Risk of Precancer Determined by HPV Genotype Combinations in Women with Minor Cytologic Abnormalities

Persistent infection with carcinogenic human papillomaviruses (HPV) is a necessary cause of cervical cancer. It has been shown that the risk of progression to cancer varies substantially between individual carcinogenic types, and infections with HPV16 are associated with elevated risk of high-grade lesions [cervical intraepithelial neoplasia grade 3 (CIN3); refs. 1–5]. In recently updated cervical cancer screening guidelines, testing for HPV16 and 18 (the second most important HPV type and a major cause of adenocarcinoma) has been recommended as a triage for HPV-infected, cytology-negative women (6). It has also been suggested that HPV genotyping could be used for management of women with minor cytologic abnormalities [atypical squamous cells of undetermined significance (ASCUS) and low-grade squamous intraepithelial lesions (LSIL)] to decide who should be referred to immediate versus delayed colposcopy (5). Furthermore, it has been proposed that HPV45 be added to HPV genotyping assays. Following up on previous studies conducted in ALTS that showed the strong risk of precancer associated with HPV16 (5), we compared the risk stratification achieved with HPV genotyping combinations 16 versus 16/18 versus 16/18/45 in HPV-positive women with ASCUS and LSIL.

The ASCUS–LSIL Triage Study (ALTS) was a randomized trial directed by the National Cancer Institute (NCI; NIH, Bethesda, MD) that compared 3 triage strategies for women with ASCUS or LSIL. Details of the design, methods, and primary results of ALTS have been published extensively elsewhere (7–9). Briefly, women with ASCUS or LSIL cytology were recruited to participate in the study at 4 clinical centers. A total of 5,060 women enrolled in the study from January 1997 to December 1998. At enrollment, the ALTS participants were referred to colposcopy depending on the study arm. In the immediate colposcopy arm, all women had colposcopy at, or soon after, the enrollment regardless of enrollment test results. In the HPV triage arm, women were referred to colposcopy if the enrollment HPV test was positive (56.4%) or missing (3.9%), or if the enrollment cytology was high-grade squamous intraepithelial lesion (HSIL), although cytology added almost no referrals. In the conservative management arm, women were referred to colposcopy if enrollment cytology was interpreted as HSIL. The ALTS participants were followed at 6-month intervals for 2 years. At the semiannual follow-up visits, regardless of randomization arm, colposcopic examinations were triggered only by HSIL cytology. At the exit visit, all women were scheduled for a colposcopic examination. Throughout the trial, women with histologic diagnosis of CIN grade 2 or worse (CIN2+) as defined by the clinical center pathologists were treated by loop electrosurgical excision procedure (LEEP) or more extensive surgery if needed. At exit, women with persistent lower-grade lesions as well were offered LEEP to maximize safety after follow-up ended. NCI and local Institutional review boards approved the study.

The enrollment examination included a pelvic exam with the collection of cells for cytology and HPV DNA testing as well as high-resolution photography of the cervix for visual screening (Cervicography, National Testing Laboratories). After liquid-based ThinPrep (Cytyc Corporation) cytology, slides were prepared and 4-mL aliquots of the residual PreservCyt samples were used for HPV DNA testing by Hybrid Capture 2 (Qiagen Corporation). As mentioned, clinical management was based on the clinical center pathologists' cytologic and histologic diagnoses. In addition, all cytology and histology slides were sent to the Pathology Quality Control Group for independent review. Pathology Quality Control histologic diagnoses were masked to cytology results and used in this data analysis to avoid center-specific variation.

Among women with ASCUS cytology, aliquots of the archived standard transport medium (STM) specimens were tested as previously described (10) using Linear Array (Roche Molecular Systems), an L1-based PCR assay that uses a primer set designated PGMY09/11. Among women with LSIL, HPV genotyping on aliquots of the STM specimen was conducted by Line Blot Assay (LBA; Roche Molecular Systems), a research-use-only version of Linear Array (11, 12). The 2 assays, Linear Array and LBA have shown comparable performance in these specimens (10). Women were considered infected with a carcinogenic HPV if they tested positive to one or more of the 13 carcinogenic HPV genotypes (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 68; ref. 13) plus HPV66. Although HPV66 is classified as only “possibly carcinogenic” (13) historically, it has been combined with other carcinogenic HPV genotypes in pooled probe HPV assays and is therefore included in this analysis.

