Managing atypical squamous cells of undetermined significance (ASCUS): Human papillomavirus testing, ASCUS subtyping, or follow-up cytology?

Managing atypical squamous cells of undetermined significance (ASCUS): Human papillomavirus testing, ASCUS subtyping, or follow-up cytology? Sara A. Hughes, MD,a Deqin Sun, MS,b Cheryl Gibson, MD,c Bronya Bellerose, MT, ASCP,c Lynda Rushing, MD,d Hao Chen, BS,b Bernard L. Harlow, PhD,a David R. Genest, MD,b Ellen E. Sheets, MD,a and Christopher P. Crum, MDb Boston and Cambridge, Mass, and Burlington, Vt OBJECTIVE: This study related morphologic subtype, human papillomavirus status, and a second cytologic examination to the follow-up biopsy-proven high-grade squamous intraepithelial lesion (HSIL; grade II or III cervical intraepithelial neoplasia) after a cytologic diagnosis of atypical squamous cells of undetermined significance (ASCUS). STUDY DESIGN: Seven hundred four liquid-based cervical cytology specimens were classified as normal, “ASCUS, favor reactive” (AFR), “ASCUS, not otherwise specified,” “ASCUS, favor low-grade squamous intraepithelial lesion,” “ASCUS, favor HSIL” (AFHS), low-grade squamous intraepithelial lesion, and HSIL. Human papillomavirus typing used polymerase chain reaction–restriction fragment length polymorphism analysis. A longitudinal review of the cytologic and histologic records of ASCUS cases with ≥1 follow-up test or biopsy ascertained the frequency of a follow-up diagnosis of biopsy-proven HSIL (grade II or III cervical intraepithelial neoplasia). RESULTS: Three hundred eighty-six cases (208 ASCUS, 68 normal, 86 with low-grade squamous intraepithelial lesions, and 24 with HSIL) were evaluated. High-risk human papillomavirus (HRHPV positive) was lowest with normal cytology (13%), highest with HSIL (71%), and was present in 29.8% of ASCUS cases, ranging from 22.2% (AFR) to 75% (AFHS). Most ASCUS tests (64%) were followed by a negative cytologic or histologic examination. Overall, 3.8% and 11% of ASCUS and HRHPV-positive ASCUS had histologic outcomes of HSIL. AFHS had the highest (25%) and AFR had the lowest (1.1%) proportion of HSIL outcomes. Sensitivity, specificity, and positive predictive values of human papillomavirus testing for biopsy-proven HSIL were 87.5%, 72.5%, and 11.3%, respectively. CONCLUSION: HSIL and AFHS are distinguished by the highest frequency of HRHPV types and higher rates of HSIL outcome. The remaining categories of ASCUS are heterogeneous with respect to human papillomavirus type and HSIL risk, and the value of subclassification of these entities is dependent on the practice. A human papillomavirus detection system based on polymerase chain reaction–restriction fragment length polymorphism identifies a smaller percentage of high-risk human papillomaviruses than mixed probe–based methods, probably because of the more precise exclusion of cross-reacting low-risk human papillomavirus. Negative HRHPV findings by either system show a markedly reduced risk of an HSIL outcome. However, the relative advantage of human papillomavirus testing over follow-up cytology will be influenced by the frequency of negative follow-up cytologic examination and sensitivity of liquid-based preparations in a given practice. (Am J Obstet Gynecol 2002;186:396-403.)

Key words: HPV testing, polymerase chain reaction, Papanicolaou smear, cervical neoplasia

From the Department of Obstetrics and Gynecology, Division of Gynecologic Oncology,a and the Department of Pathology, Division of Women’s and Perinatal Pathology,b Brigham and Women’s Hospital, Planned Parenthood of New England,c and the Division of Gynecologic Pathology and Cytology, Pathology Services, Inc.d Received for publication May 30, 2001; revised August 24, 2001; accepted October 30, 2001. Reprint requests: Christopher P. Crum, MD, Department of Pathology, Brigham and Women’s Hospital, 75 Francis St, Boston, MA 02115. Copyright 2002, Mosby, Inc. All rights reserved. 0002-9378/2002 $35.00 + 0 6/1/121626 doi:10.1067/mob.2002.121626

396

During the past several decades, cytologic screening with Papanicolaou tests has improved the detection of precancerous cervical lesions, and a decrease in the incidence of and mortality rates from invasive cervical cancer has been attributed in part to the increased detection of these precursor lesions. Although the management algorithms for a cytologic diagnosis of low-grade squamous intraepithelial lesion (LSIL) or high-grade squamous intraepithelial lesion (HSIL) have been formulated, the

