Role of P16(INK4a) expression in identifying CIN2 or more severe lesions among HPV-positive patients referred for colposcopy after abnormal cytology
p16 is strongly overexpressed in dysplastic cervical cells because of the transforming activity of the E7 oncogene of all high-risk human papillomavirus (HR-HPV) types and may be easily revealed by immunochemistry: p16 may, therefore, be considered a surrogate marker for the activated oncogene expression of HR-HPV in dysplastic cervical cells.
HPV and p16INK4a testing were performed in a consecutive series of 283 patients with abnormal cytology referred to colposcopy assessment or follow-up. Triage of patients to colposcopy by HPV or HPV and p16 testing was simulated, and the relative sensitivity, specificity, and positive predictive value (PPV) of HPV and p16 testing for > CIN2 lesions was determined as well as the cost balance of the two triage types.
Compared to current protocol, triage by HPV testing reduced the number of colposcopies by 44.2%, but also reduced the > CIN2 detection rate by 10.7%, and was associated with a cost of € 54.16 per assessed woman and of € 613.20 per > CIN2 detected. Compared with current protocol, triage by HPV and p16 testing combined reduced the number of colposcopies by 73.1%, but reduced > CIN2 detection rate by 21.5%, and was associated with a cost of € 54.73 per woman assessed and of € 704.09 per > CIN2 detected.
Triage by HPV and p16 improves considerably the PPV of diagnostic assessment, but decreases > CIN2 detection rate, and is associated with substantially higher costs. Further decrease of molecular immunochemistry testing due to technological progress may allow HPV and p16 testing to become a cost effective procedure in the future. Cancer (Cancer Cytopathol) 2006. © 2006 American Cancer Society.
The protein p16INK4a (henceforth referred to as p16) is a cellular protein involved in cell-cycle regulation, and its expression is tightly controlled in normal cells. In normal, nondysplastic cells, p16 protein is expressed at a very low level (p16 can be expressed physiologically in a few cells, especially those undergoing squamous metaplasia, during this transdifferentiation process) and is almost undetectable by immunochemistry. On the contrary, due to the transforming activity of the E7 oncogene of all high-risk human papillomavirus (HR-HPV) types,1 p16 is strongly overexpressed in dysplastic cervical cells and may be easily revealed by immunochemistry. Therefore, p16 may be considered a surrogate marker for the activated oncogene expression of HR-HPV in dysplastic cervical cells.2–8
p16 has been shown to be associated with HPV-infected high-grade lesions but its positive predictive value (PPV) and sensitivity in prospective follow-up for relevant outcomes (> CIN2) has yet to be determined.9 To assess the accuracy of p16 in identifying high-grade cervical lesions, we performed p16 assay and HR-HPV testing on a consecutive series of women referred to colposcopy for abnormal cytology. The results were compared with colposcopy and biopsy findings. The possible role of p16 testing to select women with abnormal cytology and positive HPV testing to colposcopy is discussed.
MATERIALS AND METHODS
The studied series consisted of 283 patients consecutively referred to colposcopy within the Florence (Italy) District screening program for cervical cancer. According to the ongoing protocol, referral to colposcopy was prompted by 1) cytologic evidence of low-grade squamous intraepithelial lesion (LSIL) or more severe, 2) persisting atypical squamous cells of undetermined significance (ASCUS) at 6-month repeat cytology10 or 3) surveillance after loop excision for CIN 2–3.9 As the PPV of high-grade squamous intraepithelial lesions (HSIL) is sufficiently high to justify colposcopy assessment, we limited our study to patients referred to colposcopy for less severe cytology findings (LSIL or less), commonly associated with a low PPV for high-grade lesions. Cytologic reports in the studied series of patients was ASCUS in 130, LSIL in 124, and negative (followed up cases) in 29, respectively. Patient ages were < 30 years in 52, 30–39 years in 133, 40–49 years in 83, and > 49 years in15 patients, respectively.
HPV and P16 testing were performed in the whole series prior to colposcopy assessment: cervical material was collected using ThinPrep® (Cytic Corp., Boxborough, MA), allowing for multiple slide preparation and residual fluid. Laboratory operators performing the testing were blinded to the colposcopy assessment outcome.
From each specimen, 2 mL of residual ThinPrep fluid was used, and DNA extraction was carried out using a QIAmp DNA Mini Kit (Qiagen Corporation, Venlo, the Netherlands) according to the manufacturer's protocol. Polymerase chain reaction (PCR) analysis was performed according to a previously described protocol,11–13 using primers for the E6/E7 region of high risk HPV types (HPV 16, 18, 31, 33, 35, 45, 52 and 58). For a quality control of DNA extraction, the primer set PC04 and Gh20 was employed to amplify a 268-base pair (bp) fragment of the human beta globin gene in all specimens. In each PCR reaction, negative and positive controls were introduced.
