Lead-time bias: why earlier diagnosis can fake longer survival

Here is a statistic that sounds like great news and often is not: people whose cancer is caught by screening survive longer after diagnosis than people whose cancer is caught later. It seems obvious that earlier is better. But part of that longer survival can be an illusion baked into how the clock is started, and it has a name: lead-time bias.

Picture someone whose cancer will kill them in 2030 no matter what. If a scan finds it in 2020, their survival from diagnosis is ten years. If instead it is found when symptoms appear in 2027, their survival from diagnosis is three years. Same person, same death date, same treatment outcome, but screening made the survival number more than triple. Nothing about their life was extended. The diagnosis just moved earlier.

This is why survival since diagnosis is a treacherous way to judge whether screening works. Earlier detection lengthens that number automatically, whether or not the screening actually postpones death. It is one of the main reasons a screening programme can look wonderful in the raw survival statistics and still not help people live longer.

The fix is to ask the right question. The measure that cannot be faked by lead time is whether screening lowers the death rate in the whole population, comparing people invited to screen against people who were not, ideally in a randomized trial. When you read that a screening test improves survival, check whether that means fewer people are dying, or just that the diagnosis clock started earlier. They are very different claims.