Since the 1960s, colposcopy of the cervix with directed biopsies has been an accepted diagnostic standard for women with abnormal cytology and positive high-risk human papillomavirus (HPV) results. These directed biopsies play a crucial role in determining which women should undergo further therapy compared with close follow-up. Colposcopy has undoubtedly contributed to reducing the morbidity and mortality of cervical cancer in the United States and many other parts of the world. However, several recent studies have assessed the accuracy and sensitivity of colposcopy for the detection of high-grade cervical disease (cervical intraepithelial neoplasia [CIN] grade 2 or worse and CIN grade 3 or worse) and many of these have demonstrated that the sensitivity of colposcopy may be quite limited with a range of 30–70%.1–3
Investigators have also demonstrated improved sensitivity with colposcopy when additional biopsies were taken by the clinical provider,4 but the question remains whether sensitivity is further improved if there are no visible lesions and a random biopsy is performed. This subanalysis of data from a recent, large U.S. Food and Drug Administration registration trial for a novel HPV diagnostic test sought to determine whether a random cervical biopsy taken at the squamocolumnar junction can detect additional high-grade cervical disease.
MATERIALS AND METHODS
A total of 47,208 women aged 21 years or older undergoing routine cervical cancer screening were enrolled in the baseline phase of the ATHENA (Addressing the Need for Advanced HPV Diagnostics) trial between May 2008 and August 2009 at 61 clinical centers across the United States. Study inclusion and exclusion criteria were as follows: 21 years of age or older; not pregnant; intact uterus; willing to undergo colposcopy and biopsy within 12 weeks, if required; no treatment for CIN in the preceding 12 months; and no current or planned participation in a clinical trial for HPV treatment. The current analysis focused on the overall population of women 25 years or older, independent of cytology results, who had undergone colposcopy (n=7,100) and who had valid HPV tests and cervical biopsy results. The overall population was further stratified by abnormal cytology results (atypical squamous cells of undetermined significance or greater, n=2,012) and negative cytology (negative for intraepithelial lesion or malignancy, n=5,088). The protocol was approved by the institutional review boards of all study sites, and all women provided written informed consent before undergoing any study procedures.
The ATHENA study design has been previously described in detail.5,6 Briefly, after providing written informed consent, all women at enrollment underwent cervical sampling for both liquid-based cytology and high-risk HPV testing using three assays. Cytology evaluation was conducted at four accredited clinical laboratories in the United States and results of the cytology evaluation were reported using the 2001 Bethesda System terminology. Triage to colposcopy was based on the analytically sensitive first-generation HPV tests: AMPLICOR HPV Test that detects 13 high-risk HPV types (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 68; optical density cutoff 0.2 or greater) and the modified LINEAR ARRAY HPV Genotyping Test that detects 16 high-risk HPV types (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, 68, 73, and 82). All analyses were conducted using results from the second-generation cobas HPV Test that simultaneously detects a total of 14 high-risk HPV types: HPV 16 individually, HPV 18 individually, and 12 pooled high-risk HPV genotypes (31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68). A separate β-globin control is also included in the cobas HPV Test. All HPV tests were performed at the same four study laboratories that conducted the cytology evaluation; Roche Molecular Systems served as a fifth site for cobas HPV testing. To provide adjudication for those cases that were cobas HPV Test-negative and where high-grade disease was detected, immunohistochemistry for p16INK4A was performed using the CINtec Histology Kit according to the manufacturer's instructions.7
Women with atypical squamous cells of undetermined significance or greater cytology results or those 25 years or older with positive high-risk HPV results were referred for colposcopy; in addition, a randomized subset (1:35) of women 25 years or older who were negative for both cytology and HPV were also referred to colposcopy so accuracy could be adjusted for verification bias. Participants and colposcopists were masked to both the patients' cytology and HPV test results. Colposcopy with biopsy, endocervical curettage, or both were performed within 12 weeks of enrollment according to a standardized protocol that included biopsy of all visible cervical lesions; a colposcopic lesion was defined as an acetowhite lesion suspected of representing neoplasia.8 In women with a satisfactory colposcopy but without visible cervical lesions, a single random biopsy at the squamocolumnar junction was performed per protocol, and all women with an unsatisfactory colposcopy underwent an endocervical curettage. Biopsy and endocervical curettage specimens were reviewed in a blinded manner by a panel of three expert pathologists (central pathology review) and diagnosed using standard criteria and CIN terminology.