From each specimen, 2 ml of residual ThinPrep fluid was used for a cytospin preparation; after cytocentrifugation (5 min at 1000 rpm), slides were air dried for 10 minutes, then treated with spray fixation reagent, containing polyethylene glycol, and immunostained within 24 hours. Before they were immunostained, all spray-fixed specimens were incubated in 50% volume/volume alcohol, followed by one washing step in deionized water. For immunostaining, CINtectm p16 Cytology kit (Dako Cytomation, now Dako A S, Glostrup, Denmark) was used, according to the manufacter's protocol. In brief, smears were treated with 3% hydrogen peroxide and then submitted for epitope retrieval at 95–99 °C for 40(±1) minutes; after cooling, the p16 antibodies were applied for 30 (±1) minutes and then a reagent for observation and substrate-chromogen solutions were added. Hematoxylin was used as counterstain. The methodology differs from other studies on p16,2, 4 but we used the same monoclonal antibody and believe that the results are comparable. The choice of cytospin preparation was essentially aimed at a more efficient use of the residual ThinPrep fluid. Before the study, we made a comparison of p16 testing on cytospin and ThinPrep preparations on limited numbers of negative and positive samples (data not shown), and we observed no differences.
Slides were read by two investigators blinded to final outcome, and a minimum of 500 cells in different fields were analyzed. A negative result was defined if no cells immunoreactive to the p16 antibody were in evidence. Slides showing positive staining for p16 were categorized on the basis of the percentage of positive cells as: < 5%, 5–10 % or > 10%. The cellular staining site was also evaluated and categorized as 1) nuclear, 2) cytoplasmic, or 3) nuclear plus cytoplasmic.
The frequency of high-risk HPV and p16 testing was investigated in the whole series, as well as its distribution by cytology report and final outcome. Final outcome was defined according to colposcopy-directed biopsy result (< CIN2 or > CIN2) and was assumed to be negative in the presence of negative colposcopy, indicating no biopsy. In particular, we assessed the proportion of detected > CIN2 lesions which were HPV and p16 positive, to assess the sensitivity, specificity, and PPV of HPV and p16 testing when used to triage to colposcopy those patients with abnormal cytology. Observed differences were tested by the chi-square test, with statistical significance set at P < 0.05.
We attempted a cost analysis of three different scenarios, namely 1) current selection to colposcopy on the basis of abnormal cytology/follow-up, 2) triage by HPV testing prompting colposcopy only in HPV-positive patients, or 3) triage by HPV and p16 testing, prompting colposcopy only in HPV- and p16-positive patients. Comparisons between scenarios were done in terms of cost per assessed woman and cost per > CIN2 lesion detected. Itemized unitary costs were based on National Health Service (Italy) current tariffs.
Table 1 shows the distribution of patients by final outcome (negative, < CIN2, > CIN2) according to age and HPV and p16 testing. Overall, 28 > CIN2 (CIN2 = 22, CIN3 = 6) cases were detected among 283 patients (PPV = 9.8). The PPV of different cytologic categories was 6.8% for ASCUS, 14.5% for LSIL and 3.4% for a negative (followed up cases) report, respectively. HPV positivity rate was 44.2% among < CIN1, 69.1% among CIN1, and 89.2% among > CIN2 patients (chi-square for trend < 10−6). The sensitivity, specificity, and PPV of HPV testing for > CIN2 were 89.2% (25 of 28), 47.8% (122 of 255), and 15.8% (25 of 158), respectively.
Table 1. Distribution of Studied Cases by Final Colposcopy Outcome (Negative/<CIN2 or >CIN2) According to Age, Cytological Report, Positive HPV Testing (Data in Parentheses), and p16 Testing
|Age category in yrs|| || || |
| < 30||24||19||9|
| > 50||15||0||0|
|Cytology|| || || |
| ASCUS||91 (31)||30 (24)||9 (9)|
| LSIL||59 (37)||47 (30)||18 (15)|
| Neg (follow-up)||24 (9)||4 (2)||1 (1)|
|HPV testing|| || || |
| Negative||97||25||3 (CIN2)|
|Total||174||81||28 (CIN2 = 22, CIN3 = 6)|
|p16 testing (on HPV positives)|| || || |
| Negative||56||23||3 (CIN2=3)|
| Positive||19||31||22 (CIN2 = 16, CIN3 = 6)|
|p16+|| || || |
| < 5% cells||8||7||8|
| 5–10% cells||7||14||6|
| > 10% cells||4||10||8|
Because of inadequate material, p16 testing was not possible in 4 (final outcome: negative = 2, CIN1 = 2) of 158 HPV positive cases. Table 1 shows results of p16 testing in the remaining 154 HPV positive cases. p16 positivity rate was 25.3% among < CIN1, 57.4% among CIN1, and 88.0% among > CIN2 patients (chi-square for trend < 10−6). All CIN3 cases were positive at p16 testing. Sensitivity for > CIN2 was 88.0% (22 of 25), specificity was 61.2% (79 of 129), and PPV was 30.5% (22 of 72). The number of p16 positive cells did not correlate with the probability of > CIN2 (< 5% = 36.3%, 5–10% = 27.2%, > 10% = 36.3%) and was no further considered as a relevant variable. At immunocytochemistry, the large majority of p16-positive cases showed both nuclear and cytoplasmic staining: the site of cellular staining showed no association with final outcome (data not shown), and this variable was also discarded as of no diagnostic value. There was no background staining for p16 in any slide, although some staining was rarely observed in endocervical and inflammatory cells.
