The seminar is cancelled due to cancelled flights at Brussels airport.
Abstract
Screening behavior depends on factors like previous screening history and family members’ behaviors, which can act as both confounders and intermediate variables on a causal pathway from screening to disease risk. Conventional analyses that adjust for these variables can therefore lead to incorrect inferences about the causal effect of screening: a positive screen might increase the subsequent frequency of screening, making it appear positively associated with cancer, even though the effect is really beneficial. Inverse propensity score weighting provides a way to analyze observational data as if it were a sequence of trials in which screening is applied at random rather than self-selected. To assess this approach, I simulated family data under plausible models for the underlying disease process and for screening behavior. The goal was to assess the performance of alternative methods of analysis, as well as to assess whether a targeted screening approach based on individual’s risk factors like a genetic risk index, would lead to a greater reduction in cancer incidence in the population than a uniform screening policy. Simulation results indicate that there can be a substantial positive bias in the estimated effect of screening on subsequent cancer risk when using conventional analysis approaches, and that this bias is eliminated by using propensity score weighting. In contrast, standard analyses of data on a large case-control study of colonoscopy and colorectal cancer from Germany show a very substantial protective effect of screening, but inverse propensity score weighting makes this effect even stronger. This is due in part to some strong determinants of screening propensity that are also risk factors for cancer. Targeted screening approaches based on either fixed risk factors or family history yield somewhat larger reductions in cancer incidence and fewer screens needed to prevent one cancer than population-wide approaches, but the differences may not be enough to justify the additional effort required.