HPV viral load in self-collected vaginal fluid samples as predictor for presence of cervical intraepithelial neoplasia

The data was derived from women between 30 and 60 years of age participating in two randomized intervention studies that were conducted in Uppsala County, Sweden, between 2013 and 2015 [3, 23]. The first randomized study included women between 30 and 49 years of age and had as aim to compare the detection rate of CIN2+ lesions based on histology in women performing repeated self-sampling of vaginal fluid for hrHPV testing with women following the regular screening program based on Pap smear cytology [3]. Only women in the intervention arm that followed the protocol of that study were included in the present analysis. The second randomized study included women between 50 and 60 years of age and had as aim to compare the detection rate of CIN2+ based on histology in women performing repeated self-sampling of vaginal fluid for hrHPV test with women sampled by medical personnel on the cervix for hrHPV test [23]. In the present study we only included women from the previous two studies that had performed self-sampling for hrHPV test according to study protocol. Women that were hrHPV-positive in their first sample were informed of the test result within 2 weeks after their sample was returned to the laboratory. These women were also informed that they would be asked to repeat the sampling in 4–6 months, and that they could contact a gynecologist in case of questions or symptoms. Women that were hrHPV positive in two consecutive HPV tests were referred to colposcopy and eventual biopsies. Women that were hrHPV negative in their first or second HPV test were referred back to the regular screening program. The follow-up period was 18 months from date of invitation.

The procedure for self-sampling of vaginal fluid was based on using a silicon brush and the indicating FTA elute micro card™ and regular mail both for distribution of the sampling kit and return of the sample, as previously described [2]. The women were instructed to perform self-sampling of vaginal fluid using the Rovers® Viba-brush (Rover Medical Devices B.V., Oss, The Netherlands) and apply the vaginal fluid sample to the indicating FTA elute micro card™ (GE Healthcare, United Kingdom, art. no WB129308). Self-samples of vaginal fluid were returned to the HPV laboratory at Uppsala University by regular mail. The samples were processed using a dedicated automated laboratory system (easyPunch STARlet, Hamilton Robotics, Bonaduz, Switzerland) which collects each card, takes a photograph of the sample deposition area, identifies the regions with the highest amount of cellular material using a machine learning software, and then takes 4 punches from that area with a 3-mm diameter knife. The punches are deposited in a single well in a 96-well microtiter plate. DNA extraction from the punches was performed as described earlier [24].

HPV testing was performed using the clinically validated, real-time PCR-based, hpVIR test [25, 26]. This test detects and quantifies HPV16, 31, 35, 39, 51, 56 and 59 as individual genotypes, HPV18 and 45 in one group and HPV33, 52 and 58 as a second group. The test also detects and quantifies a human single copy nuclear gene (HMBS), which serves as a control for that the samples contain sufficient amounts of cellular material for the test to be informative, and a reference to which the HPV copy number can be related, i.e. for normalization of the HPV copy number. The limit of detection (LOD) for hpVIR was 10 copies per PCR for both the nuclear single copy gene HMBS and HPV.

Gynecological examinations and colposcopies with biopsies were performed at the Clinic of Obstetrics and Gynecology, and histology at the clinic of Pathology and Cytology, both at Uppsala University Hospital, Uppsala. The outcome was the number of women with CIN2+ based on histology according to the SNOMED classification code diagnosed during the 18 months follow-up period from date of invitation. We also included squamous cell carcinoma, adenocarcinoma and adenosquamous carcinoma as outcomes.

Statistical calculations were performed and figures generated using R version 3.4.3 [27]. Significance levels for comparison between transient infections, persistent infections without CIN2+ and persistent infections with CIN2+ were calculated using the two-sided rank-based Spearman test (Wilcoxon) and p-values were adjusted for multiple testing using Bonferroni correction, and q-values< 0.05 considered significant. Odds ratios (OR) with 95% confidence intervals were calculated with the “oddsratio” function from the “epitools”-package (version 0.5–10) [28], and p-values were calculated with Fishers exact test. Receiver operating characteristic (ROC) curves were generated for women with transient infections and women with persistent infections and CIN2+ lesions. A general linear model was built using 50% of the samples for a 5-fold cross-validation training using the “caret”-package (version 6.0–78) in R [29]. The remaining 50% of the samples were used as test set to which the model from the training set were applied. The training set and the test set were chosen to contain the same frequency of cases and controls. ROC (reporter operator characteristic) curves, AUC (area under ROC curve), sensitivity and specificity were then generated using the pROC-package (version 1.10.0) [30].

