BCBR - High yield topics

Dr T Siva Ilango
 Three levels of review are essential

• Scientific review: novelty, rationality, justification

• Ethics review: Human subjects protection

• Regulatory review: Foreign funding, sample shipment,

intellectual property, exchange of visitors
Random error representing wrong result due to
chance: unknown sources of variation that can distort
findings in either direction
• Can be minimized by increasing sample size and
increasing precision

Systematic error signifying wrong result due to bias -

mostly due to variation that would distort the results in
one direction
• Can be minimized by improving study design
Confounders: Affect both study variable as well as
outcome
• Effect can be minimized by proper study design and
through stratified analysis

Effect modifiers: Can alter [generally negatively] the

relationship between the study variable and outcome
by independently affecting outcome
• Good to be aware of them through adequate
literature review and not to include them in the study
• Translational research:
Bench to Bedside, designed to improve health
outcomes
Basic biology/clinical trials to techniques and tools

• Implementation Research:
Application of research findings in Practice, including
Policy
 The life cycle of research

• Identify data needs

• Spell out the research question
• Formulate the study objectives
• Plan the analysis
• Prepare data collection instruments
• Collect data
• Analyse data
• Draw conclusions
• Formulate recommendations
• Inform the stakeholders
 Research question sets out

What the investigator wants to know

NOT
☒What the investigator might do or
☒What the results of the study might ultimately
contribute to that particular field of science
 Steps in conceiving a research question

• Review of state-of-art information

• Raise a question
• Decide worth investigating by peer-review
• Define measurable exposures & outcomes
• Sharpen the initial question
• Refine the question by specifying details
Good research question should pass the ‘so what?’ test

•Feasible
•Interesting
•Novel
•Ethical
•Relevant
Purely descriptive questions DO NOT require
hypothesis
 Characteristics of good hypothesis

Simple
•One exposure
•One outcome
Specific
•No ambiguity about study participants/variables
Stated in advance
•Written at outset
•Focused on primary objective
Objectives for descriptive vs. analytical studies

Descriptive: Estimating a quantity

•Use the verb“Estimate”
•E.g., Estimate prevalence of physical activity

Analytical: Testing a hypothesis

•Use the verb“Determine”
•E.g., Determine whether exercise reduces blood sugar
level
 Literature review

• Searching PUBMED
• Boolean Query
• Lung or Infections yield more results than
‘and’ ‘not’

MeSH: Medical Subject Headings

(NLM controlled Vocabulary Thesarus)
 Prevalence – (P)

• Number of existing cases (old and new) in a defined

population at a specified point of time

# people with disease at a specified time

P = -------------------------------------------------- x 10n
Population at risk at the specified time

• In some studies the total population is used as an

approximation if data on population at risk is not available
 Scenario

• Population of 150 persons

• Follow up for one year
• 25 had a disease of interest at the beginning
• Another 15 new cases developed during the year
Calculate:
•Point prevalence at the start of the period
P = C/N = 25 /150 = 0.17 (17 %)

•Period prevalence for the year

PP = (C+I)/N = (25+15)/150 = 0.27 (27 %)
 Prevalence

Increase Decrease
•Long duration •Shorter duration
•Low cure rate •High cure rate
•Low case fatality •High case fatality
•Increase in new cases •Decrease in new cases
•Immigration of patients •Emigration of patients
•Improved detection •Improved cure rate
•Emigration of healthy people •Immigration of healthy people
 Incidence – (I)
•Number of new cases in a given period in a
specified population
•Time, (i.e., day, month, year) must be specified
•Measures the rapidity with which new cases are
occurring in a population

Can be expressed:
•In absolute numbers
•In terms of cumulated incidence
•In terms of incidence density
• Cumulated Incidence – Attack rate, indicate RISK
• Incidence Density- Incidence rate, Person-time
observed
• I – Evaluate trends in diseases, impact on primary
prevention programs
• P=I*D
• Case Fatality : Expressed as ratio/proportion,
Deaths/cases
• P is a static measure While I is dynamic measure
• Mortality Vs Fatality
• Case reports and case series are useful for
uncommon clinical manifestations
• Ecological studies can be used to relate group level
data and generate hypothesis
• Cross sectional surveys help to measure the burden
or magnitude of health condition

• Cross sectional surveys – Prevalence Studies

• Cohort Studies : Relative Risk = a/a+b % b/b+d
• Incidence study
• Eg: Framingham Heart Study, Aniline Dye industry

• Case Control Studies : Odds ratio = ad/bc

• Smoking causes Lung cancer
 Cohort study – Strengths and weaknesses
Strengths
•Allows calculation of incidence
•Examine multiple outcomes for a given exposure
•Clarity of temporal sequence
•Good for investigating rare exposures

Weakness
•May have to follow large numbers of subjects for a long time.
•Expensive and time consuming.
•Not good for rare diseases.
•Not good for diseases with a long latency.
•Differential loss to follow up can introduce bias.
 Case control study: Strengths and weaknesses

Strengths
•Good for examining rare outcomes or outcomes with long latency
•Relatively quick to conduct, inexpensive
•Requires comparatively few subjects
•Multiple exposures or risk factors can be examined