Hughes et al 397

Volume 186, Number 3 Am J Obstet Gynecol

management of women with diagnoses of nondiagnostic squamous atypia (ASCUS) has posed a greater challenge. ASCUS rates in laboratories vary and average between 5% and 10%.1 Approximately 5% to 15% are found to harbor HSIL (grade II or III cervical intraepithelial neoplasia) at the time of colposcopic-directed biopsy.2 Efforts to more efficiently identify those women whose ASCUS confer a greater risk of HSIL have centered on 3 strategies, including cytologic follow-up examination, human papillomavirus (HPV) testing, and management of patients based on the degree of concern registered by the cytopathologist in the diagnostic report. Current guidelines recommend that atypia on a Papanicolaou test, classified as either LSIL or ASCUS, may be followed by immediate referral to colposcopy with endocervical curettage and biopsy or, alternatively, by follow-up with a Papanicolaou test within 3 to 6 months. Colposcopic referral is recommended for HSIL.3 Molecular epidemiologic studies have elucidated the central role that specific HPV types play in the development of cervical neoplasia. A wide range of “high-risk” HPV (HRHPV) types (types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 68, MM3, and MM7), ranging from those frequently (types 16 and 18) to those occasionally (types 52, 56, 58, 59, 68, MM3, and MM7) associated with cervical cancer have been identified.4 Based on previous reports, 1 or more of these types have been associated with more than 95% of women with cervical cancer and HSIL, 75% of those with LSIL, and up to 55% of those with ASCUS.4-6 However, questions remain about the advantages gained by HPV testing over repeat Papanicolaou test, particularly with the increased sensitivity afforded by liquid-based preparations.7,8 Moreover, the high rate of HPV positivity in all patients with abnormal cytologic results has generally been associated with a low positive predictive value. It therefore remains unclear whether colposcopic referral rates will be lowered enough by HPV testing to justify this test on economic grounds. Despite problems in interobserver reproducibility for ASCUS classification, an association has been shown between grades of ASCUS interpretation and grade II or III cervical intraepithelial neoplasia biopsy outcomes. Three studies found that a diagnosis of SIL shown by biopsy specimens ranged from 9% to 11% among ASCUS tests qualified as “ASCUS, favor reactive” (AFR) versus 15% to 37% among tests characterized as “ASCUS, favor SIL/HSIL” (AFS).9-12 One study also showed a trend toward increased HRHPV positivity in high-risk ASCUS groups AFS versus AFR or “ASCUS, not otherwise specified” (ANOS) (17.4%, 8.8%, and 5.7%, respectively).13 This study was conducted in a single referral laboratory that serves a range of office practices in an effort to clarify the potential roles of ASCUS qualification, HPV typing, and follow-up cytologic examination in the prediction of HSIL based on biopsy reports.

Fig 1. Simulated autoradiograph depicting predicted fragment sizes of high-risk human papillomavirus (HPV) types after digestion of polymerase chain reaction–amplified HPV DNA. HPV types are designated on the bottom, with m4 = MM4, m7 = MM7, and m9 = MM9. Size fragments based on digested pBR 322 are on the left.

Methods Case selection. After approval from the institutional review board (Brigham and Women’s Hospital, Boston, Mass), we identified a series of 918 patients whose liquidbased cervical cytologic samples (ThinPrep test, Cytyc Corp, Boxborough, Mass) were collected at Pathology Services, Inc (Cambridge, Mass) between 1997 and 2000. Cases of ASCUS diagnosed with the liquid-based preparations were consecutively culled during this time period. Cases diagnosed as LSIL and HSIL were selected for comparison and were obtained from samples consecutively collected in a 6-month period between November 1997 and June 1998. Cases determined to be within normal limits (including benign cellular changes) were obtained from consecutive samples obtained in March 1998 and March 1999. Those classified as cases of normal, ASCUS, LSIL, and HSIL were selected to undergo HPV testing. Screening for history of abnormal cervical cytologic findings was not done at the time of case selection. At the time of initial diagnosis, ASCUS was further characterized by 1 of 4 cytopathologists as AFR, ANOS, AFLS, or “ASCUS, favor HSIL” (AFHS) according to the following general criteria: (1) AFR, mild nuclear enlargement and absence of coarse chromatin similar to that seen in a reactive process; (2) AFLS, perinuclear halos with mild nuclear atypia or binucleation, or rare cells otherwise fulfilling the criteria for LSIL; (3) AFHS, cells with a high nuclear cytoplasmic ratio but with nuclear features not sufficient for a diagnosis of HSIL, or rare cells otherwise fulfilling the criteria for HSIL; (4) ANOS, reserved for cellular changes not

398 Hughes et al

Fig 2. Actual autoradiograph after gel electrophoresis of amplified DNA. Lanes marked with an S show amplified host DNA only. The required host DNA products of approximately 230 and 200 base pairs are enclosed in the white box in the center of the image. HPV positive samples are numbered with 1 = HPV66, 2 = HPV 53, 3 = HPV 16 + 53, 4 = novel (undigested), 5 = HPV 58, 6 = HPV 53, 7 = HPV 16, 8 = HPV 6, and 9 = HPV 56. Unsatisfactory samples (showing no high molecular weight products) are depicted by a U. Positive (p, HPV 16) and negative (n, buffer only) are on the right. Molecular weight markers are on the far right.

