REVIEW ARICLE www.ijcmr.

com

Identifying and Eliminating Bias in Interventional Research Studies

– A Quality Indicator
C. Santhosh Kumar1, Shweta Somasundara Yale2

intensely on some research assignments while neglecting

ABSTRACT others – that tends to report a poor quality or extremely low
This review article deals with highlighting the importance of number of high quality research papers.
identification and elimination mechanisms of important biases Hence, the authors considered to present this review with the
in interventional research studies.In simple terms, Bias means a following objectives:
systematic error that can occur in the event of any phase of the • To become familiar with various types of bias in
research, during planning, implementing, data collection, analysis experimental research study.
and also during publication stage. An in-depth knowledge • To discuss how bias influences experimental research study
regarding the bias allows researchers and readers to critically
and highlights some of its sources.
and independently review the scientific literature and avoid
• To use the knowledge of such biases that may help us
interventions which are suboptimal or potentially harmful. A
thorough understanding of bias has a stronger implication towards recognize them and minimize their impact on the planning
conducting good research and publishing high quality articles, of research and health-related decisions.
which are very much essential for the practice of evidence-based METHODS
practice.
A search strategy was done in 04 electronic databases and
Keyword: Identifying and Eliminating Bias e-books, for English-language source, published over the
period 1995-2015 for the topics of bias in interventional
studies, strategies to overcome those biases, strengthening of
INTRODUCTION interventional studies by elimination of bias in experimental
In scientific terms Bias is "any factor or process that tends to research study. Hand searching was additionally conducted
deviate the results or conclusions of a trial systematically away in relevant research methodology books. The intent of this
from the truth". It can be simply defined as "the deviation from literature search was to identify and review the potential sources
the truth".1 of bias in interventional studies and methods to overcome for
Bias is not uncommon in interventional trials, but to a greater conducting quality based studies in this review.
extent in Indian studies compared with those western trials, RESULTS
says one of the systematic review and comparative empirical
The different types of biases that occur in interventional studies
analysis of randomized controlled trial reports published in
are broadly categorized into biases in Randomized Controlled
selected Chinese, Indian, and European or North American
Trials [RCT] and Non-Randomized Studies [NRS].
medical journals.2
Further, the review deals with discussing different biases in
The possible reasons for trials appearing biased, which reflect interventional studies, its sources and highlighting elimination
underlying inadequacies in the design and conduct of the trials strategies of some of the important biases.
are:
1. Inadequate knowledge of the researcher in accurate I. Biases that can arise, even before the trial is conducted
designing and conduction of trials. 1. Choice-of-Question Bias
2. Indian higher education looks research as a differential It is one of the most unrecognized types of bias that occur in
component in the academic functioning. It is not RCTs. This bias is concealed within the question that the study
intends to answer. This bias may not have a stronger impact on
considered as a lucrative career option. Apart from this,
the strength of the study but it may affect the generalizability of
resource constraints, lack of commitment, lack of proper
the study outcomes.6
encouragement, etc., are the impediments that are affecting
This bias can take many forms:
the quality of research in our institutions of higher
i. Hidden agenda bias:
education.3
3. Career in pharmacology, physiology and other basic
sciences are not rewarding in India. So the doctors who Senior Lecturer, 2Reader, Department of Public Health Dentistry,
1

opt for these branches are usually from the bottom of the RRDCH, Bangalore, India
talent pool. That leads to the poor quality of basic medical
research in India. Corresponding author: Dr. C. Santhosh Kumar, Department of Public
4. Those researchers and journal editors in India not adopting Health Dentistry, RajaRajeswari Dental College and Hospital, No.14,
Ramohalli Cross, Kumbalagodu, Bangalore: 560074, India
the CONSORT reporting guidelines4, are sole responsible
for the rejection of papers in International Journals and How to cite this article: C. Santhosh Kumar, Shweta Somasundara
publishing of poor quality trials respectively. Yale. Identifying and eliminating bias in interventional research studies
5. Research organizations conducting and reporting a trial in – a quality indicator. International Journal of Contemporary Medical
favor of the funder, budgeting their efforts5 – working more Research 2016;3(6):1644-1648.

