Human Papillomavirus Type–Specific 18-Month Risk of High-Grade Cervical Intraepithelial Neoplasia in Women with a Normal or Borderline/Mildly Dyskaryotic Smear

This is the first large population-based screening trial in which high-grade cervical intraepithelial neoplasia (CIN) risk for 14 oncogenic HPV types has been determined for all hrHPV-positive baseline smears. Four HPV types could be distinguished that were associated with increased high-grade CIN risk (i.e., HPV16, HPV18, HPV31, and HPV33). In the subset of women with normal cytology at baseline, HPV16- and HPV18-positive women had an increased risk of high-grade CIN compared with women positive for other hrHPV types also after multiple normal smears. Besides, cytology was less predictive for risk of high-grade CIN in HPV16/18–positive women than in other hrHPV-positive women. This seems to warrant more aggressive management for HPV16/18–positive women compared with non-HPV16/18 hrHPV-positive women, particularly if the cytologic reading is normal and therefore may have important implications for cervical screening management.

Persistent infection with high-risk human papillomavirus (hrHPV) is the central cause of cervical cancer (1) and several studies have shown that the detection rate of early cancer or clinically relevant precursor lesions can be enhanced by hrHPV DNA testing along cervical cytology (2-6). However, only few hrHPV infections progress to cancer or its closest precursor CIN grade 3 (7).

Distinguishing hrHPV-positive women with an elevated risk of CIN grade 3 or worse (≥CIN3) may result in a better management of hrHPV-positive women. At least to a certain extent, this may be accomplished by an hrHPV type analysis as recent meta-studies have shown that only a few HPV types, including HPV16 and HPV18, cover the vast majority of invasive cancer cases (8, 9). Moreover, the risk of high-grade CIN may vary across types. In a recent cross-sectional study, elevated prevalences of HPV16 and HPV33 were found in moderately dyskaryotic or worse smears with underlying ≥CIN3 compared with normal smears (10) and an increased risk of ≥CIN3 posed by HPV16 was found in a cohort with equivocal or mildly abnormal cytology at enrollment (11). Besides, by HPV16 and HPV18 typing of enrollment smears in a routine screening cohort study, increased risks were found for those two types (12). Together, these study results suggest that the risk posed by different hrHPV types for ≥CIN3 varies across cytologic categories.

The aim of this study was to assess the hrHPV type–specific 18-month risk of high-grade CIN in women with normal cytology and women with borderline/mild dyskaryosis (BMD) participating in a population-based screening program (13). Here, particular attention was paid to the role of baseline cytology. All women were typed for 14 hrHPV genotypes (9).

From January 1999 to December 2002, 44,102 women of ages between 30 and 60 years and eligible for cervical screening were recruited to participate in the Population-Based Screening Amsterdam trial (13). In this two-armed trial, effectiveness of cytology and hrHPV DNA testing (intervention group) was compared with the effectiveness of cytology (control group, hrHPV DNA results blinded). Written informed consent was provided by all women and the study was approved by the Medical Ethics Committee of the Vrije University Medical Centre (no. 96/103A) and the Ministry of Public Health (VWS no. 328650).

Women in the intervention group (n = 21,996) were referred for immediate colposcopy when the cytologic result was moderate dyskaryosis or worse (>BMD; high-grade squamous intraepithelial lesion according to Bethesda 2001), were returned to the next screening round when the smear result was normal and the hrHPV DNA test result was negative, and were followed with cytologic and hrHPV DNA testing at 6 and 18 months when the smear result was borderline or mild dyskaryosis (BMD; translates into Bethesda 2001 atypical squamous cells of undetermined significance/high-grade squamous intraepithelial lesion/low-grade squamous intraepithelial lesion; ref. 14) or when the smear result was normal and the hrHPV DNA result test was positive. Women with follow-up cytologic and hrHPV DNA testing were referred for colposcopy when the 6-month smear result was >BMD, when the baseline and 6-month smear result were BMD and the 6-month smear was hrHPV positive, or when the 18-month smear was hrHPV positive and/or interpreted as >BMD. Women in the control group (n = 22,106) followed the current Dutch screening guidelines; they were referred for immediate colposcopy when the smear result was >BMD, were returned to the next screening round at 5 years when the result was normal, and were followed with cytology at 6 or 18 months when the result was BMD. Women with follow-up cytology were referred for colposcopy when the 6- or 18-month result was ≥BMD. All women in the control and intervention group were tested for hrHPV at baseline. A flow chart of the screening management of women who were advised to return for repeat testing is presented in Fig. 1.