We compared positivity and risk stratification among carcinogenic HPV-positive women for 3 combinations of genotypes: (i) HPV16, (ii) HPV16 and HPV18, and (iii) HPV 16, 18, and 45. For each group, we calculated the cumulative risk given any positive versus negative genotype results over 2 years follow-up of CIN2+ by clinical center diagnosis and CIN3 and cancer (CIN3+) by Pathology Quality Control. Specifically, we compared HPV16 with the other 13 types as a group, HPV16 or HPV18 compared with the other 12 types as a group, and HPV16 or HPV18 or HPV45 compared with the other 11 types as a group. Results were stratified by enrollment cytology result (ASCUS or LSIL) and age (< 30 vs. 30 or older). We examined whether the findings differed by study arm. Also, because women aged 18 to 21 years are no longer recommended for routine screening, we considered whether excluding them from the analysis would change our findings (6). Supplementary Figures present ancillary analyses using the outcome of detection of CIN2+ by clinical center and CIN3+ by Pathology Quality Control at enrollment as well as the risk for all women (HPV-positive and HPV-negative) enrolled in ALTS. Positivity was compared using probability tests of significance. Risks of precancer were compared using χ² test statistics.

Overall, 55.6% of women with ASCUS cytology (Table 1) and 73.5% of women with LSIL cytology (Table 2) were positive for any carcinogenic HPV type. HPV16 positivity ranged from 8.0% among women 30 years and older with ASCUS cytology (Table 1) to 22.5% among women less than 30 years with LSIL cytology (Table 2). For women with ASCUS cytology, the addition of HPV18 compared with HPV16 alone resulted in almost one-third more referrals among both women younger than 30 years (28.0% vs. 21.9%, P < 0.01) and women 30 years or older (10.8% vs. 8.0%, P = 0.01; Table 1). Adding HPV45 slightly increased the overall referral rate to 32.2% among women younger than 30 years (P < 0.01) and 13.4% among women aged 30 years and older (P = 0.02). For women with LSIL cytology, the incremental positivity for adding HPV18 and HPV45 to HPV16 was similar to what was observed among women with ASCUS.

To show the extent of risk stratification provided by genotyping carcinogenic HPV-positive women, we plotted the absolute 2-year risk of CIN2+ and CIN3+ among women testing positive and negative for the different genotype combinations, stratified by referral cytology. Figure 1 shows women aged less than 30 years, whereas Fig. 2 shows women aged 30 years or older. In all strata, HPV16 showed the greatest difference between risk in test positives and test negatives, a measure of risk stratification. Adding HPV18 and HPV45 decreased the risk in women testing positive, whereas the risk in test negatives remained largely unchanged. Notably, for women aged 30 years or older and HPV16 positive, the 2-year risk of a CIN3+ diagnosis was 48.6% [95% confidence interval (CI), 32.5–64.8], an elevated risk possibly warranting consideration of treatment in certain populations with limited access to follow-up.

Positivity tended to confer equal risk of high-grade neoplasia across age strata (Figs. 1 and 2). Although the absolute risk for CIN3+ did not differ by age for women with ASCUS, for women with LSIL the CIN3+ risks given HPV16 positivity was higher, albeit nonsignificantly, in older versus younger women (48.6% in women aged 30 years and more vs. 37.9% in women under 30 years, P = 0.21). The risks were similar by age when adding HPV18 (P = 0.78), HPV45 (P = 0.94), and all carcinogenic types (P = 0.92).

