Infection by high-risk (HR) anogenital human papillomaviruses (HPV) is a necessary cause of cervical cancer (CC) [1]. HPV16 and 18 account for 70% of cases worldwide, while the prototypic low-risk (LR) types, HPV6 and 11, cause the most frequent sexually-transmitted infection (condylomata acuminata). Acquisition of anogenital HPV is associated with sexual debut [2]. Most infections resolve within 2 years in immunocompetent individuals, and less than 1 % of initial HR HPV infections can progress to CC over an extended period of time (10–20 years) [3–5].

Vaccination against HR HPV targeting girls prior to sexual debut is expected to reduce CC burden, even in those who are not vaccinated thanks to herd immunity [6]. Three HPV vaccines are currently available: Cervarix® targeting HPV16 and 18, Gardasil® targeting HPV6, 11, 16, and 18, and Gardasil-9® targeting HPV6, 11, 16, 18, 31, 33, 45, 52, and 58. Cervarix® and Gardasil® could prevent 70% of CC cases and Gardasil-9® almost 90%. Proving HPV vaccines’ effectiveness against CC is difficult owing to the long delay between initial infection and cancer. Surrogate markers therefore have been proposed to determine vaccines’ effectiveness on a shorter term, such as population-based continuous monitoring of high grade precursor lesions (cervical intraepithelial lesions grade 3 or more, CIN3+) [7]. Accordingly, a CIN3+ obligatory reporting system is being evaluated for HPV vaccine monitoring in Switzerland [8].

Data obtained from large cohorts of women several years after implementation of Cervarix® [9, 10] or Gardasil® (reviewed by Garland et al. [11]) have shown that both vaccines are efficient at reducing the frequency of precursor lesions associated with the vaccine genotypes. However, even the nonavalent vaccine will not be able to prevent all CC cases, and a large fraction of older women are presently not vaccinated. Successful prevention of CC therefore will still rely on screening for years to come. Cervical cancer screening modalities will need to take into consideration the progressive reduction of abnormal lesions harbouring the most carcinogenic HPV genotypes. This reduction is expected to affect negatively the clinical specificity of cytology [12]. The knowledge of HPV burden in the transition from the pre-vaccine to the vaccine era therefore is important in countries such as Switzerland where cytology is used for primary screening and HPV testing as an adjunct in an opportunistic setting.

Gardasil® has been used in Switzerland and offered to schoolgirls aged 11–14 since 2008, yet its impact is not known in Switzerland. The aims of our work were to examine the proportion of vaccine-type HPV among anogenital HPV in sexually-active 18-year-old Swiss females five years after vaccine implementation, and the temporal impact of Gardasil® on this proportion in outpatients participating to CC screening from 1999 to 2015.

Most (n = 562; 81.4%) of the 690 subjects had an intercourse at least once in the previous five years (mean age at sexual debut of 16.0 ± 1.4 years, Additional file 1) and less than 5 % reported sexual debut occurring before age 14. The number of sexual partners was one (n = 189; 33.6%), two to five (n = 240; 42.7%) and more than five (n = 62; 11%). Usage of condom was systematic for 115 subjects (20.5%). The majority of subjects had not been treated for STI (>95%), nine (1.3%) reported suffering from an immune-related disease and 203 reported being smokers (29.4%).

Five hundred and forty-nine (79.6%) subjects accepted self-sampling for HPV and CT analyses. Acceptance was found in association with “having sexual intercourse” (P < 0.001) and “being smoker” (P = 0.02) by multivariable analysis (Additional file 4, panel A). However, only 327 (59.6%) eventually provided a self-collected sample resulting in a low proportion (8.8%) of women among the entire population to be examined for HPV and CT. The working status “being student” was significantly associated with having actually provided a sample (P = 0.04) by multivariable analysis (Additional file 4, panel B). The vast majority (99.7%) considered the self-collection device user-friendly.

The distribution of the 217 genotype-sample combinations in 18-year-old sub-study-1 subjects in 2013 is graphically displayed according to their vaccination status in Fig. 1 a. The proportion of the four vaccine-type HPV (n = 9) was 4% (9/217), with HPV16 (n = 6) ranking 14th, HPV6 (n = 3) 20th, HPV11 and HPV18 (n = 0) 23rd.

