Survival and Lead-time Bias - Mortality Correction?

[bonemann]

There's a Uworld question that tests understanding of lead-time bias in screening.
The stem describes a new diagnostic marker that appears to increase survival by 3 months.
Then a subsequent mortality at 6 months shows no effect.

I get why its lead-time bias, and understand that mortality rate would show if there is an actual change in rate of dying and thus effectiveness of the screen.

However, the explanation says that to test for lead-time bias, a mortality rate should be tested that's longer than the perceived survival benefit, ie if the survival benefit is 3 months one should look at mortality rate for more than 3 months.
what I don't get is why its necessary to follow patients for longer than the apparent survival benefit time in order to check the actual benefit, if any.

---

Members don't see this ad.

 

[sahell]

This intuitively makes sense to me so my explanation may not be the best. The apparent mortality benefit is because of earlier diagnosis, so we would want to see longer follow up to check if it has any survival benefit compared to the control group. A diagram may help:

new test: Onset--------------(3mo)+++++++++++++++(6mo)+++++++++++++++Dead(9mo)
old test: Onset---------------(3mo)--------------------------(6mo)+++++++++++++++Dead(9mo)

+ for after diagnosis.
If we compare the mortality 3 months after diagnosis, the new test seems to have benefit. If you can make out from the crude illustration, there is no actual survival benefit.