only two sites were
studied.

3.4.2 Quality of the data gathering instrument

Quality of data collection instrument refers to its validity and reliability of measuring
concepts of interest.

3.4.2.1 Validity and reliability

 72

Validity of a research instrument refers to whether an instrument is measuring what it is

supposed to be measuring (Polit & Beck 2012:336; Shaughnessy et al 2012:164). For
this study, the concepts being measured have been clearly defined (see section 1.4),
hence there was a match between concepts intended to be measured and what was
actually measured. Though data was collected retrospectively, the operational
definitions used were identical to that used in clinical practice as outlined by the working
guideline of HIV treatment. Polit and Beck (2012:332) state ―Reliability refers to the
accuracy and consistency of information obtained in a study.‖ Since this study was
retrospective and based on existing patient medical records a number of potential
measurement biases were mostly avoided. These include interviewer bias, recall bias,
and Hawthorn effect (Jacobsen 2011:58,70,83). One particular variable that needed to
be measured and has influence on the outcome variable, that is immunologic treatment
failure, was adherence. Self-reported adherence could be unreliable because of social
desirability bias. For this reason, the investigator measured adherence to appointment
to drug refill. This measure of adherence prevented misclassification error with respect
to adherence than if one relied on self-report to assess adherence. All monitoring and
evaluation tools have their own standard operating procedure which helped to uphold
consistency of data (LoBiondo-Wood & Haber 2010:279). Apart from this, data was
cross checked between medical records and registers as well as pharmacy records to
ensure consistency and accuracy of data.

3.5 ETHICAL CONSIDERATIONS

Ethics can be defined as standards that guide how a person or a professional should
behave or act (Polit & Beck 2012:154). According to the Belmont Report (Polit & Beck
2012:151) there are three main domains of ethics that must be addressed while
conducting research. These include principles of beneficence, respect for human dignity
and justice, which are discussed below in line with what was done to uphold each while
this study was being carried out (Polit & Beck 2012:152; Zaidi 2014:75).

The principle of beneficence is about maximizing benefit and minimizing harm at all
times while conducting research (LoBiondo-Wood & Haber 2010:250; Zaidi
2014:83,85). In this study there was no interaction of the researcher with study subjects
as this was a retrospective study based on existing medical records. Therefore, the
possibility of harm being inflicted during the study was remote.
 73

The next important principle stated in the Belmont Report is about the respect for
human dignity. This principle entails the researcher to respect confidences of research
subjects, and recognize that they have the right to decide for themselves (Zaidi
2014:75). It means they have the right to information about the study and the right to
decide to participate in a study or end it anytime they feel. The researcher cannot
influence their decision in any way. This is ensured through the implementation of
informed consent whenever it applies (Jacobsen 2011:154). In this study, once again,
since there was no interaction between study subjects and investigator or data collector,
there was no need for informed consent. (LoBiondo-Wood & Haber 2010:250)

The last major principle in the Belmont Report is the principle of justice. This principle
deals with the right to fair treatment and the right to privacy (Zaidi 2014:75). In this study
the privacy of patients‘ information was kept in confidence. Access to patients‘ medical
records (including electronic records) was limited to data clerks assigned and already
working in the facility. Medical records are kept in medical record room under normal
circumstances which is secured and this was maintained throughout the study. Since
random selection of patients‘ records was implemented, all eligible patients had equal
opportunity to be studied. (LoBiondo-Wood & Haber 2010:251).

3.5.1 External review

It is important for every researcher to make a risk-benefit analysis for participants when
conducting a research. However, there may be bias in balancing risk and benefit as a
result of the researcher being over enthusiastic to conduct a high quality study or as a
result of benefit one may get out of the findings of a study. For this reason, this
judgment should not be left for the researcher alone and it is important for external body
to review the protocol of a study (Jacobsen 2011:161; Polit & Beck 2012:165;
Shaughnessy et al 2012:61). For this study, ethical approval was sought from UNISA
College of Human Sciences, National Research Ethics Review Committee of Ethiopia,
ethics committee of Southern Nations Nationalities and People‘s Regional Health
Bureau of Ethiopia, and the directors of each hospital where the HIV clinics were being
used for the purpose of this study.
 74

3.6 SCOPE AND LIMITATIONS

The scope of this study was limited in many ways.

 The study focused on patients in ART care in Southern Nations Nationalities and
People‘s Region and no other regions in Ethiopia.
 The study was done among patients in selected hospitals and not health centres.
 Site selection was not random.
 Existing patient medical records were used as data source with all its limitations.

3.7 CONCLUSION

This chapter provided an in-depth discussion of the research methodology that was
applied in this study. The details of research design and research methods were
covered. The next chapter will cover the results of the study.
 75

CHAPTER 4

PRESENTATION AND DESCRIPTION OF THE RESEARCH FINDINGS

4.1 INTRODUCTION

Chapter three addressed the research methodology used in this study. This included
the research design, research methods, data management and design, data and design
quality, and ethical considerations. This chapter is dedicated to present and discuss the
research findings. The aim of the study was to describe the prevalence of immunologic
failure and identify its determinants among people living with HIV in selected hospitals.

The specific objectives included:

 To assess the pattern of immunologic monitoring in selected hospitals in

southern Ethiopia
 To assess the pattern of immunologic response of patients on HAART in
selected hospitals in southern Ethiopia
 To describe immunologic treatment failure and identify its predictors among
patients on HAART in selected hospitals in southern Ethiopia

4.2 DATA MANAGEMENT AND ANALYSIS

In this section data management are described.

4.2.1 Data management

A self-designed document checklist was used for the purpose of data collection. The
checklist consisted three sections: demographic data, clinical data, and immunologic
data.

The documents of people living with HIV on antiretroviral therapy enrolled and started
on ART from September 1, 2005 to December 31, 2012, those aged five or more years
at ART initiation, those with two or more CD4 cell count tests, and those with at least six
 76

months of follow-up were included in the study. Documents of patients transferred-in

from other facilities after starting HAART were excluded from the study. Follow-up
information was abstracted for all patients until end of August 2013.

After applying the eligibility criteria, there were 1,321 patient records. Since the study
sample was 1,304, the researcher used all the available patient records for analysis.
Data cleaning was done with Epi InfoTM 3.5.1 statistical software (United States Centers
for Disease Control and Prevention 2007). Data on patient follow-up weight was missing
in 43% of cases for which reason the variable was excluded during analysis. The data
was analysed using Stata/IC 12.0 (StataCorp LP) statistical software. The results of
analysis are presented in the next section.

4.3 RESEARCH RESULTS

The results of the study are presented in this section. Both descriptive and inferential
statistics are presented as follows.

4.3.1 Sample characteristics

In the following section, the study sample is described with respect to demographic,
clinical characteristics, follow-up time, and follow-up status.

4.3.1.1 Demographic characteristics

Demographic data collected in this study include age and gender which are presented
as follows.

4.3.1.1.1 Age

The age distribution characteristic of patients in the study sample is displayed in table
4.1 and 4.2. The mean age for both male and female patients was 32 years with a
standard deviation of 9.6. The median age was 30 years. The minimum and maximum
ages of those in the study sample were five and 71 years, respectively. Out of the total
patients in the study (N=1,321), paediatric patients between five and 14 years of age
accounted for 3.94% (n=52), while adults accounted for 96.06% (n=1,269) of it (see
 77

table 4.2). This is very close to the national figure for the proportion of patients under 15
years of age on antiretroviral therapy, which is 4.8% (FHAPCO & FMOH 2012:33).

TABLE 4.1 DESCRIPTIVE STATISTICS FOR CONTINOUS VARIABLES OF PATIENTS

ON ANTI-RETROVIRAL THERAPY (N=1321)

Standard
Variable Number Mean Minimum Median Maximum
deviation
Age 1321 32 9.6 5 30 71
Baseline CD4 cell count 1321 163 100 1 156 1221
Follow-up CD4 cell count 5241 415 234 5 372 2024

4.3.1.1.2 Gender

The gender distribution of the study sample is presented in table 4.2. Female patients
accounted for the majority of those in the study sample at 62.91% (n=831) while males
made up 37.09% (n=490) of the sample. This is in line with the proportion of women
with HIV in Ethiopia which, according the FHAPCO and FMOH (2012:32), is estimated
at 60% of the total population in 2011.

TABLE 4.2 CHARCTERISTICSOF PATIENTS ON ANTI-RETROVIRAL THERAPY,

(N=1,321)
Variable Category Frequency, n Per cent,%
Age in years 5-14 52 3.94
>14 1269 96.06
Gender Female 831 62.91
Male 490 37.09
WHO Stage I or II 535 40.50
III or IV 786 59.50
Baseline CD4 (cell <100 378 28.61
3
count/mm ) 100-199 547 41.41
200-349 359 27.18
>349 37 2.80
Tuberculosis No 1086 82.21
Yes 235 17.79
Adherence Good 1080 81.76
Poor 241 18.24
Final follow-up Active 960 72.67
status Dead 44 3.33
Drop 62 4.69
Lost 4 0.30
Stop 2 0.15
Transferred out 249 18.85

4.3.1.2 Clinical information

This section provides data on clinical characteristics of those in the study sample.
 78

4.3.1.2.1 Baseline WHO Stage

Out of the total number of patients‘ files examined in this study, those patients at WHO
Stages I or II during initiation of anti-retroviral therapy accounted for 40.5% (n=535)
while those at advanced disease stages (WHO Stages III or IV) accounted for 59.5%
(n=786) (see table 4.2). This is similar to what was reported in the SNNP region by
Mulissa et al (2010:online) who reported that the proportion of patients who started
antiretroviral therapy who were at WHO stages III or IV at HAART initiation being 62.5%
(n=1,369) at one hospital in the SNNP. However, it was different from what Tsegaye
and Worku (2011:104) reported. In their study, 81.2% (n=4,580) of patients in WHO
stages III or IV had started ARVs in eight hospitals in the SNNP region. This finding is
higher than that reported in this study.

Figure 4.1 Baseline CD4 cell count distribution for the study sample (N=1,321)

4.3.1.2.2 Baseline CD4 cell count

 79

The mean CD4 cell count at the point of initiation of anti-retroviral therapy was 163
cells/mm3, while the standard deviation was 100. The median CD4 cell count at initiation
of anti-retroviral therapy was 156 cells/mm3. The minimum and maximum baseline CD4
cell counts were one and 1,221 cells/mm3 respectively (see table 4.5). The distribution
of baseline CD4 cell counts as well as the expected normal distribution for those in the
study sample is shown in figure 4.1. These findings show a distribution skewed to the
right. Table 4.2 shows the percentage of patients with a baseline CD4 cell count
between 100 and 199 which accounts for the highest proportion of 41.41% (n=547),
followed by those below 100, 28.61% (n=378), and those between 200 and 349,
27.18% (n=359). Of the total sample only 2.8% (n=37) had a CD4 cell count of 350 and
above. The proportion of patients with a baseline CD4 cell count of ≥200 was similar to
what was reported by Tsegaye and Worku (2011:104) who found that 29.2%of a total
sample of 5,181 had CD4 cell count above that (n=1,516).

4.3.1.2.3 Diagnosis of tuberculosis during follow-up

Out of 1,321 patients, 235 were diagnosed as having tuberculosis during and after
starting anti-retroviral therapy (see table 4.2). This put the prevalence of tuberculosis
after initiation of anti-retroviral therapy at 17.79%. Reported prevalence of tuberculosis
varied among those on HAART in different studies done in Ethiopia. Alene, Nega and
Taye (2013:online) report a higher prevalence in their study namely that 28% of a
sample of 470 patients on ART followed for a maximum of five years in northern
Ethiopia, while it was 21% in a multisite study in southern Ethiopia involving 6,230
patients by Yirdaw, Jerene, Gashu, Edginton, Kumar, Letamo, Feleke, Teklu, Zewdu,
Weiss and Ruff (2014:online). In yet another study done by Ali and Klotz (2012:200)
who followed patients for two years, it was 14% (n=143). In the latter study, the
prevalence was smaller probably because of the short duration of follow-up. The
prevalence of tuberculosis was even smaller (3.3%) in another study in SNNP
among185 patients followed up for three to 68 weeks (Jerene, Næss & Lindtjørn
2006:online).

4.3.1.2.4 Adherence to medication refill

Of 1,321 patients, 18.24% (n=241) exhibited poor adherence to anti-retroviral therapy,

the measure for adherence being adherence to appointment for their medication refill.
 80

Patients who had used up all their medication, and had not reported to the clinic for a
refill in a timely fashion, were declared to have poor adherence to treatment. For this
study cohort, 81.76% (n=1,080) had good adherence to picking up their medication refill
on time (see table 4.2). This is less than what was reported by Chalker et al
(2010:online), who found adherence to medication refill (termed as adherence to
appointment date in their study) to be 86.3%. This higher percentage could, however,
be attributable to the shorter duration of follow-up in that study, which was one year.

4.3.1.2.5 Final follow-up status

The final follow-up status, that is, whether patients were still in follow-up or not at the
time of the last observation, is presented in this section with respect to demographic
and baseline clinical characteristics. In general, as shown in Table 4.2, there were
72.67% (n=960) patients who were still on follow-up at the time of the last observation.
This was higher than what was reported by Mulissa et al (2010:online) which found it to
be 63.1% (n=901). The percentage of patients transferred for follow-up to other facilities
was 18.85% (n=249). The percentage of patients who died, dropped out, were lost from
follow-up, or stopped treatment was 8.48% (n=112) (see table 4.2).

