Validity of Case-Control Studies to Estimate Benefits from Cancer Screening

Background: Observational studies are increasingly used to estimate the effectiveness of cancer screening. However, the validity of such studies remains uncertain. Methods: 98,678 individuals were included in the Norwegian Colorectal Cancer Prevention trial, a randomized sigmoidoscopy screening trial. We used the trial data and designed case-control studies nested among individuals randomized to sigmoidoscopy and compared the results with those from intention-to-treat and per-protocol analyses to test the validity of the case-control design to estimate screening effect on colorectal cancer incidence and mortality, and all-cause mortality. We analyzed the case-control studies with the following methodologies: The logistic regression method, where we adjusted for self-selection in multivariable regression models including age, sex, comorbidity, socioeconomics and ethnicity; and the self-selection method where we adjusted for self-selection by different unscreened groups: regional - and national reference population. The trial is registered with ClinicalTrials.gov, NCT00119912. Findings: In intention-to-treat analysis of the randomized trial, sigmoidoscopy screening reduced colorectal cancer mortality by 19% (relative risk (RR) 0.81; 95% CI 0.67-0.99). When the same data were analyzed with a case-control design, the reductions varied between 44% (RR 0.56; 95% CI 0.33-0.97) and 67% (odds ratio (OR) 0.33; 95% CI 0.22-0.51) depending on the method used. While the intention-to-treat analysis showed no screening benefit on all-cause mortality (RR 1.00; 95% CI 0.97-1.03), the case-control studies suggested a benefit between 22% (RR 0.78; 95% CI 0.71-0.85) and 57% (OR 0.42; 95% CI 0.39-0.45) depending on the method used. Interpretation: Observational studies may overestimate the benefits of cancer screening and should therefore not guide screening policies. Clinical Trials: ClinicalTrials.gov, NCT00119912 Funding: The study received intramural funding from the University of Oslo and Oslo University Hospital. No external funding was provided. Declaration of Interest: None to declare. Ethical Approval: The NORCCAP trial was approved by the Ethics Committee of South-East Norway (project ID S-98052) and the Norwegian Data Protection Authority (Project ID 98/1408). The current study was approved by the NORCCAP steering committee.