Weaknesses
•Susceptible to recall bias
•Selection of an appropriate comparison group may be difficult
•Rates of disease in exposed and unexposed individuals cannot be
determined
 Phases in clinical trials and objectives
Trial phase End-points/ objectives Sample size and Participants
•Phase I Safety Up to 50
Acceptability Healthy volunteers

•Phase II Long-term safety 100 to 500

Dose and schedule Low risk
Early indications of efficacy

•Phase III Effectiveness 1000 and more

High risk
•Phase IV Post-marketing surveillance 1000 and more
Community based
 Blinding
Eliminates,
1. Co-intervention
2. Biased outcome ascertainment

Single - Participants
Double – Participants and Investigator
Triple – Participants, Investigator, Analyst
 Advantages & Disadvantages of RCTs

Advantages
• The only effective method known to control selection bias
• Controls confounding bias without adjustment
• Facilitates effective blinding in some trials
• Maintains advantages of cohort studies

Disadvantages
• May be complex and expensive
• Lack representativeness – volunteers differ from population of
interest
• Ethical challenges are immense
• Major Limitation of a Cross Sectional Study –
Does not establish disease etiology

• Relative Risk : is a Proportion

 Validity of studies

• External Validity - Generalizability

• Internal Validity – Accurate estimate

• Accuracy – Validity and Precision

 Threats to Validity of studies

• Selection Bias
• Information Bias
• Confounding

Errors in Estimation
Random – Chance
Systematic - Bias
 Selection Bias

• Designing stage - use Incidence cases, population

study design

• Data collection stage – Blinding

• Analysis Stage – Sensitivity Analysis

 Information Bias

• Recall Bias
• Investigator Bias
• Prevarication – systematic distortion of truth
by subjects
 Dealing with information bias

• Precise operational definitions of variables

• Detailed measurement protocols
• Repeated measurements on key variables
• Training, certification and re-certification
• Data audits (of interviewers, of data centers)
• Data cleaning – visual, computer
• Re-running all analyses prior to publication
 Dealing with confounding
Coffee /Smoking/MI
Design stage

•Restriction
•Matching
•Randomization
(Experimental studies)

Analysis stage

•Stratification
•Multivariate analysis
 Qualitative Research – EMIC

• In depth interviews – Sensitive issues

• FGD – Group interaction, more info in less time
• Participant Observation – Researcher becomes
participant in social event

• Triangulation – Use of multiple methods, sources,

theories to generate hypothesis
Quantitative Qualitative

Numbers Text

Objective (reliability) Subjective (Credibility)

ETIC EMIC

Deductive Inductive

Generalizable across Requires conversion in

cultures abstract
 Types of Data

Qualitative
•Nominal
•Eg. Color of Eyes
•Ordinal
•Eg. Stages of disease condition

Quantitative
•Discrete
•Eg. Family size
•Continuous
•Eg. Height / Weight
 Measurement of variables

• Central Value/tendency – Mean/Median/Mode

• Measure of Dispersion – Range, Interquartile range,
Mean Deviation, variance/Standard Deviation
• Square of SD is variance
• Coefficient of Variation (CV) – SD expressed as
percentage of Arithmetic Mean. (SD/AM*100)
• Mode is used for Nominal Data
• Mean and SD are commonly used
• Sample needs to be representative in terms of
time, place and person

• Sampling unit
• Sampling frame
• Sampling scheme
 Why do we sample populations?

•Obtain information from large populations

•Ensure the efficiency of a study
•Obtain more accurate information

Type of samples
Non-Probability Samples (convenient,
subjective)
Probability samples
 Probability Samples

• Simple Random
• Systematic – every ‘n’ individual
• Stratified – One from every strata
• Clusters – Sampling unit is group of subjects
• Multistage – Several sampling units
 Steps in Estimating Sample Size
•Identify major study variable
•Determine type of estimate (%, mean, ratio,...)
•Indicate expected frequency of factor of interest
•Decide on desired precision of the estimate
•Decide on acceptable risk that estimate will fall
outside its real population value
•Adjust for population size
•Adjust for estimated design effect
•Adjust for expected response rate
• Alpha – Significance Level – rejecting null
hypothesis – Confidence Level (1-alpha)– Type
1 error

• Beta – Failing to reject – Power – (1-Beta) -

Type 2 error
• Sample size for Mean : z square Sigma
square/d square
• Sample size for Proportion:n = z square pq/d
square
• In Case control Study: Higher the OR, lower
the sample size
• 10% rule: increase the sample size 10% for
each confounder/variable added
• The design effect can be calculated after study
completion, but should be accounted for at the design
stage.

• The design effect is 1 (i.e., no design effect) when

taking a simple random sample.

