Recruitment bias in studies of semen and other factors affecting pregnancy rates in fertile men.

BACKGROUND Recruitment bias is possible in population studies of semen quality because few men volunteer. We examine differences between Australian couples with natural conceptions who agreed or declined to participate in such a study. METHODS Women pregnant between 16 and 32 weeks gestation participating in a retrospective time to pregnancy (TTP) study were each requested to recruit their eligible (on the basis of age, place of his birth and of his mother's birth) male partner to complete additional questionnaires, have a physical examination and provide blood and two semen samples. RESULTS From 2061 women who completed the TTP questionnaire (response rate, 98%) there were 928 eligible male partners of whom 225 (24%) were responders. There were significant socio-demographic and self-reported exposure differences between responders and non-responders in particular, female professional occupation, knowledge of the fertile phase, pelvic inflammatory disease, non-smoker at time of conception and wine consumption per week were more frequent in the responders. There was no evidence of a bias for the subfertile being more likely to volunteer for the study. Mean TTP for planned pregnancies for responders and non-responders were 3.3 and 3.8 cycles (P = 0.319), respectively, and the cycle specific pregnancy rates were not significantly different after covariate adjustment by Cox regression. CONCLUSIONS The present study confirms that participation rates are low in studies of semen quality. Although the expected higher participation of subfertile couples was not confirmed, there remains considerable potential for bias and other problems that could invalidate this type of study.