1644
International Journal of Contemporary Medical Research
Volume 3 | Issue 6 | June 2016 | ICV: 50.43 | ISSN (Online): 2393-915X; (Print): 2454-7379
Kumar, et al. Identifying and Eliminating Bias

It occurs when a trial is mounted, not in order to answer a rather than the important long-term outcomes.
question, but rather to demonstrate a pre-required answer. 5. Selection Bias
ii. Cost and convenience bias: Randomization is an important protocol in RCTs which
It occurs when a study is done on a basis of what we can ensures that all the study participants are provided with equal
afford to study, or what is convenient to study, rather than opportunity to be selected for each study groups.8
what we really want to study. It can seriously compromise Selection bias can occur if some potentially eligible individuals
what we choose to study.
are selectively excluded from the study, because the investigator
iii. Funding availability Bias:
knows the group to which they would be allocated if they
It occurs where studies tend to concentrate on questions that
participated.
are more readily fundable, often for a vested or commercial
How can selection bias be reduced?
interest.7
• Selection bias can be reduced by concealing the
2. Bureaucracy Bias randomization sequence from the investigators at the time
• In simple terms it can be called as Institutional Review of obtaining consent from potential trial participants.
Board (IRB) bias. • Allocation concealment is a very simple maneuver that
• It most commonly occurs when IRB are unduly constrictive, can be incorporated in the design of any trial and that can
and non-permissive for the study of important concepts. always be implemented.
• It also occurs when IRB unduly allows and encourages • Allocation concealment defined “as an important technique
studies which are scientifically invalid, but having the which protects the randomization mechanism, ensuring
potential hold to get the funds or name to the institution.6 that the patient is completely unaware of the treatment been
II. Biases that can arise, during the actual course of the trial rendered before entering into the study”.9
1. Population Choice Bias • Despite its simplicity as a maneuver and its importance to
This bias can occur when the sample is drawn multiple times reduce bias, allocation concealment is rarely reported, and
from the same population and it can have profound impact on perhaps rarely implemented in RCTs. If, however, allocation
the external validity of randomized trials. concealment was not carried out, the majority of RCTs are
In certain conditions, the sampling is done with a specific gender at risk of exaggerating the effects of the interventions they
predilection (gender bias) or towards a particular age group were designed to evaluate.
(age bias), the outcomes of such study may not be generalizable • Sometimes, the researchers do tend to access the allocation
to the study population. codes, which are kept in sealed opaque envelopes. The
There are subgroups of population choice bias like informed most commonly used methods are powerful lights or high
consent bias, literacy bias and language bias wherein the intense steam to open the envelope and later reseal it, before
investigators may intentionally avoid eligible patients just others notice it. This may cause selection bias into RCTs.6
because they do not comprehend the consent form.6 6. Ascertainment Bias
2. Intervention choice Bias Ascertainment bias occurs when the results or conclusions
It occurs when the type of the intervention chosen by the of a trial are systematically distorted by knowledge of which
investigator can affect the study outcomes widely.6 intervention each participant is receiving.
Ascertainment bias can be introduced by:
i. Complexity bias
• The person administering the interventions,
It can occur when a trial is used to study complex interventions,
• The person receiving the interventions (the participants),
with a number of components, or where outcomes may
• The investigator assessing or analyzing the outcomes,
depend on multiple contingencies outside of the control of the
• The report writer who describes the trial in detail.
investigator (e.g. the skill of the surgeons or the resources of the
i. Participant ascertainment bias: If participants know
community).
that they have been allocated to the placebo group,
3. Control group bias they are likely to feel disappointed and less willing to
This bias may appear when the intervention group is compared report improvement at each of the study time points.
with control group of poor design, which may erroneously ii. Observer bias: If the people in charge of assessing
project the outcomes to be more (or less) effective. Comparing and recording the outcomes know which patients are
an interventional group with a placebo clarifies the intervention allocated to each of the study groups, they could,
is effective or not. But, it does not reveal the experimental consciously or unconsciously, tend to record the
intervention provides better outcomes or not compared to the outcomes for patients receiving the new drug in a more
existing ones.6 An obvious way to make an intervention appear favorable way than for patients receiving placebo.
to be more effective than it really is would be to choose an
ineffective comparison group. How can Ascertainment bias be reduced?
The best way to protect a trial against ascertainment bias is by
4. Outcome choice Bias keeping the people involved in the trial unaware of the identity
i. Measurement bias of the interventions for as long as possible. This is called
It occurs in those RCTs that evaluate outcomes which are easy blinding or masking.
to measure, rather than the outcomes those are relevant. Ascertainment bias can widely be reduced by blinding all
ii. Time term bias the concerned people involved in the trials: the intervention
It occurs in those RCTs where short-term outcomes are measured providers, the interventions receivers and those concerned with