Analyses with the CIN2+ endpoint (Figs. 1 and 2) and those restricting to endpoints detected at enrollment (Supplementary Figs. S1 and S2) showed similar findings. In addition, results were similar across study arms and when limiting to women aged 21 years and older. When carcinogenic HPV-negative women were included in the genotype-negative strata, the risk among genotype-negative women was much lower (Supplementary Figs. S3 and S4), potentially warranting delayed colposcopy. However, it is unlikely that HPV genotyping would be conducted without prior carcinogenic HPV testing in clinical practice; therefore, we did not expand on this analysis here.

Testing for HPV genotypes has recently been integrated in cervical cancer screening algorithms for HPV-positive/cytology-negative women. The approach has also been proposed for management of women with minor cytologic abnormalities. In this study, we analyzed the risk stratification provided by HPV genotyping for 3 different type combinations HPV16, HPV16/18, and HPV16/18/45 in women with ASCUS and LSIL referral cytology. We observed substantial risk stratification for 2-year cumulative CIN3+ for all 3 genotype combinations that was highest for HPV16 alone and decreased with adding additional genotypes. For example, among women aged 30 years and older with HPV-positive/ASCUS, the risk difference between HPV16-positive and HPV16-negative women reached 21%, whereas the risk difference between HPV16/18/45-positive and HPV16/18/45-negative women was 13%.

To be useful for clinical management, the risk estimates provided by HPV genotyping need to lead to different management decisions, such as immediate referral to colposcopy versus delayed colposcopy or repeat cytology/HPV testing. A previous analysis of ALTS data considered a 10% 2-year risk of CIN3+ to be a threshold for referral to colposcopy (14). As we show here, women with ASCUS cytology who tested negative for HPV16, HPV16/18, or HPV16/18/45 (but still positive for the remaining carcinogenic types) had a risk between 7.9 and 8.8, with the CIs including the 10% threshold (Figs. 1 and 2). Likewise, the risk in women with LSIL referral cytology was 10% or higher when they tested negative for the different genotyping combinations.

Despite the impressive risk stratification of genotyping in our study, the risk among women testing negative for the genotypes was not low enough to warrant delaying colposcopy evaluation, if currently accepted risk thresholds are applied (14). Thus, according to risk thresholds established in ALTS, both genotype-positive and -negative women would be referred to immediate colposcopy. Although the absolute risk varies in other populations, it is likely that the relative differences would result in similar management recommendations.

Until recently, the only U.S. Food and Drug Administration (FDA)-approved HPV test was Hybrid Capture 2 (Qiagen Corp.), a pooled probe test that detects the presence of one or more of 13 HPV DNA genotypes. During the past years, the FDA approved additional HPV DNA tests that provide both testing for pooled carcinogenic types as well as individual detection of the 2 most carcinogenic HPV genotypes, HPV16 and HPV18 (cobas HPV test; Roche Molecular Systems; ref. 15 and Cervista; Hologic; refs. 16, 17).

The test performance for HPV16 and 18 using enrollment colposcopically guided biopsy results mirrors that of the recent ATHENA trial. In the ATHENA trial where women with an ASCUS cytology were enrolled, the absolute risk of CIN2+ and CIN3+ at enrollment was 24.4% (95% CI, 18.7%–31.3%) and 15.9% (95% CI, 11.2%–22.0%), respectively (15). The absolute risks of CIN2+ and CIN3+ among women testing positive for carcinogenic HPV but negative for both HPV16 and HPV18 were 8.6% (95% CI, 6.0%–12.1%) and 4.4% (2.7%–7.2%), respectively. These risks are similar to the those observed among women in ALTS with ASCUS cytology and enrolled in the immediate colposcopy or HPV triage arms: testing HPV16 or 18 positive conferred a 25.8% (95% CI, 22.0%–29.9%) and 15.1% (95% CI, 12.1%–18.6%) risk of CIN2+ and CIN3+, respectively, at enrollment, whereas testing carcinogenic HPV positive but negative for both HPV16 and HPV18 conferred a 10.6% (95% CI, 8.5%–13.0%) and 3.8% (95% CI, 2.6%–5.5%) risk for CIN2+ and CIN3+, respectively, at enrollment. The risks for precancer at enrollment reported in a study of the Cervista HPV16/18 test were lower among women testing HPV16 or 18 positive at 17.1% (95% CI, 13.0%–22.2%) and 6.6% (95% CI, 4.2%–10.3%) risk of CIN2+ and CIN3+, respectively (16). Similarly, the risks when testing negative for HPV16 or HPV18, but positive for carcinogenic HPV by Cervista, were lower at 4.0% (95% CI, 2.5%–6.1%) and 1.0% (95% CI, 0.3%–2.3%) of CIN2+ and CIN3+, respectively. Both, overall differences in the population or different HPV genotyping test characteristics could explain the lower absolute risk observed in the Cervista trial.

