The ATHENA (Addressing THE Need for Advanced HPV Diagnostics) HPV study evaluated the clinical usefulness of the cobas HPV Test (Roche Molecular Systems, Pleasanton, CA) for high-risk human papillomavirus (HR-HPV) testing (14 HR types) and individual HPV-16/HPV-18 genotyping in women undergoing routine cervical cytology screening in the United States. For the study, 47,208 women were recruited, including 32,260 women 30 years or older with negative cytology. All women with positive results for HR-HPV (n = 4,219) plus a subset of HR-HPV– women (n = 886) were referred for colposcopy and biopsy. The overall prevalence of HR-HPV was 6.7% and of HPV-16/HPV-18 was 1.5%. Cervical intraepithelial neoplasia grade 2 (CIN 2) or worse was found in 1.2% of women examined. The estimated absolute risk of CIN 2 or worse in HPV-16+ and/or HPV-18+ women was 11.4% (95% confidence interval [CI], 8.4%–14.8%) compared with 6.1% (95% CI, 4.9%–7.2%) in HR-HPV+ and 0.8% (95% CI, 0.3%–1.5%) in HR-HPV– women. These analyses validate the 2006 American Society of Colposcopy and Cervical Pathology guidelines for HPV-16/HPV-18 genotyping, which recommend referral to colposcopy of HPV-16/HPV-18+ women with negative cytology.
Almost all high-grade cervical cancer precursors (cervical intraepithelial neoplasia [CIN] grades 2 and 3 [CIN 2 and 3]) and invasive cervical carcinomas are caused by persistent infection with 1 of 14 high-risk genotypes of human papillomavirus (HR-HPV).1,2 This causal relationship has led to the introduction of sensitive molecular tests for HR-HPV into clinical practice, and testing for HR-HPV is now routinely used in the United States for managing women with mild cervical cytologic abnormalities (atypical squamous cells of undetermined significance [ASC-US]).
HR-HPV testing is also used as an adjunct to cervical cytology when screening women 30 years or older.3 However, when used for screening, the clinical usefulness of HR-HPV testing is limited by the fact that HR-HPV infections are relatively common among women without CIN 2, CIN 3, or invasive cervical cancer (CIN 2 or worse). Rather than referring all HR-HPV+ women 30 years or older with negative cytology (negative for intraepithelial lesions or malignancy [NILM]) for colposcopy, current management guidelines recommend repeating the cervical cytology and HR-HPV test in 12 months.3 However, this conservative follow-up approach is somewhat problematic because it negates much of the benefit of the high clinical sensitivity of HR-HPV testing and can lead to a significant delay in the treatment of women with cervical cancer and a false-negative cervical cytologic result. Moreover, this strategy creates uncertainty for women who are HR-HPV+ and are deferred to follow-up rather than having information that guides immediate investigation for high-grade disease. These limitations may explain, in part, the relatively low utilization of HR-HPV testing by clinicians in the United States when screening women 30 years or older.4,5
One approach to limiting the number of HR-HPV+ referrals would be to stratify (ie, triage) women with positive results for HR-HPV so that only women at greatest risk for CIN 2 or worse would undergo colposcopy and intensive follow-up. It is well recognized that among the 14 HR-HPV genotypes, HPV-16 and HPV-18 confer the greatest risk for CIN 2 or worse because these 2 genotypes are associated with approximately two thirds of all invasive cervical carcinomas.6,7 This suggests that genotyping for HPV-16 and/or HPV-18 could be used to identify a subset of HR-HPV+ women with a particularly elevated risk of having CIN 2 or worse. Although the clinical usefulness of using genotyping for HPV-16 and/or HPV-18 to triage HR-HPV+ women with NILM cytologic results has been recently recognized in clinical management guidelines, the data supporting the use of HPV genotyping in this manner are relatively limited.8
The ATHENA (Addressing THE Need for Advanced HPV Diagnostics) HPV study enrolled women 21 years or older undergoing routine cervical cytology screening in the United States. The trial was designed to evaluate the performance of a polymerase chain reaction–based HPV assay (cobas HPV Test, Roche Molecular Systems, Pleasanton, CA) that detects 14 HR-HPV genotypes and provides 3 separate results: a pooled result for 12 HR-HPV genotypes (31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68) and individual genotyping results for HPV-16 and HPV-18.
