The primary goal of this study was to compare the clinical performance of an optimized and rigorously controlled immunocytochemical (ICC) assay for p16INK4a to high-risk (HR) human papillomavirus (HPV) detection by polymerase chain reaction (PCR) as diagnostic adjuncts for cytology specimens from colposcopy patients.
The study included 403 cervical cytology specimens collected within 3 months of colposcopy. The colposcopic impression and cervical biopsy diagnosis served as the standards for correlation with cytological, p16INK4a, and HPV data. p16INK4a was evaluated using an immunoperoxidase-based assay that was linear over 4 logs for the detection of HeLa-spiked positive control cytology specimens, using a threshold for positive test results that was based on receiver operating characteristic curve analysis. HR-HPV was detected by multiplex PCR using genotype-specific primers.
In all combined diagnostic categories (negative for intraepithelial lesion and malignancy, atypical glandular cells, atypical squamous cells of undetermined significance, atypical squamous cells cannot exclude high-grade squamous intraepithelial lesion, low-grade squamous intraepithelial lesion, and high-grade squamous intraepithelial lesion), the p16INK4a ICC and HR-HPV assays, respectively, had sensitivity of 81.7% and 83.3% (P = .81) and specificity of 78.1% and 50.9% (P < .001) for the detection of underlying ≥grade 2 cervical intraepithelial neoplasia (CIN) lesions on biopsy. Furthermore, the positive predictive value of p16INK4a ICC was greater than that of HR-HPV for patients with biopsies ≥CIN-2 (41.2% and 24.2%, respectively, P = .001).
Although the Papanicolaou (Pap) test is the most effective screening tool for cancer that has ever been devised, cervical cytopathology remains limited by poor specificity for underlying clinically significant lesions in cases with low-grade abnormalities. Over 4 million cases are diagnosed as atypical squamous cells of undetermined significance (ASC-US), atypical squamous cells cannot exclude high-grade squamous intraepithelial lesion (ASC-H), low-grade squamous intraepithelial lesion (LSIL), or atypical glandular cells (AGC) in the United States each year, and require further evaluation to identify the relatively small proportion of patients who have clinically significant lesions (≥grade 2 cervical intraepithelial neoplasia [CIN]/androgen insensitivity syndrome [AIS]) on cervical biopsy. In most cases, however, further evaluation, including human papillomavirus (HPV) triage and/or colposcopy, does not find high-grade lesions in patients who were initially identified based only on the diagnosis of low-grade cytologic abnormalities.1-4 High-risk (HR)-HPV testing alone does not discriminate between transient infections in patients with low-grade lesions versus cases that may result in persistent high-grade lesions.5, 6 Therefore, alternative molecular diagnostic adjuncts are needed to improve the specificity and positive predictive value (PPV) of cervical cytology to detect high-grade dysplasia.
Over the past 5 years, data from numerous studies have supported the hypothesis that p16INK4a, a cyclin-dependent kinase inhibitor and G1/S-phase cell cycle checkpoint regulator, is a surrogate marker of HPV E7-mediated functional inactivation of retinoblastoma protein and is consistently overexpressed in CIN-2/3, squamous cell carcinoma (SCC), AIS, and in adenocarcinoma of the cervical mucosa.7-9 Although p16INK4a has been detected in high-grade squamous intraepithelial lesion (HSIL),10-15 few studies have correlated p16INK4a results in cases with NILM (negative for intraepithelial lesion and malignancy), ASC-US, AGC, LSIL, or ASC-H cytology to the results of colposcopy. Furthermore, p16INK4a staining in cervical cytology specimens has not been directly compared with p16INK4a localization in the corresponding tissue biopsies.
The present study was designed to develop a p16INK4a cytology assay with optimal sensitivity and specificity performance characteristics for the detection of underlying high-grade lesions, to determine the correlation between p16INK4a cytology and p16INK4a biopsy test results, and to determine the correlation between p16INK4a and HR-HPV detection with histologic diagnoses in cases with NILM, ASC-US, ASC-H, AGC, LSIL, and HSIL cytology.