• The design effect varies using cluster sampling; it is

usually estimated that the design effect is 2 in
immunization cluster surveys.
 Study Subjects

• Representative
• Adequate size
• Acceptable cost
 Selection of study population - steps
1 . TARGET POPULATION
Specific set of inclusion criteria Establish the demographic and clinical characteristics
of subjects
2 . ACCESSIBLE POPULATION
Specific set of inclusion criteria
geographically and temporally convenient
3 . SUBSET OF POPULATION
parsimonious set of exclusion criteria Eliminate subjects who are unethical or
inappropriate to study
4 . SAMPLING
Sampling technique
Estimated Sample size
5 . RECRUITMENT STRATEGY
large enough sample to meet study needs, minimizes bias due to non-response and
loss to follow-up.
 Common reasons for study failures

1.Badly defined research question and objectives

2.Unrealistic timescales - too short or too long
3.Inappropriate and incompetent staff: Lack of
direction, motivation and training
4.Inadequate monitoring, failure to respond to
contingent situations and carry out mid-course
corrections
 Tools to collect data

• Abstraction forms – registers

• Structured observation guide – Checklist of items
• Questionnaire – Self administered, Interviewer
administered

• Questions: Open, open with closed answers, Closed

(Dichotomous, Multiple, quantitative) and semi-open
questions (Others., specify……)
 Six steps in data collection

1.Draft question-by-question guide

2.Train staff members who will collect data
3.Initiate data collection and ensure quality
4.Review collected data for quality and completeness
5.Debrief to trouble shoot difficulties
6.Validate
 Checks to conduct

Completeness
•Did the field worker fill all items?
Readability
•Is the writing readable?
Consistency
•Do the answer make sense?
•Is there internal consistency?
Basic structure of a database
•Lines represent records
•Columns represent variables

Structure of variable in the database

Integer, Numeric, Alpha-numeric, dates
 Individual and aggregated databases

Individual databases
•Each record is an observation

Aggregated database
•Records contain counts

Normalized database
•Only one count by record
•Facilitates further aggregation

Mother and Daughter Databases

Examine outcome/exposure association

•Based on a priori hypotheses

Compare groups for frequency of exposures
using appropriate measure of association

•Based on prior knowledge

•Based on study design

 Conduct advanced analysis

•Dose-response
•Stratifications
•Multivariate analysis – Logistic Regression

Data Dredging – use of data mining to uncover

patterns in data that can be presented as
statistically significant.
Use software with data management & analysis
tools.
e.g., EpiInfo*, Avoid Spreadsheets
Group consent is desirable [e.g. tribal studies]

Audio consent and video consent: Mandatory

for investigational new drug [IND] trials in India
 Importance of scientific review

Explores the scientific novelty, rationality and relevance

1.Justification for conducting the trial in the context of
national priorities
2.Scientific merits of the research project and
feasibility: Review of toxicological studies, laboratory
and animal data
3.Technology transfer and capacity building at sites
 Objectives of regulatory review

•Evaluate pre-clinical trials data

•Assess in-country regulatory requirements for drug/ vaccine/

product import
•Ensure national requirements for special situations -genetically
engineered products, stem cell research, research on reproductive
technologies, organ transplantation etc.
•Sample shipment and transfers, transfer of raw data: IPR issues
•Exchange of scientists or visitors
•Budget: Foreign funding
•Research in border or high-security areas
 The main challenges perceived by international investigators and
sponsors include

•Delayed approval
•Quality of ethics review
•Shipment of samples: import and export
•Overall deficiency of duly trained investigators and centers
•Clause of compensation even for clinical trial participants
•Recent requirement of audio visual recording of consent
process for IND [investigational new drug] trials, only in
specified situations
 Impediments in clinical trial participation

At the level of patients

•Don’t know about clinical trials
•Don’t have access to clinical trials
•May be afraid or suspicious of research
•Can’t afford to participate
•May not want to go against health care provider’s wishes

At the level of health care providers

•Lack awareness of appropriate clinical trials
•Be unwilling to “lose control” of a person’s care
•Believe that standard therapy is best
•Be concerned that clinical trials add administrative burdens
 The seven steps of a successful protocol

• Identify topic, question and objectives

• Outline a one-page concept paper
• Prepare dummy tables
• Write draft protocol
• Prepare instruments and annexes
• Submit to peer review
• Seek ethics committee review
Outline of the one-page, bullet-style, concept paper

• Background and justification

• Objectives
• Methods
• Expected benefits
• Key references
• Budget
 EQUATOR network

• Enhancing quality and transparency of Health

Research
 Components of Publication Ethics

1.Ethics review/ Breach of confidentiality

2.Fabrication and Falsification

3.Authorship
4.Plagiarism

5.Ethics related to submission

6.Conflict of interest
 Ethics review/ Breach of confidentiality

• Trials-Registration with ‘Clinical Trial Registry of

India’ (CTRI)

• Informed consent/ assent- A must when you are

conducting research with human participants

• Data confidentiality-Without institutional permission,

you can not share your data with someone for
analysis who is working in a different institution
• If the research results not generated from the study
(fabrication) or generated by manipulating data
(falsification)

• Simultaneous submission-submitting the same

manuscript simultaneously at the different journals

• Duplicate publication-Submitting a new manuscript

with same hypothesis, methods, data, discussion and
conclusion
Redundant publications ('salami' publications)
• Publishing similar manuscripts/reports based on the
same experiments

• Use anti-plagiarism software tools like ‘URKUND’,

‘Ithenticate’ or ‘Turnitin’ etc. (as per University
norms)