Materials and Methods
The ATHENA study is a prospective trial designed to evaluate the performance of a new HR-HPV test (cobas HPV Test) among women undergoing routine cervical cancer screening in the United States. The trial is being conducted in 2 phases: a baseline (cross-sectional) phase and a 3-year follow-up (longitudinal) phase that is due to be completed in December 2012. Only data from the completed cross-sectional phase are reported here. The primary objective of the ATHENA study was to evaluate the performance of the cobas HPV Test in women 21 years or older with ASC-US cytology, and results are reported elsewhere.9 Preplanned secondary objectives included a comparison of the risks of CIN 2 or worse among women 30 years or older with NILM cytology and positive for HPV-16 and/or HPV-18 with the risks for women who were positive for the 12 other HR-HPV genotypes or were HR-HPV–. The current analysis is restricted to the subset of women 30 years or older with NILM cytology. The protocol was approved by the institutional review board of all study sites, and all women provided written informed consent before undergoing any study procedures.
In total, 47,208 nonpregnant women 21 years or older undergoing routine cervical cancer screening were enrolled between May 2008 and August 2009 at 61 clinical centers in 23 states across the United States. Of these women, 46,887 (99.3%) met the study inclusion/exclusion criteria: intact uterus, willing to undergo colposcopy and biopsy if required, no treatment for CIN in the preceding 12 months, and no current or planned participation in a clinical trial for HPV treatment.
Study Visit 1: Enrollment Visit, All Participants
After informed consent and demographic and medical history were obtained, 2 liquid-based cervical cytology samples (ThinPrep, Hologic, Bedford, MA) were obtained from the cervix of each participant using a spatula and an endocervical brush. One sample was used for liquid-based cytology and for HR-HPV testing. The second liquid-based cytology sample was reserved for additional testing.
Study Visit 2: Colposcopy Visit, Selected Participants
Based on the results of the cervical cytology and HR-HPV testing performed on cervical samples obtained at the enrollment visit, a subset of the enrollees was selected for colposcopy as follows. Before reporting screening test results to the clinical centers, results were entered into a subject selection and randomization database that generated a list of women selected for colposcopy. This subset (n = 5,726) included all women 25 years or older with NILM cervical cytology and a positive HR-HPV test by first-generation Roche HPV tests (ie, positive for HR-HPV by the AMPLICOR HPV Test and/or LINEAR ARRAY High Risk HPV Genotyping Test [LA], Roche Molecular Systems, Pleasanton, CA) and 1,041 randomly selected women 25 years or older with NILM cytology who were negative for HPV by the AMPLICOR and LA tests. Randomization was done in a block size of 35 using SAS software, version 9.1.3 (SAS Institute, Cary, NC). Results of the second-generation Roche HPV test (cobas HPV Test) were not used to select women for colposcopy because the test cutoff value had not been finalized at the time that enrollment into the ATHENA study was initiated.
Colposcopy with biopsy and/or endocervical curettage (ECC) was performed in nonpregnant women selected for colposcopy within 12 weeks of the enrollment visit according to a standardized protocol that included biopsy of all cervical lesions and a random biopsy sample at the squamocolumnar junction in women without cervical lesions and a satisfactory colposcopy. All women with an unsatisfactory colposcopy underwent ECC. Participants and colposcopists were blinded to enrollment visit test results until the colposcopy visit was completed. Biopsies and ECCs were reviewed in a blinded manner by a consensus panel of 3 gynecologic pathologists and diagnosed using standard criteria and the CIN terminology, as previously described in detail.9,10 The consensus panel diagnosis was used for all analyses. Women who underwent colposcopy and did not meet the primary study end point of histologically diagnosed CIN 2 or worse by consensus pathology were eligible for the follow-up phase (not described herein).