readily segregated into the other categories. We performed a retrospective review of the cytologic and histopathologic records subsequent to the HPV-typed index test. A diagnosis of HSIL was based on biopsy reports. HPV testing. After collection of material for the Papanicolaou test, the remaining liquid collection medium from the 918 patients in each diagnostic category was saved for HPV testing. One milliliter of collection medium was centrifuged, and the cell pellet was incubated in a proteinase K digestion buffer as described previously.13 After overnight incubation, Chelex resin (Biomed Laboratories, Hercules, Calif), was added and the specimens were boiled for 10 minutes and centrifuged for 5 minutes at 10,000 rpm. Two to five microliters of supernate was analyzed by polymerase chain reaction (PCR) with consensus primers (MY09 and MY11) designed to amplify a 450–base pair segment of DNA that encoded the major capsid (L1) protein from a wide range of papillomaviruses.14 After PCR amplification in the presence of 32p-dCTP, a 10X digestion buffer was added and the samples were incubated with PstI, RsaI, and HaeIII, as described previously.15 Digested samples were electrophoresed on 8% polyacrylamide gels and, when positive, the HPV type was assigned for each on the basis of the restriction fragment length polymorphism (RFLP) observed on the autoradiographs. Fig 1 is a “simulated” autoradiograph in which the known digestion fragment sizes (horizontal lines) for each high-risk HPV PCR product (bottom) are plotted. The plots are on a logarithmic scale

March 2002 Am J Obstet Gynecol

similar to the properties of electrophoretic migration. Fig 2 is an autoradiograph of an actual gel. Positive controls consisted of isolated HPV 16 DNA. Negative controls consisted of buffer solution. In HPV-negative cases, template (input DNA) integrity was confirmed if a doublet of PCRamplified human genomic DNA of approximately 230 and 200 base pairs was identified after restriction digest and gel electrophoresis. Cases were classified as HPV positive (p, Fig 2), including HRHPV (as described earlier) or low-risk HPV types (6, 11, 30, 34, 42, 43, 44, 49, 53, 54, 61, 64, 66, and 69) or novel types, satisfactory but HPV negative (S, Fig 2), and unsatisfactory (negative for both HPV and genomic DNA amplification; U, Fig 2).16 Interpretation of gels and assignment of HPV type was done by one of us (C. P. C.), with no knowledge of the cytologic diagnosis. To both verify the relationship between specific restriction polymorphisms and known HPV types and to resolve the HPV type when novel restriction polymorphisms were encountered, 100 selected samples were reamplified and the HPV product was gel-purified and subjected to DNA sequencing in the Harvard Cancer Center Sequencing Facility. Sequence data were compared to reference sequence information in the National Center for Biotechnology Information database and HPV types were assigned on the basis of this information. Statistical analysis. A power analysis was performed to ensure that sufficient numbers of cases were included to provide meaningful comparisons between normal samples and ASCUS cases with respect to HPV status. Proportions of patients who were positive for HPV were calculated for each category (normal, AFR, AFLS, AFHS, ANOS, LSIL, and HSIL). Comparisons, including the association between HSIL outcomes and ASCUS subtype were analyzed with use of the Fisher exact test. Similarly, sensitivities, specificities, positive predictive values, and negative predictive values for the detection of HSIL were calculated for HRHPV testing among all ASCUS and among select ASCUS categories. Sensitivity, specificity, positive predictive values, and negative predictive values were also calculated for ASCUS subtyping alone. An outcome was considered to be positive if HSIL was proven with a biopsy. Results Liquid-based cervical cytologic specimens from 918 subjects were selected for HPV testing on the basis of their cytologic results. These cases included normal, 4 subtypes of ASCUS (ANOS, AFR, AFLS, and AFHS), LSIL, and HSIL. Of the 918 patients, 704 (76.7%) were included because of the presence of either HPV DNA or host DNA products of sufficient size to guarantee template integrity (Table I). No significant differences in HPV status were observed between the entire group and

Hughes et al 399

Volume 186, Number 3 Am J Obstet Gynecol

Table I. Distribution of HPV positivity among all women and the subset of women with follow-up cytologic examinations or biopsies according to cytologic diagnosis All women Cytologic diagnosis Normal ASCUS AFR ANOS AFLS AFHS LSIL HSIL

Subset with follow-up data

No.

HPV positive (%)

HRHPV positive (%)

105 426 155 135 111 25 126 47

26.6 48.8 32.1 46.9 72.2 60.0 88.0 91.4

15.2 29.1 18.5 27.7 40.9 56.0 48.4 80.8

No. 68 208 90 49 65 4 86 24

HPV positive (%)

HRHPV positive (%)

19.1 53.8 41.1 53.1 70.8 75.0 80.2 95.8

13.2 29.8 22.2 30.6 36.9 75.0 44.2 70.8

HPV, Human papillomavirus status; HRHPV, high-risk human papillomavirus; ASCUS, atypical squamous cells of undetermined significance; AFR, ASCUS, favor reactive; ANOS, ASCUS, not otherwise specified; AFLS, ASCUS, favor low-grade squamous intraepithelial lesion; AFHS, ASCUS, favor high-grade squamous intraepithelial lesion; LSIL, low-grade squamous intraepithelial lesion; HSIL, high-grade squamous intraepithelial lesion.

