HPV cervical infections and serological status in vaccinated and unvaccinated women

Study population

Table 1 highlights key demographic characteristics of the population studied. Vaccination status is well balanced in the cohort, with 73 women (49%) being vaccinated. Among these, n = 63 (86%) received three doses, mostly with the tetravalent Gardasil vaccine (n = 62, 85%) at a median age of 15 years old. This vaccine coverage is high compared to the French national average in the corresponding cohorts, with estimates of 33.3%, 24.7%, and 5.4% for 14 years old girls born in 1993, 1994, and 1995, respectively [13]. However, most participants were students (86%) and had a high education level (40% have a BSc or a MSc), which has been reported to be associated with vaccine uptake [14].

The number of new sexual partners reported over the last twelve months was high compared to other studies [15]. However, this is consistent with the longitudinal cohort inclusion criteria that participants should report at least one new sexual partner over the last year [11]. Smoking status was in line with the national average, which is 35.3% in women between 18 and 25 years [16] and so was the contraception method used, with twice as many hormonal than non-hormonal contraception users [17]. Finally, the median age at sexual debut was slightly lower than the national average, which is between 17 and 18 years [18]. The only factor that differed slightly between the vaccinated and unvaccinated populations, although non-significantly after a Bonferroni correction, was age at menarchy, which was slightly lower in the vaccinated group.

In the following, all variables listed in Table 1 are used as cofactors in the statistical tests but we only report significant effects in the main text (detailed analyses are shown in Appendix).

Cytology

Women were screened for cervical lesions at the first visit using liquid cytology (see the Methods). In the analysis, following national recommendations, Atypical Squamous Cells of Undetermined Significance (ASCUS) were re-qualified as ‘normal’ if the sample was found to be HPV negative, and as Low-grade Squamous Intraepithelial Lesion (LSIL) if HPV positive. Twelve women (8.1%) were diagnosed with LSIL, which is consistent with French data [19], and none were found to have higher grade lesions.

The only factor associated with an increased risk of detection of a LSIL was the reported number of partners over the last twelve months (OR=1.35, 95% CI [1.02;1.81]). However, this effect could be related to our requalification of HPV positive-ASCUS as LSIL, since, as we will see, HPV detection is associated with this variable. Eight of the twelve women originally diagnosed with ASCUS were not vaccinated, but this effect was not significant (Supplementary Table S2).

Prevalence and genotypes of HPVs

66% of the women were positive for alphapapillomavirus detection using the DEIA test in at least one of their two visits, and 47% were positive at the two visits one month apart (Table 2). This is high compared with estimates in many other countries (e.g. 24% for women less than 25 years old in [20]). However, since one of the inclusion criteria of the study was to report a new sexual partner over the last twelve months, this likely selects for a more sexually active and thus more exposed population. Vaccination status had the most significant effect on HPV detection with a strong protective effect, with an odds ratio (OR) of 0.45 and 95% confidence interval (CI) of [0.22;0.89]. The reported number of partners over the last twelve months and the participant’s age at inclusion were also associated with increased HPV detection with OR of 1.23 (95% CI [1.08;1.44]) and 1.22 (95% CI [1.02;1.47]) respectively. Counter-intuitively, participants who reported no use of contraception presented a lower odds ratio for HPV detection, although their number of new sexual partners over the last twelve months was not significantly different from that of the other participants (p-value of 0.89).

The most frequent genotypes in the study population were HPV51, HPV53, and HPV66 (Figure 1). Except for the conspicuously lower prevalence of HPV16, this is consistent with the most prevalent types in France in normal cytologies being HPV16, HPV51, HPV54, and HPV53, although across all ages [20].

• Download figure
• Open in new tab

Figure 1. HPV genotype relative prevalence at the two visits according to vaccination status.

Samples that were EIA-positive but LiPA25-negative are indicated as non-typable. We also highlight differences between genotypes that were present at both visits (dark color) or at a single visit (light color). HPVs here classified as ‘high-risk’ belong into groups 1 and 2A in the IARC carcinogenicity scheme [21].