International Journal of Contemporary Medical Research 1645

ISSN (Online): 2393-915X; (Print): 2454-7379 | ICV: 50.43 | Volume 3 | Issue 6 | June 2016
Kumar, et al. Identifying and Eliminating Bias

assessment and reporting the outcomes.7,9 can cause this bias.10

The strategies that can be used to reduce ascertainment bias can I. Biases that occurs during reporting of a trial
be applied during at least two periods of a trial: 1. Withdrawal Bias
a. During the time of Data collection • This Bias can happen due to incorrect management of data
b. After data have been collected pertaining to patients’ refrainment, withdrawal mechanism
Strategies to reduce ascertainment bias during data and protocol violations.
collection phase • Any researcher would expect that all the trial participants
The best strategy to reduce ascertainment bias during should follow the protocol, provide data on all study
data collection is with the use of placebos. Placebos are outcomes at each point in time and ensure to complete the
interventions believed to be inactive, but otherwise identical trial. However, dropouts are most commonly encountered
to the experimental intervention in all aspects other than the in many studies.
postulated specific effect. One of the best comparisons in any • Dropouts can happen because of some participants tend to
trials are the Placebos, which are very easy to develop and refrain away from the study before the trial is completed or
apply in drug trials, and it is important that the placebo should inappropriate following of the protocol or because certain
resemble in taste, smell and appearance of the active drug, and study outcomes are incorrectly measured or even with the
should be delivered using same procedure as for the active drug. problems of multiple repeated measures.6
• On occasion, it is impossible to know the status of
Strategies to reduce ascertainment bias after data collection
participants at the times when the missing information
phase
should have been collected. Example: Relocation of
This bias can occur easily after data collection, which can be
participants without informing the investigator or failing to
controlled by keeping anonymity of the study groups, with
contact for an unknown reason. Those outcomes measured
the people involved with data analyzing and reporting the trial
and analyzed excluding these participants, and if it is
outcomes.
related to the interventions or the treatment rendered, can
In any trial, the coding of the study groups should be done prior to
cause bias.6
the time of providing the data to the statistical analysis, wherein
Reduction Strategies that can be used to eliminate withdrawal
the results thus obtained will contain the same codes and further
bias
the similar codes are followed until the trial reporting stage. The
• The first strategy is intention-to-treat analysis, which deals
codes remain undisclosed until all the process of analysis and
with including all the study participants in the data analyses,
reporting of the trial is completed.6
randomly allocated to their respective groups, irrespective
Selection Bias Ascertainment Bias of whether the participants completed the study or not.
[Bias due to lack of alloca- [Bias due to lack of blinding] • The second strategy is sensitivity analysis, which deals
tion concealment]
with accounting the worst possible outcomes or time points
Allocation concealment Blinding process helps to pre-
with worst results on one end or similar confinement of
helps to prevent selection vent ascertainment bias by
bias, protects the random- protecting the randomization best possible outcomes or time points in the group that
ization sequence before and mechanism, even after the al- shows the best results on the other end with reference to the
until the interventions are location to the study groups is dropouts. This is followed by sensitively analyzing the data
given to study participants done. It may or may not facil- for possibility of the results that may support or refute the
and can be always imple- itate to implement in certain initial analysis results, which includes the missing data.11
mented. conditions.7
2. Selective Reporting Bias
7. Contamination Bias A major and common source of bias in an RCT is selective
The control group subjects may mistakenly receive the maneuver reporting of results, describing those outcomes with positive
of interest or be affected by an extrinsic maneuver, which results, or which favor the studied intervention.
diminishes the differences in outcomes of the experimental and The sub-categories of this bias are:
control groups. • Social desirability bias in which the items that are desired,
are more likely to be reported.
8. Compliance Bias
• Data dredging bias / Interesting data bias:Following the
Sometimes, there can be erroneous outcomes which can impact
data analysis, the researcher may get influenced with those
the efficacy of the treatment rendered to the patients, and it could
outcomes which are of more concern / interesting to them
be possibly due to non-compliance to the treatment regimen.10
and subsequently report them, leaving behind the lesser
9. Bogus control Bias important ones.6
When subjects allocated to the experimental maneuver group Steps to reduce biases that occur during the course and
expire or withdraw before the maneuver is administered and are reporting of a trial
reallocated to the control group or are omitted, the experimental • Double blinding subjects and investigators when possible,
and control groups are no longer matched and the differences to prevent knowledge of exposures from influencing the
between may be biased toward the experimental group. detection of outcome events.
10. Proficiency Bias • Robust instrument development and validation process for
Absence of establishing the equilibrium with respect to the data collection.
experimental interventions or treatments rendered to subjects • Hide the identity of the subjects from the data collector

