The selective uptake of screening by healthy participants and its impact on the evaluation of screening effectiveness in non-randomized studies have been discussed, but hardly studied. We quantified excess mortality among cervical screening non-participants compared to participants. Based on Danish healthcare registers, we determined women's participation in cervical screening in 1990-1993 (one screening round) and 1990-1997 (two screening rounds). Women were followed until end of 2010. We computed hazard ratios (HR) comparing non-participants' and participants' risk of death, and analyzed the impact of age, calendar period of screening evaluation, screening intensity, length of follow-up and cause of death. After one screening round, the 17-year HR of death in non-participants was 1.61 (95% CI: 1.59-1.63), with an increasing trend over calendar time. After two rounds, regular non-participants had a HR of 2.09 (95% CI: 2.05-2.14) compared to regular participants. The HR for human papillomavirus (HPV)-related cancers other than cervical cancer was 3.80 (95% CI: 2.67-5.41). Younger women, whose coverage rates were higher, had higher all-cause mortality HRs. Women screened more frequently than recommended had the same mortality as those screened as recommended. Acute illness did not seem to be a major reason for non-participation, as the excess risk of death was not higher in the first years following screening evaluation. Non-participants in cervical screening had substantially higher all-cause mortality than participants, and a particularly increased risk of HPV-related causes of death. These results indicate that improper control for the selective uptake of cervical screening may result in overestimating its effectiveness.
Keywords: cervical cancer; cohort study; mortality; participation; screening; selection bias.
© 2013 UICC.