TABLE 4.3 FINAL FOLLOW-UP STATUS BY DEMOGRAPHIC AND BASELINE

CHARACTERISTICS OF STUDY PATIENTS ON ANTI-RETROVIRAL
THERAPY (N=1,321)
Variable Category Total, n Number Active (%) Pearson Chi Square P- value
Age in years 5-14 52 38 (73.08) 0.0045 0.947
>14 1269 922 (72.66)
Gender Female 831 599 (72.08) 0.3932 0.531
Male 490 361 (73.67)
WHO Stage I or II 535 392 (73.27) 0.1623 0.687
III or IV 786 568 (72.26)
Baseline CD4 <100 378 270 (71.43) 2.8564 0.414
3
(cell count/mm ) 100-199 547 409 (74.77)
200-349 359 257 (71.59)
>349 37 24 (64.86)

The following section presents findings concerning the comparison of the final follow-up
status with respect to demographic and baseline clinical features.

4.3.1.2.5.1 Final follow-up status by age group

 81

The percentage of patients actively on follow-up ages 5-14 years was 73.08% (n=38),
while that for those more than 14 years of age was 72.06% (n=922) (see table 4.3). This
indicates that there was no difference in the percentage of patients who were actively
on follow-up based on age group (p-value > 0.05).

4.3.1.2.5.2 Final follow-up status by gender

The percentage of patients actively on follow-up was 72.08% (n=599) for females while
for males it was 73.67% (n=361) (see table 4.3). This indicates that there was no
significant difference in the percentage of patients who were actively on follow-up based
on gender (p-value > 0.05).

4.3.1.2.5.3 Final follow-up status by baseline WHO Stage

The percentage of patients actively on follow-up, for patients at baseline WHO Stages I
or II, was 73.27% (n=392), while for those at baseline WHO Stages III or IV was 72.26%
(n=568) (see table 4.3). This indicates that there was no significant difference in the
percentage of patients who were actively on follow-up based on baseline WHO Stages
(p-value > 0.05).

4.3.1.2.5.4 Final follow-up status by baseline CD4 cell count

The highest percentage of patients actively on follow-up was recorded for patients with
a baseline CD4 cell count between 100 and 199, 74.77% (n=409), while the lowest
percentage was for patients with baseline CD4 cell counts of more than 349, at 64.86%
(n=24). Of patients receiving the treatment, those with baseline CD4 cell counts below
100 and baseline CD4 cell counts between 200 and 249, the percentages of those
actively on follow-up was comparable 71.43% (n=270) and 71.59% (n=257)
respectively. Overall, the percentages of patients active on follow-up from the different
categories of baseline CD4 cell count just described was comparable (p-value > 0.05)
(see table 4.3).

4.3.1.2.6 Follow-up time

 82

All in all, patients were observed for a minimum time of 0.51 years and a maximum of
8.36 years. The mean and median duration of follow-up were 4.06 and 3.89 years
respectively. Patients were observed for a collective time of 5,366 years from ART
initiation to last visit.

4.3.2 Objective 1: assess the pattern of immunologic (CD4 cell count) monitoring
in selected hospitals in southern Ethiopia

The pattern of timely CD4 cell count monitoring for this study sample is depicted
graphically in figure 4.2.The ‗expected CD4 cell count tests‘ represents the number of
follow-up CD4 cell count tests that should have been determined had the national
monitoring and evaluation guidance for determination of CD4 cell count been
implemented correctly as described in the methods section. This number is seen to get
smaller with increase in follow-up time as there were progressively fewer patients on
follow-up. For example, the number of patients active on follow-up for seven years or 84
months was 136, while there were 1,262 patients who were active for at least one year
(see table 4.4).

Figure 4.2 Pattern of CD4 cell count monitoring by follow-up time (N=1,321)
 83

The ‗actual CD4 cell count tests received‘ represents the actual number of CD4 cell
count tests determined at designated follow-up times.

TABLE 4.4 DETERMINATION OF TIMELY FOLLOW-UP CD4 TESTING BY FOLLOW-UP

PERIOD IN MONTHS AMONG PATIENTS ON ANTI-RETROVIRAL THERAPY
(N=1,321)

Follow-up Month Expected number of Number of follow-up CD4 Per cent (%) CD4
CD4 cell count tests cell count tests provided cell count tests
provided
0 1321 1321 100.00
6 1317 610 46.31
12 1262 417 33.04
24 1053 179 17.00
36 842 132 15.68
48 628 108 17.20
60 459 68 14.81
72 271 40 14.76
84 136 10 7.35
Total follow-up CD4 cell 7289 2885 40.58
count tests

Although both the number of patient‘s on follow-up (or expected number of CD4 cell
count tests) as well as the number of timely follow-up CD4 cell count tests determined
decreased progressively, the proportion of timely follow-up CD4 cell count tests
provided decreased much faster. The actual percentage of follow-up CD4 cell count
tests provided is shown in table 4.4. It can be seen that all patients had baseline CD4
cell count testing at month zero. The number of timely CD4 cell count tests performed
decreased to 46.31% (n=610) at six months of follow-up while it was 33.04% (n=417) at
one year after ART initiation. It decreased and reached 7.35% (n=10) for patients who
were in follow-up for seven years. Overall, of the expected 7,285 baseline and follow-up
CD4 cell count tests, only 2,885 tests were done in a timely fashion. This was 40.58%of
the total expected, which is very low.

Berhan (2011b:199) reports low follow-up CD4 cell count monitoring as well among
patients studied in southern Ethiopia. According to that study, at six and 24 months after
initiating HAART, follow-up CD4 cell count monitoring was done for 68% and 37% of a
total sample of 1,163 patients. The findings from the current study indicate that the level
of monitoring for the same follow-up time interval from ART initiation was much lower, at
46.31% (n=610) and 17% (n=179) respectively. In a recent study in Malawi
(Palchaudhuri, Tweya & Hosseinipour 2014:43), only 5,361 ART patients out of 17,737
 84

(30.2%) on follow-up received follow-up CD4 cell count testing. And out of 15,924
patients who were eligible for two tests, only 1,006 (6.3%) received testing as per
guideline. The reasons for poor follow-up CD4 cell count testing could be multifactorial:
providers not requesting tests, patients not coming for testing, lack of awareness from
the patients‘ side, or a breakdown of machines or lack of reagents at facilities. This is
consistent with the opinions of Were et al (2011:152) and Zachariah et al (2011:39).

The following sections present an analysis on the pattern of follow-up CD4 cell count
monitoring with respect to whether there were more or less timely follow-up CD4 cell
count tests based on demographic and baseline clinical characteristics.

4.3.2.1 Pattern of CD4 cell count monitoring by age

Table 4.5 shows a comparison of CD4 cell count monitoring by age group. Among
patients ages 5-14 years, follow-up CD4 cell count tests were done for 35.97% (n=100),
and for those patients older than 14 years of age, 39.73% (n=2,786) had follow-up tests
(p value>0.05). This shows that there was no substantial difference in provision of
timely follow-up CD4 cell count testing by age group.

TABLE 4.5 DETERMINATION OF TIMELY FOLLOW-UP CD4 CELL COUNT TESTING

BY DEMOGRAPHIC AND BASELINE TREATMENT VARIABLES AMONG
PATIENTS ON ANTI-RETROVIRAL THERAPY (N=1,321)
Variable Category Total expected Number with Per cent timely Chi P value
number of CD4 timely follow- follow-up CD4 square
cell count tests up CD4 cell cell count test
count test done
Age in years 5-14 278 100 35.97 1.5812 0.209
>14 7012 2786 39.73
Gender Female 4578 1806 39.45 0.0993 0.753
Male 2712 1080 39.82
WHO Stage I or II 2897 1150 39.70 0.0233 0.879
III or IV 4393 1736 39.52
Baseline CD4 <100 2163 860 39.76 0.2358 0.972
3
(cell count/mm ) 100-199 3089 1220 39.49
200-349 1843 726 39.39
>349 195 80 41.03

4.3.2.2 Pattern of CD4 cell count monitoring by gender

Table 4.5 shows a comparison of CD4 cell count monitoring by gender. Among female
patients, 39.45% (n=1,806) of expected follow-up CD4 cell count tests were done, while
 85

it was done for 39.82% (n=1,080) of male patients (p value>0.05). This shows that there
was no significant difference in provision of timely follow-up CD4 cell count testing by
gender.

4.3.2.3 Pattern of CD4 cell count monitoring by baseline WHO Stage

Table 4.5 shows a comparison of CD4 cell count monitoring by baseline WHO Stage
groups. Follow-up CD4 cell count tests were done for 39.7% (n=1,150) of patients at
baseline WHO Stage I or II. It was almost the same for those with baseline WHO Stage
III or IV. This shows that there was no difference in provision of timely follow-up CD4
cell count testing by baseline WHO Stage (p value>0.05).

4.3.2.4 Pattern of CD4 cell count monitoring by baseline CD4 cell count

Table 4.5 shows a comparison of CD4 cell count monitoring by baseline CD4 cell count
groups. Patients with a baseline CD4 cell count of more than 349 had the highest
percentage of follow-up CD4 cell count testing namely 41.03% (n=80), while for the
other groups, it was between 39% and 40% (n=860, 1,220, and 726 for those with
baseline CD4 cell count<100, between 100 and 199, and between 200 and 349
respectively). Overall, there was no association between follow-up CD4 cell count
testing and baseline CD4 cell count (p value>0.05).

4.3.3 Objective 2: assess the pattern of immunologic (CD4 cell count) response
for patients on HAART in selected hospitals in southern Ethiopia

Figure 4.3 shows the pattern of immunologic (CD4 cell count) response after the
initiation of anti-retroviral therapy (N=1,321). The total number of follow-up CD4 count
tests done was 5,241 over the study period of eight years. It can be seen that the
median CD4 cell count for those who made up the study sample showed a rapid
increase over the first six months. CD4 cell count recovery continued after that, but with
slight subsequent change, followed by a levelling off after 60 months (five years) of
follow-up. This is similar to what was described by Tsegaye and Worku (2011:107),
Reda et al (2013:online), Kassa et al (2013:online), and more recently by Mulu, Liebert
and Maier (2014:online).
 86

The pattern of immunologic response was further analysed by demographic (age and
gender) and baseline clinical characteristics (baseline WHO Stage and baseline CD4
cell count) presented in the following sections taking as an outcome the percentage of
those patients in whom the follow-up CD4 cell count achieved was more than or equal
to 350 (see table 4.7). The median follow-up CD4 cell count was used as guide to
decide what cut off to use. Since the median follow-up CD4 cell count was 371, the
researcher chose 350 as an approximation for ease of interpretation and
communication of results.

Figure 4.3 Box plot for magnitude of baseline and follow-up CD4 cell count among
cohort of patients on anti-retroviral therapy by follow-up month (N=1,321)

4.3.3.1 Pattern of CD4 cell count response by age

In table 4.6 the follow-up CD4 cell count tests amounted to 5,241 for the total study
sample of 1,321 patients during the study period.

The percentage of patients whose age was between five and fourteen inclusive and
who had a follow-up CD4 cell count of more than or equal to 350 was 88.48% (n=146).
 87

This was in contrast to the percentage of those whose age was more than 14 years,
among whom 53.49% (n=2,715) had such a follow-up CD4 cell count. This finding
suggests that the percentage of those whose follow-up CD4 cell count values was more
than or equal to 350was greater for those ages 5-14 years (p value<0.05) (see table
4.6).

TABLE 4.6 MAGNITUDE OF FOLLOW-UP CD4 CELL COUNT BY DEMOGRAPHIC AND

BASELINE TREATMENT VARIABLES AMONG PATIENTS ON ANTI-
RETROVIRAL THERAPY (N=1,321)
Variable Category Total, n Follow-up Per cent with Chi square P value
CD4 cell follow-up CD4 cell
count ≥350 count ≥350
Age in years 5-14 165 146 88.48 78.9601 0.000
>14 5076 2715 53.49
Gender Female 3330 2006 60.24 117.6648 0.000
Male 1911 855 44.74
WHO Stage I or II 2079 1193 57.38 10.8558 0.001
III or IV 3162 1668 52.75
Baseline CD4 <100 1569 512 32.63 568.0106 0.000
3
(cell count/mm ) 100-199 2224 1253 56.34
200-349 1300 983 75.62
>349 148 113 76.35

4.3.3.2 Pattern of CD4 cell count response by gender

The percentage of female patients with a follow-up CD4 cell count more than or equal to
350 was 60.24% (n=2,006). This was in contrast to male patients, for whom the
percentage of follow-up CD4 cell count test result of more than or equal to 350 was
44.74% (n=855). This finding suggests that the percentage of patients with higher
follow-up CD4 cell count values was greater for female patients (p value<0.05) (see
table 4.6). This finding is consistent with the findings of Kassa et al (2013:online) who
report a steeper increase in follow-up CD4 cell counts results for female patients than
those for males.

4.3.3.3 Pattern of CD4 cell count response by baseline WHO Stage

Table 4.6providesa comparison by percentage of follow-up CD4 cell count levels of

≥350 by baseline WHO Stage groups. The percentage of patients whose baseline WHO
Stage was I or II who had follow-up CD4 cell counts of more than or equal to 350 was
57.38% (n=1,193). This was in contrast to those whose baseline WHO stage was III or
IV, 52.75% (n=1,668) of whom had follow-upCD4 cell count ≥350. This finding suggests
 88

that the percentage of those whose follow-up CD4 cell count values was higher was
greater for those at baseline WHO Stage I or II (p value<0.05).

4.3.3.4 Pattern of CD4 cell count response by baseline CD4 cell count

Table 4.6 also provides a comparison by percentage of those with follow-up CD4 cell
counts by baseline CD4 cell count groups. Patients with baseline CD4 cell count below
100 had the lowest percentage of follow-up CD4 cell count values ≥350, 32.63%
(n=512). The next lowest percentage was for patients with a baseline CD4 cell count
between 100 and 199, among whom 56.34% (n=1,253) had follow-up CD4 cell count
≥350. The other two baseline CD4 cell count groups had comparable percentages with
at around 76% each for those with a baseline CD4 cell count of 200-349 (n=983) and
>349 (n=113) respectively. Overall, the percentage of patients with follow-up CD4 cell
count tests more than or equal to 350 was statistically different among the different
baseline CD4 cell count groups (p value<0.05). This finding is supported by Kassa et al
(2013:online) who report better immune restoration among patients with baseline CD4
cell counts higher than 200 as compared to those with counts less than 200. Kanters et
al (2013:online) also report the baseline CD4 cell count to be a strong predictor of
immunological response, stating that only those with CD4 cell counts above 200 were
able to achieve follow-up CD4 cell counts of more than 500.

