Questions: Screening Program Evaluation and Implementation

Question 1

A new cancer screening program reports that screened patients survive an average of 8 years after diagnosis, while unscreened patients survive only 4 years after diagnosis. A critic argues this comparison is misleading. What is the most fundamental problem with this comparison?

• A. The screened group is likely wealthier and has access to better supportive care
• B. Screening moves the diagnosis earlier in time, so survival measured from diagnosis will appear longer even if the actual date of death is unchanged
• C. Four-year survival is too short a follow-up period to assess cancer outcomes
• D. The comparison is only valid if both groups received the same treatment

Answer: (B)

This is lead time bias. If a cancer takes 10 years from detectable onset to death, screening at year 0 gives an apparent 10-year survival; clinical diagnosis at year 4 gives 6-year survival — yet both patients die at the same calendar date. The 4-year 'advantage' is entirely the lead time. The correct way to assess benefit is cause-specific mortality in randomized trials comparing screened vs. unscreened populations, not survival from diagnosis within the screened group.

Question 2

A prostate cancer screening trial finds that screened men have significantly better outcomes than unscreened men. Critics note that the screened population is enriched with slow-growing, indolent tumors. This represents which threat to validity?

• A. Lead time bias, because slow-growing tumors are detected earlier in their natural history
• B. Overdiagnosis bias, because any detected tumor is counted as a screening success
• C. Length bias, because slow-growing tumors spend more time in the detectable-but-asymptomatic window and are disproportionately captured by periodic screening
• D. Selection bias, because men who accept screening may be healthier overall

Answer: (C)

Length bias arises from the mechanics of periodic screening: a tumor that grows slowly spends more time in the 'screen-detectable but not yet symptomatic' phase, making it far more likely to be caught by any given screen. Fast-growing aggressive tumors progress to symptoms (or death) between screens. The result is that a screened cohort is systematically enriched with indolent disease, making outcomes look better regardless of any benefit from early detection.

Question 3

If a randomized trial shows that cause-specific mortality is identical in screened and unscreened groups, this proves that the screening test has poor sensitivity.

• T. True
• F. False

Answer: False

Cause-specific mortality being identical means screening did not reduce deaths from the disease, but there are multiple explanations: lead time bias (earlier detection but no change in outcome), length bias (enriching for indolent disease), overdiagnosis (detecting disease that would never have killed), or ineffective treatment of screen-detected cases. Poor test sensitivity is only one of many possible explanations, and the trial result doesn't distinguish between them.

Question 4

A rigorous evaluation of a screening program requires randomized trial evidence comparing cause-specific mortality in screened versus unscreened populations, rather than comparing survival from diagnosis within the screened group alone.

• T. True
• F. False

Answer: True

Comparing survival from diagnosis within the screened group is confounded by both lead time bias (earlier detection inflates apparent survival time) and length bias (screened populations are enriched with less aggressive disease). Only a randomized trial that tracks cause-specific mortality across groups — assigning screening randomly and following both groups long enough — eliminates these biases, because both groups have the same underlying disease spectrum and the calendar date of death is the outcome.

Question 5

Explain the difference between lead time bias and length bias. Why can both make a screening program appear effective even when it provides no real benefit?

Model answer: Lead time bias occurs because screening detects disease earlier, so survival measured from diagnosis is longer even if the patient dies at the same calendar date — the 'survival benefit' is just the lead time, not extended life. Length bias occurs because slow-growing tumors spend more time in the detectable-but-asymptomatic window, so periodic screening preferentially catches indolent cases; the screened cohort has better outcomes because it contains less aggressive disease, not because screening helped. Both can produce apparent benefit in observational comparisons without any actual reduction in mortality.

The key distinction: lead time bias is about *when* diagnosis occurs relative to a fixed death date; length bias is about *which cases* get detected. A rigorous trial with randomized allocation and cause-specific mortality outcome is needed to control for both simultaneously.