Assessing generalizability of the findings of sigmoidoscopy screening trials: the case of SCORE trial.

Abstract

Several reports indicated that volunteers enrolled in preventive trials tend to show a different profile, with respect to sociodemographic characteristics, health-related behaviors, or medical history, compared with the source population. We conducted an incidence and mortality follow-up within a cohort of subjects who had been mailed a recruitment questionnaire in the SCORE trial of sigmoidoscopy (FS) screening for colorectal cancer (CRC) to assess the impact of self-selection in the study of volunteers willing to be screened on the outcomes estimates and on the generalizability of the results. We compared baseline demographics, CRC risk, and overall mortality at 11-year follow-up of responders declaring their interest in screening, with those of nonresponders and of responders not interested in screening using logistic regression and Cox proportional hazards multivariable models. Both subjects who volunteered in the trial and those who refused were better educated than nonresponders. Men and people younger than age 60 years were more likely to volunteer among responders. At 11-year follow-up, interested responders showed a similar CRC risk as nonresponders, while CRC mortality was substantially reduced (hazard ratio [HR] = 0.70, 95% confidence interval [CI] = 0.54 to 0.91). All-cause mortality was reduced both among interested (HR = 0.61, 95% CI = 0.57 to 0.65) and uninterested responders (HR = 0.81, 95% CI = 0.76 to 0.86). The implementation of an FS population-based screening program would result in a similar reduction in CRC incidence, as observed in the SCORE trial, and likely in a larger impact on CRC mortality.