In conclusion, the association between demographic and baseline clinical

characteristics with magnitude of follow-up CD4 cell count was evident in this study. In a
study done in eastern Ethiopia by Reda et al (2013:online), there was no such
association evident between the variables stated above and CD4 cell count increment.
The findings were different, however, probably because different outcome parameters
were used in the form of CD4 cell count increment in the above mentioned study in
contrast to achieving a follow-up CD4 cell count of greater than or equal to 350 used in
the current study. In a more recent study in northern Ethiopia by Mulu et al (2014:online)
that used a CD4 cell count of 200 as cut off amongst 100 adults and 100 children, no
association was found in contrast to this study. The findings of the current study are,
however, comparable to a study done in Uganda among 325 patients on ART
(Crawford, Wakabi, Magala, Kibuuka, Liu & Hamm 2014:online) where they found out
that age was inversely related to CD4 cell count recovery and female patients had
higher baseline CD4 cell count as well as higher recovery during follow-up.
 89

4.3.4 Objective 3: describe immunologic treatment failure and identify its

predictors among patients on HAART in selected hospitals in southern
Ethiopia

Of the 1,321 patients‘ medical records reviewed, 2.95% (n=39) discontinued their
treatment for considerable time. Such patients would have needed additional follow-up
before treatment failure diagnosis could be made, for which reason they were not
evaluated for treatment failure (see table 4.7).

TABLE 4.7 IMMUNOLOGIC TREATMENT FAILUREAMONG PATIENTS ON ANTI-

RETROVIRAL THERAPY (N=1,321)
Variable Category Frequency, n Per cent,%
Immunologic failure (first instance) No 1056 79.94
Yes 226 17.11
Not Evaluated 39 2.95
Immunologic failure (at last immunologic evaluation) No 1135 85.92
Yes 147 11.13
Not Evaluated 39 2.95

4.3.4.1 Prevalence of immunologic failure

After excluding those not evaluated, the prevalence of patients that ever experienced
immunologic treatment failure was found to be 17.6% (95% confidence interval: 15.6%-
19.8%) (n=226). The prevalence of immunologic failure at the last immunologic
evaluation was 11.5% (95% confidence interval: 9.8%-13.3%) (n=147). This is almost
comparable to what was reported in Ethiopia by Bacha et al (2012:online), a percentage
of immunologic failure of 8.2% (n=79), as well as Workneh et al (2009:78), who reported
immunologic failure percentage of11.5% (n=11). Direct comparison is not possible,
though; since both studies focused on paediatric patients with a mean age of six years,
and follow-up duration was shorter in the second study (a mean follow-up duration of
13.7 months). In other sub-Saharan African countries, studies point to a higher
prevalence of immunologic failure, at a range of 10%-32%, even when the follow-up
duration was shorter compared to that of this study (El-Khatib et al 2011:online; Jaka et
al 2009:6; Rawizza et al 2011:1286; Reynolds et al 2009:3).

4.3.4.2 Immunologic failure by criteria of diagnosis

 90

Table 4.8 provides information on the type of criteria that was met to diagnose
immunologic failure. Among 226 patients with immunologic failure, looking further into
the criteria that led to the diagnosis indicates that the majority namely 57.96% (n=131)
were diagnosed with failure based on the criteria ‗drop by 50% of follow-up CD4 cell
count from peak value‘. Diagnosis by the criteria ‗drop to or below baseline CD4 cell
count‘ accounted for 34.51% (n=78) of patients. The criteria used for the diagnosis of
immunologic failure for the remaining 7.52% (n=17) of cases was a follow-up CD4 cell
count persistently below 100 CD4 cells/mm3 of blood.

TABLE 4.8 DIAGNOSTIC CRITERIA FOR FIRST INSTANCE IMMUNOLOGIC FAILURE

AMONG STUDY PATIENTS ON ANTI-RETROVIRAL THERAPY IN (n=226)

Diagnostic criteria Frequency Per cent

Follow-up CD4 cell count dropped by at least 50% from peak value 131 57.96
3
CD4 cell count persistently below 100/mm 17 7.52
Follow-up CD4 cell count dropped to or below baseline 78 34.51
Total 226 100.00

4.3.4.3 Immunologic failure at the final immunologic evaluation (n=226)

Table 4.9 shows that, among patients with first instance immunologic failure (that is
patients who ever experienced immunologic failure), follow-up immunologic evaluation
was not done for 23.89% (n=54) of the patients. Of the remaining patients, immunologic
treatment failure diagnosis persisted and was recorded at last evaluation in 54.07%
(n=93), while 45.93% (n=79) of patients did not fulfil the immunologic failure criteria
anymore. This may be due to improvement in adherence to medication after the first
treatment failure diagnosis. This statement is consistent with the views expressed by
the WHO (2013:183).

TABLE 4.9 FOLLOW-UP IMMUNOLOGIC EVALUATION AND FINAL IMMUNOLOGIC

STATUS OF PATIENTS WHO EVER HAD IMMUNOLOGIC FAILURE (n=226)

Immunologic failure at last evaluation

Follow-up CD4 cell count testing * * *
Yes, number (% ) No, number (% ) Total, number (% )
No 54 (100.00) 0 (0) 54 (100%)
Yes 93 (54.07) 79 (45.93) 172 (100%)
Total 147 (65.04) 79 (34.96) 226 (100%)
*
All percentages are row percentage.
 91

4.3.4.4 Patients on second line HAART among those with confirmed

immunologic failure

Of the 93 patients with confirmed immunologic failure, 7.53% (n=7) were put on second
line anti-retroviral therapy (see table 4.10). The rest (92.47% and n=86) remained on
first line therapy for as long as they were studied. Among all patients who were followed
(N=1,321), the prevalence of patients on second line HAART treatment was 1.67%
(95% confidence interval: 1.0% - 2.4%) (n=22).

This finding is comparable to what was reported by Assefa et al (2011:online), which in

their study found that among 7,451 patients, 2.13% (95% confidence interval: 1.8% to
2.5%) at the second year of HAART treatment were on second line HAART.

TABLE 4.10 FREQUENCY DISTIBUTION OF PATIENTS ON SECOND LINE TREATMENT

AMONG THOSE WITH CONFIRMED IMMUNOLOGIC FAILURE (n=93)

On second line HAART Frequency, n Per cent, %

No 86 92.47
Yes 7 7.53
Total 93 100.00

4.3.4.5 Predictors of immunologic failure

Predictors of immunologic failure were examined with survival analysis. Specifically, the
time from onset of treatment to first instance immunologic failure was examined. Firstly,
descriptive survival analysis was done. Then, Kaplan-Meier survival curves were
examined in univariable analysis for effect of independent variables on immunologic
treatment failure. The log-rank test was used to statistically compare different groups.
This was then followed by multivariable analysis using Cox proportional hazards
regression modelling. All results were adjusted for site level effects by stratification.
Patients not evaluated for treatment failure were excluded (n=39). This left97%
(n=1,282) of patients to be included in survival analysis.

4.3.4.5.1 Event free survival

Follow-up time for patients included in the study from start of observation to diagnosis of
immunologic failure or censoring ranged between 0.51 and 8.36 years; the mean and
 92

median duration of follow-up time were 3.61 and 3.42 years respectively. The patient
data scrutinized amounted to a total of 4,634 person-years at censoring or at diagnosis
of first instance of immunologic failure (see table 4.11).Table 4.12 and figure 4.4
summarize event free survival, the event being the first instance of immunologic
treatment failure. It can be deduced that there had been a more or less steady drop in
event free survival with time. At the end of the second year of follow-up, the probability
of being free of immunologic failure was 0.89 (95% CI: 0.87 – 0.91), while at end of five
years it was 0.80 (95% CI: 0.77 - 0.82). The confidence interval widens after this, as the
number of patients who had reached more than five years of follow-up progressively
decreased, leading to more uncertainty of estimates.

TABLE 4.11 DESCRIPTION OF FOLLOW-UP TIME IN YEARS FOR STUDY PATIENTS

ON ANTI-RETROVIRAL THERAPY I (n=1,282)

No. of
Total time at risk Mean Minimum Median Maximum
subjects
Failures 226 543.08 2.40 0.51 1.80 7.92
Censored 1056 4091.23 3.87 0.51 3.72 8.36
Total 1282 4634.32 3.61 0.51 3.42 8.36

TABLE 4.12 LIFE TABLE FOR IMMUNOLOGIC FAILURE AMONG PATIENTS ON ANTI-
RETROVIRAL THERAPY (n=1,282)

Follow-up Number of Immunologic Censored Event free Standard 95%

Interval in Patients at Failure Survival Error Confidence
Years Beginning Interval
of Interval
0-1 1282 67 55 0.9466 0.0063 0.9326 - 0.9577
1-2 1160 60 184 0.8934 0.0090 0.8744 - 0.9097
2-3 916 33 172 0.8579 0.0105 0.8358 - 0.8772
3-4 711 21 179 0.8289 0.0119 0.8041 - 0.8509
4-5 511 18 143 0.7950 0.0139 0.7662 - 0.8206
5-6 350 16 140 0.7495 0.0171 0.7141 - 0.7812
6-7 194 8 93 0.7089 0.0214 0.6646 - 0.7484
7-8 93 3 71 0.6719 0.0290 0.6114 - 0.7251
 93

Figure 4.4 Kaplan-Meier survival functions for survival free of immunologic

treatment failure (n=1,282)

4.3.4.5.2 Univariable analysis

In this section, results of univariable survival analysis are presented.

TABLE 4.13 STRATIFIED LOG-RANK TEST FOR EQUALITY OF FAILURE FUNCTIONS

BY COVARIATES AMONG PATIENTS ON ANTI-RETROVIRAL THERAPY
(n=1,282)
Variable Category Events Events Chi square P value
Observed Expected
Age(in years) 5-14 10 7.59 0.79 0.3742
>14 216 218.41
Gender Female 133 142.33 1.65 0.1989
Male 93 83.67
WHO Stage I or II 77 91.20 3.71 0.0540
III or IV 149 134.80
Baseline CD4 <100 60 69.79 51.19 0.0000
3
(cell count/mm ) 100-199 78 98.03
201-349 71 54.04
>349 17 4.14
Tuberculosis No 223 222.87 0.01 0.9358
Yes 3 3.13
Adherence Good 187 186.20 0.02 0.8896
Poor 39 39.80
 94

Figure 4.5 Kaplan-Meier survival functions by age group (n=1,282)

4.3.4.5.2.1 Age and immunologic failure

Figure 4.5 shows the Kaplan-Meier survival function by age. The probability of event
free survival was similar for the age groups under comparison since their survival
curves remained close to one another. The log rank test for equality of these functions
adjusted for site indicated that the two age groups under comparison had a similar
probability of event free survival (p value>0.05) as shown in table 4.13.

4.3.4.5.2.2 Gender and immunologic failure

Figure 4.6 shows the Kaplan-Meier survival function by gender. The probability of event
free survival for was different, especially after two years. The probability of event free
survival was higher for female patients after two years, reaching around 0.68 at year
eight as compared to 0.63 for male patients. However, overall, the log rank test of
equality of these functions adjusted for site indicated that the difference in probability of
 95

event free survival was statistically not significant (p value >0.05) as shown in table
4.13.

Figure 4.6 Kaplan-Meier survival functions by gender (n=1,282)

4.3.4.5.2.3 Baseline WHO Stage and immunologic failure

Figure 4.7 shows Kaplan-Meier survival function by baseline WHO stage. The
probability of event free survival for patients at WHO stage III or IV was lower
throughout follow-up as compared to those at WHO stages I or II reaching around 0.65
at year eight for the former group in contrast to slightly above that for the latter group.
The log rank test of equality of these functions adjusted for site indicates that the
difference in probability of event free survival was statistically not significant at the 5%
level (p value= 0.054), but was significant at the 10% level, as shown in table 4.13.
 96

Figure 4.7 Kaplan-Meier survival functions by WHO Stage (n=1,282)

4.3.4.5.2.4 Baseline CD4 cell count and immunologic failure

Figure 4.8 shows the Kaplan-Meier survival function by baseline CD4 cell count group.
The probability of event free survival for patients with a baseline CD4 cell count ≥350
was higher, followed by those whose baseline CD4 cell count was <100 or 200-349.
Patients with a baseline CD4 cell count between 100 and199 had a curve slightly higher
than the rest. The log rank test for equality of these functions adjusted for site indicates
that the difference in probability of event free survival was statistically significant for
those in the various baseline CD4 cell count groups (p value < 0.05), as shown in table
4.13. The log-rank test for trend was also statistically significant (p value < 0.05).
 97

Figure 4.8 Kaplan-Meier survival functions by baseline CD4 cell count (n=1,282)

4.3.4.5.2.5 Tuberculosis disease and immunologic failure

Figure 4.9 shows Kaplan-Meier survival function by diagnosis of tuberculosis disease

after the first six months on HAART. The probability of event free survival for those
diagnosed with tuberculosis as compared to those without tuberculosis was very similar.
The log rank test of equality of these functions adjusted for site indicates that the
difference in probability of event free survival was not statistically significant (p value
<0.05) as shown in table 4.13. But in general, there were far too few events of
immunological failure recorded among patients with tuberculosis, so generalization
should be made cautiously.
 98

Figure 4.9 Kaplan-Meier survival functions by diagnosis of tuberculosis disease

(n=1,282)

4.3.4.5.2.6 Adherence to medication refill and immunologic failure

Figure 4.10 shows the Kaplan-Meier survival function by adherence to medication refill.