MATERIALS AND METHODS
Patients were selected from women who received care at the women's healthcare clinics of the University of Colorado Hospital in Aurora, Colorado and the Denver Health Hospital in Denver, Colorado between March 2002 and December 2007. The basis for the referral was a previous abnormal cytology test result (≥ASC-US). A total of 1878 remnant cervical cytology specimens, including 574 NILM, 64 AGC, 681 ASC-US, 257 LSIL, 93 ASC-H, 207 HSIL, and 2 cases classified as cervical carcinomas were collected from the cytology laboratory at the Department of Pathology at the University of Colorado in Denver. NILM cases included specimens with normal and/or reactive cellular changes. An additional 251 remnant specimens of ASC-US were collected from the cytopathology laboratory at Denver Health Medical Center. Surgical pathology databases were reviewed to identify all cases that had undergone colposcopic examination at the same time or up to 3 months after collection of the cervical cytology specimen. Each remaining vial was examined by trypan blue staining and manual counting, to confirm that there was a residual volume equal to at least 2 mL and that all specimens included at least 50,000 squamous cells. Cases without corresponding colposcopy results or insufficient volume or cellularity were excluded from further consideration. The final study population consisted of 403 cervical cytology specimens, including 95 NILM (including cases with normal cytology and reactive cellular changes), 164 ASC-US, 28 AGC, 34 ASC-H, 42 LSIL cases, and 40 HSIL cases.
Colposcopic biopsy of visible lesions, and/or endocervical curettage (ECC) when clinically indicated, was performed in 327 cases. Cases that had negative colposcopy and did not undergo cervical biopsy or ECC were classified as having no evidence of CIN-2 or greater lesions. Cases that had positive cytology (AGC, ASC-H, or HSIL) but biopsies that were negative for high-grade dysplasia or AIS were followed by electrocautery or cold knife cone biopsy within 3 months of colposcopy. Furthermore, pathology case records were followed for a minimum of 12 months after the date of the colposcopy to record subsequent diagnoses of HSIL and/or biopsy-confirmed lesions ≥CIN-2 in patients who had negative colposcopy or negative (<CIN-2) cervical biopsies. Of note, p16INK4a data were not shared with clinical care providers and had no impact on patient referral for colposcopy.
All Pap-stained cytology slides and H & E-stained cervical biopsy slides were reviewed by a cytopathologist (K.R.S.), who was blinded to the original diagnosis and to the HPV status (where applicable). In 73 cases where there was discordance with the original diagnosis (≥single step discordance), the cytology slides, as well as the corresponding tissue biopsy slides, were re-evaluated concurrently by 3 pathologists (K.R.S., M.S., and D.F.) using a multihead microscope, to arrive at a final consensus diagnosis, again without knowledge of the original diagnosis of the slides or HPV status. This study was conducted with informed patient consent under a protocol that was reviewed and approved by the Colorado Multiple Institutional Review Board and by the Human Subjects Research Committee of Denver Health Medical Center.
Cytology Specimen Preparation
Cervical cytology specimens were collected using a broomlike device, were placed into ThinPrep (Hologic, Marlborough, Mass) vials containing PreservCyt transport medium, and were processed for routine staining, screening, and diagnosis. After the final diagnostic reports were released, the residual fluid cytology specimens and corresponding paraffin-embedded tissue blocks of the colposcopic biopsies were deidentified and entered into the study.
p16INK4a Detection in Cervical Cytology Specimens
Replicate slides of ThinPrep cervical cytology specimens were processed using a ThinPrep 2000 Processor to prepare slides for p16INK4a. Endogenous peroxidase activity was quenched by incubation in 0.6% H2O2 in 100% methanol, and antigen retrieval was performed by heating in a decloaking chamber (Biocare Medical, Walnut Creek, Calif) for 10 minutes at 120°C in 1× citrate buffer (pH 6.0). Permeabilization of the cellular specimens was done by incubation in 0.25% Triton X-100 for 30 minutes, and the slides were incubated for 1 hour at room temperature with a primary mouse monoclonal antibody to p16INK4a (NeoMarkers 4, Fremont, Calif) at a dilution of 1:25 (8 μg/mL). After incubation with the primary antibody, slides were processed by an indirect avidin-biotin–based immunoperoxidase procedure using a biotinylated horse antimouse antibody (R.T.U. Vectastain Universal Elite ABC kit; Vector Laboratories, Burlingame, Calif), incubated overnight at 4°C in 0.25% Triton X-100, developed using 3, 3′-diaminobenzidine (DakoCytomation, Carpentaria, Calif), and counterstained with hematoxylin. Antibody specificity was evaluated by Western blot analysis of the p16INK4a antibody against p16INK4a-positive cell lines (HeLa, C4-1, C33a, and SiHa) and a p16INK4a-negative cell line (SW962). The sensitivity of the p16INK4a assay to detect abnormal cells was evaluated by spiking HeLa cells into normal cytology specimens, and the number of p16INK4a-positive HeLa cells was linear over 4 log orders of magnitude. The interassay and intra-assay reproducibility of p16INK4a detection in 10 normal/HeLa spiked specimens and in 10 HSIL specimens demonstrated <20% variability across triplicate sample means. Positive controls specimens (NILM spiked with 1000 and 100 HeLa cell equivalents) were included in each staining run. Negative controls were performed on all cases using an equivalent concentration of a subclass-matched immunoglobulin G1κ (Becton Dickson PharMingen, San Jose, Calif), generated against unrelated antigens, in place of the p16INK4a antibody.
