length-time bias

**Definition:** A selection-type bias in screening studies. Screening tests preferentially detect diseases with a long detectable preclinical phase (slow progression). Fast-growing, aggressive tumors have a short preclinical window, so they are more likely to become symptomatic in the interval between screens (interval cancers) and escape detection.

**Consequence:** The screen-detected cohort is enriched for indolent disease with inherently better prognosis. This makes screening appear to improve survival or outcomes even if it doesn't change the natural history. It can also contribute to overdiagnosis (detecting disease that would never have caused symptoms or death), e.g., low-grade DCIS, low-risk prostate cancer, and some thyroid cancers.

**Mechanism/example:** Imagine two tumors: one slow-growing (detectable for 5 years before symptoms) and one fast-growing (detectable for 6 months). An annual or biennial screen has a much higher chance of catching the slow one. Screen-detected cases therefore have better survival, which reflects tumor biology rather than a benefit of screening.

**Lead-time bias vs length-time bias (high-yield comparison):** - Lead-time bias: screening detects disease earlier, so survival time from diagnosis looks longer, but the time of death is unchanged. The cause is earlier diagnosis. Survival from diagnosis is inflated. - Length-time bias: screening preferentially detects slower-growing disease. The cause is differential detection by disease speed. Prognosis of the detected group is inflated. - Both make survival look better without proving mortality benefit.

**How to avoid or address both biases:** - Use disease-specific mortality (or all-cause mortality) as the endpoint rather than survival time or 5-year survival. - Use randomized controlled trials comparing screened vs unscreened populations with analysis by intention to screen (all participants in the screening arm analyzed, whether or not tumors were screen-detected). - Compare mortality rates in the whole population rather than survival among diagnosed cases. - Lead-time bias can also be addressed by adjusting survival for the lead time, but this is difficult.

**Related concepts:** - Overdiagnosis: detection of "pseudo-disease" that would never become clinically relevant (an extreme form of length-time bias). - Interval cancers: aggressive cancers presenting between screens. - Selection bias: length-time bias is a subtype in the screening context. - Healthy volunteer (healthy screenee) bias: another screening-related bias, because those who participate in screening tend to be healthier and more health-conscious.

**Exam tip:** If a stem describes a screening test that seems to improve survival because it picks up slowly progressive cases, answer length-time bias. If it describes earlier diagnosis with the same date of death, answer lead-time bias.