Conventional cytologic testing was done on smears taken with Cervex brush. Detection of hrHPV DNA in the scrapes was done by GP5+/6+ PCR enzyme immunoassay using a cocktail of 14 high-risk types (i.e., 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68; ref. 15). GP5+/6+ PCR-positive cases were subsequently typed by reverse line blotting (16).

From the women enrolled in the Population-Based Screening Amsterdam study, we included all hrHPV-positive women with normal cytology from the intervention group (n = 763) and with BMD from the intervention (n = 185) and control (n = 196) groups. Women negative on reverse line blot typing (n = 57) were excluded, leaving 713 hrHPV-positive women with normal cytology and 374 hrHPV-positive women with BMD for the analyses. We calculated 18-month type-specific cumulative absolute risks of CIN grade 2 or worse (≥CIN2) and ≥CIN3. Kaplan-Meier estimates were used and 95% confidence intervals (95% CI) for the cumulative risk were obtained by bootstrapping (17). If all women positive for a certain HPV type were censored (i.e., no cases of ≥CIN2 or ≥CIN3) or the sample size was very small (<10), a binomial reference distribution was assumed for the estimated risk. Time was set equal to the target referral date (6 or 18 months), reasoning that high-grade lesions were already present at the time of referral. CIN cases that were not histologically confirmed within 3 years after the baseline smear (13 of 179) were censored as they might have been developed after the referral date. Censoring was also applied to women who were lost to follow-up. Preferential risk of ≥CIN2 and ≥CIN3 was tested for each HPV type separately using the exact stratified log-rank test (18). Data were stratified in three age categories corresponding to the first round in nationwide screening, the second round, and rounds 3 to 7 (i.e., 29-33, 34-38, and 39-63 years). Exact P values were obtained via simulation. Because HPV16 is considered the most prevalent HPV type, risk estimates and log-rank tests for non-HPV16 types were calculated after discarding HPV16-positive cases. To examine the effect of coexisting types, analyses were repeated for single infections only. Type-specific relative risks of baseline cytology (BMD versus normal) on ≥CIN2 and ≥CIN3 risk were estimated by Cox regression. The effects of age cohort on ≥CIN2 and ≥CIN3 risk were tested by the log-rank test. Calculations were done with SPSS9.0 and Matlab7.0.

The mean age of hrHPV-positive women with normal cytology at baseline was 38.3 years (range, 29-60 years) and the mean age of hrHPV-positive women with BMD at baseline was 36.2 years (range, 29-59 years). In women of the intervention group with normal cytology, 23.1% (165 of 713) did not respond to a follow-up invitation at 6 months and 28.0% (146 of 522) did not respond to a second follow-up invitation at 18 months. Among the responders, 1 (0.1%) squamous cell carcinoma, 1 (0.1%) adenocarcinoma in situ, 27 (3.8%) CIN3 cases, and 29 (4.1%) CIN2 cases were detected during follow-up. For women with BMD, the nonresponse rates at follow-up invitations at 6 and 18 months were 9.9% (37 of 374) and 28.8% (53 of 184), respectively. In women of the intervention group with BMD, 1 (0.6%) adenocarcinoma, 2 (1.1%) squamous cell carcinoma, 1 (0.6%) adenocarcinoma in situ, 29 (16.0%) CIN3 cases, and 21 (11.6%) CIN2 cases were detected during follow-up, and in women of the control group with BMD, 1 (0.5%) adenocarcinoma, 1 (0.5%) squamous cell carcinoma, 28 (14.5%) CIN3 cases, and 24 (12.4%) CIN2 cases were detected.