Criteria to score p16INK4a test results were established by evaluation of a series of 20 NILM cases and 20 biopsy-confirmed HSIL cytology specimens from colposcopy patients. Each slide was screened with a 20× objective, using an ocular micrometer, and the total number of p16INK4a-stained cells present per slide was calculated for each case. p16INK4a immunocytochemistry showed strong nuclear and cytoplasmic staining in morphologically abnormal cells. Faint staining was occasionally seen in benign squamous and glandular cells and was interpreted as nonspecific background. Although most cases that were classified as NILM had no p16INK4a-positive cells, in a few cases, p16INK4a was detected in scattered benign-appearing metaplastic or glandular cells. Staining was detected in cells that were morphologically diagnostic of HSIL in 20 of 20 HSIL cases in the training set. In all but 2 of these HSIL cases, the number of p16INK4a-positive HSIL cells exceeded 30 cells per slide. Quantitative criteria were thus developed to define test results using the following scoring system:
p16INK4a positive: Strong staining in ≥30 metaplastic, koilocytotic, or cytologically equivocal cells (consistent with ASC-US or atypical glandular cells of undetermined significance [AGUS]), excluding patchy staining in scattered normal-appearing endocervical cells, or strong staining in cells that are morphologically consistent with HSIL or SCC.
p16INK4a negative: Strong staining in <30 metaplastic, koilocytotic, or cytologically equivocal cells (consistent with ASC-US or AGUS), or patchy staining in scattered normal-appearing endocervical cells.
A “gold standard” series of cytology specimens that included 20 cases each of NILM, ASC-US, LSIL, and HSIL was developed by consensus diagnosis of all 3 pathologists (K.R.S., M.S., and D.F.), and these cases were then used as a training set to ensure the utilization of uniform criteria for cytologic diagnoses and the scoring of p16INK4a test results. The training set was also used to evaluate intrareader and inter-reader reproducibility of the p16INK4a scores. p16INK4a-stained slides from the “gold standard” set of cases were independently reviewed and scored as positive or negative for p16INK4a by each of the 3 participating pathologists, using criteria established as described above, and the p16INK4a scores were compared with the diagnoses of the Pap-stained cytology slides and the corresponding cervical biopsies and ECCs. There was complete intraobserver and interobserver reproducibility in the scores of 157 (98%) of 160 slides between the 3 pathologists who scored the cases on 2 separate occasions (with at least a 30-day interval between readings). The kappa coefficients for the 2 different reviews by each pathologist had highly significant P values (P < .0001), and there was no statistically significant difference between kappa values for the 3 different pathologists.
p16INK4a Detection in Cervical Biopsies
Formalin-fixed, paraffin-embedded tissue blocks were sectioned at 5 μm, mounted on charged glass slides (Superfrost Plus, Fisher Scientific, Pittsburgh, Pa), and deparaffinized. Endogenous peroxidase activity was blocked by 15-minute treatment with 3.0% hydrogen peroxide. Antigen retrieval and immunohistochemical staining, including negative controls, were performed as described above, except that the primary antibodies were used at a dilution of 2 μg/mL for 2 hours. A section of cervical SCC was included in each staining run as a positive control. p16INK4a staining criteria in tissue specimens were based on the detection of nuclear and/or cytoplasmic staining, staining intensity, proportion of cells stained, staining pattern (diffuse, focal, and sporadic), and the histologic distribution of cells in the tissue. A confluent, intense pattern of staining, seen typically in CIN-2/3, was considered as a positive test result. Negative test scores were assigned to sections with either negative or weak sporadic staining.