Figure 4.10 Kaplan-Meier survival functions by adherence (n=1,282)

 99

The probability of event free survival for those with good and poor adherence to
medication refill remains close to one another. The log rank test of equality of these
functions adjusted for site indicates that the difference in probability of event free
survival was not statistically significant (p value <0.05) as shown in table 4.13.

4.3.4.5.3 Multivariable analysis with Cox proportional hazards regression modelling

First, results from individual regression analysis are presented in table 4.14. Age and
baseline CD4 cell count were treated as continuous variables throughout the regression
modelling. It can be seen that baseline CD4 cell count and baseline WHO Stage were
significant at 5% level of significance at individual level.

TABLE 4.14 INDIVIDUAL COVARIATE COX PROPORTIONAL HAZARDS REGRESSION

ANALYSIS FOR PREDICTORS OF IMMUNOLOGIC FAILURE AMONG
PATIENTS ON ANTI-RETROVIRAL THERAPY (n=1,282)

Covariate Coefficient Standard Wald test Wald P value 95% confidence interval
error
Age 0.0058159 0.0070915 0.82 0.412 -0.0080832 - 0.0197150
Gender 0.1542121 0.1355448 1.14 0.255 -0.1114507 - 0.4198749
Stage 0.2957729 0.1416056 2.09 0.037 0.0182310 - 0.5733148
Baseline CD4 0.0029675 0.0004540 6.54 0.000 0.0020776 - 0.0038574
cell count
Tuberculosis -0.0216643 0.5867462 -0.04 0.971 -1.1716660 - 1.1283370
Adherence* 0.1125241 0.4546164 0.25 0.805 -0.7785077 - 1.0035560
* Adherence to medication refill

4.3.4.5.3.1 Step 1

In this step, a multivariable model containing all variables significant in the individual
analysis at the 20-25 % level were included, as shown in table 4.15.

TABLE 4.15 INITIAL MULTIVARIABLE MODEL FOR PREDICTORS OF IMMUNOLOGIC

FAILURE AMONG PATIENTS ON ANTI-RETROVIRAL THERAPY (n=1,282)
Covariate Coefficient Standard error Wald test Wald P value 95% confidence interval
WHO Stage 0.2827662 0.1419603 1.99 0.046 0.0045292 - .5611118
Baseline CD4 0.0028975 0.0004497 6.44 0.000 0.0020161 - .0037790
cell count
 100

4.3.4.5.3.2 Step 2

TABLE 4.16 PARTIAL LIKELIHOOD RATIO TEST FOR THE EFFECT OF REMOVAL OF
COVARIATES ONE AT A TIME FROM THE INITIAL MULTIVARIABLE
MODEL (n=1,282)
Models -2 x log likelihood -2 x log likelihood for Degree of Likelihood P value
for initial model model without freedom ratio Chi
covariate square
WHO Stage 2660.15 2664.22 1 4.07 0.0436
Baseline CD4 2660.15 2688.58 1 28.43 0.0000
cell count

In this step, p-values from the Wald tests of the individual coefficients were used to
identify covariates that might be deleted from the model. The deletion was made one
variable at a time. The p-value of the partial likelihood ratio test was used to determine
whether or not the deleted covariate made any significant change in the model. The
finding is as shown in table 4.16. Deletion of both variables results in statistically
significant change, for which reason both were kept in the model.

4.3.4.5.3.3 Step 3

Since no variable was removed during Step 2, checking for the effect of confounding
was not made at this stage.

4.3.4.5.3.4 Step 4

In this step, variables left out from inclusion in the initial multivariable model were added
one at a time and checked to see if that made any significant change in the model or
changed the coefficients in the model. The partial likelihood ratio test revealed that the
addition, one at time, of the above mentioned covariates didn‘t result in significant
change at p value of 0.05 (see table 4.17),nor did any covariate added affect the
coefficient of variables in the model by more than 20%. For this reason, all added
variables were left out (see tables 4.18 and 4.19). So, the preliminary main effects
model consisted of only the WHO Stage and the baseline CD4 count.
 101

TABLE 4.17 PARTIAL LIKELIHOOD RATIO TEST FOR THE EFFECT OF THE ADDITION
OF COVARIATES NOT INCLUDED IN STEP 1, ONE AT A TIME, TO THE
INITIAL MULTIVARIABLE MODEL (n=1,282)

Models -2 x log -2 x log Degree Likelihood P value

likelihood for likelihood for of ratio Chi
initial model model with freedom square
covariate
Age 2660.15 2658.37 1 1.78 0.1826
Gender 2660.15 2658.24 1 1.91 0.1671
Tuberculosis 2660.15 2660.14 1 0.01 0.9030
Adherence* 2660.15 2660.08 1 0.07 0.7935
* Adherence to medication refill

TABLE 4.18 CHANGES IN THE COEFFICIENT OF WHO STAGE DURING ADDITION OF

COVARIATES EXCLUDED DURING STEP 1 OF MODEL BUILDING
(n=1,282)
Models Coefficient in Step 1 Coefficient in Step 4 Percentage change of coefficient
Age 0.2827662 0.2644602 -6.47%
Gender 0.2827662 0.2577927 -8.83%
Tuberculosis 0.2827662 0.2820564 -0.25%
Adherence* 0.2827662 0.2801304 -0.93%
* Adherence to medication refill

TABLE 4.19 CHANGES IN THE COEFFICIENT OF BASELINE CD4 CELL COUNT

DURING ADDITION OF COVARIATES EXCLUDED DURINGSTEP 1 OF
MODEL BUILDING (n=1,282)
Models Coefficient in Step 1 Coefficient in Step 4 Percentage change of coefficient
Age 0.0028975 0.0030015 3.59%
Gender 0.0028975 0.0029719 2.57%
Tuberculosis 0.0028975 0.0029002 0.09%
Adherence* 0.0028975 0.0029016 0.14%
* Adherence to medication refill

4.3.4.5.3.5 Step 5

In this step, the scale of continuous covariates found in the model was checked. There
was only one continuous variable, namely the baseline CD4 cell count. The scale of this
variable was checked graphically by plotting Martingale residuals from a model
excluding the baseline CD4 cell count against it. The Lowess curve, as shown in figure
4.11, indicates that it was linear after a CD4 cell count value of 200 but slightly curved
upwards for the lowest CD4 cell count values. This suggests its scale may be non-
linear. For this reason, fractional polynomials analysis was carried out.
 102

Figure 4.11 Martingale residual vs baseline CD4 cell count (n=1,282)

TABLE 4.20 SUMMARY OF A FRACTIONAL POLYNOMIALS ANALYSIS OF THE

BASELINE CD4 CELL COUNT (n=1,282)

Baseline CD4 cell Degree of freedom Deviance Deviance difference * Powers

P value
Count
Not in model 0 2993.500 33.177 0.000
Linear 1 2965.176 4.853 0.183 1
M=1 2 2965.176 4.853 0.088 1
M=2 4 2960.323 -- -- 2,3
*
P-value from deviance difference comparing reported model with m = 2 mode

Table 4.20 shows findings from fractional polynomials analysis. The findings suggest
that a two term model did not differ statistically from the linear model. In figure 4.12, it
can be seen that there are differences in the two graphs for CD4 cell count values
below 200 and above 500. Had the two term model been a better fit, the Grambsch,
Therneau and Fleming smoothed plot and fractional polynomial function would have
aligned better. For this reason, the baseline CD4 cell count was used as a linear
variable without transformation.
 103

Figure 4.12 Plot of the Grambsch, Therneau and Fleming smooth and fitted two-term
(2, 3) fractional polynomial model (n=1,282)

4.3.4.5.3.6 Step 6

In this step, interaction between the baseline CD4 cell count and the WHO stage was
checked in the final model. As shown in table 4.21, the interaction term failed to make
any statistically significant change in the model. For this reason, it was not included in
the preliminary final model.

TABLE 4.21 PARTIAL LIKELIHOOD RATIO TEST FOR THE ADDITION OF

INTERACTION TERM BETWEEN THE BASELINE CD4 CELL COUNT AND
THE WHO STAGE TO THE PRELIMINARY MULTIVARIABLE MODEL
(n=1,282)
Interaction -2 x log -2 x log likelihood Degree of Likelihood P value
likelihood for for model with freedom ratio Chi
initial model interaction term square
WHO Stage, 2660.15 2660.09 1 0.06 0.8052
Baseline CD4
cell count

4.3.4.5.3.7 Step 7
In this step, the preliminary final model containing the baseline CD4 cell count and the
WHO stage was checked for collinearity between covariates, fulfilment of the
 104

proportional hazards assumption, and regarding the existence of influential records, as

well as the goodness-of-fit test.

4.3.4.5.3.7.1 Collinearity between covariates

TABLE 4.22 COLLINEARITY BETWEEN COVARIATES IN THE FINAL MODEL (n=1,282)

Variables WHO Stage Baseline CD4 cell count
WHO Stage 1.0000
Baseline CD4 cell count -0.0047 1.0000

The correlation between the Baseline CD4 cell count and the WHO Stage was -0.0047.
There was no significant correlation.

4.3.4.5.3.7.2 Proportional hazards assumption

TABLE 4.23 TEST OF THE PROPORTIONAL HAZARDS ASSUMPTION OF THE

PRELIMINARY MULTIVARIABLE MODEL (n=1,282)
Variables rho Chi square Degree of freedom P value
WHO Stage 0.04934 0.55 1 0.4575
Baseline CD4 cell count -0.02933 0.18 1 0.6749
Global test 0.69 2 0.7095

Figure 4.13 Graphs of the scaled Schoenfeld residuals and their Lowess smooth
curves for the WHO Stage (n=1,282)
 105

Figure 4.14 Graphs of the scaled Schoenfeld residuals and their Lowess smooth
curves for the baseline CD4 cell count (n=1,282)

Global and individual covariate tests of the proportional hazards assumption indicated
there was no evidence that would lead to rejecting the assumption as shown in table
4.20 (p value>0.05). This evidence was supported by the smoothed scaled Schoenfeld
residuals plots (see figures 4.13 and 4.14) that show randomly scattered residuals as
well as smoothed curves with slope very close to zero for both covariates.

4.3.4.5.3.7.3 Checking influential records

In this step, regression diagnostic statistics were examined to assess whether there
were outliers, and if there were, what their influence on estimates of parameters, and
their influence on the fit of the model would be. Figure 4.15 depicts that there were four
records with results that were located apart from the rest. Table 4.24 shows all their
values. Baseline CD4 cell and WHO Stage values for these records were analysed from
clinical care point of view and were found to be normal. Delta-beta (Δβ) values and the
percentage change of the respective standard errors are displayed in table 4.25a and
4.25b for the five highest differences. It can be seen that none of these records had a
Δβ greater than the standard error of the respective parameter. This indicates that there
was no influential record affecting the coefficient of covariates in the model.
 106

Figure 4.15 Graph of the score residuals for baseline CD4 cell count (n=1,282)

TABLE 4.24 EXAMINING EXTREME VALUES FOR POTENTIALLY INFLUENTIAL

VARIABLES (n=1,282)
Score Baseline CD4 cell count WHO Stage Age Id
-1262.357 1221 III or IV 20 20541
-350.3844 680 III or IV 38 10976
302.6999 996 I or II 25 20458
447.6042 768 III or IV 25 21418

TABLE 4.25a HIGHEST DIFFERENCES IN PARAMETER ESTIMATES OF VARIABLES

INCLUDED IN THE PRELIMINARY MODEL, APPROXIMATE DELTA-BETAS
FOR THE BASELINE CD4 (n=1,282)
ID number of Δβ Standard error for % change of Δβ of
deleted observation Baseline CD4 cell count the standard error
11242 -0.0149249 0.1419612 11%
10518 -0.0146230 0.1419612 10%
10492 -0.0146215 0.1419612 10%
11018 -0.0145682 0.1419612 10%
10505 -0.0145401 0.1419612 10%
 107

TABLE 4.25b HIGHEST DIFFERENCES IN PARAMETER ESTIMATES OF VARIABLES

INCLUDED IN THE PRELIMINARY MODEL, APPROXIMATE DELTA-BETAS
FOR THE WHO STAGE (n=1,282)
Id number of Δβ Standard error for % Change of Δβ of
deleted WHO Stage the standard error
observation
20541 -0.0002530 0.0004497 56%
10976 -0.0000700 0.0004497 16%
20824 -0.0000460 0.0004497 10%
20160 -0.0000443 0.0004497 10%
20358 -0.0000439 0.0004497 10%

4.3.4.5.3.7.4 Goodness-of-fit test

The fit of the model was evaluated by using Cox-Snell residuals. It can be seen that the
two curves fit well for small values but match poorly at large values. This may be
because of the large number of censored records present in the data Collett (2003:240),
which was around 83% for this study (n=1056).

Figure 4.16 Cumulative hazard plot of the Cox-Snell residual (n=1,282)

4.3.4.5.3.8 Summary of regression analysis to predict immunologic failure (n=1,282)

 108

The final model is shown in tables 4.26a and 4.26b. Only the WHO Stage and baseline
CD4 cell count remained in the model. Their coefficients are shown in the first table
while their hazard ratios are shown in the second one. The WHO Stage is a categorical
variable, the reference in the model being WHO Stage I or II. Hence, patients at WHO
Stage III or IV had a hazard function 33% higher than the reference. Baseline CD4 cell
count was treated as a continuous variable. Hence, it means a single unit increase in
baseline CD4 cell count increased the hazard of immunologic failure by 0.3% while an
increase by 100 increased the hazard of immunologic failure by 33.6%
(100x0.0028975
(2.718 ). Age, gender, tuberculosis status, and adherence to medication

refill appointment were not included in the model since they failed to predict
immunologic treatment failure.