Cytology and HR-HPV Testing
Sample processing and testing were performed at 5 clinical laboratories in the United States. Two 2-mL aliquots were removed from the first liquid-based cervical cytology specimen (PreservCyt Solution, Hologic) before preparation of the cytology slide. One 2-mL aliquot was used for HR-HPV testing with first-generation HPV tests (both AMPLICOR HPV Test and LA test) according to the manufacturer’s instructions. The AMPLICOR HPV Test detects 13 HR-HPV types (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 68; cutoff, ≥0.2 optical density), and the LA test detects 16 HR-HPV types (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, 68, 73, and 82). The other 2-mL aliquot was tested at 1 of 5 laboratories using the second-generation HPV test (cobas HPV Test) that detects 12 pooled HR-HPV genotypes (31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68) and HPV-16 and HPV-18 individually and includes a separate β-globin control. Cycle threshold cutoff values for the cobas HPV Test were established by using samples from the first approximately 29,000 women enrolled in the ATHENA study; subsequent cross-validation of the test cutoff was achieved by using samples from the remaining approximately 18,000 participants.
The remainder of the liquid-based cervical cytology specimen was used to prepare a slide for cytologic evaluation (ThinPrep Pap Test, Hologic) and processed and evaluated without computerized imaging, according to the standard procedures in 4 clinical laboratories and reported using the Bethesda System.11
Only the data for women 30 years or older with a valid NILM cytology result and valid results for all 3 HPV assays (AMPLICOR, LA, and cobas HPV Test) are included in this analysis. The absolute risks and relative risks of high-grade cervical disease with the respective 95% confidence intervals (CIs) were determined for different categories of HPV DNA result (by cobas HPV Test) in women with NILM cytology. HPV results of genotypes 16 and 18 were analyzed as individual results and as combined 16/18 results (genotype 16 and/or 18). Genotype 16+ results included cases positive for genotype 16 alone, with or without genotype 18, and with or without the 12 other HR-HPV types present. Genotype 18+ results included cases positive for genotype 18 alone, with or without a positive result for the 12 other HR-HPV genotypes, and cases negative for genotype 16. The 12 other HR-HPV+ samples were positive only for these 12 HR-HPV types. Verification bias–adjusted estimates for absolute risk and relative risk in women with NILM cytology were obtained by calculating the projected number of women with and without disease who would have been found if all women had undergone colposcopy. CIs for absolute and relative risks were estimated by bootstrapping (1,000 times).12
Enrollment and Demographics
Among the 47,208 subjects enrolled into the trial, a total of 32,260 were 30 years or older and had valid HPV test results and an NILM cytology result at their enrollment visit. Of these women, 4,219 were classified as HR-HPV+ using the first-generation HPV tests and referred for colposcopy on that basis. An additional 886 women with NILM cytology who were HR-HPV– were randomized to undergo colposcopy to adjust for verification bias Figure 1. Of the 5,105 women referred for colposcopy, 4,422 (86.6%) underwent the procedure and 4,258 (83.4%) had valid results on their cervical biopsies/ECC. Among all women referred for colposcopy, the same percentage (86.6%) from each HPV group (HPV+ and women randomized from the HPV– group) completed the colposcopy visit. The mean age of the women in the NILM population included in the analysis was 44.9 years Table 1. Approximately one third were 30 to 39 years old, and 37.5% were postmenopausal. Most women (90.8%) had undergone cervical cytologic screening within the past 5 years.
Prevalence of HR-HPV
Among the 32,260 women 30 years or older with NILM cytology, the overall prevalence of HR-HPV (14 types) detected using the cobas HPV Test was 6.7% and the overall prevalence rates for HPV-16, HPV-18, and 12 other HR-HPV genotypes were 1.0%, 0.5%, and 5.2%, respectively Table 2. The overall prevalence of HR-HPV (14 types), as well as HPV-16 and HPV-18 individually, decreased with increasing age. Among women 50 to 59 years old, the prevalence of HR-HPV positivity was 5.3% and the prevalence of HPV-16 and HPV-18 combined was only approximately 1% (Table 2).