the subset that received follow-up cytologic examination. A progressive and significant increase in association with high-risk HPV types was observed from normal (15.2%) to HSIL (80.8%). Table II summarizes the breakdown in HPV types associated with different cytologic interpretations, with attention to low-risk, unclassified, and high-risk HPV types. A consistent association between high-risk HPV types was observed with all cytologic categories, ranging from normal to HSIL. Both low- and high-risk HPV types increased in frequency from normal to LSIL. However, no low-risk HPV types were associated (in the absence of high-risk or unclassified HPV types) with either AFHS or HSIL, both of which had the strongest association with high-risk HPV types (Table II). The percentage of unclassified HPV types varied from 4.8 (normal) to 12.6 (AFLS) and was generally higher in the more advanced cytologic categories (SIL or favor SIL). On the basis of the similarity in frequency of the high-risk HPV types, categories could be grouped roughly into 3 pairs: AFR/ANOS (20.0/25.9), AFS/LSIL (41.4/48.4) and, to a lesser extent, AFHS/HSIL (56.0/80.8; Table II). We studied a subset of those patients that had both successful PCR DNA amplification and 1 or more follow-up biopsies or tests in the institution within 1 year. Of the 386 women who fulfilled the above criteria, there were 208 ASCUS (90 AFR, 49 ANOS, 65 AFLS, 4 AFHS), 68 normal, 86 LSIL, and 24 HSIL cases. The mean age of this group was 36 years. The remaining 323 patients with successful PCR DNA amplification had subsequent tests or biopsies at other institutions or had no recorded followup. Within the subset studied, 28.2% of all ASCUS cases had previous ASCUS or LSIL shown by a cytologic examination or biopsy and 2.4% had HSIL before the index test. These rates are comparable to those observed in previous studies.1 The mean time to diagnosis of HSIL was 2.6 months (0 to 7 months).

To examine the relationship of HRHPV testing in the risk stratification of ASCUS cases, we analyzed the frequency of HSIL outcomes by HRHPV status determined by a PCR-based assay. Eight cases (3.8%) of ASCUS were documented to have HSIL, all of which were HPV positive and 7 of which were positive for HRHPV, yielding an 87.5% sensitivity of HRHPV testing among ASCUS cases (Tables III and IV). Table III also shows that the specificity of the test was fair (72.5%) and the positive predictive value low, with only 11.3% HRHPV-positive ASCUS having HSIL outcomes. The strength of the relationship between HRHPV and HSIL was stronger than that between all HPV types and HSIL (data not shown). Although sensitivity was 100% when HPV (either low- or high-risk types) was detected, specificity was only 48% (96 of 200) and the positive predictive value was 7.1% (8 of 112). We also showed the ability of use of the ASCUS subtype qualification to identify those cases of ASCUS at greatest risk for the presence of HSIL. The distribution of cytologic and histologic outcomes according to ASCUS subtype is depicted in Table IV. The majority (64%) of ASCUS cases reverted to normal at a follow-up cytologic examination. As discussed earlier, biopsy findings showed that only 8 cases (3.8%) of ASCUS had HSIL. The proportion of HSIL outcomes was highest in the AFHS group and lowest in the AFR group (25% versus 1.1%; P = .08). However, the sensitivity and specificity of ASCUS subtyping alone in the detection of HSIL were both low (37.5% and 67%, respectively) when outcomes of AFHS/AFLS (AFS), a presumably higher-risk category, were compared with outcomes of ANOS/AFR (Table III). However, only 4 of 69 cases in the AFS category were AFHS. We next examined the ability of HPV typing used in conjunction with ASCUS subtype qualification to stratify those ASCUS cases with the highest likelihood of HSIL outcome. Table IV depicts the diagnostic performance of

400 Hughes et al

March 2002 Am J Obstet Gynecol

Table II. Percentage of high-risk, low-risk, and unclassified and multiple HPV types in each cytologic category* HPV High risk Low risk Unclassified Mixed Negative

Normal (105)

AFR (155)

ANOS (135)

AFLS (111)

LSIL (126)

AFHS (25)

HSIL (47)

15.2% 6.6% 4.8% 2.9% 73.4%

20.0% 7.1% 5.8% 1.9% 67.9%

25.9% 8.8% 7.4% 4.4% 53.1%

41.4% 18.0% 12.6% 7.2% 27.8%

48.4% 30.9% 8.7% 11.9% 12.0%

56.0% 0.0% 8.0% 0.0% 36.0%

80.8% 0.0% 10.6% 4.2% 8.6%

*The first 3 categories (high risk, low risk, and unclassified) are scored according to “dominant” HPV type ranked in descending order from high risk, to low risk to unclassified risk. Mixed infections, although listed separately, are always included in the first 3 categories and scored according to dominant HPV type.

