Explain in detail about bias that may occur in randomised control trial and how to prevent it from latest edition of park

Reading File

Finding Sources

Reading File

I now have all the key content from Park's textbook. Let me compile a comprehensive, detailed answer based directly on what Park's Textbook of Preventive and Social Medicine states.

Bias in Randomised Controlled Trials (RCTs) - Park's Textbook of Preventive and Social Medicine

The following is a detailed account of bias in RCTs as described in Park's Textbook of Preventive and Social Medicine (the latest edition available in the library).

What is Bias in an RCT?

Bias refers to any systematic error that distorts the true relationship between an intervention and its outcome. Park defines it as a departure from the truth that leads to incorrect conclusions about the effect of an intervention. In RCTs, bias can arise at multiple stages - from selection of participants, through randomization, follow-up, and final assessment.

1. Selection Bias

Nature

Selection bias occurs when "like is not compared with like." If the study group and control group differ in some important characteristic at the outset, any observed difference in outcome may be due to that pre-existing difference rather than the intervention itself.

Classic Example from Park

A UK study of 5,174 home births and 11,156 hospital births found perinatal mortality of 5.4/1000 in home births vs. 27.8/1000 in hospital births. This apparently suggested hospitals were more dangerous. But this was a spurious (artificial) association - hospitals attract high-risk women with complications, so the higher mortality was due to that selection bias, not inferior care. The two groups were not comparable at baseline.

How Randomization Prevents It

Randomization is the single most important tool against selection bias. Park calls it the "heart" of a controlled trial.
It ensures the investigator has no control over who goes into which group, meaning every individual has an equal chance of being allocated to either the study or control group.
Randomization distributes both known and unknown confounding variables equally between groups - matching can only control for factors already known to be important, whereas randomization handles even unrecognized factors.
Stratified randomization: When a particular variable (e.g., age, sex) is known to affect outcome, the entire study population is stratified into sub-groups by that variable first, then randomization is carried out within each sub-group. This ensures balanced distribution of that variable across groups.
After randomization, it is always desirable to check that the two groups are basically similar in composition.

2. Bias Due to Confounding

Nature

Confounding occurs when a third variable (the confounding variable) is related both to the exposure and the outcome, thereby creating a misleading association. Park gives the example of altitude and endemic goitre - both are related through iodine deficiency, the true confounding factor.

How to Prevent It

Randomization (as above) distributes confounders equally.
Matching - used in analytical (non-experimental) studies when randomization is not possible. But matching can only control for known confounders.
Statistical adjustment during analysis.

3. Observer Bias (Ascertainment Bias)

Nature

Observer bias occurs when the investigator measuring the outcome of a trial is influenced by knowing which procedure or therapy a patient received. Knowing the group allocation may, even subconsciously, affect how the investigator interprets measurements, physical signs, or outcomes. Park specifically terms this "observer bias."

Prevention - Blinding

Randomization alone cannot guard against observer bias. The solution is blinding:

(a) Single Blind Trial The participant does not know whether they belong to the study group or control group. This prevents participant-driven bias but does not prevent the investigator from being influenced.

(b) Double Blind Trial Neither the doctor (investigator) nor the participant is aware of group allocation or treatment received. This is the most frequently used method of blinding. It prevents both participant and observer bias simultaneously.

(c) Triple Blind Trial The participant, the investigator, AND the person analyzing the data are all blind to group allocation. This is the ideal approach, as it prevents bias at every level - data collection, measurement, and analysis. However, it is less commonly applied in practice.

Park notes: "Ideally, of course, triple blinding should be used; but double blinding is the most frequently used method when a blind trial is conducted."

When death is the outcome being measured, blinding is not considered essential, as the outcome is objective and not subject to interpretation.

4. Subject Variation / Participant Bias (Placebo Effect)

Nature

Participants may subjectively feel better or report improvement simply because they know they are receiving a new form of treatment - regardless of whether the treatment is actually effective. Park calls this "subject variation." It is essentially the placebo effect influencing self-reported outcomes.

Prevention

Single or double blinding - participants are kept unaware of which group they belong to.
Use of a placebo in the control group - the control group receives an identical-looking but inactive substance so that the psychological effect of receiving treatment is equal in both groups.

5. Evaluation Bias (Reporting Bias)

Nature

Even when objective measurements are made, the investigator may subconsciously give a favourable report of the trial's outcome - particularly if they designed the study or have a stake in its success. This is a form of bias in the final evaluation and write-up.

Prevention

Triple blind design (person analyzing the data is also blinded).
Pre-registration of trial outcomes before the study begins.
Independent data safety monitoring boards.

6. Attrition Bias (Losses to Follow-Up)

Nature

During the follow-up phase, some participants inevitably drop out - due to death, migration, or loss of interest. Park calls this "attrition." If attrition is substantial and is not random (i.e., if the people who drop out differ systematically from those who remain), the results will be biased and it may be difficult to generalize findings to the reference population.

Prevention

Choose a stable population whose cooperation is assured from the outset.
Make every effort to minimize losses to follow-up.
Analyze using intention-to-treat analysis (all subjects analyzed in their originally assigned group regardless of what actually happened).

7. Volunteer Bias / Participation Bias

Nature

People who agree to participate in a study are likely to differ from those who decline, in ways that may affect the outcome. Volunteers may be more health-conscious, more educated, or more motivated than non-volunteers, which can make the study population unrepresentative of the reference population.

Prevention

Ensure participants are representative of the reference population.
Use random sampling to select the study population from the reference population.

Summary Table: Types of Bias in RCT and Their Prevention

Type of Bias	Stage	Cause	Prevention
Selection bias	Allocation	Non-random group assignment	Randomization (simple/stratified)
Confounding bias	Design/Analysis	Third variable affecting both groups	Randomization; matching; statistical adjustment
Observer bias	Assessment	Investigator knows group allocation	Double/triple blinding
Subject variation (placebo effect)	Follow-up	Participant knows group allocation	Single/double blinding; use of placebo
Evaluation bias	Analysis	Investigator subconsciously favors outcome	Triple blinding; independent analysis
Attrition bias	Follow-up	Unequal/systematic dropout	Stable population; minimize losses; intention-to-treat
Volunteer bias	Enrolment	Participants differ from non-participants	Random sampling; representative population

Key Principle from Park

"Randomization is an attempt to eliminate bias and allow for comparability... By a process of randomization, hopefully, these factors will be distributed equally between the two groups. Randomization is the 'heart' of a controlled trial."

And:

"Randomization cannot guard against these sorts of bias [observer bias, subject variation, evaluation bias], nor the size of the sample. In order to reduce these problems, a technique known as 'blinding' is adopted, which will ensure that the outcome is assessed objectively."

Source: Park's Textbook of Preventive and Social Medicine (latest edition in library), Chapter on Experimental Epidemiology - Sections on Randomization (pp. 94-95), Assessment (pp. 95-96), and Spurious/Indirect Association (pp. 99-100).

This is a shared conversation. Sign in to Orris to start your own chat.