Multiple hrHPV infections were less prevalent in women with normal cytology than in women with BMD (18.0% versus 24.6%; P = 0.011, Fisher's exact test). The women with invasive cancer (n = 6) were positive for only one hrHPV type; four contained HPV16, one HPV18, and another one HPV31. Loss to follow-up was not type specific (P > 0.05 for each type, Fisher's exact test).

The cumulative 18-month ≥CIN2 and ≥CIN3 risks are presented in Table 1. Overall risks were 13% (95% CI, 10-17) and 6% (95% CI, 4-9), respectively. HPV16-positive women had a ≥CIN2 risk of 27% (95% CI, 20-35) and a ≥CIN3 risk of 14% (95% CI, 9-21). In the subset of hrHPV-positive women without HPV16, overall risks of ≥CIN2 and ≥CIN3 were 7% (95% CI, 5-11) and 3% (95% CI, 1-6). HPV18-positive women had significantly elevated ≥CIN2 (18%; 95% CI, 7-35) and ≥CIN3 (9%; 95% CI, 2-23) risks compared with HPV18-negative women. HPV31-positive women had a significantly elevated ≥CIN3 risk of 7% (95% CI, 2-16) compared with HPV31-negative women and had a significantly elevated ≥CIN2 risk of 20% (95% CI, 6-44). Risk estimates remained similar when excluding women with multiple infections.

The cumulative 18-month risks of high-grade CIN in women with BMD are presented in Table 2. Overall risks of ≥CIN2 and ≥CIN3 were 34% (95% CI, 29-39) and 20% (95% CI, 16-25). In HPV16-positive women, ≥CIN2 risk was 48% (95% CI, 38-58) and ≥CIN3 risk was 37% (95% CI, 28-48). In the subset of hrHPV-positive women without HPV16, overall ≥CIN2 and ≥CIN3 risks were 27% (95% CI, 21-33) and 12% (95% CI, 8-17). HPV31-positive women had an elevated ≥CIN3 risk of 27% (95% CI, 14-46) compared with HPV31-negative women and had elevated ≥CIN2 and ≥CIN3 risks of 49% (95% CI, 29-71) and 22% (95% CI, 9-44). Different from the results for women with normal cytology at baseline, HPV18 was not associated with an elevated risk of ≥CIN2 or ≥CIN3. Risk estimates remained similar when excluding women with multiple infections.

Of all women with ≥CIN2 and normal cytology at baseline (n = 58), 14 (24%) women also had normal cytology at the first recall at 6 months and 9 (16%) women had normal cytology at both 6 and 18 months. These figures indicate that women with an hrHPV infection at baseline may have a substantial risk of high-grade CIN even when the baseline smear is followed by one or two normal smears. In subsequent analyses, we assessed whether the risk of high-grade CIN after multiple normal smears was associated with HPV types 16 and 18. Cumulative 18-month risk estimates of ≥CIN2 and ≥CIN3 in women with an hrHPV-positive normal smear at baseline and a normal repeat smear at 6 months are presented in Table 3.

Overall 18-month risk estimates of ≥CIN2 and ≥CIN3 for this group of hrHPV positive women were 6% (95% CI, 4-10) and 2% (95% CI, 1-4). Women with HPV16 at baseline had an elevated ≥CIN2 risk of 13% (95% CI, 7-24; P = 0.004) compared with hrHPV-positive women without HPV16. Presence of HPV16 was not associated with an elevated ≥CIN3 risk (P = 0.548). When positive for HPV18, and after exclusion of those positive for HPV16, ≥CIN2 and ≥CIN3 risks were 14% (95% CI, 5-37; P = 0.007) and 5% (95% CI, 1-28; P = 0.037), indicating an increased risk for HPV18 as well. Inspection of the hrHPV status at 6 months revealed that all four CIN3 cases were detected after two hrHPV-positive normal smears and 1 of 10 CIN2 cases was hrHPV-negative at 6 months. The latter case was HPV18 positive at baseline. In case the baseline smear was positive for hrHPV types other than types 16 and 18 and the 6-month smear was still hrHPV positive, 18-month ≥CIN2 and ≥CIN3 risks were only 2% (95% CI, 1-9) and 1% (95% CI 0-8).