HPV testing among ASCUS groups of greater severity (AFS) compared with HPV testing of all ASCUS tests combined. One might predict an improved predictive value on the basis of the exclusion of HRHPV-positive cases with normal follow-up cytologic findings. What was observed, however, was that the sensitivity of HRHPV testing improved slightly (100%), but both specificity and the positive predictive value declined. We also evaluated the diagnostic performance of testing for HRHPV among lower grades of ASCUS (ANOS/AFR), a presumably lower-risk population. Within the group of AFR/ANOS, the sensitivity of HPV testing was lower (80%) but not markedly different from sensitivity to HSIL detection that used HPV testing alone. Similarly, neither specificity nor predictive value was altered by examination of the HPV status of lower-risk ASCUS groups versus examination of all ASCUS tests combined. Therefore, in an ASCUS population that principally excludes AFHS, it appears that ASCUS subtype qualification does not enhance the ability of HPV testing to selectively triage those patients with ASCUS at risk for underlying HSIL. In 133 cases of ASCUS (63.9%), the next test or biopsy had normal results. In 21 cases, the first or only follow-up sample was a biopsy. In 62 cases, multiple normal followup cytologic findings were recorded. In 5 of 67 cases (7.5%) with multiple tests and no initial biopsy, a succeeding abnormal (ASCUS or worse) cytologic finding was recorded, including 1 HSIL. Comment The management of nondiagnostic squamous atypia (ASCUS) in cervical cytologic samples has traditionally been limited to 2 options: follow-up colposcopy and repeated cytologic examination.3 After the creation of a specific term for these atypias (ASCUS), an extensive body of literature confirmed the lack of reproducibility of this diagnosis, the high variability of its use in cytologic practice, and the small but definite (approximately 10%) risk of a significant underlying preinvasive squamous neoplasm (grade II or III cervical intraepithelial neoplasia). Efforts to enrich the diagnosis of ASCUS for an underlying HSIL have included both ASCUS qualifiers and HPV testing. Qualifiers of ASCUS, particularly AFHS, have

been shown to confer a significantly higher risk of biopsyproven HSIL.11,17 However, this diagnosis is uncommon relative to other ASCUS qualifiers (AFR or ANOS) and will cull a minority of cases with ASCUS destined for biopsy-proven HSIL. The emergence of HPV testing as a potential management tool has been spurred by at least 3 sequential developments. First, the application of a specific term (ASCUS) to nondiagnostic squamous atypia intensified a management dilemma that had existed for years.18 The second was the development of a highly sensitive nonradioactive assay (Hybrid Capture II [HC-II], Digene Corporation, Silver Spring, Md)) with a high negative predictive value.4 The third development was a large National Cancer Institute–sponsored clinical trial that explored the potential application of HC-II to ASCUS and LSIL triage.19 The value of HPV testing among patients with low-grade cervical cytologic abnormalities (LSIL/ASCUS) was initially included in this analysis but dismissed because of a high rate of cases that scored positive (over 80%) with the probe set in the HC-II assay.20 The current study differed from the ALTS trial in several ways. First, the study population was derived from a single practice that employed 4 cytopathologists. Second, HPV type was determined by RFLP analysis of PCR products rather than a probe-based (such as Hybrid Capture II) or combination PCR and probe-based (reverse line hybridization) method.16 Third, the practice studied reported a relatively low rate of ASCUS (about 5%) relative to the national average, with rates of 1.9% during 1997, 2.2% during 1998, and 3.4% during 1999. Fourth, ASCUS categories were subdivided into AFR, ANOS, and AFLS or AFHS. Finally, HPV positive rates for ASCUS were compared with normal subjects and patients with LSIL and HSIL. Moreover, patients were not randomized into treatment or follow-up groups. This study supports previous reports that linked AFHS with a significantly higher percentage of both associated high-risk HPV types and HSILs shown by follow-up biopsy findings. In previous studies, the frequency of (H)SIL outcomes has been shown to rise markedly with increasing ASCUS grade, ranging from as low as 5% to 29% among AFR to as high as 27% to 53% among AFS cases.9-12

Hughes et al 401

Volume 186, Number 3 Am J Obstet Gynecol

Table III. Sensitivity, specificity, and predictive value of HPV positivity and ASCUS subtype on the likelihood of progression to HSIL among women diagnosed with ASCUS cervical cytology

No. of cases Percentage with HSIL HRHPV positive (%) Sensitivity Specificity Positive predictive value Negative predictive value

