Estimating the direct effect of human papillomavirus vaccination on the lifetime risk of screen‐detected cervical precancer

Abstract Birth cohorts vaccinated against human papillomavirus (HPV) are now entering cervical cancer screening. Assessment of (pre)cancer (CIN3+) risk is needed to assess the residual screening need in vaccinated women. We estimated the lifetime (screen‐detected) CIN3+ risk under five‐yearly primary HPV screening between age 30 and 60, using HPV genotyping and histology data of 21,287 women participating in a screening trial with two HPV‐based screening rounds, 5 years apart. The maximum follow‐up after an HPV‐positive test was 9 years. We re‐estimated the CIN3+ risk after projecting direct vaccine efficacy for the bivalent and the nonavalent HPV vaccines, assuming life‐long protection. The lifetime CIN3+ risk was 4.1% (95% confidence interval 3.5‐4.9) and declined by 53.5% and 70.5% after bivalent vaccination without and with cross‐protection, respectively, translating into a residual lifetime CIN3+ risk of 1.9% (1.4‐2.4) and 1.2% (0.9‐1.5). The CIN3+ risk declined by 88.5% after nonavalent vaccination, translating into a residual lifetime CIN3+ risk of 0.5% (0.2‐0.7). The latter risk increased to 1.6% when vaccine protection only lasted until the first screening round at age 30. Among HPV‐positive women with abnormal adjunct cytology, the nine‐year CIN3+ risk was 16.9% (8.7‐32.4) after nonavalent vaccination. In conclusion, HPV vaccination will lead to a strong decline in the lifetime CIN3+ risk and the remaining absolute CIN3+ risk will be very low. Primary HPV testing combined with adjunct cytology at five‐year intervals still seems feasible even after nonavalent vaccination, although unlikely to be cost‐effective. Our results support a de‐intensification of screening programs in settings with high vaccination coverage.


Mathematical formulas for estimating risks
Let be the set of all women with an HPV infection detected in screening round i (i=1,…,7).
Note that HPV infections in round 2,…,7 are incident infections preceded by a negative HPV test result in the previous round. For the m-th woman in the data, let be the set of all HPV types detected in the cervical smear. is equal to the empty set ∅ if the m-th woman is negative for all high-risk HPV types. The event that any type is detected in the m-th woman can be denoted by ≠∅ using as indicator function for any event . For the -th HPV type ( = 1, … , 14) within the set [HPV16, HPV18, HPV31, HPV33, HPV35, HPV39, HPV45, HPV51, HPV52, HPV56, HPV58, HPV59, HPV66, HPV68], is the HPV type-specific risk of developing CIN3+ within 9 years.
By Assumptions III and IV in the Methods section, the risk of CIN3+ in HPV-positive women in Round 1 Among women with an HPV infection detected in Rounds 2 to 7, the CIN3+ risk is equal to Risks 1 and 1 are defined in a similar way as 1 , and risks 2 and 2 are defined in a similar way as 2 . More specifically, for 1 and 2 the type-specific CIN3+ risk in 1 and 2 is replaced by the typespecific risk of both CIN3+ and abnormal cytology in women positive for the -th HPV-type, and for 1 and 2 the type-specific risk is replaced by the type-specific risk of abnormal cytology in women positive for the -th HPV-type.
Denote by the vaccine efficacy of the -th HPV genotype. Then, after vaccination, the probability of a positive HPV result in screening round ( = 1, … , 7) becomes After vaccination, the risk of developing CIN3+ after a positive result in Round 1 becomes , and the CIN3+ risk after a positive result in Rounds 2 to 7 becomes 1 , 2 , 1 and 2 were re-estimated in a similar way as shown for 1 and 2 .    Figure S1: Effect of vaccination on the probability of HPV infection.

Supplementary Figures
The probabilities of HPV infection are estimates needed for the estimation of the lifetime risks of CIN3+ and CIN2+. The probability of a prevalent HPV infection in Round 1 is shown in blue and the probability of an incident HPV infection in Rounds 2 to 7 is shown in red. Separate estimates are presented for the no vaccination scenario and for the three scenarios bivalent/quadrivalent (2/4vHPV), bivalent with cross-protection (2vHPV + cross), and nonavalent (9vHPV) vaccination.