TABLE 4.26a COEFFICIENTS OF VARIABLES IN FINAL MULTIVARIABLE MODEL FOR

PREDICTORS OF IMMUNOLOGIC FAILURE AMONG PATIENTS ON ANTI-
RETROVIRAL THERAPY (n=1,282)
Covariate Coefficient Standard error Wald test Wald P value 95% confidence interval
WHO Stage 0.2827662 0.1419603 1.99 0.046 0.0045292 - 0.5611118
Baseline CD4 0.0028975 0.0004497 6.44 0.000 0.0020161 - 0.0037790
cell count

TABLE 4.26b HAZARD RATIO OF VARIABLES IN FINAL MULTIVARIABLE MODEL FOR

PREDICTORS OF IMMUNOLOGIC FAILURE AMONG PATIENTS ON ANTI-
RETROVIRAL THERAPY (n=1,282)
Covariate Hazard Standard error Wald test Wald P value 95% confidence interval
ratio
WHO Stage 1.326937 0.1883735 1.99 0.046 1.004645 - 1.752620
Baseline CD4 1.002902 0.0004510 6.44 0.000 1.002018 - 1.003786
cell count

After an extensive search of the PubMed database, only one published study in
northern Ethiopia by Assefa, Gelaw, Getnet and Yitayew (2014:online) was found that
described prevalence and predictors of immunologic failure using the WHO (2010a:50)
immunologic failure criteria to define immunologic failure. According to that study, the
prevalence of immunologic failure was 22.2% (n=89) which is slightly higher than the
current study. It identified CD4 cell count <100, male gender and tuberculosis as
predictors of immunologic failure. A number of issues make it difficult to make direct
comparison with this study. Follow-up CD4 cell count values not falling within two
months of follow-up months that are multiples of six, for example, 6, 12, 18, etc., were
excluded during their analysis. There is no mention of the number of CD4 cell counts
excluded. This may bias the findings of the study because the WHO criteria doesn‘t put
 109

such limitations on the timing of follow-up CD4 cell count testing apart from the fact that
immunologic treatment failure assessment couldn‘t be made before having ART for at
least six months. In addition, 122 charts were excluded because of missing and
incomplete information that may result in selection bias.There is no mention of control to
make sure that tuberculosis cases diagnosed prior to occurrence of immunologic failure
were considered during regression analysis to predict immunologic failure. In fact, in
that study, the median time for the diagnosis of immunologic failure for those patients
with TB was six months as compared to 9.5 months for the occurrence of incident TB.
This clearly shows that at least a considerable number of TB cases occurred after
occurrence of treatment failure and can only be taken as consequence of immunologic
failure rather than cause for it. This is in direct violation of the Bradford Hill criteria (Hill
1965:297) that stipulates that temporal relationship must hold in order to consider
cause-effect relationship between two variables; i.e., exposure must always precede
outcome. There are other studies that assessed predictors of immunologic failure but
comparison was not possible because different immunologic failure criteria were used.
For instance, Kassa et al (2013:online) used the following criteria to define immunologic
failure: increase of CD4 cell count by fewer than 50cells/μl at month six, and the criteria
of less than 100 cells/μl at months 18 and 24 of HAART. Anude et al (2013:online),
used the criteria of a fall of follow-up CD4 cell count to baseline or below, or an increase
in the CD4 cell count of less than 50 cells/mm3 at one year after HAART initiation. All
these criteria are different from that stated by the WHO.

4.4 OVERVIEW OF RESEARCH FINDINGS

A total of 1,321 documents of patients which had had two or more CD4 cell count tests
and at least six months of follow-up were reviewed. The study was conducted at two
hospitals in Ethiopia in the time frame September 1, 2005 to August 31, 2013.

The findings show that the median age of those in the study sample was 30 years (a
range of 5-71 years). Females accounted for 63% (n=831) of patients. Sixty per cent of
patients (n=786) were in the WHO stage III or IV at ART initiation. The median baseline
CD4 cell count at ART initiation was 156 cells per mm 3 (a range of 1-1,221 cells mm3).
After the initiation of HAART, tuberculosis was diagnosed in 17.79% (n=235) of
patients. Eighteen per cent (n=241) of patients showed poor adherence as they could
not adhere to their appointments for medication refill. By the end of the follow-up, 73%
 110

(n=960) of patients were still on HAART in the same facility where treatment was
initiated. There was no difference with respect to being actively on follow-up with regard
to baseline covariates.

Taking into account only the time patients were on follow-up, the pattern of timely
immunologic monitoring was, in general, inadequate. There was no predilection for
determination of timely follow-up CD4 cell count test by age, gender, baseline WHO
stage or baseline CD4 cell count (p value > 0.05 for all).

Immunologic (CD4 cell count) recovery was observed among patients on antiretroviral
therapy. The magnitude of follow-up CD4 cell count increase was rapid in the first few
years of follow-up after starting antiretroviral therapy and reached a plateau thereafter.
Children of less than 14 years of age, female patients, those who were at baseline
WHO clinical stages I or II, as well as patients with a higher baseline CD4 cell count
were able to achieve a higher proportion of follow-up CD4 cell count greater than or
equal to 350 than their respective counterparts (p value < 0.05 for all).

The prevalence of patients who experienced immunologic failure at least once was
17.6% (n=226) (95% confidence interval: 15.6%-19.8%).But, with further evaluation and
follow-up CD4 cell count determination this percentage dropped to 11.5% (n=147) (95%
confidence interval: 9.8%-13.3%). Having a follow-up immunologic evaluation resulted
in a decrease of the number of patients who fulfilled immunologic failure criteria, with
almost 46% of those with follow-up CD4 cell count evaluation not fulfilling the criteria
anymore. Diagnosis of treatment failure was mostly based on the criteria ‗drop in follow-
up CD4 cell count by 50% from follow-up peak‘ (58%, n=131) followed by the criteria
‗drop of follow-up CD4 cell count to or below baseline CD4 cell count‘ (35%, n=78).
There were very few patients on second line HAART among those with confirmed
immunologic failure (7.53%, n=7) even though there were a total of93patients who were
confirmed to have had immunologic treatment failure.

The median duration of follow-up from HAART initiation to diagnosis of first instance
immunologic failure was 1.8 years (range: 0.51-7.92). For those who never experienced
treatment failure, the median time of follow-up was 3.72 years (0.51-8.36).
 111

During univariable as well as multivariable analysis, being at WHO Stage III or IV or

having a higher CD4 cell count at baseline were associated with being at level of
increased hazard for immunologic failure. None of the other covariates (age, gender,
tuberculosis, adherence to picking up medication refill on time) predicted immunologic
treatment failure.

4.5 CONCLUSION

In this chapter, analysis techniques, and results of the study were presented.
Comparison of findings with other studies was made as much as possible but was
limited to few a few studies because of lack of similarity in variable selection, study
design, and study setting.
 112

CHAPTER 5

CONCLUSIONS AND RECOMMENDATIONS

5.1 INTRODUCTION

Chapter 4 addressed data analysis, presentation and description of the research

findings. In this chapter interpretation of the findings, conclusions, recommendations
and limitations of the study are presented.

The aim of the study was to describe the prevalence of immunologic failure and identify
its determinants among people living with HIV receiving treatment at selected hospitals.

The objectives of this study were to:

 Assess the pattern of immunologic (CD4 cell count) monitoring in selected

hospitals in southern Ethiopia
 Assess the pattern of immunologic (CD4 cell count) response for patients on
HAART in selected hospitals in southern Ethiopia
 Describe immunologic treatment failure and identify its predictors among patients
on HAART in selected hospitals in southern Ethiopia

5.2 RESEARCH DESIGN AND METHOD

A self-designed document checklist was used for the purpose of data collection. The
checklist consisted of three sections: demographic data, clinical data and immunologic
data in the form of CD4 cell count tests.

The documents of people living with HIV on antiretroviral therapy enrolled and started
on ART between September 1st, 2005 and December 31st, 2012 in the selected study
sites, those with age five or older at ART initiation, those with two or more CD4 cell
count tests and those with at least six months of follow-up were included in the study.
Documents of patients transferred into the study sites from other facilities after starting
 113

HAART were excluded from the study. Follow-up information was abstracted for all
eligible patients until August 31st, 2013.

After applying the eligibility criteria, there were 1,321 patient records to draw data from.
Since the ideal for the study sample was 1,304, the researcher used all available patient
records for analysis. Data on follow-up weight was missing in a significant number of
records, for which reason, that factor was excluded from analysis. Data cleaning was
done with Epi InfoTM 3.5.1 statistical software. The data was analysed using Stata/IC
12.0 (StataCorp LP). Descriptive statistics were presented in tables and graphs. The
Chi Square test and Cox proportional hazards regression were used to analyse data.
The P value was set at 0.05.

5.3 CONCLUSIONS

This section contains the conclusions of the study in relation to the study‘s objectives.

5.3.1 Objective 1: assess the pattern of immunologic (CD4 cell count) monitoring
in selected hospitals in southern Ethiopia

5.3.1.1 Pattern of immunologic monitoring

The pattern of timely immunologic monitoring was, in general, inadequate. There was a
correlation between the length of follow-up and the inadequacy of CD4 cell count
determination: at the 6th month of follow-up, 46% (n=610) of patients still under care had
their CD4 cell count determined, while at one year, only 33% (n=417) did so; it went
below 20% after two years of follow-up and below 10% (n=10) at the seventh year of
follow-up. There was no preference for determining whether there was follow-up CD4
cell count testing in a timely fashion of those studied by age, gender, baseline WHO
stage or baseline cell CD4 cell count (p value>0.05 for all).

5.3.2 Objective 2: assess the pattern of immunologic (CD4 cell count) response
for patients on HAART in selected hospitals in southern Ethiopia

5.3.2.1 Pattern of immunologic response

 114

Immunologic recovery was observed among patients on treatment. The magnitude of

follow-up CD4 cell count increased progressively in the first few years of follow-up after
starting antiretroviral therapy and reached a plateau thereafter. Children of less than 14
years of age, female patients, those at baseline WHO stage I or II, as well as patients
with higher baseline CD4 cell count were able to achieve a higher proportion of follow-
up CD4 cell count greater than or equal to 350 than their respective counterparts (p
value <0.05 for all). This signifies that earlier initiation of HAART would result in better
immunological recovery in the form of an increased CD4 cell count reaching the normal
range soon after treatment initiation.

5.3.3 Objective 3: describe immunologic treatment failure and identify its

predictors among patients on HAART in selected hospitals in southern
Ethiopia

5.3.3.1 Prevalence and predictors of immunologic failure

Despite the adequacy of group level immunologic recovery, immunologic failure was
observed at an individual level. The proportion of patients who ever experienced
immunologic failure was 17.6% (95% confidence interval: 15.6%-19.8%).However with
further evaluation and follow-up CD4 cell count determination, it dropped to 11.5%
(n=147) (95% confidence interval: 9.8%-13.3%). Since a significant number of patients
who were re-evaluated after immunologic failure didn‘t fulfil immunologic failure criteria,
it could be possible that the actual percentage of patients with immunologic failure
would have been even lower than this, had there been proper follow-up immunologic
evaluation. This also means that it may not be wise to rush to change antiretroviral
medication to second line antiretroviral medication before proper evaluation or follow-up
has been made. This could be done by repeating the CD4 cell count, addressing
adherence issues, or utilizing more sensitive measures of treatment failure like viral load
testing.

During multivariable analysis, two factors were found to be important in predicting

immunologic failure: the baseline WHO Stage and the baseline CD4 cell count. In
general, the findings were that the higher the baseline CD4 cells count, the higher the
risk or hazard of immunologic failure; the same was true for baseline WHO stage. This
means, patients fulfilling these criteria should be prioritized for immunologic
 115

assessment. With the recent changes in the eligibility criteria for the initiation of
antiretroviral treatment for adults and adolescents raising the CD4 cell count cut off to
below 500 (FMOH 2013:2), the proportion of patients who experience a diagnosis of
immunologic failure could become higher.

5.4 RECOMMENDATIONS

Immunologic treatment failure was prevalent in this study setting. Since immunologic
monitoring was at the same time inadequate, the true level of immunologic failure could
not be determined, though it was expected to be lower since with on-going evaluation
less number of patients fulfilled the immunologic failure criteria. Also, since there are
variables collected in routine care that are significant predictors of immunologic failure,
they should be incorporated in a treatment plan that identifies and manages such
patients. The following section discusses such plan that could be used by health care
providers to screen and manage patients with presumptive immunologic failure based
on the literature review and findings of the study.

5.4.1 Treatment plan to identify and manage patients with immunologic failure

Table 5.1 is a sample documentation tool to screen for and manage patients with
respect to immunologic failure. This tool would be placed in each patient‘s medical
document. Risk assessment will be done once at enrolment since the predictors
identified are determined at baseline. The cut-off to identify those at greater and lower
risk for immunologic failure using baseline CD4 cell count as criteria has been made
350 cells/mm3 in order to simplify the tool. This is based on findings displayed in table
4.12 and figure 4.9 that show a greater risk of immunologic failure for patients with
baseline CD4 counts of more than or equal to 350. Patients with a baseline CD4 cell
count between 200 and 349 were at risk as well, compared to those with a count below
200, but the magnitude of difference was small as can be seen in figure 4.9. For
baseline WHO staging, it is clear that those at WHO stages III or IV are at greater risk.