All subtypes

ASCUS AFS

AFR/ANOS

208 3.8% 29.8% 0.875 0.725 0.113 0.993

69 4.3% 39.1% 1.0 0.636 0.111 1.0

139 3.6% 25.2% 0.80 0.769 0.114 0.99

In this study, there was a striking difference between AFHS and the other forms of ASCUS in HPV types, specifically the near absence of low-risk or novel HPV types in AFHS, similar to HSIL (Table II). However, as in other studies, AFHS comprised a minority of ASCUS cases (5.8% versus 66.8% for AFR/ANOS), and most HSILs were derived from ASCUS categories other than AFHS. Excluding AFHS, the other subdivisions of ASCUS (AFR, ANOS, and AFLS) are less distinct with respect to risk of HSIL outcome. This is compounded further by recent studies that showed poor interobserver reproducibility in subqualification of ASCUS.13,21 Nevertheless, there is considerable support in this study for these subcategories as entities with respect to diagnostic implications, HPV status and, to a lesser degree, HSIL outcome. First, in this study and a previous study from this group, the percentage of cases associated with high-risk HPV was significantly higher in the AFLS than ANOS and AFR groups, approaching that of LSIL.13 Moreover, difficulties in distinguishing LSIL and AFLS should be expected, inasmuch as essentially the same criteria are used to identify the 2 entities. When assigned equivalent diagnostic value (ie, “either LSIL or AFLS”), LSIL/AFLS can be distinguished from AFR/ANOS with very good interobserver reproducibility (κ = .63).13 Second, several reports have documented significant differences in HSIL outcome between AFR and AFHS.9-12 Therefore, although AFR is poorly reproduced, its use signifies a degree of confidence on the part of the cytopathologist that the risk of HSIL outcome is low. Moreover, in this study AFR had the lowest rate of HSIL outcome. Third, ANOS is the most perplexing category, signifying the greatest degree of uncertainty on the part of the cytopathologist, an intermediate frequency of HRHPV between AFR and AFLS and, in this study, the highest rate of HSIL outcome of the 3 ASCUS categories, excluding AFHS. This underscores the complexity of these ASCUS subgroups, the subdivision of which may or may not be clinically useful, depending on the practice. Conversely, combining these 3 groups into a single ASCUS category, as has been proposed recently,* may or may not convey the same information, depending on the nature of the individual practice.

In contrast to the more recent ALTS trial data in which the HRHPV rate was 50% by HC-II, this study scored only 29.1% of ASCUS cases as positive for HRHPV. One explanation for the difference between the rates of HRHPV detection in this study and the ALTS trial is the age difference in the populations studied: the mean age of this population was older (36 years) compared with the age of the subjects in the ALTS trial (29 years).6 Two other explanations include differences in sensitivity between HPV testing methods used and their relative specificity for high-risk HPV types. This study used the conventional MY09/MY11 primers originally designed by Manos et al.14 A recent study reported improvements in sensitivity for high-risk HPV types by use of a modification of the above primer system applied to a population with a high HPV prevalence rate.16 The improved system increased both the detection of certain individual high-risk HPV types and multiple infections. Therefore, although the HPV detection rates in this study for LSIL and HSIL categories (88% and 91% respectively) were similar to other studies that used more sensitive assays, it is likely that the proportion of high-risk HPV types is underestimated. Concerning specificity, in a previous study from the ALTS group, the “high-risk” HPV probe set used in the HC-II assay (HPV types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 68) detected not only these HPV types but also HPV types associated with cancer risk (MM7 and MM4) and HPV types not associated with cancer risk (6, 11, 40, 42, 53, 55, 70, and MM8).20 In a preliminary analysis of ThinPrep samples tested by both techniques, we confirmed that HC-II identifies some low-risk HPV types (D. Sun and C. P. Crum, unpublished data June 2000). HPV types 53 and 66 and others of unknown type were also detected by HC-II. Depending on the study, HPV-66 has been designated to have a high or low risk.22-24 In this study, none of the 15 cases that were positive for HPV-66 was associated with an HSIL, AFHS, or a follow-up of HSIL shown by biopsy or cytologic examination. The distinction of HRHPV from other HPV types in this study

*See http://bethesda2001.cancer.gov.

402 Hughes et al

March 2002 Am J Obstet Gynecol

Table IV. The distribution of cytologic and histologic outcomes according to ASCUS subtype ASCUS All subtypes AFR ANOS AFLS AFHS

No of cases

Normal

ASCUS

208 90 49 65 4

133 (64%) 72 (80%) 30 (61.2%) 14 (44.6%) 2 (50%)

27 (13%) 6 (6.7%) 7 (14.3%) 14 (21.5%) 0

HSIL CIN II or III (Papanicolaou test) (Bx) 8 (3.8%) 3 (3.3%) 2 (4.1%) 2 (3.1%) 1 (25%)

8 (3.8%) 1 (1.1%) 4 (8.2%) 2 (3.1%) 1 (25%)

LSIL (Papanicolaou test) 20 (9.6%) 6 (6.7%) 4 (8.2%) 10 (15.4%) 0

CIN I (Bx) 12 (5.8%) 2 (2.2%) 2 (4.1%) 8 (12.3%) 0

CIN, Cervical intraepithelial neoplasia; Bx, biopsy.

