Misleading and avoidable: design-induced biases in observational studies evaluating cancer screening-the example of...

Observational studies evaluating the effectiveness of cancer screening are often biased due to nonalignment at time zero, which can be avoided by target trial emulation (TTE). We aimed to illustrate this by evaluating site-specific effectiveness of screening colonoscopy regarding colorectal cancer (CRC) incidence. Based on a German health care database, we assessed the effect of screening colonoscopy vs no screening colonoscopy in preventing CRC in the distal and the proximal colon over 12 years of follow-up in 55-69-year-old persons. We compared four different study designs: cohort and case-control study, each with/without alignment at time zero. In both analyses with time zero-alignment, screening colonoscopy showed a rather similar effectiveness in reducing the incidence of distal and proximal CRC (cohort analysis: 32% (95% CI, 27%-37%) vs 28% (20%-35%); case-control analysis: 27% vs 33%). Both analyses without alignment suggested a difference by site: Incidence reduction regarding distal and proximal CRC, respectively, was 65% (61%-68%) vs 37% (31%-43%) in the cohort analysis and 77% (67%-84%) vs 46% (25%-61%) in the case-control analysis. Violations of basic design principles can substantially bias the results of observational studies. In our example, it falsely suggested a much stronger preventive effect of colonoscopy in the distal vs the proximal colon. Our study illustrates that TTE avoids such design-induced biases.