Those patients who have a baseline CD4 cell count more than or equal to 350 and/or
those at baseline WHO stages III or IV are at risk of immunologic failure and would
need prioritization for assessment as well management. Assessment for immunologic
failure would be conducted at time of every follow-up CD4 cell count determination.
 116

TABLE 5.1 IMMUNOLOGIC FAILURE RISK ASSESSMENT AND MANAGEMENT TOOL

IMMUNOLOGIC FAILURE RISK ASSESSMENT AND MANAGEMENT DOCUMENTATION TOOL

3
Immunologic failure risk at baseline: 1. Baseline CD4 cell count ≥350/mm ? Yes No High risk for immunologic treatment failure if
2. Baseline WHO Stage III or IV? Yes No ‘Yes’ to either question
Follow-up date Month Adherence Antiretroviral medication CD4 cell Immunologic failure Decision (Use if there is immunologic
3
on ART (G, F, P) regimen count/mm (Y or N) failure)
0 = Re-evaluate with CD4 cell count
1 = Re-enforce adherence
2 = Re-evaluate with viral load
3 = Regimen change to second line
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
*Adherence= G (Good) if missed doses<3 per month; F (Fair) if missed doses 3-5 and 3-9 for 30 and 60 doses regimen per month respectively;
P (Poor) if missed doses >5 or >9 for 30 and 60 doses regimen per month respectively. (FMOH 2008a:72)
**Immunologic failure criteria: 1. 50% below maximum follow-up CD4 cell count; 2. at or below baseline CD4 cell count; 3. persistently below 200. The presence of
any one of these criteria in patients who have been taking antiretroviral treatment for at least six months and is still on treatment is used to diagnose it. (FMOH
2008a:159)
 117

There would be a reminder to give a follow-up CD4 cell count test emphasized by
shaded rows at months 0, 6, 12, 18, 24, and so on at and after the initiation of ART
based on the current WHO recommendation (2013:132). Assessment for antiretroviral
medication intake needs to be carried out as well, since it is one of the reasons for CD4
cell count decline. If a patient fulfills immunologic failure criteria while on the medication
and after at least six months of treatment, then that would be documented on a
specified column and a decision would be made on what to do next. The decision could
be to repeat the CD4 cell count, to address adherence issues, to confirm treatment
failure with viral load testing, or to change the treatment regimen to second line
antiretroviral medication if failure is confirmed. Adherence issues could be addressed by
proper counseling on the consequences of poor adherence for the patient, and the risk
of transmission to others, as well as the means to handle it.

5.4.2 Recommendations to health care providers

In order to identify patients with immunologic failure or to know the true magnitude of
the problem, proper CD4 cell count monitoring has to be done. This study indicates that
immunologic monitoring has been poor in the study facilities. Hence, health care
providers need to assess their CD4 cell count determination practices at facility level
and take corrective measures whenever possible. In addition to that, the low proportion
of patients on second line HAART indicates that, even in patients who had adequate
CD4 cell count monitoring and fulfilled immunologic failure criteria, providers failed to
manage them by shifting their antiretroviral regimen to a second line medication. Again,
heath care providers need to work more on the proper implementation of treatment
guidelines. In the case of resource constraints in the form of limited capacity to do CD4
cell count monitoring, health care providers should give priority to those at a high risk of
immunologic failure as outlined in this study and to those with immunologic failure.
Regimen changes based on immunologic criteria alone should be thought out carefully
as a number of factors including adherence could easily affect it. There was no
predictive effect of adherence to medication pick-upon immunologic treatment failure.
Hence, providers should not rely on it to assess the risk of immunologic failure.

5.4.3 Recommendations to program managers

 118

The reported magnitude of immunologic failure reasserts the need for the continuation
and scaling up of the ongoing effort to address development of drug resistance to
antiretroviral medications and transmission of drug resistance. Since immunologic
monitoring is one of the most important elements of this process, the reason its
implementation is suboptimal and how it can be corrected should be evaluated.

5.4.4 Recommendations for further research

This study merely identified the problem of poor immunologic determination. Therefore,
further studies need be made to address barriers to proper CD4 cell count
determination. In the face of confirmed immunologic failure, very few patients were
managed by shifting antiretroviral treatment regimen to second line. Hence, providers‘
knowledge, attitude, and practice to managing patients with immunologic failure, which
could be one determinant, should be evaluated in order to design an intervention to
address observed gap. Also, the proposed tool to monitor and manage immunologic
criteria needs to be validated and piloted before implementation.

5.5 CONTRIBUTIONS OF THE STUDY

The study may contribute to patients and providers at the study sites and beyond since
it identified gaps in the implementation of immunologic monitoring and management of
immunologic failure and potential interventions to alleviate the challenges. Providers in
study facilities will be made aware of the prevalence of immunologic failure among
patients under their care in addition to the gap in immunologic monitoring. Due to a
large number of records of patients and difficulty in reviewing immunologic monitoring,
an immunologic failure screening tool will be provided for health care providers to adapt
and conduct proper CD4 cell count monitoring and screening of patients for
immunologic failure.

For the wider health care system, the estimate of the prevalence of immunologic failure
for children older than five years, adolescents, and adults provides an estimate for the
specified age groups for Ethiopia using the WHO immunologic failure diagnostic criteria.
This would help in planning for better management of patients on HAART as it provides
an estimate of the magnitude of the problem. Most importantly it gives some idea as to
the proportion of patients who may require advanced treatment monitoring in the form of
 119

viral load testing. Finally, this study is an example of how data collected from routine
program monitoring can be used to shed light on important questions that may in turn
help to improve the quality of care.

5.6 LIMITATIONS OF THE STUDY

Despite the overall positive contributions of this study for clinical practice and program
implementation, it has important limitations. Since this was a retrospective study based
on existing medical records, the effect of possibly erroneous entries as well as
incomplete entries of data cannot be underestimated. For instance, these reasons have
led the researcher to drop the variable weight. In addition, the study could not assess
the effect of clinical disease progression on immunologic failure except for tuberculosis
as treatment staging was not well implemented.

Not all patients were active on follow-up at last observation because of treatment
interruption or transfer to other facility for follow-up. For this reason, immunologic failure
could only be measured for as long as patients were on follow-up. This effect was
accounted for during survival analysis, which is specifically, designed to analyze such
data. Moreover, there was no association between baseline covariates and being active
on follow-up at last observation making the effect of incomplete observations minimal.

Finally, since only two hospitals were selected for convenience, findings couldn‘t be
generalized for the whole region.

5.7 CONCLUDING REMARKS

This chapter concludes the study. The most important finding is that immunologic failure
was found to be prevalent in the study setting. A number of recommendations were
made in order to better implement identification and management of immunologic failure
among patients on antiretroviral therapy. Proactive implementation of these and other
existing measures should be made timely so that patients benefit sooner than later.
 120

BIBLIOGRAPHY

Aber, V, Aboulker, JP, Babiker, AG, Bragman, K, Breckenridge, AM, Carbon, C,

Charreau, I, Chene, G, Collis, P, Cooper, D, Darbyshire, JH, Dormont, J, Fiddian, P,
Flepp, P, Gazzard, B, Goebel, FD, Hooker, M, Lange, J, Luthy, R, Peto, TEA, Reiss, P,
Seligmann, M, Stone, AB, Thomis, J, Vella, S, Walckenaer, G, Warrell, D, Weller, IVD,
Wilber, R, Yeni, P, Yeo, J & Withnall, R. 1996. Delta: a randomised double-blind
controlled trial comparing combinations of zidovudine plus didanosine or zalcitabine with
zidovudine alone in HIV-infected individuals. Lancet 348: 283-291.

Alemayehu, YK, Bushen, O & Muluneh, AT. 2009. Evaluation of HIV/AIDS clinical care
quality: the case of a referral hospital in North West Ethiopia. International Journal for
Quality in Health Care 21(5):356–362.

Alemu, AW & San Sebastián, M. 2010. Determinants of survival in adult HIV patients on
antiretroviral therapy in Oromiyaa, Ethiopia. Global Health Action 3: 5398.
From: www.globalhealthaction.net/index.php/gha/rt/printerFriendly/5398/6222
(accessed on January 2013).

Alene, KW, Nega, A & Taye, BW. 2013. Incidence and predictors of tuberculosis among
adult people living with human immunodeficiency virus at the University of Gondar
referral hospital, Northwest Ethiopia. Biomedical Central Infectious Diseases 13:292.

Ali, K & Klotz, SA. 2012. The immune reconstitution inflammatory syndrome with
tuberculosis: a common problem in Ethiopian HIV-infected patients beginning
antiretroviral therapy. Journal of International Association of Physicians AIDS Care
11(3):198-202.

Althoff, KN, Justice, AC, Gange, SJ, Deeks, SG, Saag, MS, Silverberg, MJ, Gill, MJ,
Lau, B, Napravnik, S, Tedaldi, E, Klein, MB & Gebo, KA. 2010. Virologic and
immunologic response to HAART, by age and regimen class. AIDS 24:2469–2479.

Anude, CJ, Eze, E, Onyegbutulem, HC, Charurat, M, Etiebet, M, Ajayi, S, Dakum, P,

Akinwande, O, Beyrer, C, Abimiku, A & Blattner, W. 2013.Immuno-virologic outcomes
 121

and immunovirologic discordance among adults alive and on anti-retroviral therapy at

12 months in Nigeria. Biomedical Central Infectious Diseases 13:113.

Assefa, A, Gelaw, B, Getnet, G & Yitayew, G. 2014. The effect of incident tuberculosis
on immunological response of HIV patients on highly active anti-retroviral therapy at the
University of Gondar hospital, northwest Ethiopia: a retrospective follow-up study.
Biomedical Central Infectious Diseases14:468.

Assefa, Y, Kiflie, A, Tesfaye, D, Haile-Mariam, D, Kloos, H, Edwin, W, Laga, M & Van

Damme, W. 2011. Outcomes of antiretroviral treatment program in Ethiopia: retention of
patients in care is a major challenge and varies across health facilities. Biomedical
Central Health Services Research 11:81. doi: 10.1186/1472-6963-11-81. From:
http://www.biomedcentral.com/content/pdf/1472-6963-11-81.pdf (accessed on 25
February 2013).

Azmeraw, D & Wasie, B. 2012. Factors associated with adherence to highly active
antiretroviral therapy among children in two referral hospitals, northwest Ethiopia.
Ethiopian Medical Journal 50(2):115-124.

Azzoni, L, Foulkes, AS, Liu, Y, Li, X, Johnson, M, Smith, C, Kamarulzaman, A,

Montaner, J, Mounzer, K, Saag, M, Cahn, P, Cesar, C, Krolewiecki, A, Sanne, I &
Montaner, LJ. 2012. Prioritizing CD4 count monitoring in response to ART in resource-
constrained settings: a retrospective application of prediction-based classification.
Public Library of Science Medicine 9(4): e1001207. From:
http://www.plosmedicine.org/article/info%3Adoi%2F10.1371%2Fjournal.pmed.1001207
(accessed 26 February 2013).

Bacha, T, Tilahun, B & Worku, A. 2012. Predictors of treatment failure and time to
detection and switching in HIV-infected Ethiopian children receiving first line anti-
retroviral therapy. Biomedical Central Infectious Diseases 12:197. From:
http://www.biomedcentral.com/1471-2334/12/197 (accessed 25 February 2013).

Bärnighausen, T, Chaiyachati, K, Chimbindi, N, Peoples, A, Haberer, J & Newell, ML.

2011. Interventions to increase antiretroviral adherence in sub-Saharan Africa: a
systematic review of evaluation studies. Lancet Infectious Diseases 11: 942-951.
 122

Barré-Sinoussi, F, Chermann, JC, Rey, F, Nugeyre, MT, Chamaret, S, Gruest, J,

Dauguet, C, Axler-Blin, C, Vézinet-Brun,F, Rouzioux, C, Rozenbaum, W & Montagnier,
L. 1983. Isolation of a T-lymphotropic retrovirus from a patient at risk for acquired
immune deficiency syndrome (AIDS). Science220(4599):868-71.

Bélec, L & Bonn, JP. 2011. Challenges in implementing HIV laboratory monitoring in
resource-constrained settings: how to do more with less. Future Microbiology
6(11):1251-1260.

Berhan, Y. 2011a. Change in absolute CD4 count with antiretroviral treatment in

resource limited setting: a retrospective cohort study in children in Ethiopia. Ethiopian
Medical Journal 49(4):289-298.

Berhan, Y. 2011b. Evaluation of adherence to national guideline for clinical follow-up of

HIV infected children using the proxy hematological, biochemical and anthropometric
indicators of care. Ethiopian Medical Journal 49(3):199-209.

Beyene, KA, Gedif, T, Gebre-Mariam, T & Engidawork, E. 2009.Highly active

antiretroviral therapy adherence and its determinants in selected hospitals from South
and Central Ethiopia. Pharmacoepidemiology and Drug Safety 18(11):1007-1115.

Boruett, P, Kagai, D, Njogo, S, Nguhiu, P, Awuor, C, Gitau, L, Chalker, L, Ross-Degnan,

D, Wahlström, R & Tomson, G. 2013. Facility-level intervention to improve attendance
and adherence among patients on anti-retroviral treatment in Kenya: a quasi-
experimental study using time series analysis. Biomedical Central Health Services
Research 13:242.

Bruce, N, Pope, D & Stanistreet, D. 2008.Quantitative methods for health research: a

practical interactive guide to epidemiology and statistics. Chichester: John Wiley &
Sons.

Burns, N & Grove, SK. 2009.The practice of nursing research: appraisal, synthesis, and
generation of evidence. 6th edition.St Louis: Elsevier/Saunders.
 123

Chalker, JC, Andualem, T, Gitau, LN, Ntaganira, J, Obua, C, Tadeg, H, Waako, P,

Ross-Degnan, D &International Network for the Rational Use of Drugs Initiative on
Adherence to Antiretrovirals. 2010. Measuring adherence to antiretroviral treatment in
resource-poor settings: the feasibility of collecting routine data for key indicators.
Biomedical Central Health Services Research 10:43. From:
http://www.biomedcentral.com/1472-6963/10/43 (accessed February 1 2013).