The HPV genotype-sample combinations (n = 3619) observed before vaccine implementation (1999–2007) in the sub-study-2 outpatients were distributed by year and patients’ age (<26, 26–30 and >30), and ranked by decreasing occurrences within the youngest age group (<26, n = 1438 genotype-sample combinations) (Fig. 1 b). The proportion of the four vaccine-type HPV (n = 369) was 25.7%, with HPV16 (n = 212) ranking first, HPV6 (n = 74) fifth together with HPV58, HPV18 (n = 64) 10th, and HPV11 (n = 19) 21st. Restricting the outpatients’ population to the youngest age (<26) with normal cytology (n = 87; 50 HPV-negative) gave a similar proportion (25%) (n = 14 out of 56 genotype-sample combinations); HPV16 (n = 7) ranked first (data not shown). To test whether this proportion decreased after vaccine implementation, as suggested by the low proportion observed in the sub-study-1 subjects, the yearly proportion of the four vaccine-type HPV was calculated from 1999 to 2015 for each of the three age groups of outpatients, and its evolution examined by segmented logistic regression. Model adjustment was made to evaluate the year at which the regression coefficient of the model changed (breakpoint year, Additional file 5). No significant variation was observed for the patients aged >25. In contrast, a significant reduction in the proportion of the four vaccine-type HPV was observed for the patients aged <26 starting in 2010, two years after the vaccination campaign was initiated (breakpoint year 2009, OR = 0.82, P < 0.0001) (Fig. 2 and Additional file 5). The same analysis performed after further stratifying the youngest age group (<21, 21–25) confirmed this result (OR = 0.73, P = 0.05 and OR = 0.77, P < 0.001, respectively; Additional file 5). To establish the percentage of reduction of the vaccine-type HPV, the average yearly proportions of the vaccine-type HPV in the pre-vaccine era (1999–2007) was compared with that in the post-vaccine era at the plateau (2013–2015, fig. 2) for the youngest outpatients (<26). It was 25.5% (95% CI: 20.8%–30.1%) in the pre-vaccine era and significantly reduced by 56% (P = 0.0028 unpaired t test) in the post-vaccine era to 11.3% (95% CI: 10.2%–12.5%). Restricting the vaccine genotypes to HPV16 and HPV18, excluding HPV6 and HPV11 from the calculations, gave similar results; average yearly proportions of 21.0% (95% CI: 16.6%–25.3%) compared with 8.7% (95% CI: 6.7%–10.8%), 59% reduction (P = 0.0048).

This study is the first to investigate the impact of Gardasil® after its introduction in 2008 in Switzerland. We first assessed vaccination coverage and demographics as well as HPV genotypes in an 18-year-old women population targeted by the vaccine five years earlier in the canton of Vaud. The participation rate was low with 18.5% returned questionnaires and 8.8% self-collected samples. The low participation rate is a major weakness of this sub-study that necessitated to verify the representativeness of the participants. Reassuringly, our sub-study population was comparable to the Swiss women population of similar age based on risk factors for HPV such as hormonal contraception and use of condoms (Swiss Federal Statistical Office) [19] and on demographics of adolescents in Switzerland (SMASH-02 report) [20]. The education levels of the participants were also comparable to the general population of the same age in the canton of Vaud. Thus the proportions of students (57.5% ± 3.7%) and subjects without formation (11.1% ± 2.3%) overlapped those of the 18-year-old population registered December 31st 2013 (53.1% ± 1.5%, 8.8% ± 0.8%, respectively) while apprentices (31.3% ± 3.5% vs. 38.0% ± 1.4%, respectively) were slightly underrepresented (personal communication: Mabillard, H, October 2016, canton of Vaud Statistical Office). Vaccinees and non vaccinees showed similar rates of CT infections, consistent with HPV vaccination not influencing sexual behaviour as reported by others in England [21] and in Canada [22]. In addition the CT prevalence rate (4.6 ± 1.2%) was similar (5.5 ± 0.9%) to that of a previous survey of sexually-active women aged <25 in the same area of Switzerland [13], consistent with sexually-active subjects having preferentially provided a self-collected sample.

In the canton of Vaud, 69.5% of the schoolgirls aged 14 have been vaccinated (≥1 dose) against HPV during the 2010–2011 school year [23]. The vaccination rate reported by the subjects in our study is somewhat higher at 77.5%. This is most likely due to catch-up vaccination offered to women aged up to 19. In contrast, only 56% of the vaccinees reported having had the three-dose regimen compared with 96% in the 2010–2011 survey. This difference most likely reflects some degree of unreliability of the self-reported vaccination status (regimen in particular). This mode of reporting was preferred to consultation of vaccination books by nurses to minimize costs and ensure anonymous treatment of the files according to our ethical approval.