The biopsy data were entered directly into the database by the clinical site and the data were source-verified by the site monitor. The pathology data were entered from case report forms by the Data Management Department at Roche Molecular Systems and quality control was performed on 100% of the entries to verify that the data from the case report forms were entered correctly into the database.
For the purposes of this analysis, we defined four categories of biopsies: 1) directed: visible lesion present and biopsy taken per protocol; 2) random: no visible lesion but squamocolumnar junction visualized and biopsy taken at the transformation zone; 3) endocervical curettage: squamocolumnar junction not visualized or partially visualized and endocervical curettage performed; and 4) treatment: excisional procedure (loop electrosurgical excision procedure or conization) performed based on the biopsy community reading of CIN 2 or worse when the central pathology review diagnosis was less than CIN 2 and the subsequent central pathology review reading on the excisional specimen was CIN 2 or worse. Results from the endocervical curettage or the treatment biopsy categories are not included in the subanalyses here but will be reported in future publications. Analyses evaluating the contribution of biopsy to disease detection were performed on a per-patient rather than a per-biopsy basis. Because each patient without a lesion would, by study design, have at most one random biopsy and because the average for number of biopsies per patient was 1.7, the difference in denominators would make the comparisons between directed and random biopsy difficult unless results were expressed on a per-patient basis. SAS software was used for statistical calculations.
Of the 40,901 women aged 25 years or older enrolled in the ATHENA study, 2,839 (6.9%) underwent a random biopsy and 2,796 had valid biopsy results (Fig. 1). Characteristics for these 2,839 women are shown in Table 1. The mean age of women undergoing colposcopy with a random biopsy was 39.5 years compared with 37.4 years for directed biopsy. The majority of women who had a random biopsy were white, had previously undergone cervical cancer screening within the past 5 years, and were not immunocompromised. Among women who had biopsies, 33.9% were classified as smokers and approximately equal percentages had directed compared with random biopsies (Table 1). A total of 2.3% of study participants who had biopsies were immunized, again with similar distribution between directed and random.
The total number of cervical biopsies (including endocervical curettage if done with a biopsy) performed in women 25 years or older in this trial was 12,015 and the mean number of cervical biopsies per patient was 1.7. Approximately 23.9% of women underwent cervical biopsies, endocervical curettage, or both.
Overall, there were 2,853 random biopsies performed when a cervical lesion was not identified colposcopically (six women had more than one random biopsy performed), and this comprised 31.2% of all cervical biopsies (directed and random) performed in the ATHENA trial. Among the total 9,149 biopsies, 1.4% and 1.5% of biopsies were read as inadequate for directed (86/6,296) and random (43/2,853) biopsies, respectively.
The histopathologic distribution among the 2,839 women who had a random biopsy performed was as follows: 90.0% (2,552) negative or normal; 5.7% (163) CIN 1; 1.3% (36) CIN 2; 1.4% (41) CIN 3; and 0.14% (4) CIN 3 or worse (Table 2).
In the overall population of women 25 years or older, the random biopsy contributed a total of 20.9% (81/388) and 18.9% (45/238) of the CIN 2 or worse and CIN 3 or worse cases, respectively (Table 2). In women with abnormal cytology, 19.7% (39/198) and 22.4% (28/125) of CIN 2 or worse and CIN 3 or worse, respectively, were identified in those without a visible lesion who underwent a random biopsy (Table 2). In the population of women with negative for intraepithelial lesion or malignancy cytology, the random biopsy detected 22.1% (42/190) and 15.0% (17/113) of the CIN 2 or worse and CIN 3 or worse cases, respectively. When stratified by HPV status, a higher percentage of HPV 16- or HPV 18-associated disease (13.1%) was detected overall with the random biopsies compared with disease attributed to the 12 other high risk genotypes (3.5%, P≤.001), more notably in the abnormal cytology compared to the negative for intraepithelial lesion or malignancy population (Table 2). When stratified by age, the percentage of high-grade disease detected with the random biopsy in all three populations was highest in women 25–29 years, and this trend decreased with age (Table 3).
In women who were cytology-negative and HPV-negative, there were eight random cervical biopsies consistent with CIN 2 or worse and, of note, five of these were negative for p16, suggesting the possibility that these represented false-positive histology results.