1646
International Journal of Contemporary Medical Research
Volume 3 | Issue 6 | June 2016 | ICV: 50.43 | ISSN (Online): 2393-915X; (Print): 2454-7379
Kumar, et al. Identifying and Eliminating Bias

when possible. technology in health care.

• Create a division of labor by having a different person 2. Resource allocation bias: It happens when readers’
record data than performs the maneuver. exhibit strong inclination for certain types of provision
• Maintain good contact to avoid attrition from the study. of resources. It is more widely seen in health care sector,
IV. Biases that can occur during dissemination of the trials originating from its potential stake holders ranging from
1. Publication bias consumers to policy makers.6
Publication bias may occur if any journal is more inclined to 3. Trial design bias: It occurs when a study that uses a
publish only the studies with positive outcomes or those with design supported, publicly or privately, by the reader (e.g.
a consumer advocate overrating an RCT that takes into
good study designs. This cannot be identified within a single
account patient preferences).
study but rather it can be elicited better in systematic reviews
4. Flashy title bias: It happens when the study results are
and meta-analysis. This can lead to over-emphasis of the
overvalued based on their attractive titles (especially
outcomes and may mislead the readers.12
by the journalists) or undervalued by the academicians,
The failure to write and publish negative results is not a random
considering as undue sensational in the field.6
event, but is heavily influenced by the direction and strength
of research findings, whereby manuscripts with statistically Bias in Non-Randomized Experimental Studies
significant (positive) results are published preferentially over Bias may be present in findings from Non-Randomized
manuscripts reporting non-significant (negative) results. experimental studies in many of the same ways as in poorly
designed or conducted randomized trials. For example,
How to reduce publication bias?
the indistinct exclusion criteria, absence of monitoring the
• Publication bias can be eliminated through compulsory
standardized protocols during intervention and outcome
registration of trials at inception, and publication of the
assessments and lack of blinding are the most probable causes
results of all trials.
for bias, irrespective of whether the trial is randomized or non-
• Establishing the equilibrium between numbers of studies
randomized.14
published with positive and negative results.
• Non-randomized experimental study are susceptible to the
• Sensitization of the researchers should be done regarding
same bias as RCTs.
the importance of negative studies being published, with
• Selection Bias – caused by inadequate selection of
journals giving equal space for the publication of the same.
participants.
• Evaluation of studies should be based on the internal
• Performance Bias – caused by inadequate measurement of
validity of the study rather than the conclusions.13
intervention.