All tests were performed for both hrHPV copy number, representing the amount of virus per sample (and PCR) and hrHPV titer, i.e. hrHPV copy number normalized by HMBS copy number, representing the amount of HPV per number of cells in the sample.

This study included women from two previous randomized studies whom had a positive hrHPV screening test and a follow-up hrHPV test in 4–6 month based on a second self-sample [3, 23]. From these two studies a total of 667 women performed two consecutive hrHPV tests, had histology available, and were considered eligible for the present analysis. In these women, a total of 752 hrHPV infections were detected in the primary screening test, with HPV16 being the most prevalent type and also the HPV type most likely to be present in the two consecutive hrHPV tests (HPV16 persistence; 76%, non-HPV16 persistence; 20–61%) (Table 1). Among women with persistent infection, 191 had or developed a CIN2+ histology during the 18-month follow-up period. The analysis was performed for both hrHPV copy number and hrHPV titer, i.e. hrHPV copy number normalized for amount of human cellular material.

The women were divided into three groups depending on hrHPV status and histology; I) transient infections (i.e. those that cleared the infection in the 4–6 months between the first and second hrHPV test), II) persistent infections (those with two consecutive positive hrHPV tests) but no CIN2+ lesion within the 18 months follow-up period, and III) women with persistent infection and CIN2 + .

HPV16 copy number and HPV16 titer in the primary screening test were both significantly higher in women with persistent infection, with and without CIN2+, as compared to women with transient infection (Bonferroni adjusted p-value = 1.42E-02 and 5.33E-03) (Fig. 1 A-B, Table 2). Similarly, both copy number and titer of the HPV33/52/58 group were significantly higher in women with persistent infection, with and without CIN2+, as compared to women with transient infection (Bonferroni adjusted p-value = 9.72E-07 and 1.82E-04) (Table 2). In the follow-up sample, HPV16 titer was significantly higher in women with persistent infection and CIN2+ as compared to women with persistent infection without CIN2+ (Bonferroni adjusted p-value = 2.29E-03) (Fig. 1B, Table 2).

Considering all hrHPV types, i.e. total hrHPV viral load in the sample, the copy number and titer in the baseline test was higher in women with persistent infection with or without CIN2+, in comparison to women with transient infection (Bonferroni adjusted p-value = 1.34E-08, 3.85E-07 and 5.82E-07, 4.73E-06) (Fig. 1 C-D, Table 2). There was no significant difference in total hrHPV copy number or titer in the follow-up sample between women with CIN2+ and those without CIN2+ (Fig. 1 C-D, Table 2).

Women with HPV16 infections were divided into four quartiles based on HPV copy number or titer in the primary screening test. First, we compared course of infection for women in the 1st titer quartile with women in the 2nd, 3rd and 4th titer quartiles, which gave an OR for persistent infection of 2.4, 3.9 and 3.9 for HPV16 infection and an OR of 2.2, 3.5 and 3.1 for all hrHPV titers (Table 3). For HPV copy number the OR was 1.9, 2.3 and 4.5 for HPV16 infection and 2.2, 3.5 and 3.1 for total hrHPV copy number, for the 2nd, 3rd and 4th titer quartiles, respectively. In comparison to the first quartile, the OR for distinguishing between a transient HPV16 infection and a persistent infection with CIN2+ was OR = 3.0, 3.4 and 3.1 for the 2nd, 3rd and 4th titer quartiles (Table 3). For total hrHPV titer, as compared to the first quartile, OR = 2.4, 2.9 and 2.3 for the 2nd, 3rd and 4th quartile (Table 3). For copy number, the OR were 2.1–3.6 for HPV16 and 1.4–2.4 for total hrHPV viral load (Table 3).

The samples with transient infections were used together with the samples with persistent infection and CIN2+ lesions to build a general linear model, using 50% of the samples for 5-fold cross-validation training. The remaining 50% of the samples were used as validation set to which the model from the training set was applied. The training and the validation sets were chosen to contain the same frequency of cases and controls. ROC curves for HPV16 copy number showed an AUC of 0.67 (0.53–0.81, 95% confidence interval) with best point sensitivity SE = 0.77 and specificity SP = 0.60 (Fig. 2 a). For HPV16 titer, AUC was 0.67 (0.54–0-80, 95% confidence interval) with SE = 0.62 and SP = 0.68 (Fig. 2 b). Total hrHPV viral copy number had an AUC of 0.64 (0.56–0.71, 95% confidence interval) with SE = 0.57 and SP = 0.65, and total hrHPV titer had an AUC of 0.61 (0.54–0.69, 95% confidence interval) and SE = 0.60 and SP = 0.62.