Our testing scheme also allows to assess whether the same HPV is present in the two visits (at V1 and V2), which we refer to as ‘sustained’ infections, as opposed to ‘transient’, i.e. detected only once. As shown in Table 2, sustained infections were found in 64 women (that is 44% of the study population). This means that for most of the women who were positive twice for DEIA (94%), the repeated infection was a sustained infection. Note that for HPV18, all the infections detected were transient, whereas for HPV56 most infections were sustained. Reporting a higher number of partners over the last twelve months was associated with increased probability (OR=1.22, 95% CI [1.06;1.43]) of having a sustained infection. We also found a significant effect of Body Mass Index (BMI) towards lower probability of presenting a sustained infection (OR=0.87, 95% CI [0.76;0.98]). The effect of vaccination on presenting a sustained infection was only borderline significant with a 95% confidence interval of [0.23;1.04] for the odds ratio. However, when performing the same analysis only with oncogenic HPVs, the odds ratio for presenting a sustained infection in vaccinated women becomes 0.33 (95% CI [0.16;0.64]) (Supplementary Table S5). In this case, the odds ratio associated with the number of partners remains significant, but not the BMI effect.

Using HPV genotype information, we analysed coinfections, which we define as the detection of the same two (or more) HPV genotypes at the two visits spread one month apart. The prevalence of coinfections among all participants was 12.3%. The number of partners was positively associated with coinfections (OR=1.30, 95% CI [1.10;1.58]), while the BMI (OR=0.78, 95% CI [0.58;0.98]) and vaccination status (OR=0.18, 95% CI [0.03;0.65]) were negatively associated with coinfections (Table 2). This pattern is consistent with our results on HPV detection and sustained infections. The link with the number of partners is consistent with earlier results [15]. Note that contrary to single infections, the effect of vaccination is strongly significant without restricting the analysis to oncogenic HPVs.

Type-specific serological status

We estimated IgM and IgG serological status specific for ten different HPVs at inclusion. As expected, IgG seropositivity was very strongly associated with vaccination status (Figure 2 and Table S8). This was true for the vaccine types but also for six other oncogenic HPVs (although to a lower extent for HPV33).

• Download figure
• Open in new tab

Figure 2. Type-specific HPV seroprevalence at inclusion.

The panels show the IgG (left) and IgM (right) for vaccinated (red) and unvaccinated (blue) participants. Figures refer to the corresponding HPV, as in ‘IgG 6’ for HPV6. The antibodies detected target the L1 virus protein.

Some vaccinated participants were seronegative for the vaccine types. More precisely: one vaccinated woman presented no IgG against HPV16 and HPV18, but she only received one single dose of the vaccine at age 19; five women with a complete Gardasil vaccination schedule were seronegative for HPV18; and one woman vaccinated with Cervarix presented no immunity against HPV6 and HPV11 (which is expected given that these two types are not targeted by this vaccine). These results are consistent with the literature, especially with the lower antibody titers against HPV18 compared to those against HPV16.

To further analyse the results, we compared participants depending on the number of HPVs for which they were seropositive. For the whole population, vaccination was the only significant effect (Table S8). Focusing on unvaccinated participants, we found that participants who reported not using any contraception method were seropositive for a lower number of HPVs (Supplementary Table S9). This was consistent with these same participants having a lower risk of HPV detection (Table 2).

For IgMs, the main factor associated with being seropositive to multiple types was HPV detection at the first visit and not at the second visit one month later (Table S10). Vaccination status did not affect IgM seroprevalence, which is consistent with earlier studies [22].

Vaccination and HPV prevalence

We found that certain HPVs were disproportionately less often detected in vaccinated women. This is the case, as expected, for vaccine-genotypes such as HPV16 (OR = 0.17, although this difference was not significant after the Bonferroni correction on testing for 25 genotypes, p-value = 0.415). We also found a similar protective effect against HPV31. Unexpectedly, the protective effect was most significant for HPV51 (OR = 0.11, p-value=0.001). The result remained significant even when we ignored transient infections and restricted the data to sustained infections. Although this result was unexpected (HPV51 is not closely related genetically to any of the vaccine types), it echoes an earlier result found on the large international PATRICIA cohort study [23].

To rule out potential issues associated with the typing test, we performed an additional PCR analysis specific for HPV51 on all the 149 inclusion visits (V1) using another protocol [24] targeting the E7 gene instead of the L1 gene targeted by the DEIA and LiPA₂₅ assays [25]. We confirmed that vaccination status was associated with a lower prevalence of HPV51, with an odds ratio of 0.31 (95% CI [0.09,0.9], p-value = 0.023 with Fisher’s exact test).