HR-HPV Detection by Multiplex Polymerase Chain Reaction by Using Genotype-Specific Primers
DNA was extracted from cytology samples using proteinase K (Roche Applied Science, Indianapolis, Ind; 200 μg/mL final concentration) in 100 μL digestion buffer (10 mM Tris HCl, pH 8.5, 1 mM ethylenediaminetetraacetic acid, 0.5% Tween 20) at 55°C for 3 hours. Multiplex polymerase chain reaction (PCR) for HR-HPV detection was performed as described previously by Nishiwaki et al16 with minor modifications. Briefly, PCR was performed using 4 μL of digested sample as the DNA template, 25 μL of 2× Qiagen (Valencia, Calif) multiplex PCR master mix, 16 μL of nuclease-free distilled water, and 5 μL of the total primer mix at a final concentration of 0.2 μM. The HPV multiplex primer mix (Eurofins MWG Operon, Huntsville, Ala) consisted of primers to detect 16 different HPV genotypes (genotypes 6, 11, 16, 18, 30, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 66) along with another primer pair to amplify a 100-bp aminolevulinate synthase 1 internal positive control. Amplification was performed by incubation at 95°C for 15 minutes, followed by 44 cycles of denaturation for 30 seconds at 94°C, 1.5 minutes of annealing at 70°C, and 1 minute of elongation at 72°C in a PTC-225 Peltier Thermal cycler apparatus (MJ Research, Waltham, Mass). HeLa and SiHa DNA-positive controls and reagent aqueous control samples were included with each PCR run.
Individual HPV Typing by Capillary Electrophoresis
Ten-μL aliquots of the PCR products were processed by capillary electrophoresis at the genomics core facility of Stony Brook University Medical Center, using the QIAxcel DNA Screening Kit (2400) (Qiagen). The HPV type(s) in each sample were analyzed using the amplicon size of each HPV type normalized to the 100-bp internal control in each sample and the 2 positive controls used in each reaction run. In addition, PCR reactions were performed to confirm each of the HPV types using individual primers in at least 3 samples with a single infection for each HPV type.
Comparisons were made between the p16INK4a test results and the detection of high-risk HPVs with the final cytological and histologic diagnoses using the McNemar test for paired data. For the purpose of statistical analysis, patients who had no evidence of a lesion on colposcopic evaluation were grouped together with cases that had only biopsies or ECCs with normal histology. The performance of a combined test approach was evaluated by serial analysis using cases that first tested positive for HR-HPV. To test differences in proportions of positive tests, either the chi-square test or the Fisher exact test was used. All analyses were done using NCSS2007 (NCSS, Kaysville, Utah) or SAS statistical software (version 9.2; SAS Institute, Cary, NC). Statistical significance was set at a P value <.05.
p16INK4a Cytology Test
p16INK4a-positive results were observed in 1 (1.1%) of 95 NILM, 53 (32.3%) of 164 ASC-US, 8 (28.6%) of 28 AGC, 17 (40.4%) of 42 LSIL, 13 (38.8%) of 34 ASC-H, and 32 (80%) of 40 of HSIL specimens (Table 1) (Fig. 1). High-grade cervical dysplasia (≥CIN-2) was diagnosed by colposcopic biopsy in some cases with p16INK4a-positive cytology, including 9 (17.0%) of 53 ASC-US, 1 (12.5%) of 8 AGC, 7 (41.2%) of 17 LSIL, in 9 (69.2%) of 13 ASC-H, and 26 (81.1%) of 32 HSILs. By contrast, lesions ≥CIN-2 were rarely detected by cervical biopsy in cytologic specimens from patients who were negative for p16INK4a. The single p16INK4a-positive NILM specimen was from a patient who had been referred for colposcopy on the basis of a prior cytologic diagnosis of HSIL. This patient did not undergo further clinical evaluation.
