Non-response bias in a surveillance program for asbestos-related lung cancer

Abstract

Objectives: In a cohort study non-response might lead to a biased selection of cohort members and may affect the validity and reliability of the study outcome. To detect the possible effects of a non-response bias on study results, we evaluated the reasons for non-participation and the differences of respondents and non-respondents in a health surveillance program for power industry workers, formerly exposed to asbestos. Methods: A cohort of former power plant workers was formed to participate in an early detection program for lung cancer. We evaluated the results of 1,019 individuals (mean age 66 yr), of which 839 took part in at least one examination, 180 refused to participate or did not respond. To obtain the reasons for non-response, we interviewed the cohort members by telephone or we requested them by mail to complete and return a brief questionnaire. Further sources of information were the communal registration offices and local health offices. Results: The main reasons for non-participation were refusal (35%), illness (23.3%), death (16.7%) and difficulties with traveling (13.3%). It was impossible to make contact with or obtain an explanation from 11.7%. In a logistic regression model we demonstrated that advanced age and a long travel distance from the study center negatively affected the participation rate (p<0.001). There was no difference between respondents and non-respondents regarding prevalence (p=0.559) and incidence of lung cancer (p=0.882). Conclusion: We concluded that in our cohort non-participation did not cause a selection bias in terms of lung cancer rates.