QUICK-REFERENCE BOX — The Traps
- Bias = systematic (non-random) error baked into how the study was done — more data won't fix it.
- Confounding = a third factor linked to both exposure and outcome, mixing up the association.
- Randomization is the only tool that controls unknown confounders.
- Lead-time bias and length-time bias make screening look better than it is.
- Intention-to-treat analysis preserves randomization against attrition bias.
1. Overview
The Core Idea
Bias is systematic error that distorts a result in a particular direction; unlike random error, a bigger sample doesn't fix it. Confounding is a specific problem where a third variable muddles the exposure–outcome link. Screening programs carry their own biases and must meet established criteria to be worthwhile.
2. Types of Bias
| Bias | What happens | Guard |
|---|
| Selection bias | Systematic differences in who is included/retained | Careful sampling; minimize loss to follow-up |
| Information/measurement bias | Errors in classifying exposure or outcome | Blinding; objective, standardized measures |
| Recall bias | Cases remember exposures differently (case-control) | Objective exposure records |
| Observer bias | Assessor's expectations color measurement | Blinded outcome assessment |
| Attrition bias | Differential dropout between groups | Intention-to-treat analysis |
3. Confounding
The Classic Setup
A confounder is associated with both the exposure and the outcome and is not on the causal pathway between them. Example: coffee drinking appears linked to lung cancer, but smoking (linked to both coffee and cancer) is the confounder. Unaddressed, confounding creates a spurious or distorted association.
4. Controlling Confounding
Design & Analysis Tools
- Randomization (design) — the only method that also balances unknown confounders.
- Restriction and matching (design) — limit or pair on the confounder.
- Stratification and multivariable adjustment (analysis) — account for measured confounders.
5. Screening Biases
Why Screening Can Look Falsely Good
- Lead-time bias: earlier diagnosis lengthens apparent survival without changing the true course.
- Length-time bias: screening preferentially catches slow, indolent disease, overstating benefit.
- Overdiagnosis: detecting disease that would never have caused harm.
6. Wilson-Jungner Screening Criteria
Is Screening Worthwhile?
- The condition is an important health problem with a recognizable early/latent stage.
- There is a suitable, acceptable, and accurate test.
- There is an effective, accepted treatment that works better when started early.
- Diagnosis and treatment facilities are available, and screening is cost-effective and continuous.
7. Key Pearls
High-Value Points
- Bias is systematic and not fixed by sample size; random error is.
- Only randomization controls unknown confounders.
- Lead-time and length-time bias flatter screening programs.
- Blinding combats information/observer bias; ITT combats attrition bias.
- Wilson-Jungner criteria decide whether screening is justified.
8. Common Mistakes to Avoid
Misreadings & Better Practice
Frequent errors with bias and screening.
| Mistake | Why it's wrong | Better reading |
|---|
| "A bigger sample fixes the bias." | Bias is systematic. | Only better design fixes bias. |
| Crediting screening for longer survival alone. | Lead-time bias. | Look for reduced mortality, not just survival time. |
| Adjusting only for measured confounders in observational data. | Unknown confounders remain. | Prefer randomized evidence for causal claims. |
9. Board-Style High-Yield Summary
Key Takeaways
- Bias = systematic error (selection, information/recall/observer, attrition); not fixed by sample size.
- Confounding = third variable linked to exposure and outcome; controlled by randomization, restriction, matching, stratification, adjustment.
- Randomization uniquely handles unknown confounders.
- Lead-time and length-time bias inflate apparent screening benefit.
- Wilson-Jungner criteria justify a screening program; ITT preserves randomization.
10. References
- 1.Wilson JMG, Jungner G. Principles and practice of screening for disease. WHO. 1968.
- 2.Sackett DL. Bias in analytic research. J Chronic Dis. 1979.
- 3.Grimes DA, Schulz KF. Bias and causal associations in observational research. Lancet. 2002.
Back to Clinical Guideline Hubs