The absolute risk of detection of CIN 2 or worse and CIN 3 or worse by random and directed biopsies was calculated and stratified by high-risk HPV result (Figs. 2 and 3). In all populations, the absolute risk was highest for women positive for HPV 16 or HPV 18, presumably reflecting the higher prevalence of HPV 16 or 18–associated disease compared with the other genotypes. The risk of detecting CIN 2 or worse with the random biopsy in women positive for HPV 16 or 18 was 13.1% (95% CI 9.8–17.4%), 24.7% (95% CI 16.8–34.8%) and 8.6% (95% CI 5.6–13.1%) for the overall, abnormal cytology, and negative for intraepithelial lesion or malignancy populations, respectively (Fig. 2). By comparison, the risk in women positive for the 12 other genotypes was 3.5% (95% CI 2.5–5.0%), 8.8% (95% CI 5.3–14.2%) and 2.3% (95% CI 1.4–3.7%) for the overall, abnormal cytology, and negative for intraepithelial lesion or malignancy populations, respectively (Fig. 2). Similar higher absolute risks for CIN 3 or worse were seen in random biopsies from women positive for HPV 16 or 18 compared with the other 12 high-risk genotypes across all three populations. In the overall population, the absolute risks for HPV 16 or 18 and 12 other high-risk were 8.2% (25/305, 95% CI 5.6–11.8%) and 1.7% (14/820, 95% CI 1.0–2.8%), respectively (Fig. 3).
When compared with the absolute risks of the directed biopsy, the risks associated with random biopsy are somewhat lower, as would be expected. However, the risk of CIN 2 or worse associated with random biopsy in women with negative cytology who tested positive for HPV 16 or HPV 18 was similar to the risk of directed biopsy for women who were positive for the 12 other genotypes.
In this post hoc analysis, approximately 20% additional CIN 2 or worse and CIN 3 or worse was identified in women who did not have a visible lesion on colposcopy and had a random biopsy performed at the squamocolumnar junction. This absolute increase of disease was most notable in women who were HPV 16- or 18-positive. These results point to the surprisingly high rate of additional disease detection achieved with random biopsies. This analysis is strengthened by several factors, including a standardized, blinded, and “unprejudiced” protocol, adjudicated pathology review, and, most importantly, the large number of women undergoing colposcopy.
Over the past decade, numerous investigators have demonstrated the limited sensitivity and accuracy of colposcopy, variable interobserver agreement regarding digitized cervical images and biopsy sites, and effect of the absolute number of biopsies on disease detection.1,2,4,9–13 In a study by Massad et al, colposcopies were performed on more than 2,800 women. Only one third of women had exact agreement between the colposcopic impression and histology, and the overall strength of correlation (κ value) was poor at 0.20. The sensitivity of colposcopy for CIN 2 or 3 was 56%.1
In an ancillary study based on the Atypical Squamous Cells of Undetermined Significance—Low-Grade Squamous Intraepithelial Lesions Triage Study, Gage et al analyzed factors influencing the sensitivity of colposcopy in 408 women undergoing colposcopy. Only 70% of women with a diagnosis of CIN 3 or worse had a colposcopically directed biopsy of CIN 2 or worse. Interestingly, sensitivity was significantly improved when two or more biopsies were performed, and there was essentially no difference in sensitivity by colposcopist experience (ie, general gynecologist, gynecologic oncologist, gynecologic oncology fellow, and nurse practitioner).4
In a seminal study by Pretorius et al3 in 2004, random biopsy detected additional CIN 2 or worse, particularly in women with high-grade cytology. In another study by Pretorius et al,14 the value of random biopsies performed in cervical quadrants without visible disease in a low risk-population (Southern California Permanente Medical Group, Fontana) was evaluated. Approximately one fifth of CIN 3 or worse cases were diagnosed after further evaluation of random cervical biopsies (with and without endocervical curettage) that were consistent with CIN 3 or worse.
Despite this additional disease detection, it is unclear whether CIN 2 or worse lesions detected with random biopsies are biologically and clinically similar to lesions colposcopically identified and biopsied. Without lengthy longitudinal follow-up, it is nearly impossible to adequately address this question; however, it is quite possible that these nonvisible lesions represent clinically quiescent disease or disease that has yet to achieve sufficient size or density to be visible colposcopically. On the other hand, do we have sufficient evidence to indicate that these lesions are not clinically important or relevant? Additionally, one would suspect that multiple random biopsies would increase the abnormality rate; however, it is unclear by how much.
At the present time in the United States, there is no “standard” for colposcopy regarding the number of biopsies that should be taken. Clinical practice can range from taking one biopsy from the worst visible lesion to performing four quadrant biopsies, irrespective of the presence or absence of colposcopic abnormalities. Thus, it is highly challenging to understand the contribution of random biopsies in clinical practice when no true clinical standard exists for colposcopy. What should be the benchmark: biopsy of all visible colposcopic abnormalities or systematic four quadrant biopsies?
