Predictors of non-compliance with baseline sigmoidoscopy in the PLCO screening trial

A55

In observational (e.g. case-control) studies of cancer screening efficacy, the possibility of confounding is great, as those who choose to be screened often have a substantially different underlying risk of mortality from the cancer of interest than those who do not (i.e. self-selection bias). The objective of this study was to evaluate this potential bias by exploring the association between a number of established colorectal cancer (CRC) risk factors and compliance with a screening invitation, using data from the sigmoidoscopy intervention in the PLCO cancer screening trial. The study population included individuals randomized into the intervention arm of the PLCO trial. We excluded those with a self-reported history of colorectal disease (including ulcerative colitis, Crohn’s Disease, Gardner’s syndrome, familial polyposis, or colorectal polyps), because endoscopic surveillance is recommended in this group, and this could affect their compliance with study sigmoidoscopy. Similarly, we excluded those non-compliers who either died or were diagnosed with colorectal cancer within three months of randomization, since these individuals may have been involved with other more pressing healthcare issues and/or outside-of-study colorectal examinations that made sigmoidoscopy screening unwarranted. The final sample size for this analysis was 69,246, of whom 10,024 (14 percent) did not attend the T0 sigmoidoscopy. As measured by relative risks (RR) in multivariate poisson regression models, non-compliance was greater in females (RR 1.6, 95% confidence interval (CI) 1.6, 1.7), those of non-white race (RR 1.4, 95% CI 1.3, 1.4), current smokers (vs. never-smokers, RR 1.4, 95% CI 1.3,1.5), and those who were underweight (vs. normal weight, RR 1.3, 95% CI 1.1, 1.5). There was also a trend towards increasing non-compliance with age (RR 70+ vs. <60 years 1.2, 95% CI 1.2, 1.3). Non-compliance was inversely associated with education (RR college graduate vs. less than high school 0.7, 95% CI 0.7, 0.8), fruit and vegetable consumption (RR 10+ vs. <2 servings/day 0.5, 95% CI 0.4, 0.7), and vigorous physical activity (RR 4+ vs. 0 hours/week 0.7, 95% CI 0.6, 0.7), and with post-menopausal hormone use among women (RR 0.7, 95% CI 0.7, 0.8). Weaker or null associations with non-compliance were observed in those with a family history of CRC, NSAID users, obese individuals, and alcohol users. These associations were not modified by colorectal screening test use prior to entering the trial. Because those who participate in a randomized trial of cancer screening are likely on average to be more health-conscious than the population as a whole, these observed associations may represent a conservative estimate of the strengths of association as compared to the general population. Despite this, we observed a number of demographic and lifestyle factors that were associated with compliance with an invitation for sigmoidoscopy screening. Furthermore, factors associated with non-compliance tended to be those that are associated with an increased risk of colorectal cancer; therefore, these results suggest that failing to account for these factors in observational studies of screening efficacy would tend to result in an overestimate of the benefit of sigmoidoscopy.