Lead-time bias: why earlier diagnosis can look like longer survival

• The idea
• Related traps
• What a fair evaluation looks like
• The takeaway

Screening programs are often judged by survival: “patients whose cancer was found by screening lived longer after diagnosis.” That comparison has a built-in trap called lead-time bias.

§The idea

Survival is usually measured from the date of diagnosis. Screening moves the date of diagnosis earlier. If it does nothing else, survival from diagnosis still gets longer, simply because the clock started sooner.

Picture someone whose disease would cause symptoms at age 67 and death at age 70. Without screening, they are diagnosed at 67 and survive 3 years. With screening, the disease is found at 64. They still die at 70, but now their “survival after diagnosis” is 6 years. Survival doubled; the person gained nothing.

§Related traps

• Length bias. Screening at fixed intervals is more likely to catch slow-growing disease, which has a better outlook anyway. Fast-growing cases tend to appear between screens. Screen-detected cases therefore look healthier for reasons that have nothing to do with the screening itself.
• Overdiagnosis. Some screen-detected abnormalities would never have caused symptoms. Counting these people as “survivors” makes outcomes look better while exposing them to treatment they didn’t need.

§What a fair evaluation looks like

The usual answer is a randomized trial that compares mortality in the whole population invited to screening with mortality in a similar population that was not invited. Deaths per 1,000 people over a fixed period don’t depend on when diagnosis happened, so lead time can’t inflate them.

Useful questions when reading a claim about screening:

1. Is the outcome survival from diagnosis, or deaths in the whole invited population?
2. Did the number of diagnoses go up a lot? That can be a sign of overdiagnosis.
3. Were the screened and unscreened groups comparable before screening, for example in health habits and access to care?

§The takeaway

Five-year survival after diagnosis can rise even when nobody lives a day longer. To judge screening, look at mortality in the whole population, not at survival from diagnosis.

More research explainers: https://www.crushsci.com