Despite the large number of women undergoing colposcopy in the ATHENA trial and the standardized colposcopy protocol, including the requirement of random biopsy, there are some limitations and weaknesses associated with this analysis. Again, the primary objective of the ATHENA trial was not to address the value of random cervical biopsy. Investigators were limited to one random biopsy, and a random biopsy was not performed in women in whom the squamocolumnar junction was not visualized. Additionally, these results may not be fully generalizable because the investigators were blinded to cytology and HPV results, which is not consistent with standard clinical practice.
Nonetheless, these results call attention to disease that is clearly missed on colposcopy. In particular, the absolute risk of lesions detected in the absence of visible lesions in women testing positive for HPV 16 or 18 justifies taking a random biopsy in these populations, in accordance with proposed risk assessment models.15 The practice of colposcopy has undergone scientific scrutiny over the past decade, and it is clear that this procedure is far from perfect and likely misses clinically significant (and perhaps clinically insignificant) disease. Whether taking additional biopsies, random biopsies, or both, colposcopists should strive to optimize disease detection and standardize this procedure based on current published evidence.
1. Massad LS, Collins YC. Strength of correlations between colposcopic impression and biopsy histology. Gynecol Oncol 2003;89:424–8.
2. Massad LS, Jeronimo J, Katki HA, Schiffman M; National Institutes of Health/American Society for Colposcopy and Cervical Pathology Research Group. The accuracy of colposcopic grading for detection of high-grade cervical intraepithelial neoplasia. J Low Genit Tract Dis 2009;13:137–44.
3. Pretorius RG, Zhang WH, Belinson JL, Huang MN, Wu LY, Zhang X, et al.. Colposcopically directed biopsy, random cervical biopsy, and endocervical curettage in the diagnosis of cervical intraepithelial neoplasia II or worse. Am J Obstet Gynecol 2004;191:430–4.
4. Gage JC, Hanson VW, Abbey K, Dippery S, Gardner S, Kubota J, et al.. Number of cervical biopsies and sensitivity of colposcopy. Obstet Gynecol 2006;108:264–72.
5. Wright TC Jr, Stoler MH, Behrens CM, Apple R, Derion T, Wright TL. The ATHENA human papillomavirus study: design, methods, and baseline results. Am J Obstet Gynecol 2012;206:46.e1–11.
6. Stoler MH, Wright TC Jr, Sharma A, Apple R, Gutekunst K, Wright TL, et al.. High-risk human papillomavirus testing in women with ASC-US cytology: results from the ATHENA HPV study. Am J Clin Pathol 2011;135:468–75.
9. Stoler MH, Vichnin MD, Ferenczy A, Ferris DG, Perez G, Paavonen J, et al.. The accuracy of colposcopic biopsy: analyses from the placebo arm of the Gardasil clinical trials. Int J Cancer 2010;128:1354–62.
10. Jeronimo J, Schiffman M. Colposcopy at a crossroads. Am J Obstet Gynecol 2006;195:349–53.
11. Cantor SB, Cárdenas-Turanzas M, Cox DD, Atkinson EN, Nogueras-Gonzalez GM, Beck JR, et al.. Accuracy of colposcopy in the diagnostic setting compared with the screening setting. Obstet Gynecol 2008;111:7–14.
12. Moss EL, Dhar KK, Byrom J, Jones PW, Redman CW. The diagnostic accuracy of colposcopy in previously treated cervical intraepithelial neoplasia. J Low Genit Tract Dis 2009;13:5–9.
13. Wentzensen N, Zuna RE, Sherman ME, Gold MA, Schiffman M, Dunn ST, et al.. Accuracy of cervical specimens obtained for biomarker studies in women with CIN3. Gynecol Oncol 2009;115:493–6.
14. Pretorius RG, Belinson JL, Azizi F, Peterson PC, Belinson S. Utility of random cervical biopsy and endocervical curettage in a low-risk population. J Low Genit Tract Dis 2012;16:333–8.
© 2014 by The American College of Obstetricians and Gynecologists. Published by Wolters Kluwer Health, Inc. All rights reserved.
15. Castle PE, Sideri M, Jeronimo J, Solomon D, Schiffman M. Risk assessment to guide the prevention of cervical cancer. Am J Obstet Gynecol 2007;197:356.e1–6.