• Identify funding sources and possible conflicts of interest.
• Detection Bias – caused if the assessment of outcomes is
Variants of publication Bias:
not standardized or blinded.
i. Language bias • Attrition biases – caused by inadequate handling of
Bias which may arise due to predilection of certain authors incomplete outcome data because of drop-outs.15
for submitting and publishing their papers by journals in • Reporting Bias – caused by selective outcome reporting
different languages, based on the direction of their results. The study designs classified as NRS, and their varying
E.g.: Presumption that studies with positive results are more susceptibility to different biases, makes it difficult to produce
published in English.6 a generic robust tool that can be used to evaluate risk of bias. 16
ii. Time lag bias 19th Cochrane Colloquium and VI International Conference
Bias that occurs when the speed of publication depends on the on Patient Safety, held at Madrid during 19 – 22 October 2011
direction and strength of the results of the trial. In general, it has dealt with development and validation of a new Instrument
seems that trials with ‘negative’ results take twice as long to be – Risk of Bias Assessment tool for Non-Randomised Studies
published as ‘positive’ trials. (RoBANS). And discussion concluded that RoBANS is a valid
tool designed to assess the Risk of Bias of Non-Randomised
V. Biases that can occur during uptake phase of the trials Studies.
The following are some of the biases which are most common The Cochrane’s RoB tool and GRADE (Grading of
and pertinent: Recommendations Assessment, Development and Evaluation),
1. Relation to the author bias / rivalry Bias endorses RoBANS, henceforth it can be incorporated into
Bias that can occur by under-rating the strengths or exaggerating RevMan and GRADEpro, which appears to be useful to
the weaknesses of studies published by a rival. undertake systematic reviews.17 A Cochrane Risk Of Bias
2. Clinical practice Bias Assessment Tool: for Non-Randomized Studies of Interventions
(ACROBAT-NRSI) is another robust tool that can be used for
It takes place when readers judge a study according to whether it
evaluating the risk of bias in the results of non-randomized
supports or challenges their current or past clinical practice (e.g.
studies of interventions that compare the health effects of two
a clinician who gives lidocaine to patients with acute myocardial
or more interventions.18
infarction underrating a study that suggests that lidocaine may
increase mortality in these patients)11. CONCLUSION
VI. Miscellaneous Types Bias is an ever-present and insidious problem in research study
1. Technology bias: Bias which relates to judging a study design and execution, and while no study design is exempt from
according to the reader’s attraction or aversion for bias, some are more prone to particular types.The main reason

International Journal of Contemporary Medical Research 1647

ISSN (Online): 2393-915X; (Print): 2454-7379 | ICV: 50.43 | Volume 3 | Issue 6 | June 2016
Kumar, et al. Identifying and Eliminating Bias