Absolute Risks of CIN by HR-HPV Test Result
Referral of HR-HPV+ women with NILM cytology was based on HPV status determined by using the first-generation HPV tests (AMPLICOR and LA tests), whereas the absolute risk of cervical disease was estimated based on the results obtained with the second-generation cobas HPV Test. It should be noted that the AMPLICOR test is more sensitive than the cobas HPV Test from an analytic perspective, as indicated by limit of detection values expressed as cells per milliliter.13,14 Thus, among the 4,258 women who underwent colposcopy and had a valid biopsy result, only 1,744 women were HR-HPV+ using the cobas HPV Test and the remaining 2,514 women were HR-HPV–. Of note, among women with negative results by the cobas HPV Test and who had combined AMPLICOR and LA test positive results, the crude prevalence of CIN 2 or worse was comparable with the prevalence of CIN 2 or worse for women with negative results on the AMPLICOR and LA tests (0.9% and 0.8%, respectively), thereby confirming the clinical validation of the appropriate cutoff for the cobas HPV Test.
Based on the consensus pathology review diagnosis, the samples for 3,898 of 4,258 evaluable women who underwent colposcopy were classified as within normal limits, while 229 had biopsy-confirmed CIN 1, 51 had biopsy-confirmed CIN 2, and 80 had biopsy-confirmed CIN 3 or worse. Cases classified as CIN 3 or worse included 5 cases of adenocarcinoma in situ. Verification bias–adjusted estimates of the overall absolute risk among women 30 years or older with NILM cytology at ATHENA enrollment were 1.2% (95% CI, 0.6%–1.8%) for CIN 2 or worse and 0.5% (95% CI, 0.3%–0.9%) for CIN 3 or worse Table 3. The estimated absolute risk was dependent on HR-HPV status, and HPV-16+ women had the highest absolute risks for CIN 2 or worse and CIN 3 or worse. The estimated absolute risk for CIN 3 or worse ranged from 0.3% (95% CI, 0.02%–0.7%) in women who were negative for HR-HPV to 11.7% (95% CI, 7.9%–15.8%) in women who were positive for HPV-16.
Estimated absolute risks for all cobas HPV Test results stratified by age are shown in Table 3. Absolute risk in women who were HR-HPV+ showed a modest decrease with age across all combinations of results. However, the decrease in risk was more marked among women who were HPV-16+ and/or HPV-18+ compared with women who were not HPV-16+/HPV-18+ (ie, positive for 12 other HR-HPV types).
Relative Risks of CIN by HR-HPV Test Result
Verification bias–adjusted estimates for relative risk of CIN 2 or worse and CIN 3 or worse by HR-HPV status in women 30 years or older with NILM cytology are provided in Table 4. The relative risks for CIN 2 or worse and CIN 3 or worse were highest in women found to be HPV-16+ using the cobas HPV Test; the relative risk for CIN 2 or worse was 16.3 (95% CI, 8.2–51.3) and for CIN 3 or worse was 42.0 (95% CI, 15.5–695.3), compared with HR-HPV– women. Furthermore, compared with HR-HPV– women, the relative risks for CIN 2 or worse and CIN 3 or worse were intermediate in HPV-18+ women: 8.4 (95% CI, 2.9–26.6) and 20.5 (95% CI, 4.3–372.4), respectively. The relative risks were also intermediate in women with positive results for 12 other HR-HPV types: 5.5 (95% CI, 2.8–16.4) and 8.7 (95% CI, 3.2–148.0), respectively.
The ATHENA study is the first trial to evaluate the clinical usefulness of HR-HPV DNA testing and HPV-16/HPV-18 genotyping among a large cohort of women undergoing routine cervical cytologic screening in the United States. The current analysis focused on women 30 years or older with NILM cytology results. The main findings are that the absolute and relative risks of CIN 2 or worse and CIN 3 or worse in women with NILM cytology are impacted not only by HR-HPV (14 types) status, but even more so by whether HPV-16 or HPV-18 is detected.
Prior cross-sectional15–17 and prospective studies18–20 have clearly demonstrated that HR-HPV status is an important predictor of the current and future detection of CIN 2 or worse in women with NILM cytology. Based on this finding, current US cervical cancer screening guidelines recommend extending the screening interval to at least 3 years in women 30 years or older with NILM cytology who are HR-HPV– and retesting women who are HR-HPV+ at 12 months using cervical cytology and HPV tests. Women who are HPV– with low-grade squamous intraepithelial lesions or greater on the repeated cytology and women who are persistently HR-HPV+ should be referred for colposcopy.3 Therefore, it is not unexpected that in the current analysis, HR-HPV status had a marked impact on the risks of CIN 2 or worse and CIN 3 or worse in women 30 years or older with NILM cytology; however, data have not been previously reported for HPV-16/HPV-18.