The addition of carcinogenic types HPV 18 and HPV45 to HPV16 as a possible triage test for ASCUS and LSIL cytology results resulted in higher positivity and reduced risk stratification. In general, our analysis did not suggest a benefit of adding HPV45 to genotyping assays because the risk among test positives was reduced compared with HPV16 and HPV16/18 without providing greater reassurance among test negatives. Although the incremental benefit of adding HPV18 was also limited in our study, HPV18 is the second most common genotype in cervical cancers, and therefore, warrants inclusion in genotyping assays, despite the noted deficit of cervical precancers associated with this type.

We found that changes in positivity and predictive value were not uniform across age groups. Women younger than 30 years with ASCUS were more likely to test positive across all HPV genotype combinations although the age-stratified absolute risk for CIN3+ was statistically similar. The noted exception was in women with LSIL cytology where the risk associated with HPV16 alone was nonsignificantly higher in older versus younger women (age < 30).

We showed that HPV genotyping provides substantial additional risk stratification among HPV-positive women with mild cytologic abnormalities. However, in ALTS, the risk in women testing negative for the HPV genotypes was close to or above the threshold that is widely considered for referral to colposcopy. We confirmed that HPV16 provides the biggest risk stratification, whereas the additional benefit of HPV18 and HPV45 is limited. Although the high attribution of HPV18 to invasive cancers warrants inclusion of that type in typing assays, there does not seem to be benefit of including HPV45 in genotyping assays. It is possible that testing for HPV45 offers some additional sensitivity to detect adenocarcinomas, although this study was not sufficiently powered to address this question and the tradeoff is a much higher referral rate and lower positive predictive value.

J.C. Gage has commercial research support from Qiagen Gaithersburg and Roche Corp. C.M. Wheeler has commercial research support from Roche Molecular Systems and GSK. P.E. Gravitt is a consultant/advisory board member of Qiagen. P.E. Castle is employed as an executive director in Global Cancer Initiative and Immunexpress, is a director of clinical research, has honoraria from speakers' bureau from Roche, has ownership interest (including patents) in OncoHealth, and is a consultant/advisory board member of BD, GE Healthcare, Cepheid, and Merck. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J.C. Gage, N. Wentzensen

Development of methodology: J.C. Gage, C.M. Wheeler, N. Wentzensen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Schiffman, D. Solomon, C.M. Wheeler, P.E. Gravitt, P.E. Castle

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.C. Gage, M. Schiffman, P.E. Gravitt, P.E. Castle, N. Wentzensen

Writing, review, and/or revision of the manuscript: J.C. Gage, M. Schiffman, D. Solomon, C.M. Wheeler, P.E. Gravitt, P.E. Castle, N. Wentzensen

Study supervision: M. Schiffman, D. Solomon

The authors thank ALTS Group Investigators for their help in planning and conducting the trial.

This work is supported by the intramural program of the Department of Health and Human Services, National Cancer Institute, NIH, and contracts CN-55153, CN-55154, CN-55155, CN-55156, CN-55157, CN-55158, CN-55159, and CN-55105.

Some of the equipment and supplies used in this study were donated or provided at reduced cost by Digene Corporation, Cytyc Corporation, National Testing Laboratories, Denvu, TriPath Imaging, Inc., and Roche Molecular Systems, Inc. C.M. Wheeler received funding through the University of New Mexico from Merck and Co., Inc. and GlaxoSmithKline for HPV vaccine studies and equipment and reagents from Roche Molecular Systems for HPV genotyping.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.