Population fraction of cervical neoplasia attributable to high-risk human papillomaviruses

Abstract

The efficacy of human papillomavirus (HPV) type 16 and 18 vaccines against cervical intraepithelial neoplasia grade II (CIN2+) has been verified, but the active follow-up of studies with invasive cervical cancer or cervical intraepithelial neoplasia grade III (CIN3) as primary end points are ethically not possible. Furthermore, ongoing registry-based passive follow-up studies with invasive cervical cancer as the end point will take time. To evaluate the feasibility of CIN3 as a surrogate end point, we compared high-risk (hr) HPV-associated relative risk and population attributable fraction (PAF) of CIN3 and/or squamous cell carcinoma (SCC) estimated in a large serological case-cohort HPV study. Our case-cohort comprised 83 SCC and 389 CIN3 cases and a subcohort of 7862 out of 230,998 Finnish women, who at baseline were under 32 years of age and had undergone a minimum of two pregnancies within 5 years during 1983-1997. PAFs of the case-cohort, approach-based, serologically defined and misclassification-corrected HPV16 and hrHPV (HPV types 16, 18, 31 and 33) exposures in the SCC samples were 61% (95% CI: 18-85) and 73% (95% CI: 13-93), respectively. Considerably lower HPV16 and hrHPV PAF estimates in CIN3 of 6% (95% CI: -19-35) and 36% (95% CI: -5-65), respectively, were obtained. A meta-analysis-derived, PCR-based, hrHPV-associated relative risk estimate of 20.3 in CIN2/3+ yielded a PAF estimate for hrHPV in CIN2/3+ of 86% (90% CI: 63-95) in our study population. The former, hrHPV serology-based CIN3 PAF estimates were biased owing to low sensitivity of HPV16 and/or HPV16/18/31/33 serology, most notably in cervical cancer precursor lesions, but the latter estimate overlapped with our hrHPV serology-based cervical cancer PAF estimate. CIN3 may be a valid surrogate efficacy end point for HPV vaccination studies, but the associated causality of multiple hrHPV exposures needs to be unambigously defined.