Table 1. ICC p16INK4a Expression in Cytology Specimens Versus Histologic Diagnosis
p16INK4a immunohistochemical staining was performed on all cervical biopsies to evaluate potential sources of p16INK4a-positive cells in the corresponding cytology specimens. In cases with positive p16INK4a cytology and underlying biopsies ≥CIN-2, p16INK4a was consistently positive in the tissue sections (Fig. 2B, C). In cases with positive p16INK4a on cytology but no evidence of a lesion on colposcopy or with negative biopsies (<CIN-2), CIN-1, benign intercalated endocervical cells, areas of tuboendometrial metaplasia, and rarely, squamous metaplasia appeared to be the source of the p16INK4a-positive cell population (Fig. 2A, D-2F).
HR-HPV in Cervical Cytology
The internal control aminolevulinate synthase 1 DNA target was detected in 380 (94.3%) of 403 of test specimens. The remaining 23 (5.7%) cases were scored as nonamplifiable and were excluded from HPV studies. Discrete PCR amplification products (Fig. 3) defined the presence of HPV types 16 and/or 18 in 58 (15.3%) of 380 cases and other high-risk types in 171 (45.0%) of 380 cases. A single high-risk HPV type was detected in 136 (65.7%) and infection by multiple high-risk HPVs was detected in 71 (34.3%) of 207 high-risk positive cases. The 6 most commonly detected high-risk HPVs, in descending order of prevalence, were types 51, 58, 16, 31, 52, and 18. The detection of HPVs 16 and/or 18 was statistically correlated with the diagnosis of lesions ≥CIN-2 (P < .005). High-risk HPVs were detected in 20 (23.5%) of 85 NILM, 103 (65.2%) of 158 ASC-US, 8 (29.6%) of 27 AGC, 27 (65.9%) of 41 LSIL, 21 (65.6%) of 32 ASC-H, and 27 (73%) of 37 HSIL cytology specimens (Table 2).
Table 2. HR-HPV Detection in Cytology Specimens Versus Histologic Diagnosis
p16INK4a was positive in 6 (60.0%) of 10 cytology specimens, including 1 LSIL and 5 HSIL specimens that were HR-HPV negative but had ≥CIN-2 on cervical biopsy. Conversely, p16INK4a was negative in 105 (67.3%) of 156 cytology cases that were HR-HPV positive but had no histologic evidence of ≥CIN-2. HR-HPV was detected in 7 (63.6%) of 11 cytology cases that were negative for p16INK4a but were ≥CIN-2 on cervical biopsy, including 3 ASC-US, 2 LSIL, and 2 ASC-H cases. HR-HPV was negative in 19 (27.1%) of 70 cases that were positive for p16INK4a but had no evidence of ≥CIN-2 on colposcopy, with or without histologic assessment. Fifty-one (12.65%) cytology specimens, including 31 (18.9%) of 164 ASC-US, 7 (17.5%) of 40 LSIL, 4 (14.3%) of 28 AGC, 4 (11.8%) of 34 ASC-H, and 5 (12.5%) of 40 HSIL specimens were positive for both p16INK4a and HR-HPV but were negative for lesions ≥CIN-2. Four (0.01%) cytology specimens, including 1 NILM, 1 ASC-US, 1 AGC, and 1 HSIL specimen, were negative for both p16INK4a and HR-HPV but had ≥CIN-2 on biopsy. Histologic review and p16INK4a staining confirmed the diagnosis ≥CIN-2 in 3 of 4 of those cases, but 1 case showed no staining for p16INK4a on the tissue biopsy, despite consensus panel confirmation of the diagnosis of CIN-2. Clinical follow-up was performed by review of the surgical pathology records over a minimum of 12 months to track the outcome of HSIL specimens that had negative (<CIN-2) biopsies (n = 13). Six of these cases underwent subsequent loop electrosurgical excision procedure (LEEP), 4 had repeat colposcopy with cervical biopsies (n = 3) or ECC (n = 1), and 3 had no subsequent procedures. One of these cases was p16INK4a positive and had CIN-3 on a subsequent LEEP. The remaining cases had CIN-1 (n = 5), were negative for dysplasia (n = 4), or did not have subsequent procedures (n = 3).