In women with an hrHPV-positive normal smear at baseline and two subsequent normal smears at 6 and 18 months, ≥CIN2 and ≥CIN3 rates were 4% (95% CI, 2-8) and 1% (95% CI, 0-3). ≥CIN2 rates were significantly different (P = 0.028) in women with baseline HPV16 and/or HPV18 (9%; 95% CI, 4-18) and women positive for another hrHPV type at baseline (2%; 95% CI, 0-6).

The relative risk of ≥CIN2 in women with a baseline hrHPV-positive BMD smear compared with those with an hrHPV-positive normal smear was 2.17 (95% CI, 1.40-3.37) if positive for HPV16, 1.45 (95% CI, 0.54-3.90) if positive for HPV18, and 5.22 (95% CI, 2.94-9.26) if positive for another hrHPV type. The difference between the relative risks in HPV16/18–positive women and in women positive for another hrHPV type was statistically significant (P = 0.012). For end-point ≥CIN3, relative risks in women with a baseline hrHPV-positive BMD smear compared with those with an hrHPV-positive normal smear were 2.98 (95% CI, 1.69-5.23) if positive for HPV16, 1.03 (95% CI, 0.23-4.64) if positive for HPV18, and 6.27 (95% CI, 2.50-15.75) if positive for another hrHPV type. A marginal negative effect of age on ≥CIN2 risk was found for HPV16-positive women with normal cytology at baseline (P = 0.054) and a positive age effect was found for other hrHPV-positive women with normal cytology at baseline (P = 0.024). No effect of age on ≥CIN2 was found for HPV16-positive women with BMD at baseline (P = 0.811) or for other hrHPV-positive women with BMD at baseline (P = 0.284). Age was not associated with ≥CIN3 risk.

Our results showed that HPV16-positive women with normal cytology and BMD had strongly increased 18-month absolute risks of high-grade CIN compared with hrHPV-positive women who were negative for HPV16. In the latter group, increased risks of high-grade CIN were found for HPV31 and HPV33, both after a normal and BMD baseline smear. Women with an HPV18 infection had an increased risk of high-grade CIN when the baseline smear was normal but not when the baseline smear was BMD.

We calculated that the 18-month ≥CIN3 risk in HPV16-positive women was 14% after normal cytology and 37% after BMD. In hrHPV-positive women without HPV16, ≥CIN3 risks were substantially lower and only 3% after normal cytology and 12% after BMD. Nonetheless, in HPV16-negative women, we were able to distinguish types 18, 31, and 33 from the other hrHPV types after normal cytology and types 31 and 33 also after BMD. Results are consistent with cross-sectional analyses of women with a cytologic reading of moderate dyskaryosis or worse (>BMD) where both HPV16 and HPV33 could be distinguished from other hrHPV types (10). Results are also consistent with two prospective studies (11, 12). In the first study (11) of women in a trial cohort with abnormal cytology at enrollment, HPV16, but not HPV18, was associated with a relatively high risk of high-grade CIN compared with other hrHPV types. In the second study (12), baseline smears of a routine screening population were typed for HPV16 and HPV18 and elevated risks were reported for both types. Together, the HPV types 16, 18, 31, and 33 are the most prevalent ones in high-grade CIN (19). Our data suggest that the high prevalence of these types in high-grade CIN is likely to be associated with a type-specific progression risk.

In HPV16/18–positive women, having baseline BMD instead of normal cytology was significantly less predictive for the risk of high-grade CIN than in other hrHPV-positive women. The relatively weak effect of BMD was the most pronounced in HPV18-positive women for whom the relative risk of ≥CIN3 when comparing baseline BMD to normal cytology was equal to 1.0. A plausible explanation for this observation is that HPV16/18–positive women also have a moderate risk of high-grade CIN when the smear is cytologically normal, and the added predictive value of cytology therefore is smaller for these women. We also found that after multiple normal smears, the ≥CIN2 risk was elevated in women positive for HPV16/18 at baseline and that the ≥CIN3 risk was elevated in women positive for HPV18. Apparently, a substantial portion of HPV16-positive and particularly HPV18-positive CIN cases were initially missed by cytology. A discrepancy between cytologic reading and underlying histology of HPV18-positive infections was also found in a study of young women ages 15 to 19 years (20).