In women with NILM cytology who were also HR-HPV+ (14 types), the estimated absolute risk of CIN 3 or worse at baseline was 4.1% (95% CI, 3.1%–5.0%), and the relative risk of CIN 3 or worse was 14.4 (95% CI, 5.7–227.9) compared with women who were HR-HPV–. The presence of HPV-16 or HPV-18 in women 30 years or older with NILM cytology was found to greatly magnify the risk of CIN 3 or worse, as did the presence of HPV-16 alone. The absolute risk for CIN 3 or worse in women with NILM cytology who were HPV-16+ was 11.7% (95% CI, 7.9%–15.8%); the relative risk of CIN 3 or worse compared with women who were HR-HPV– (14 types) was 42.0 (95% CI, 15.5–695.3). The absolute risk in women who were HR-HPV+ showed a modest decrease with age across all combinations of results, and the decrease was more marked among women who were HPV-16+ and/or HPV-18+ compared with women who were not HPV-16+/HPV-18+ (ie, positive for 12 other HR-HPV types). The decrease in risk of high-grade disease is consistent with the reported peak age of CIN 3 in the United States (25–29 years)21 and was also observed in the overall population of the ATHENA trial.22
Two prospective follow-up studies have shown the importance of HPV-16 in predicting the women with NILM cytology in whom high-grade CIN will develop. In the study by Khan et al,23 in women enrolled in the Kaiser Permanente health management organization (Portland, OR), the 10-year cumulative incidence rate (CIR) of CIN 3 or worse in women 30 years or older with NILM cytology at enrollment was 20.7% (95% CI, 8.6%-32.8%) among HPV-16+ women and 17.7% (95% CI, 0.0%-36.0%) among HPV-18+ women. In contrast, the 10-year CIR of CIN 3 or worse was only 1.5% (95% CI, 0.3%–2.7%) among HR-HPV+ women who were negative for both HPV-16 and HPV-18.23 In the Kaiser study, the increased detection of CIN 3 or worse in women who were HPV-18+ was delayed by approximately 2 years compared with the increase seen in women who were HPV-16+.23 A 2-year delay in the oncogenic effects of HPV-18 compared with HPV-16 could explain the somewhat lower impact of HPV-18 compared with HPV-16 observed in the current cross-sectional analysis of the ATHENA trial.
Similar data have recently been reported from a Danish prospective follow-up study. After a 12-year follow-up, the estimated probability of developing CIN 3 or worse in HPV-16+ women with NILM cytology was 26.7% (95% CI, 21.1%–31.8%), and for HPV-18+ women, it was 19.1% (95% CI, 10.4%–27.3%). In contrast with the Kaiser study, an increased risk similar to that associated with HPV-18 was also found in women who were positive for HPV-31 and HPV-33. Women with an HR-HPV other than types 16, 18, 31, or 33 had a relatively lower absolute risk of CIN 3 (6.0% [95% CI, 3.8%–8.3%]).20
In the current analysis, the verification bias–adjusted absolute risks of CIN 2 or worse and of CIN 3 or worse in women 30 years or older with NILM cytology were 1.2% (95% CI, 0.6%–1.8%) and 0.5% (95% CI, 0.3%–0.9%), respectively. Based on older data, this finding seems somewhat high, especially given that liquid-based cytology was used for screening in the ATHENA study and that only women 30 years or older are included in the current analysis. For example, a retrospective study of 128,805 women with NILM cytology enrolled in the National Breast and Cervical Cancer Early Detection Program found that the age-adjusted incidence rate of having a cervical cytology result of high-grade squamous intraepithelial lesions or worse within 3 years of the initial negative cytology was only approximately 0.2%.24 However, more recent studies that have performed colposcopy in women with abnormal cytology and women who are HR-HPV+ have found a higher prevalence of high-grade CIN in women with NILM cytology.18,25,26