Clinical Performance of p16INK4a, HR-HPV, and p16INK4a/HR-HPV in NILM, ASC-US, ASC-H, LSIL, and HSIL Cytology Specimens
p16INK4a testing of all NILM, ASC-US, ASC-H, LSIL, and HSIL cervical cytology specimens (all combined categories) was more specific than HR-HPV for the detection of underlying lesions ≥CIN-2 (P < .001) (Table 3). There were no differences, however, in the specificity of p16INK4a and HR-HPV detection in cytology specimens for underlying lesions ≥CIN-2 in AGC. For cytology cases (excluding AGC) that had negative colposcopy or negative biopsies, the p16INK4a cytology test was more likely to be negative compared with the HPV test (McNemar's test, P < .01). The sensitivities of p16INK4a and HR-HPV testing in any cytology specimen to detect underlying lesions ≥CIN-2 were nearly equivalent (81.7% and 83.3%, respectively; P = .81). There was improved PPV for p16INK4a versus HR-HPV (41.2% vs 24.2%, respectively) for underlying lesions ≥CIN-2 in the aggregate analysis of all categories and a trend for improved PPV for p16INK4a in ASC-US, LSIL, ASC-H, and HSIL, although there were no differences in the categories of NILM and AGC. Furthermore, no differences were found in the high negative predictive value of either p16INK4a or HR-HPV for the detection of underlying ≥CIN-2 in all combined categories (95.8% and 94.2%, respectively) or in any of the individual cytologic diagnostic categories.
Table 3. Clinical Performance of p16INK4a and HR-HPV Cytology Tests for the Detection of Underlying Lesions ≥CIN-2
The primary goals of this study were to develop an optimized and rigorously controlled p16INK4a cytology assay and to assess its clinical performance in comparison to HR-HPV detection, as diagnostic adjuncts for the triage of patients with abnormal cervical cytology. When cytologic diagnostic categories were evaluated in aggregate, p16INK4a and HR-HPV had similar sensitivity (81.7% vs 83.3%, P = .81), but p16INK4a was more specific than HR-HPV (78.1% vs 50.9%, P < .001) for the detection of underlying ≥CIN-2. Furthermore, the PPV of p16INK4a was superior to that for HR-HPV for underlying ≥CIN-2 (41.2% vs 24.2%; P = .001).
Over the last several years, >25 studies have been conducted to evaluate the performance of p16INK4a in cervical cytology, but none have included histologic or colposcopic correlation with cervical biopsy in an NILM test population or have included p16INK4a immunohistochemical studies on the corresponding biopsies. Previous reports used arbitrary criteria for scoring positive p16INK4a test results that ranged from any staining up to a threshold in many reports of ≥10 atypical cells.14, 17, 18 In the current study, we established the threshold of 30 stained abnormal cells/slide for positive p16INK4a test results, based on the analysis of test performance in a pilot series of NILM versus HSIL cases and in a second gold standard training series that encompassed NILM, ASC-US, LSIL, and HSIL cytology cases. By using this threshold, underlying lesions ≥CIN-2 were detected in only 5% of p16INK4a-negative cases but in 29% of p16INK4a-positive cases that had ASC-US, AGC, LSIL, or ASC-H. Although not rigorously evaluated in this study because of the low prevalence of underlying high-grade dysplasia in NILM and ASC-US cases, the selection of a lower p16INK4a threshold would tend to increase test sensitivity but decrease specificity in all diagnostic categories. As might have been anticipated, the application of more rigorous criteria for positive test results in the current study was associated with decreased sensitivity and negative predictive value, but improved specificity and PPV compared with most previous reports of p16INK4a test performance.10, 12, 13, 18
The comparison of p16INK4a status in cytology specimens to p16INK4a staining in the corresponding biopsies provided some insight into the potential causes of discordance in test results. Staining of scattered benign intercalated glandular cells, cells from tubal metaplasia, or cells from immature squamous metaplasia were detected in the biopsies of cases that had positive p16INK4a cytology test results but no evidence of high-grade dysplasia, including many cases with a cytologic diagnosis of AGC. Conversely, p16INK4a-negative cytology test results were often encountered in cases that had only minimal focal p16INK4a staining in the corresponding biopsy specimens. p16INK4a was positive in 96% of HSIL cases that had a biopsy diagnosis ≥CIN-2 but was only positive in 50% of HSIL cases that did not have high-grade lesions on cervical biopsy.