was based solely on the presence or absence of PCR product restriction profiles that have been established for high-risk HPV types. When the additional low-risk HPV types (including type 66) were taken into account, the potential number of ASCUS cases positive for HPV that would be detected by HC-II increased by 22 cases, raising the percentage of cases positive for HPV by approximately 25%, from 29% to 39%. The latter figure more closely approximates that reported for HC-II in the most recent report from the ALTS trial. Whether HC-II also detects a proportion of HPV types designated as novel (approximately 13% in this study) is not clear. Nevertheless, the potential contribution of these low-risk HPV types to high-risk HPV positives in the HC-II detection system is substantial (Table II). In this study a relatively low percentage of total and high-risk HPV-positive ASCUS cases (3.8% and 11%, respectively) proved by biopsy to have HSIL, compared with approximately 10% and 20%, respectively, recorded in the ALTS trial. Excluding possible differences in lesion classification by histologic examination, this most likely signifies the short follow-up interval of this study, the low index of AFHS cytologies, and the low rate of biopsy in the immediate follow-up period. The latter may also explain the near absence of any biopsy-proven HSILs after a follow-up cytologic examination read as normal. Irrespective of the reasons for the lower index of HSIL after ASCUS in this study, a negative result on the first followup cytologic examination identified a population with a significantly lower percentage of high-risk HPV (21% versus 40%; P = .001). Significantly, nearly two thirds of all ASCUS cases were followed by a normal cytologic examination or biopsy. Case patients with normal follow-up cytologic examinations had a low risk (less than 10%) for development of abnormal cytologic findings on subsequent samples. In both the ALTS trial and this study, the percentages of patients eligible for referral to colposcopy— by HPV result or abnormal follow-up cytology—were similar: approximately 56% and 59% in the ALTS trial and 29% and 36% in this study, respectively.6 Again, the differences in percentages that would be referred on the basis of HPV status may reflect several factors, including the more efficient exclusion of low-risk HPV types by the

PCR-RFLP method in the current study and factors unique to the population studied in this report. Whether the high rate of negative findings on repeated negative cytologic examinations justifies cytologic followup in lieu of HPV testing is not clear. The reported higher sensitivity of this test relative to follow-up cytologic examination (95% versus 85%) supports HPV testing.6 Moreover, 1 study reported a lower rate of women lost to follow-up in those who underwent HPV testing, although the relative patient compliance rates for repeated testing versus colposcopic examination remain unknown.2 In contrast, the reported increases in sensitivity with liquidbased preparations (as high as 92.9% for HSIL and 100% for carcinomas) suggest that, depending on the laboratory and methods used for cytologic preparation, sensitivity of follow-up cytologic examination may compare both more or less favorably to that observed in the ALTS trial.8 Even in the ALTS trial, rates of HRHPV-positive ASCUS varied widely between study site, ranging from 30% to 60%.6 Realistically, sites with the lowest HRHPV-positive rate would benefit the most from HPV triage, whereas those with the highest rates would benefit least. Therefore, in a given laboratory, sensitivity of cytologic screening method, index of HRHPV-positive ASCUS, and proportion of all ASCUS with a subsequent negative cytologic examination will influence the relative advantage or disadvantage imposed by HPV triage. The sensitivity and specificity of HRHPV testing in detecting HSIL were comparable to values determined in earlier studies (87.5% and 72.5%, respectively).2,6 Moreover, the presence of HRHPV in the initial test was associated with a 16-fold increase in the frequency of HSIL outcome compared to HRHPV-negative tests (11.3% versus 0.7%), a ratio similar to other reports.6 HRHPV typing is therefore of value in identifying patients who are at substantially greater risk of subsequent HSIL. The costeffectiveness of this technology will depend heavily on the length of follow-up considered to be safe in HPVnegative ASCUS and the expected clinical response to HPV-positive ASCUS. If “reflex HPV testing” (such as testing of the original liquid-based sample immediately after a diagnosis of ASCUS) becomes common practice, recording individual and laboratory HPV rates for both

Volume 186, Number 3 Am J Obstet Gynecol

ASCUS and its qualifiers may be critical to controlling the frequency with which these diagnoses are used and in determining the relative value of HPV testing for ASCUS in a given practice. For the latter, the relatively low positive predictive value of even HRHPV-positive ASCUS for HSIL (10% to 20%) mandates that testing methods used to test for HRHPV exclude low-risk HPV types. This is particularly important because if HPV testing were to be applied to large populations, a substantial number of HRHPVpositive women conceivably would be subject to colposcopy, biopsy, or even excision on the basis of infection by HPV types that confer negligible risk of HSIL or cancer.