Chernick, MR. 2011. The essentials of biostatistics for physicians, nurses, and
clinicians. New Jersey: John Wiley & Sons.

Chun, T & Fauci, AS. 2012. HIV reservoirs: pathogenesis and obstacles to viral
eradication and cure. AIDS 26:1261-1268.

Clavel, F, Mansinho, K, Chamaret, S, Guetard, D, Favier, V, Nina, J, Santos-Ferreira,

M, Champalimaud, J & Montagnier, L. 1987. Human Immunodeficiency Virus Type 2
Infection Associated with AIDS in West Africa. New England Journal of Medicine
316:1180-1185.

Cohen, CJ, Meyers, JL & Davis, KL. 2013. Association between daily antiretroviral pill
burden and treatment adherence, hospitalisation risk, and other healthcare utilisation
and costs in a US Medicaid population with HIV. British Medical Journal Open
3:e003028. From: http://bmjopen.bmj.com/content/3/8/e003028.long (accessed August
15 2013).

Collett, D. 2003. Modelling survival data in medical research. 2nd edition. London:
Chapman & Hall.

Crawford, KW, Wakabi, S, Magala, F, Kibuuka, H, Liu, M& Hamm, TE. 2014. Evaluation
of treatment outcomes for patients on first-line regimens in US President's Emergency
Plan for AIDS Relief clinics in Uganda: predictors of virological and immunological
response from RV288 analyses. HIV Medicine doi: 10.1111/hiv.12177.

Dagnew, YW. 2009. Factors influencing anti-retroviral therapy adherence in Ethiopia.

Unpublished Master of Public Health dissertation. Pretoria: University of South Africa
 124

Daniel, WW. 2009. Biostatistics: a foundation for analysis in the health sciences.9th
edition. New York: John Wiley & Sons.

Department of Health and Human Services [USA]. 2013. Guidelines for the use of
antiretroviral agents in HIV-1 infected adults and adolescents. New York: Department of
Health and Human Services.

Egger, M, May, M, Chêne, G, Phillips, AN, Ledergerber, B, Dabis, F, Costagliola, D,

D'Arminio Monforte, A, De Wolf, F, Reiss, P, Lundgren, JD, Justice, AC, Staszewski, S,
Leport, C, Hogg, RS, Sabin, CA, Gill, MJ, Salzberger, B & Sterne, JA. 2002. Prognosis
of HIV-1-infected patients starting highly active antiretroviral therapy: a collaborative
analysis of prospective studies. Lancet 360(9327):119-129.

EHNRI & FMOH. See Ethiopian Health and Nutrition Research Institute & Federal
Ministry of Health.

El-Khatib, Z, Katzenstein, D, Marrone, G, Laher, F & Mohapi, L. 2011. Adherence to

drug-refill is a useful early warning indicator of virologic and immunologic failure among
HIV patients on first-line ART in South Africa. Public Library of Science 6(3):e17518.

Ethiopian Health and Nutrition Research Institute & Federal Ministry of Health. 2012.
HIV Related Estimates and Projections for Ethiopia – 2012. Ethiopia.

Eze, EU & Uyagu, O. 2009. Prevalence of anti-retroviral treatment failure and salvage
therapy in University of Benin teaching hospital Edo state, Nigeria. Annals of Biomedical
Science 8(2):40-46.

Fawibe, AE, Olafimihan, KO, Salami, AK, Desalu, OO & Odeigah, LO. 2010. A
retrospective evaluation of pretreatment and follow-up monitoring of HIV-infected adult
patients in a resource-poor setting. Journal of International Association of Physicians
AIDS Care 9(5):289-291.

Federal HIV/AIDS Prevention and Control Office & Federal Ministry of Health. 2007.
Guidelines for management of opportunistic infections and antiretroviral treatment in
adolescents and adults in Ethiopia. Addis Ababa, Ethiopia.
 125

Federal HIV/AIDS Prevention and Control Office & Federal Ministry of Health. 2008a.
Participant manual for the basic chronic HIV care, antiretroviral therapy and prevention.
Addis Ababa, Ethiopia.

Federal HIV/AIDS Prevention and Control Office & Federal Ministry of Health.
2008b.Guidelines for Paediatric HIV/AIDS Care and Treatment in Ethiopia. Addis
Ababa, Ethiopia.

Federal HIV/AIDS Prevention and Control Office & Federal Ministry of Health. 2012.
Country Progress Report on HIV/AIDS Response. Ethiopia.

Federal Ministry of Health. 2010a. Health management information system (HMIS)

participants’ handout & instruments. Ethiopia.

Federal Ministry of Health. 2010b. National training package provider-initiated HIV

testing and counselling participant’s manual. Ethiopia.

Federal Ministry of Health. 2013. Technical updates on adopted recommendations on

adolescent and adult anti-retroviral treatment. Ethiopia.

Fischl, MA, Richman, DD, Grieco, MH, Gottlieb, MS, Volberding, PA, Laskin, OL,
Leedom, JM, Groopman, JE, Mildvan, D, Schooley, RT, Jackson, GG, Durack, DT,
King, D and the AZT collaborative working group. 1987. The efficacy of azidothymidine
(AZT) in the treatment of participants with AIDS and AIDS-related complex: a double-
blind, placebo-controlled trial. New England Journal of Medicine 317:185-191.

FHAPCO. See Federal HIV/AIDS Prevention and Control Office.

FHAPCO & FMOH. See Federal HIV/AIDS Prevention and Control Office & Federal
Ministry of Health.

FMOH. See Federal Ministry of Health.

 126

Gianotti, N, Galli, L, Bocchiola, B, Cahua, T, Panzini, P, Zandonà, D, Salpietro, S,

Maillard, M, Danise, A, Pazzi, A, Lazzarin, A & Castagna, A. 2012. Number of daily pills,
dosing schedule, self-reported adherence and health status in 2010: a large cross-
sectional study of HIV-infected patients on antiretroviral therapy. HIV Medicine
14(3):153-160.

Giday, A & Shiferaw, W. 2010. Factors affecting adherence of antiretroviral treatment

among AIDS patients in an Ethiopian tertiary university teaching hospital. Ethiopian
Medical Journal 48(3):187-94.

Glynn, M & Drake, WM. 2012.Hutchison's clinical methods: an integrated approach to

clinical practice. 23rd edition. China: Saunders Elsevier.

Grossman, Z, Meier-Schellersheim, M, Paul, WE & Picker, LJ. 2006. Pathogenesis of

HIV infection: what the virus spares is as important as what it destroys. Nature Medicine
12:289– 295.

Heeney, JL, Dalgleish, AG & Weiss, RA. 2006. Origins of HIV and the evolution of
resistance to AIDS. Science313: 462–66.

Hill, AB. 1965. The environment and disease: association or causation? Proceedings of
the Royal Society of Medicine 58:295-300.

Hosmer, DW, Lemeshow, S and May,S. 2008. Applied survival analysis regression
modeling of time-to-event data. 2nd edition. New Jersey: John Wiley & Sons, Inc.

Jaka, HM, Mshana, SE, Liwa, AC, Peck, R & Kalluvya, S. 2009. Prevalence of
immunological failure and durability of first line antiretroviral therapy at Bugando
Hospital Mwanza, Tanzania. Tanzania Medical Journal 24(2): 5-8.

Jacobsen, KH. 2011. Introduction to health research methods: a practical guide.

Subury: Jones &Bartlett.
 127

Jerene, D, Næss, A & Lindtjørn, B. 2006.Antiretroviral therapy at a district hospital in

Ethiopia prevents death and tuberculosis in a cohort of HIV patients. AIDS Research
and Therapy 3(10):10.

Joint United Nations Programme on HIV/AIDS. 2012. Together we will end AIDS.
From:
http://www.unaids.org/en/media/unaids/contentassets/documents/epidemiology/2012/J
C2296_UNAIDS_TogetherReport_2012_en.pdf (accessed 18 August 2012).

Joint United Nations Programme on HIV/AIDS. 2013a. Aidsinfo epidemiological status.

From: http://www.unaids.org/en/dataanalysis/datatools/aidsinfo (accessed 13 November
2013).

Joint United Nations Programme on HIV/AIDS. 2013b. Global report: UNAIDS report on
the global AIDS epidemic 2013.
From:
http://www.unaids.org/en/media/unaids/contentassets/documents/epidemiology/2013/gr
2013/UNAIDS_Global_Report_2013_en.pdf (accessed 3 October 2013).

Joint United Nations Programme on HIV/AIDS. 2014. The gap report. From:
http://www.unaids.org/en/resources/campaigns/2014/2014gapreport/gapreport/
(accessed 20September 2014).

Kanters, S, Nachega, J, Funk, A, Mukasa, B, Montaner, JS, Ford, N, Bucher, HC &

Mills, EJ. 2013. CD4 T-Cell recovery after initiation of antiretroviral therapy in a
resource-limited setting: a prospective cohort analysis. Antiviral Therapy doi:
10.3851/IMP2670. From:
http://www.unboundmedicine.com/medline/citation/23963170/CD4_T_cell_recovery_aft
er_initiation_of_antiretroviral_therapy_in_a_resource_limited_setting:_a_prospective_co
hort_analysis_ (accessed August 31 2013).

Kassa, D, Gebremichael, G, Alemayehu, Y, Wolday, D, Messele, T &Van Baarle, D.

2013. Virologic and immunologic outcome of HAART in Human Immunodeficiency Virus
(HIV)-1 infected patients with and without tuberculosis (TB) and latent TB infection
 128

(LTBI) in Addis Ababa, Ethiopia. AIDS Research and Therapy 10:18. From:
http://www.aidsrestherapy.com/content/10/1/18 (accessed July 24 2013).

Katlama, C, Deeks, SG, Autran, B, Martinez-Picado, J, Van Lunzen, J, Rouzioux, C,

Miller, M, Vella, S, Schmitz, JE, Ahlers, J, Richman, DD & Sekaly, RP. 2013. Barriers to
a cure for HIV: new ways to target and eradicate HIV-1 reservoirs. Lancet 381: 2109-
2117.

Kebede, A & Wabe, NT. 2012. Medication adherence and its determinants among
patients on concomitant tuberculosis and antiretroviral therapy in South West Ethiopia.
North American Journal of Medical Sciences 4:67-71.

Keiser, O, MacPhail, P, Boulle, A, Wood, R, Schechter, M, Dabis, F, Sprinz, E & Egger,

M. 2009.Accuracy of WHO CD4 cell count criteria for virological failure of antiretroviral
therapy. Tropical Medicine and International Health 14(10): 1220–1225.

Keiser, O, Tweya, H, Braitstein, P, Dabis, F, MacPhail, P & Boulle, A. 2010. Mortality

after failure of antiretroviral therapy in sub-Saharan Africa. Tropical Medicine and
International Health 15(2): 251-258.

Kestenbaum, B. 2009.Epidemiology and biostatistics: an introduction to clinical

research, edited by Adeney, KL, Weiss. New York: Springer.

Le, T, Wright, EJ, Smith, DM, He, W,Catano, G, Okulicz, JF, Young, JA, Clark, RA,
Richman, DD, Little, SJ & Ahuja, SK. 2013.Enhanced CD4+ T-Cell recovery with earlier
HIV-1 antiretroviral therapy. New England Journal of Medicine 368(3): 218–230.

Lederman, MM, Calabrese, L, Funderburg, NT, Clagett, B, Medvik, K, Bonilla, H,

Gripshover, B, Salata, RA, Taege, A, Lisgaris, M, McComsey, GA, Kirchner, E, Baum,
J, Shive, C, Asaad, R, Kalayjian, RC, Sieg, SF & Rodriguez, B. 2011. Immunologic
failure despite suppressive antiretroviral therapy is related to activation and turnover of
memory CD4 cells. The Journal of Infectious Diseases 204(8):1217–26.

Levy, JA. 2009. HIV pathogenesis: 25 years of progress and persistent challenges.
AIDS 23:147–160.
 129

Lifson, AR, Krantz, EM, Eberly, LE, Dolan, MJ, Marconi, VC, Weintrob, AC, Crum-
Cianflone, NC, Ganesan, A, Grambsch, PL & Agan, BK. 2011. Long-term CD4+
lymphocyte response following HAART initiation in a U.S. Military prospective cohort.
AIDS Research and Therapy 8:2. From:
http://www.aidsrestherapy.com/content/8/1/2/comments (accessed July 24 2013).

LoBiondo-Wood, G & Haber, J. 2010.Nursing research: methods and critical appraisal

for evidence-based practice. 7thedition. Louisiana: Mosby.

Longo, DL, Fauci, AS, Kasper, DL, Hauser, SL, Jameson, JL & Loscalzo, J.
2012.Harrison’s principles of internal medicine.18th edition. New York: McGraw-Hill.

Macmillan Dictionary Online. 2013. Sv ―adherence‖ ―HIV‖ ―patient‖ ―predict‖ ―treatment‖.

New York: Macmillan Publisher.
From: http://www.macmillandictionary.com/dictionary/british/ (accessed 1 August 2013).

Machin, D, Cheung, MB & Parmar, MKB. 2006. Survival analysis: a practical approach.
2nd edition. West Sussex: John Wiley & Sons.

Maskew, M, Fox, MP, Van Cutsem, G, Chu, K, MacPhail, P, Boulle, A & Egger, M.
2013. Treatment Response and Mortality among Patients Starting Antiretroviral Therapy
with and without Kaposi Sarcoma: A Cohort Study. Public Library of Science 8(6):
e64392. From: http://www.plosone.org/article/info:doi/10.1371/journal.pone.0064392
(accessed June 19 2013).

Mghamba, FW, Minzi, OMS, Massawe, A & Sasi, P. 2013. Adherence to antiretroviral
therapy among HIV infected children measured by caretaker report, medication return,
and drug level in Dar Es Salaam, Tanzania. Biomedical Central Paediatrics 13(1):95.
From: http://www.biomedcentral.com/1471-2431/13/95 (accessed July 1 2013).