A comprehensive review of 7 prospective European screening trials found that after 3 years of follow-up, the CIR of CIN 2 or worse in women with NILM cytology was 0.8% (95% CI, 0.4%–1.2%) and the CIR of CIN 3 or worse was 0.5% (95% CI, 0.2%–0.8%).18 North American screening studies that have not only performed colposcopy on HR-HPV+ women, but also adjusted for verification bias by performing colposcopy in a subset of women with negative screening test results, report absolute risks of high-grade CIN in women with NILM cytology that are both higher and lower than found in the ATHENA study. Kulasingam et al25 used 2 different HR-HPV tests and liquid-based cytology to screen more than 4,000 women visiting Planned Parenthood clinics in Washington State. After adjusting for verification bias, the absolute risk of CIN 3 or worse was approximately 1.5% (calculated from data in Figure 2 in their article) in women of all ages with an NILM liquid-based cytology result.25 Another study of similar design that used conventional cytology screened more than 10,000 Canadian women 30 years or older. In this study, the verification bias–adjusted prevalence of CIN 2 or worse was 0.8% (calculated from data in Figure 1 in their article) in women 30 years or older with an NILM cytology result, which is similar to the 1.2% found in the current analysis.26
There is considerable controversy as to the merits of adjusting for verification bias. In general, the proportion of women with negative results by both HPV testing and cervical cytology who undergo colposcopy in studies that adjust for verification bias is low, and, therefore, attempts to adjust for missed disease in this group tend to lead to magnification of these few cases.27 Moreover, relatively few cases of high-grade CIN are identified in women who are both HPV and cytology negative, and this can also lead to unstable results.16 Adjustment for verification bias may explain why the absolute risk of CIN 2 or worse in women who had NILM cytology and tested HR-HPV– (14 types) was 0.8% (95% CI, 0.3%–1.5%) in the current analysis. This rate is considerably higher than the expected rate of 1 in 1,000.28
The present study has a number of considerable strengths. These strengths include the large number of women who were evaluated, the fact that all women enrolled had both cytology and HPV testing performed, that liquid-based cytology was used, and that all cervical cytology was evaluated in 4 different laboratories under routine practice conditions. In addition, all colposcopy examinations and evaluations of samples were performed without knowledge of clinical information or the results of other laboratory tests, a standardized approach was taken to colposcopy with a biopsy of normal-appearing cervix obtained in women without visible cervical lesions at the time of colposcopy, and adjudicated pathology was used to determine the “gold standard.”
This study clearly defines risk for disease at baseline in women 30 years or older undergoing routine cervical cancer screening in the United States who are found to have an NILM cytology result. Risk for high-grade neoplasia is greatly elevated in HPV-16+/HPV-18+ women and is further increased when restricted to HPV-16 alone. Risk associated with HPV-16+/HPV-18 is considerably greater than the risk associated with the 12 other HR-HPV genotypes. Moreover, the absolute risk for CIN 3 or worse in HPV-16+/HPV-18+ women 30 years or older with NILM in the ATHENA study (9.8%) is approximately the same as for women 21 years or older with ASC-US who are HR-HPV+ (14 pooled types) in the ATHENA study (8.4%).9
Based on rather limited data, predominantly from the study by Khan et al,23 the 2006 American Society of Colposcopy and Cervical Pathology Consensus Conference approved a strategy incorporating HPV-16/HPV-18 genotyping when managing women 30 years or older undergoing routine cervical cancer screening and found to be HR-HPV+ with NILM cytology. With this strategy, HPV-16+/HPV-18+ women with NILM cytology are referred for immediate colposcopy, whereas women who are positive for the 12 other HR-HPV genotypes undergo repeated testing with both cytology and HR-HPV testing at 12 months.8 While the risk threshold for performing colposcopy in response to specific cytology and HPV test results is still being defined, Castle et al29 suggested that women with an absolute risk for CIN 3 or worse of 10% or more over 2 years should undergo colposcopy. Because the present study indicates that the risk of an HPV-16+ and/or HPV-18+ woman with NILM cytology having CIN 3 or worse is 9.8% at baseline, these data clearly support the recommendation for immediate referral to colposcopy in accordance with the 2006 American Society of Colposcopy and Cervical Pathology guidelines for HPV-16/HPV-18 genotyping of women with NILM cytology.
Editorial assistance was provided by Health Interactions, London, England.