REFERENCES

1. Cox JT, Lorincz AT, Schiffman MH, Sherman ME, Cullen A, Kurman RJ. Human papillomavirus testing by hybrid capture appears to be useful in triaging women with a cytologic diagnosis of atypical squamous cells of undetermined significance. Am J Obstet Gyncol 1995;172:946-54. 2. Lytwyn A, Sellors JW, Mahony JB, Daya D, Chapman W, Ellis N, et al. Comparison of human papillomavirus DNA testing and repeat Papanicolaou test in women with low-grade cervical cytologic abnormalities: a randomized trial. HPV Effectiveness in Lowgrade Paps (HELP) Study No. 1 Group. CMAJ 2000;163: 701-7. 3. Ferenczy A, Franco E, Arseneau J, Wright TC, Richart RM. Diagnostic performance of hybrid capture human papillomavirus deoxyribonucleic acid assay combined with liquid-based cytologic study. Am J Obstet Gynecol 1996;175:651-6. 4. Schiffman M, Herrero R, Hildesheim A, Sherman ME, Bratti M, Wacholder S, et al. HPV DNA testing in cervical cancer screening: results from women in a high-risk province of Costa Rica. JAMA 2000;283:87-93. 5. Kurman RJ, Henson DE, Herbst AL, Noller KL, Schiffman MH. Interim guidelines for management of abnormal cervical cytology: the 1992 National Cancer Institute Workshop. JAMA 1994;271:1866-9. 6. Solomon D, Schiffman M, Tarone R, ALTS Study Group. Comparison of three management strategies for patients with atypical squamous cells of undetermined significance: baseline results from a randomized trial. J Natl Cancer Inst 2001;93: 293-9. 7. Kaufman RH, Adam E, Icenogle J, Reeves WC. Human papillomavirus testing as triage for atypical squamous cells of undetermined significance and low-grade squamous intraepithelial lesions: sensitivity, specificity, and cost-effectiveness. Am J Obstet Gynecol 1997;177:930-6. 8. Hutchinson ML, Zahniser DJ, Sherman ME, Herrero R, Alfaro M, Bratti MC, et al. Utility of liquid-based cytology for cervical carcinoma screening: results of a population-based study conducted in a region of Costa Rica with a high incidence of cervical carcinoma. Cancer 2000;90:67-9. 9. Collins LC, Wang HH, Jawdeh A. Qualifiers for atypical squamous cells of undetermined significance help in patient management. Mod Pathol 1996;9:677-81.

Hughes et al 403

10. Gonzalez D, Hernandez E, Anderson L. Clinical significance of a cervical cytologic diagnosis of atypical squamous cells of undetermined significance: favoring a reactive process or low grade squamous intraepithelial lesion. J Reprod Med 1996;41:719-23. 11. Genest DR, Dean B, Lee KR, Sheets E, Crum CP, Cibas ES. Qualifying the cytologic diagnosis of “atypical squamous cells of undetermined significance” affects the predictive value of a squamous intraepithelial lesion on subsequent biopsy. Arch Pathol Lab Med 1998;122:338-41. 12. Vlahos NP, Dragisic KG, Wallach EE, Burroughs FH, Fluck S, Rosenthal DL. Clinical significance of the qualification of atypical squamous cells of undetermined significance: an analysis on the basis of histologic diagnoses. Am J Obstet Gynecol 2000; 182:885-90. 13. Crum CP, Genest DR, Krane JF, Hogan C, Sun D, Bellerose B, et al. Subclassifying atypical squamous cells in ThinPrep cervical cytology correlates with detection of high-risk human papillomavirus DNA. Am J Clin Pathol 1999;112:384-90. 14. Manos MM, Ting Y, Wright DK, Lewis AJ, Broker TR, Wolinsky SM. Use of polymerase chain reaction amplification for detection of genital human papillomavirus. Cancer Cells 1989;7: 209-14. 15. Lungu O, Wright TC Jr, Silverstein S. Typing of human papillomaviruses by polymerase chain reaction amplification with L1 consensus primers and RFLP analysis. Mol Cell Probes 1992; 6:145-52. 16. Gravitt PE, Peyton CL, Alessi TQ, Wheeler CM, Coutlee F, Hildesheim A, et al. Improved amplification of genital human papillomaviruses. J Clin Microbiol 2000;38:357-61. 17. Sherman ME, Tabbara SO, Scott DR, Kurman RJ, Glass AG, Manos MM, et al. “ASCUS, rule out HSIL”: cytologic features, histologic correlates, and human papillomavirus detection. Mod Pathol 1999;12:335-42. 18. The 1988 Bethesda System for reporting cervical/vaginal cytological diagnoses. National Cancer Institute Workshop. JAMA 1989;262:931-4. 19. Shiffman M, Adrianza ME. ASCUS-LSIL Triage Study: design, methods and characteristics of trial participants. Acta Cytol 2000;44:726-42. 20. Human papillomavirus testing for triage of women with cytologic evidence of low-grade squamous intraepithelial lesions: baseline data from a randomized trial: the Atypical Squamous Cells of Undetermined Significance/Low-Grade Squamous Intraepithelial Lesions Triage Study (ALTS) Group. J Natl Cancer Inst 2000;92:397-402. 21. Stoler MH, Schiffman M, Atypical Squamous Cells of Undetermined Significance–Low-grade Squamous Intraepithelial Lesion Triage Study (ALTS) Group. Interobserver reproducibility of cervical cytologic and histologic interpretations: realistic estimates from the ASCUS-LSIL Triage Study. JAMA 2001;285: 1500-5. 22. zur Hausen H. Papillomavirus infections: a major cause of human cancers. Biochim Biophys Acta 1996;1288:F55-78. 23. Kleter B, van Doorn LJ, Schrauwen L, Molijn A, Sastrowijoto S, ter Schegget J, et al. Development and clinical evaluation of a highly sensitive PCR–reverse hybridization line probe assay for detection and identification of anogenital human papillomavirus. J Clin Microbiol 1999;37:2508-17. 24. Kleter B, van Doorn LJ, ter Schegget J, Schrauwen L, van Krimpen K, Burger M, et al. Novel short-fragment PCR assay for highly sensitive broad-spectrum detection of anogenital human papillomaviruses. Am J Pathol 1998;153:1731-9.