Screening biases
Earlier diagnosis can lengthen measured survival without saving lives. Lead time, length bias and overdiagnosis must be separated from screening benefit.
Fit
When it matters, and when it does not
It matters when
- A report claims screening works because survival after diagnosis has risen.
- You are weighing a screening program's benefits against false positives and overdiagnosis.
- Screen-detected cases appear to have a better prognosis than cases found after symptoms.
It matters less when
- You only want to know how accurate the screening test is. Use a diagnostic accuracy study. Read: Diagnostic accuracy study
- The question is whether earlier treatment works once disease is found. Test the treatment itself in a randomized trial. Read: Randomized controlled trial
Before you start
What to check in your study
- The comparison groups and starting pointComparing people from diagnosis can build in lead time. Check whether invitation, attendance or diagnosis defines each group.
- What the reported outcome actually measuresSurvival after diagnosis is not the same as death risk from a common starting point.
- Follow-up from a shared starting pointA short follow-up can miss later benefit, harm or excess diagnoses that later disappear.
- Who attends and who is missedAttenders can differ from nonattenders before screening. Their outcomes alone do not measure screening benefit.
- False positives and overdiagnosisSeparate further tests after false alarms from treatment of disease that would never cause harm.
- A missing-data plan that starts from an assumptionDecide why deaths or follow-up may be missing before choosing a method. The assumption picks the method, not the share missing.
- The compared populations remain comparable over follow-up.How to check: Look for random invitation and analysis by assignment. Check missing outcomes and differences in other care.
- Deaths are classified consistently in both groups.How to check: Look for blinded outcome assessment and consider deaths from any cause as well as disease deaths.
- Extra diagnoses are distinguished from extra lives saved.How to check: Review death rates and harms separately from survival after diagnosis or stage shifts.
- Missing deaths and follow-up follow the assumption you named.How to check: Register links fail more often for people who move or emigrate, so missing completely at random (MCAR) is rarely plausible. Missing at random (MAR) means the gaps depend only on recorded factors, such as age, area and group. Under MAR, multiple imputation can use those factors, but it is not automatically better than analyzing everyone followed. If people who leave differ in their risk of dying of the disease, even given recorded factors, missing not at random (MNAR) is plausible. Then add a tipping-point analysis, which shifts the imputed deaths until the conclusion changes.
Plain language
How it works
Lead time moves diagnosis earlier without moving death, so survival from diagnosis grows. Length bias means screening finds more slow-growing disease. Overdiagnosis finds disease that would never cause symptoms or death. These changes can make screening look useful even when it does not help people.
Longer survival after diagnosis and a shift to earlier stages do not establish benefit. Compare outcomes from a common starting point across the whole eligible population. Use the linked randomized screening trial for an invitation-based comparison. Count false positives, further tests and treatment of overdiagnosed disease alongside benefit.
Worked example
A worked example
- Setting
- Two people have the same fatal cancer. Neither lives beyond age 70.
- Question
- Can screening make survival from diagnosis look longer without extending life?
- Approach
- Usual care diagnoses one person at age 67. Screening diagnoses the other at age 65.
- Result
- Survival from diagnosis is 3 years with usual care and 5 years with screening, a gain of 2 measured years.
- What it means
- Both die at 70. The apparent survival gain is lead time, not longer life.
In practice
Recent studies PubMed search · not vetted
Watch out
Common mistakes
The one to remember
Earlier diagnosis can increase survival from diagnosis without delaying death. Length bias favors slow-growing disease, and overdiagnosis adds disease that would never cause harm. Compare deaths from the disease, the usual endpoint, or deaths from any cause, from a common starting point in the whole invited population. All-cause mortality also avoids errors in assigning cause of death.
- Calling longer survival after diagnosis proof that screening saved lives.Instead: Measure deaths from a common starting point, and count all people assigned to each strategy.
- Treating every extra diagnosis as useful early detection.Instead: Consider length bias and overdiagnosis, and count treatment harms.
- Comparing attenders with nonattenders as if they were randomized.Instead: Use the linked randomized screening trial page for the invitation-based design.
Inside Mallard
How Mallard uses this
Plans do not link to this concept yet. A plan whose target is diagnostic accuracy, such as of a screening test, links to the diagnostic accuracy design instead.
concept.screening_biases, the id a plan cue records.How to report it
No reporting guideline dedicated to the evaluation of a screening program was found in a PubMed search on 2026-09-16. CONSORT covers a randomized screening trial and STARD covers the accuracy of the test itself, and neither answers the mortality question this design asks. Report disease-specific mortality in the whole invited group, with the overdiagnosis and lead-time work shown rather than asserted. A reporting guideline is not a design manual: use it to make the study transparent, not as a substitute for careful design. Check the target journal's own requirements as well, which this page cannot know.
Related Learn pages
Related designs and methods
Evidence
Sources & further reading
- Overdiagnosis in cancer (opens in a new tab) · J Natl Cancer Inst · doi 10.1093/jnci/djq099
- Are increasing 5-year survival rates evidence of success against cancer? (opens in a new tab) · JAMA · doi 10.1001/jama.283.22.2975
- Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls (opens in a new tab) · BMJ · doi 10.1136/bmj.b2393
- CONSORT 2025 statement: updated guideline for reporting randomised trials (opens in a new tab) · BMJ · doi 10.1136/bmj-2024-081123
- STARD 2015: an updated list of essential items for reporting diagnostic accuracy studies (opens in a new tab) · BMJ · doi 10.1136/bmj.h5527