of bias is the absence of rigorous methodology or the inability Studies (RoBANS): Development and Validation of
to assess the potential link between the cause and an effect in a New Instrument | The 19th Cochrane Colloquium
the target population. [Internet]. [cited 2016 Apr 20]. Available from:
An imperative objective in study outline is that the outcomes are http://2011.colloquium.cochrane.org/abstracts/b8o3-risk-
bias-assessment-tool-non-randomized-studies-robans-
substantial and generalizable to the larger population. Efforts to
development-and-validation-ne
implicate rigorous statistics to minimize the bias may divert the
18. A Cochrane Risk Of Bias Assessment Tool: for Non-
readers.Better an investigator anticipates the potential areas of Randomized Studies of Interventions (ACROBAT-NRSI)
bias in every phase of the trial to achieve a much valid results. | Cochrane Methods Bias [Internet]. [cited 2016 Apr
REFERENCES 18]. Available from: http://methods.cochrane.org/bias/
cochrane-risk-bias-assessment-tool-non-randomized-
1. Attia AM. Bias in RCTs: confounders, selection bias and studies-interventions-acrobat-nrsi
allocation concealment. Middle East Fertility Society
Journal. 2005;10:258.
2. Zhang D, Freemantle N, Cheng KK. Are randomized Source of Support: Nil; Conflict of Interest: None
trials conducted in China or India biased? A comparative Submitted: 22-04-2016; Published online: 21-05-2016
empirical analysis. J Clin Epidemiol. 2011;64:90–5.
3. K M Gangadharrao. Towards Imporving the Quality
of Research in India. Journal of Research in Arts and
Education. 2:47–50.
4. 4Sathyanarayana Rao TS, Tharyan P. Editorial policies
aimed at improving the transparency and validity of
published research. Indian J Psychiatry. 2011;53:183–6.
5. Bajpai V, Bajpai V. Rise of Clinical Trials Industry in India:
An Analysis, Rise of Clinical Trials Industry in India:
An Analysis. International Scholarly Research Notices,
International Scholarly Research Notices. 2013;2013,
2013:e167059.
6. Jadad AR, Enkin MW. Bias in Randomized Controlled
Trials. In: Randomized Controlled Trials [Internet].
Blackwell Publishing Ltd; 2007 [cited 2016 Apr 18]. p.
29–47.
7. Viera AJ, Bangdiwala SI. Eliminating bias in randomized
controlled trials: importance of allocation concealment and
masking. Family medicine-kansas city. 2007;39:132.
8. Schulz KF. Subverting randomization in controlled trials.
Journal of American Medical Association 1995;274:1456–
1458. - Google Search [Internet]. [cited 2016 May 18].
9. Schulz KF, Grimes DA. Allocation concealment in
randomised trials: defending against deciphering. The
Lancet. 2002;359:614–618.
10. Krishna R, Maithreyi R, Surapaneni KM. Research
bias: a review for medical students. J Clin Diagn Res.
2010;4:2320–2324.
11. Lewis SC, Warlow CP. How to spot bias and other potential
problems in randomised controlled trials. J Neurol
Neurosurg Psychiatry. 2004;75:181–7.
12. Sica GT. Bias in Research Studies1. Radiology.
2006;238:780–789.
13. Ahmed I, Sutton AJ, Riley RD. Assessment of publication
bias, selection bias, and unavailable data in meta-analyses
using individual participant data: a database survey. BMJ.
2012;344:d7762.
14. Deeks JJ, Dinnes J, D’Amico R, Sowden AJ, Sakarovitch
C, Song F, et al. Evaluating non-randomised intervention
studies. Health Technol Assess. 2003;7:1-173.
15. Grimshaw JM, Thomas RE, MacLennan G, Fraser C,
Ramsay CR, Vale L, et al. Effectiveness and efficiency of
guideline dissemination and implementation strategies.
Health Technol Assess. 2004;8:1-72.
16. Kim SY, Park JE, Lee YJ, Seo H-J, Sheen S-S, Hahn
S, et al. Testing a tool for assessing the risk of bias for
nonrandomized studies showed moderate reliability and
promising validity. J Clin Epidemiol. 2013;66:408–14.
17. B8O3 | Risk of Bias Assessment tool for Non-randomized

1648
International Journal of Contemporary Medical Research
Volume 3 | Issue 6 | June 2016 | ICV: 50.43 | ISSN (Online): 2393-915X; (Print): 2454-7379