Mulissa, Z, Jerene, D & Lindtjørn, B. 2010. Patients present earlier and survival has
improved, but Pre-ART attrition is high in a six-year HIV cohort data from Ethiopia.
Public Library of Science ONE 5(10): e13268. doi:10.1371/journal.pone.0013268. From:
 130

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2952597/pdf/pone.0013268.pdf
(accessed 4 March 2013).

Mulu, A, Liebert, UG&Maier, M. 2014. Virological efficacy and immunological recovery

among Ethiopian HIV-1 infected adults and children. Biomedical Central Infectious
Diseases 14:28.

Murray, JS, Elashoff, MR, Iacono-Connors, LC, Cvetkovich, TA & Struble, KA. 1999.
The use of plasma HIV RNA as a study endpoint in efficacy trials of antiretroviral drugs.
AIDS 13:797–804.

Nakanjako, D, Ssewanyana, I, Nabatanzi, R, Kiragga, A, Kamya, MR, Cao, H &

Mayanja-Kizza, H. 2013. Impaired T-cell proliferation among HAART treated adults with
suboptimal CD4 recovery in an African cohort. Biomedical Central Immunology 14:26.
From: http://www.biomedcentral.com/1471-2172/14/26(accessed 31 August 2013).

Ortegoa, C, Huedo-Medinab, TB, Vejoa, J & Llorcac, FJ. 2011. Adherence to highly
active antiretroviral therapy in Spain. A meta-analysis. Gac Sanit 25(4):282–289.

Oxford Dictionary Online. 2013. Sv ‖HAART‖ ―failure‖ ―prevalence‖. Oxford: Oxford

University Press.
From: http://www.oxforddictionaries.com) (accessed 1 August 2013)

Palchaudhuri, S, Tweya, H & Hosseinipour, M. 2014. Assessment of laboratory test

utilization for HIV/AIDS care in urban ART clinics of Lilongwe, Malawi. Malawi Medical
Journal 26 (2) 42-44.

Phillips, JC, Webel, A, Rose, CW, Corless, IB, Sullivan, CM, Voss, J, Wantland, D,
Nokes, K, Brion, J, Chen, W, Lipinge, S, Eller, LS, Tyer-Viola, L, Rivero-Méndez, M,
Nicholas, PK, Johnson, MO, Maryland, M, Kemppainen, J, Portillo, CM,
Chaiphibalsarisdi, P, Kirksey, KM, Sefcik, E, Reid, P, Cuca Y, Huang, E & Holzemer,
WL. 2013. Associations between the legal context of HIV, perceived social capital, and
HIV antiretroviral adherence in North America. Biomedical Central Public Health 13:736.
From: http://www.biomedcentral.com/1471-2458/13/736 (accessed 5 August 2013).
 131

Podlekareva, DN, Reekie, J, Mocroft, A, Losso, M, Rakhmanova, AG, Bakowska, E,

Karpov, IA, Lazarus, JV, Gatell, J, Lundgren, JD & Kirk, O. 2012. Benchmarking HIV
health care: from individual patient care to health care evaluation. An example from the
EuroSIDA study. Biomedical Central Infectious Diseases 12:229. From:
www.biomedcentral.com/1471-2334/12/229 (accessed 27 August 2013).

Polit, DF & Beck, CT. 2012.Nursing research: generating and assessing evidence for
nursing practice. 9th edition. Philadelphia: Lippincott Williams and Wilkins.

Pujades-Rodriguez, M, O‘Brien, D, Humblet, P & Calmy, A. 2008. Second-line

antiretroviral therapy in resource-limited settings: the experience of Medecins Sans
Frontieres. AIDS 22(11):1305-1312.

Rawizza, HE, Chaplin, B, Meloni, ST, Eisen, G, Rao, T, Sankale´, JL, Dieng-Sarr, A,
Agbaji, O, Onwujekwe, DI, Gashau, W, Nkado, R, Ekong, E, Okonkwo, P, Murphy, RL
& Kanki, PJ. 2011. Immunologic criteria are poor predictors of virologic outcome:
implications for HIV treatment monitoring in resource-limited settings.HIV/AIDS. 53:
1283-1290.

Reda, AA, Biadgilign, S, Deribew, A, Gebre, B & Deribe, K. 2013. Predictors of Change
in CD4 Lymphocyte Count and Weight among HIV Infected Patients on Anti-Retroviral
Treatment in Ethiopia: A Retrospective Longitudinal Study. Public Library of Science
ONE 8(4): e58595.
From: www.plosone.org/article/info:doi/10.1371/journal.pone.0058595 (accessed 20
May 2013)‎.

Reynolds, SJ, Nakigozi, G, Newell, K, Ndyanabo, A, Galiwongo, R, Boaz, I, Quinn, TC,

Gray, R, Wawer, M&Serwadda, D. 2009.Failure of immunologic criteria to appropriately
identify antiretroviral treatment failure in Uganda. AIDS 23(6): 697–700.

Roux, P, Kouanfack, C, Cohen, J, Marcellin, F, Boyer, S, Delaporte, E, Carrieri, P,

Laurent, C & Spire, B. 2011. Adherence to antiretroviral treatment in HIV-positive
patients in the Cameroon context: promoting the use of medication reminder methods.
Journal of Acquired Immune Deficiency Syndrome 57 Supplement 1:S40-43.
 132

Sabin, CA & Lundgren, JD. 2013. The natural history of HIV infection. Current Opinion
in HIV and AIDS 8(4):311-317.

Seligmann, M, Warrell, DA, Aboulker, JP, Carbon, C, Darbyshire, JH, Dormont, J,

Eschwege, E, Girling, DJ, James, DR, Levy, JP, Peto, TEA, Schwarz, D, Stone, AB,
Weller, IVD, Withnall, R, Gelmon, K, Lafon, E, Swart, AM, Aber, VR, Babiker, AG,
Lhoro, S, Nunn, AJ & Vray, M. 1994. Lancet 343(8902):871-881.

Shaughnessy, JJ, Zechmeister, EB & Zechmeister, JS. 2012. Research methods in

psychology.9th edition. New York: McGraw-Hill.

Sodora DL & Silvestri G. 2008.Immune activation and AIDS pathogenesis. AIDS 22:
436–446.

Tiyou, A, Belachew, T, Alemseged, F & Biadgilign, S. 2010. Predictors of adherence to

antiretroviral therapy among people living with HIV/AIDS in resource-limited setting of
Southwest Ethiopia. AIDS Research and Therapy 7:39. From:
http://www.aidsrestherapy.com/content/pdf/1742-6405-7-39.pdf (accessed 21 February
2013).

Touloumi, G, Pantazis, N, Chaix, M-L, Bucher, HC, Zangerle, R, Kran, AB, Thiebaut, R,
Masquelier, B, Kucherer, C, Monforte, AA & Meyer, L. 2013.Virologic and Immunologic
Response to cART by HIV-1 Subtype in the CASCADE Collaboration. Public Library of
Science ONE 8(7): e71174. From:
http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0071174
(accessed August 13 2013).

Tsegaye, E & Worku, A. 2012.Assessment of antiretroviral treatment outcome in public

hospitals, South Nations Nationalities and Peoples Region, Ethiopia. Ethiopian Journal
of Health Development 25(2):102-109.

Turner, BJ. 2002. Adherence to Antiretroviral Therapy by Human Immunodeficiency

Virus—Infected Patients. Journal of Infectious Diseases 185 (Supplement 2): S143-
S151. doi: 10.1086/340197.
 133

From: http://jid.oxfordjournals.org/content/185/Supplement_2/S143.full.pdf (accessed

February 21, 2013)

UNAIDS. See Joint United Nations Programme on HIV/AIDS.

United Nations. 2012. The Millennium Development Goals Report 2012. New York:
United Nations publishers.

Webb, P, & Bain, C. 2011. Essential epidemiology: an introduction for students and
health professionals. 2nded. New York, NY: Cambridge University Press.

Were, MC, Shen, C, Tierney, WM, Mamlin, JJ, Biondich, PG, Li, X, Kimaiyo, S &
Mamlin, BW. 2011. Evaluation of computer-generated reminders to improve CD4
laboratory monitoring in sub-Saharan Africa: a prospective comparative study. Journal
of American Medical Informatics Association 18:150-155.

Weiss, RA. 2008. 25 years of HIV research: successes and challenges. Clinical and
Experimental Immunology 152(2): 201-210.

WHO. See World Health Organisation.

Workneh, N, Girma, T & Woldie, M. 2009.Immunologic and Clinical outcomes of

Children on HAART: a retrospective cohort analysis at Jimma university specialized
hospital. Ethiopian Journal of Health Science 19(2):75-82.

World Health Organization. 2003. Adherence to long term therapy: evidence for action.
Geneva: World Health Organization.

World Health Organization. 2008. Task shifting: global recommendations and

guidelines. Geneva: World Health Organization.

World Health Organization. 2009. Towards universal access scaling up priority

HIV/AIDS interventions in the health sector: progress report. Geneva: World Health
Organization.
 134

World Health Organization. 2010a. Antiretroviral therapy for HIV infection in adults and
adolescents: recommendations for a public health approach– 2010 revision. Geneva:
World Health Organization.

World Health Organization.2010b. WHO recommendations on the diagnosis of HIV

infection in infants and children. Geneva: World Health Organization.

World Health Organization. 2012a. Service delivery approaches to HIV testing and
counselling (HTC): a strategic HTC policy framework. Geneva: World Health
Organization.

World Health Organization. 2012b. WHO HIV drug resistance report 2012. Geneva:
World Health Organization.

World Health Organization. 2013. Consolidated guidelines on the use of antiretroviral

drugs for treating and preventing HIV infection: recommendations for a public health
approach. Geneva: World Health Organization.

Yirdaw, KD, Jerene, D, Gashu, Z, Edginton, ME, Kumar, AMV, Letamo, Y, Feleke, B,
Teklu, AM, Zewdu, S, Weiss, B & Ruff, A. 2014. Beneficial effect of isoniazid preventive
therapy and antiretroviral therapy on the incidence of tuberculosis in people living with
HIV in Ethiopia. Public Library of Science ONE 9(8): e104557.

Zachariah, R, Reid, SD, Chaillet, P, Massaquoi, M, Schouten, EJ & Harries, AD. 2011.
Why do we need a point-of-care CD4 test for low-income countries? Tropical Medicine
and International Health 16(1):37–41.

Zaidi, SH. 2014. Ethics in medicine. Switzerland: Springer.

Zuckerman, AJ, Banatvala, JE, Griffiths, P, Schoub, B, & Mortimer, P. 2009. Principles
and practice of clinical virology.6th edition.London: John Wiley & Sons.
 135

ANNEXURE
Annexure A: Document checklist

ID Site Code
A. Socio demographic data
A.1. Age Category Tick one Code 1
<15
≥15
A.2. Gender Category Tick one Code 2
Female
Male
B. Clinical data
B.1. Baseline WHO Stage Category Tick one Code 3
I or II
III or IV
B.2. Baseline CD4 Count Value

B.3.1 Tuberculosis diagnosed? Category Tick one Code 4

No
Yes
B.3.2 If 'Yes', date Tuberculosis diagnosed: Day: _____ Month: _____ Year:_____
B.4.1 Adherence Category Tick one Code 5
Good
Poor
B.4.2. If 'Poor', month and year ARV pick up was missed Month: _____ Year:_____
Month: _____ Year:_____
Month: _____ Year:_____
Month: _____ Year:_____
 136

Value
B.5.1. Weight at last follow-up visit
B.5.2. Weight six months prior last follow-up visit
B.5.3. Weight loss in the last Category Tick one Code 6
six months Weight loss>10% in the last six months
Weight loss 5%-10% in the last six months
Weight loss <5% in the last six months
B.6. Date ART started: Day: _____ Month: _____ Year:_____
B.7. Last follow-up status Category Tick one Code 7
Active
Lost
Drop
Stop
Dead
Transferred
out
B.8. Last visit date: Day: _____ Month: _____ Year:_____
C. Immunologic data
C.1. Immunologic monitoring
CD4 tests determined within one month of designated follow-up month? Tick on cells that apply
0 6 12 24 36 48 60 72 84 96 108

C.2. Immunologic response

Enter CD4 count value that was determined within two months of designated follow-up month
0 6 12 18 24 30 36 42 48 54 60

66 72 78 84 90 96 102 108
 137

C.3. Immunologic treatment failure

C.3.1. Immunologic treatment failure diagnosed? Category Tick one Code 8
(If ‗No‘, go to item C.3.6) No
Yes
C.3.2. Date first immunologic treatment failure diagnosed: Day: _____ Month: _____ Year:_____
C.3.3. Immunologic treatment failure criteria used to Category Tick one Code 9
make diagnosis Follow-up CD4 Persistently <100
<50% of peak follow-up CD4 value
below baseline CD4 value
C.3.4. Follow-up CD4 present after first immunologic Category Tick one Code 10
treatment failure diagnosis? (If ‗No‘, go to item C.3.6) No
Yes
C.3.5. Immunologic treatment failure at last CD4 Category Tick one Code 11
evaluation present? No
Yes
C.3.6. Last treatment regimen containing Category Tick one Code 12
lopinavir/ritonavir? No
Yes
 138

Annexure B University of South Africa ethics clearance letter

 139

Annexure C Application to conduct study sent to Ethiopian National Research Ethics

Review Committee
140
 141

Annexure D Ethics clearance letterfromthe Ethiopian National Research Ethics Review

Committee
 142

Annexure E Southern Nations Nationalities and Peoples’ Regional State Health Bureau
support letter
 143

Annexure F Nigist Eleni Mohamod Memorial (Hosanna) hospital support letter

 144

Annexure G Butajira hospital support letter