Data

Data originates from the PAPCLEAR study, which investigates the kinetics and ecology of HPV genital infections in young women. Here, we analyse data from the inclusion visit (Visit 1) and the result visit (Visit 2) for N = 149 participants (one was removed after realising she did not meet one of the inclusion criteria). The detailed protocol of the study is described elsewhere [11]. In short, during the inclusion visit, a gynaecologist or mid-wife performed a cervical smear in PerservCyt medium for squamous intraepithelial screening via liquid cytology. The same sample was used to perform HPV detection and typing (see below). Self-sample vaginal swabs and blood samples were obtained with a nurse, respectively to test for other STIs and for anti-HPV antibody assays. Women come back one month later for a results visit, where another HPV detection test is performed. Patterns of positivity at the two visits as well as population sies are shown in Figure 4.

• Download figure
• Open in new tab

Figure 3. How vaccination affects the risk of detecting specific HPV genotypes.

If OR=1, the HPV genotype is equally prevalent in vaccinated and unvaccinated women. If OR<1, the genotype is less prevalent in vaccinated women. Genotypes are ordered from the most often prevalent at the two visits (HPV51, on top) to the least prevalent. The size of the dot indicates the magnitude of the p-value. Significant p-values after the Bonferroni correction are in yellow.

• Download figure
• Open in new tab

Figure 4: Flowchart showing the visits and the outcome of the HPV detection and typing tests.

Protocols

For DNA extraction, we centrifuged 2mL of the cervical smear at 4000rpm for 5min. We then removed the supernatant and extracted DNA from the pellet using a QIAamp DNA Mini Kit (Qiagen51306).

For HPV detection and typing, we first tested for the presence of alphapapillomaviruses using the generic DEIA test [38]. We then used the LiPA25 typing [25] on DEIA-positive samples. Both tests are based on detecting the same amplicon of the L1 viral gene. Total DNA in the sample was quantified using Qubit kit (Thermo Fisher Scientific). Samples that were DEIA-positive and LiPA25 negative were amplified using the PGMY PCR [39] and sequenced using Sanger sequencing. Samples for which the sequencing did not yield a clear sequence, most likely because of coinfections, are labelled as ‘non-typable’. For HPV51, we performed an additional PCR targeting the E7 viral gene using already published primers and protocol [24].

Protocols for circulating anti-HPV antibody

IgG and IgM antibodies against late (L1) proteins of high-risk HPV types 16, 18, 31, 33, 35, 45, 52 and 58, as well as low-risk HPV-types 6 and 11 were analysed with the multiplex serology assay using beads coated with recombinant glutathione s-transferase (GST) fusions proteins. The assay procedure has been previously described in detail [40]. Briefly, the samples were tested at a final serum dilution of 1:100 using an IgG and an IgM goat anti-Human secondary antibody. Seropositivity was defined based on standard definitions [41, 42].

Epidemiological meta-data

Participants were asked to fill in questionnaires during the visit, and additional information was collected by the gynaecologist or mid-wife during the interview. Further details are available in the detailed protocol [11]. The variables included in this analysis are shown in Table 1.

Statistics

We used either logistic regression or ordinal logistic regression models, depending on the number of levels of the response. For each model, we computed the odd ratios associated with each predictor along a 95% confidence interval. Note that while all variables were considered in our analyses, we only report the variables that are at least borderline significant.

For the difference in HPV genotype prevalence between the vaccinated and the unvaccinated populations, we used Fisher’s exact tests and applied a Bonferroni correction for multiple hypothesis testing.

All the analyses were performed in R. We used the glm function to conduct logistic regressions, and the polr function from the MASS package to perform ordinal logistic regressions. The proportional odds assumption was verified using the lrm and residuals.lrm functions from the rms package, and this assumption was verified for each of our predictors except chlamydia on the HPV presence model (which was dropped). We used the fisher.test function to realize Fisher’s exact tests.

Ethics

The PAPCLEAR trial is promoted by the Centre Hospitalier Universitaire de Montpellier and has been approved by the Comitè de Protection des Personnes (CPP) Sud Mèditerranèe I on 11 May 2016 (CPP number 16 42, reference number ID RCB 2016-A00712-49); by the Comitè Consultatif sur le Traitement de l’Information en mati re de Recherche dans le domaine de la Santé on 12 July 2016 (reference number 16.504); and by the Commission Nationale Informatique et Libertès on 16 December 2016 (reference number MMS/ABD/AR1612278, decision number DR-2016-488). This trial was authorised by the Agence Nationale de Sècuritè du Mèdicament et des Produits de Santè on 20 July 2016 (reference 20160072000007). The ClinicalTrials.gov identifier is NCT02946346. All participants provided written informed consent.