Respondent selection in a repeated survey on lifestyle within the randomized colorectal cancer screening programme

Screening for colorectal cancer (CRC) has been shown to decrease colorectal cancer mortality in randomized-controlled trials. However, screening may have an adverse impact on an individual's lifestyle. We describe here the design of a repeated survey study on lifestyle and assess response in the survey target population by randomization to CRC screening and demographic factors. The survey study population (n=10271) included Finnish men and women born in 1951 who were randomized for the CRC screening programme in 2011 and received a questionnaire on lifestyle in 2010 and 2012. We assessed responding by randomization to CRC screening, calendar time and demographic factors using the population-averaged Poisson model. Responding to survey was overall similar in 2010 and 2012. Those invited for CRC screening increased responding in time [incidence rate ratio (IRR) 1.06, confidence interval (CI) 1.03-1.09], whereas controls decreased their responding (IRR 0.97, CI 0.94-1.00). Women were more likely to respond than men (IRR 1.17, CI 1.12-1.23). Also, secondary (IRR 1.20, CI 1.13-1.27) and tertiary (IRR 1.31, CI 1.23-1.40) level education increased the response proportion compared with primary-level education. We could reliably assess the effect of CRC screening invitation and demographic factors on survey response. Although invitation to CRC screening increased responding, the self-selection was similar on both survey rounds. Self-selection should be taken into account when generalizing results from survey studies to their target population.