The percentages of women who systematically used condoms (20.5%) and contraceptive pills (64%) are similar to those of young women in a stable relationship (36.1% and 64%, respectively) (SMASH-02 report) [20]. This suggests that the majority was engaged in a stable relationship at time of questionnaire, consistent with the mean age at sexual debut (16.0 ± 1.4 years). This age is similar to that reported in Germany and in other European countries [24–26]. In addition, less than 5% of girls from our study reported having initiated sexual intercourse at 13 years or younger, in line with the recommended age (11–14) for vaccination in Switzerland.

A low percentage of returned samples (12%) was reported in a German study assessing HPV prevalence in women aged 20–25 years using the Delphi-Screener® [24]. The even lower percentage (8.8%) in our study may reflect consecutive shipment of questionnaire and self-collection devices while it was simultaneous in the German study. Self-sampling was accepted more frequently by women having sexual intercourse, and those actually providing a sample were more frequently students. This latter observation may be related to a possibly better knowledge of HPV infection by women with a higher level of education.

Global HPV prevalence increased with number of sex partners and other variables linked to undergoing sexual activity, as well as with lower degree of education. These data confirm previously published data on identified risk factors for HPV infection [4]. Four participants in the sub-study-1 vaccinated group were positive for HPV16. Three reported to have received the three recommended doses, whereas one mentioned two doses. All four reported being sexually active after the vaccination was completed. The low response rate and the overall low numbers of vaccine-type HPV-positive cases as well as some degree of unreliability of self-reporting did not allow meaningful statistical evaluation of the age at vaccination and age at first sex to test the hypothesis that HPV16-positive vaccinees may have been infected before vaccination. Taking into account the information provided by the volunteers, the presence of HPV16 in vaccinees may also reflect transient infections or vaccine failures that cannot be distinguished since HPV genotyping was performed at a single time point, thereby excluding assessment of persistent infections.

In view of the vaccination coverage well above 50%, the low proportion (4%) of the vaccine-type HPV was expected in our 18 year-old subject population compared with that (25%) in the youngest sub-study-2 outpatients during the pre-vaccination era according to a recent meta-analysis of the impact of HPV vaccination programmes and herd effect [6]. Herd immunity induced by HPV vaccination is reported by more and more researchers across the globe [11]. For instance, several studies have shown a major decline in the prevalence of vaccine-type HPV and in clinical expressions of HPV-induced illnesses like genital warts in Australian boys before their access to HPV vaccination [27].

The lower proportion of the vaccine-type HPV in sub-study-1 subjects may have resulted from using different means of sample collection (Delphi screener self-collection vs. PAP smear) and HPV genotyping assays (Anyplex II™ HPV28 vs. PGMY-CHUV), hence for this reason was not tested statistically. The strength of our study however relied on our ability to assess without bias the evolution of the yearly proportion of the vaccine-type HPV with the sub-study-2 outpatients’ data because the cervical cancer screening algorithm, sample collection and HPV genotyping method remained identical over the years (1999–2015). The youngest sub-study-2 outpatients (<26) were most likely enriched in women to whom the vaccine was offered since 2008, while the older women (26–30 and >30) were gradually less likely to have received the vaccine or have benefitted from herd immunity. Consistent with this and with the known clearance time of HPV (90% within two years), a time wise significant reduction of the yearly proportion of the four vaccine-type HPV was observed for our outpatients aged <26. The 56% reduction of the four vaccine-type HPV and the 59% reduction of HPV16 and HPV18 achieved as of five years after vaccine implementation is reminiscent of the numbers published in the US [28] [29] and in the UK [30–32]. Our data add to these studies the yearly kinetic of reduction of the vaccine-type HPV in a real-world condition of opportunistic cervical cancer screening. We could not establish the prevalence of HPV genotypes in the global population since the sub-study-2 population represented a high-risk group of screened women referred to HPV testing based on cervical abnormalities. Also, we could not evaluate with confidence the influence of vaccination on the proportion of high grade lesions, population wise, as has been published for Scotland, which shows a reduction in the proportion of CIN lesions in vaccinees under an organized screening algorithm [33]. Our data will however be useful to evaluate cervical cancer screening modalities in the vaccine era in Switzerland and in other countries with similar vaccination coverage and opportunistic screening algorithms.

The low proportion of vaccine-type HPV in 18-year-old females and its rapid decrease in young women participating to cervical cancer screening extend the success of HPV vaccination to Switzerland. Our data suggest that cervical cancer screening is now entering a stage of declining prevalence of precursor lesions associated with the two most carcinogenic HPV as proposed earlier by a modelisation study [34]. There is good evidence that the decline of HPV16 and HPV18 will negatively affect the clinical specificity of cytology but not that of high-risk HPV testing [12] with a concomitant diminution of the positive-predictive value of cytology [35]. Primary screening modalities involving HPV testing and cytology should now be re-evaluated in Switzerland.