Neyman's bias re-visited

Abstract

In case-control studies using prevalent cases, an apparent association may be spurious if the risk factor affects survival. In his description of this potential bias, Neyman disregarded competing risks. We use a compartment model to illustrate Neyman's bias and show that it can explain the apparent association only if the risk factor influences mortality from the disease being studied. Any effect of the risk factor on mortality from other causes is not relevant. The apparent protective effect of smoking in relation to Alzheimer disease is used as an example, but the result holds for any irreversible noncommunicable disease and for any dichotomous risk factor.