Compared with high-risk HPV-positive women without HPV16 or HPV18, the elevated risk of ≥CIN2 after multiple normal smears for women with HPV16 and/or HPV18 found in our study warrants a more aggressive management of these women. Because HPV18 is the most prevalent type in adenocarcinoma and HPV16 and HPV18 are responsible for the majority of invasive cervical cancers (8, 21), further endocervical inspection including for instance endocervical curettage may be justified for HPV16/18–positive women when the transformation zone is found normal by colposcopy (22).

Some methodologic issues need to be discussed. First, women belonging to the three categories (i.e., intervention group-normal cytology, intervention group-BMD, and control group-BMD) were subjected to slightly different referral strategies. However, it is unlikely that this led to detection bias in the intervention group as women in that group only returned to routine screening when the hrHPV infection cleared and the cytologic result was normal or BMD. For women in the control group with BMD, management was not based on hrHPV DNA because the result of the hrHPV test was blinded during follow-up and cytologically poorly accessible lesions might have been missed. However, distributions of histologically diagnosed lesions in women with BMD of the control and intervention group were very similar; thus, it is not likely that results were substantially affected by the referral strategies. Second, 23% of the women with normal cytology and 10% of the women with BMD were lost to follow-up at 6 months and nearly 30% of the women did not show up at the second invitation at 18 months. We accounted for loss to follow-up by applying Kaplan-Meier censored analyses. This may invalidate the results when censoring is HPV type specific, which cannot be excluded beforehand. However, we did not find an association between HPV type and censoring. Third, HPV types 59 and 68 were relatively rare; thus, it would have been difficult to find an increased risk of high-grade CIN for any of these two types. However, the four types which were significantly associated with risk of high-grade CIN also had a relatively high absolute ≥CIN2 and ≥CIN3 risk and therefore were not merely singled out because they were the most common ones.

An important strength of our prospective trial study of 44,102 women is that the included women had age ≥30 years and were eligible for screening so that the effectiveness of HPV typing in current screening can be directly assessed. In our study, the prevalence of HPV16- and HPV18-positive normal smears was 37% (262 of 713) and the prevalence of HPV16-, HPV31-, and HPV33-positive BMD smears was 52% (195 of 374). Hence, both after normal cytology and after BMD, HPV typing can be used to identify subgroups of hrHPV-positive women for which different follow-up algorithms may lead to an improvement in screening management.

In conclusion, in hrHPV-positive women, we identified four hrHPV types that were associated with a substantially increased risk of high-grade CIN. Adjunct hrHPV typing therefore enables distinguishing risk classes for high-grade CIN and may lead to considerable improvements in screening management. Identification of HPV16 or HPV18 is of utmost importance because women with a persistent infection with one of these types are at risk of prevalence of incipient high-grade CIN also when these are not detected by cytology.

K. van Groningen (Spaarne Ziekenhuis, Heemstede), W. Ruitinga (Stichting PA Laboratorium Kennemerland, Haarlem), M.E. Boon (Leiden Cytology and Pathology Laboratory, Leiden), M. van Ballegooijen (Department of Public Health and Social Medicine, Erasmus University Rotterdam), A.J.P. Boeke (Institute for Research in Extramural Medicine, Vrije University Medical Centre, Amsterdam), Prof. R.H.M. Verheijen (Department of Obstetrics and Gynaecology, Vrije University Medical Centre, Amsterdam), and F. van Kemenade and L. Rozendaal (Unit cytopathology, Vrije University Medical Centre, Amsterdam).

We thank the 242 GPs and their assistants, the District Health Authority Amstelveen, Medial, and DHV Kennemerland-Haarlemmermeer e.o.; the research analysts of the Unit Molecular Pathology, Vrije University Medical Centre, Amsterdam for hrHPV testing and HPV typing; and the cytotechnologists (Spaarne Ziekenhuis, Heemstede; Stichting PA Laboratorium Kennemerland, Haarlem; Leiden Cytology and Pathology Laboratory, Leiden; Unit Cytopathology, Vrije University Medical Centre, Amsterdam) for cytologic testing.