HPV testing was performed using a PCR-based protocol that has not yet undergone extensive clinical validation.16 This assay does, however, provide unambiguous typing data for the HPV types that commonly infect the female genital tract. The assay has been shown to detect all 16 genotypes when >102 copies of the plasmid is present in test samples,16 and our experience suggests that this assay is generally as sensitive in clinical samples as other PCR-based methods. In contrast to other methods that require RFLP analysis or reverse hybridization assays, the PCR capillary electrophoresis genotyping method is less time-consuming and provides unambiguous results when multiple HPV types are encountered. Although the Hybrid Capture 2 HPV DNA Test has been extensively validated, it could not be used for the current study because of the large volume (4 mL) of PreservCyt Solution that is required for testing without jeopardizing sample adequacy for p16INK4a immunocytochemistry. Of note, the proportion of ASC-US cases that tested positive for HR-HPV (65.2%) in the current study was similar to that observed in a large US military population study (61.2%) of patients between the ages of 18 and 22 years,19 and the prevalence of HR-HPV was found to be even higher (74%) in women aged <30 years from Sweden.20 Although data were not recorded for patient age, the high proportion of HR-HPV test results in the ASC-US group suggests that a relatively high proportion of younger women were enrolled in the current study. The high prevalence of HPV types 58 and 51 in the current study may reflect the demographics (a relatively high proportion of Hispanic patients) served by the colposcopy clinics of the University of Colorado Health Sciences Center and Denver Health Medical Center, which serve a relatively high proportion of Hispanic patients. Similarly, HPV type 58 was previously reported to have been detected in 28.5% of all HPV-positive women in a Mexican colposcopy population, and HPV type 51 was found to be the most common HPV type in patients with LSIL or HSIL-moderate dysplasia at the University of Oklahoma Medical Center.21, 22
This study is the first to focus exclusively on cytology specimens from patients who were evaluated by colposcopy. The sensitivity of any cytology test, however, could be biased by utilization of the colposcopic biopsy as the gold standard. Colposcopy has been reported to often fail to detect small ≥CIN-2 lesions and, as a result, the sensitivity of cytologic screening may be inflated by the use of the colposcopic biopsy alone as the gold standard, compared with the use of colposcopic biopsies in combination with random biopsies and endocervical curettage.23, 24 Although this effect has not been found to impact the sensitivity of HPV testing,24 it is possible that colposcopy sampling error could impact the apparent sensitivity of a p16INK4a-based screening approach.
One limitation of this study is that that the majority of cases with NILM cytology and some cases with ASC-US had negative colposcopy but did not undergo histologic assessment. Because all of these cases were referred for colposcopy on the basis of a prior abnormal cytologic test result (≥HR-HPV-positive ASC-US), it is likely that some CIN-2/3s were not detected by colposcopy. The number of NILM and ASC-US cases that had CIN-2/3 but negative colposcopy results, however, may be relatively low, considering the low prevalence of detected CIN-2 in the NILM and ASC-US groups (1.2% and 7.6%, respectively). Furthermore, none of the NILM or ASC-US cases with negative colposcopic examinations was found to have a subsequent cytologic or tissue biopsy diagnosis ≥HSIL/CIN-2. Although the failure to detect some CIN-2/3s by colposcopy could increase the apparent specificity of HR-HPV and p16INK4a cytology test results, the potential magnitude of this effect could not be determined in this study because of the expected low prevalence of underlying high-grade dysplasia in the NILM and ASC-US populations.
In summary, the p16INK4a immunocytochemical assay has similar sensitivity and superior specificity to HR-HPV testing for the detection of underlying high-grade lesions of the cervical mucosa in patients who are referred for colposcopy. The evaluation of only specimens from patients who were referred for colposcopy provided for the opportunity to correlate p16INK4a staining in the cytology specimens with the corresponding biopsy diagnoses and with the results of p16INK4a tissue staining, but these immunocytochemical data should not be extrapolated to a general screening population. Additional large-scale studies are still needed to validate the performance of the p16INK4a cytology assay as a screening test in low-risk populations. Furthermore, the determination of whether this should best be used as an independent test or in combination with HR-HPV testing for primary screening will await the outcome of large-scale prospective studies with clinical outcome correlation.
We thank David Heinz for his excellent technical assistance.
CONFLICT OF INTEREST DISCLOSURES
Supported by National Institutes of Health grants R21CA110519 and R33CA110519.