Table 2 shows the results of performance and cost analysis for three compared modalities of selecting patients to colposcopy. The current protocol was associated to a cost of € 41.34 per assessed woman and of € 417.85 per > CIN2 detected. Compared with current protocol, triage by HPV testing reduced the number of colposcopies by 44.2%, reduced > CIN2 detection rate by 10.7%, and was associated with a cost of € 54.16 per assessed woman and of € 613.20 per > CIN2 detected. Compared with current protocol, triage by HPV and p16 testing combined reduced the number of colposcopies by 73.1%, reduced > CIN2 detection rate by 21.5%, and was associated with a cost of € 54.73 per woman assessed and of € 704.09 per > CIN2 detected.
Table 2. Cost Analysis for Three Different Criteria of Selection to Colposcopy
|Punch biopsy + histology||40.00||151||6040.00||92||3680.00||58a||2320.00|
|> CIN2 detected||—||28||—||25||—||22||—|
|Cost per assessed woman||—||—||41.34||—||54.16||—||54.73|
|Cost/> CIN2 detected||—||—||417.85||—||613.20||—||704.09|
Although HPV testing has been successfully used and proposed for triaging to colposcopy those patients with minor cytologic abnormalities,11, 14, 15 its PPV is suboptimal, and a substantial proportion of patients are still referred unnecessarily to colposcopy. Identifying other molecular events associated with progression from low- to high-grade lesions is a crucial area of research,9 as it may further improve selection of HPV-positive patients really worthy of assessment (and treatment). The present study, based on a relatively large consecutive series, allowed evaluation of how HPV and p16 testing may perform when used to triage to colposcopy those patients with cytologic abnormalities associated with a moderate PPV for > CIN2.
As to the study design, the aim of the present study was to test a triage option of minor cytologic abnormalities within a screening program in which colposcopy is employed as the standard assessment modality for screened positive cases, as recommended by European guidelines. Thus, we adopted colposcopy and colposcopy-directed biopsy findings as the gold standard for the presence of high-grade dysplasia, although we are aware that some underestimation might have occurred, as the sensitivity of colposcopy has been reported to be suboptimal.16
Our findings confirmed the relatively low PPV associated with two cytology reporting categories, persistent ASCUS (6.8%) and LSIL (14.5%). They also confirmed that HPV testing is associated with a relatively good sensitivity (89.2%) for > CIN2 and also with a nonsignificant (P = 0.16) and rather limited improvement of PPV (15.8% vs. 10.6%) with respect to abnormal cytology that currently prompts colposcopy assessment. The sensitivity of HPV in the present study was comparable to literature results, although better results have also been reported. Testing for a larger range of HPV types could maximize sensitivity but would also reduce specificity and advantages of triage. Moreover, in a previous study comparing PCR (primers for types 16, 18, 31, 33, 52, 58) to Hybrid Capture II (primers for types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58 and 68), we observed almost the same sensitivity in the two methods for high-grade histologic lesions.18
Critical features of evaluating p16 positivity are debated in the literature2–5, 8, 17: our experience showed that the proportion of stained cells and the site of cellular staining are not relevant for clinical purposes, as they are associated with almost unchanged positive PPV for high-grade dysplasia. Although p16 cannot be recommended at the present time as a primary screening tool, the finding that staining features are not a relevant variable would facilitate automated screening of cervical slides to detect p16 positivity, as suggested by Trunk et al.6
p16 testing of HPV-positive cases showed a relatively good sensitivity (88.0 %) and a double PPV (30.5%) compared with HPV testing (P = 0.016). This finding compares well with previous reports suggesting a high association of p16 positivity with high-grade cervical dysplasia.2–4, 6–8 However, if triage criteria is adopted to prompt referral only in HPV- and p16-positive cases, then the > CIN2 detection rate would be reduced by 21.5% (3 HPV-negative and 3 HPV-positive or p16-negative out of 28 total > CIN2 cases). On the other hand, triage by HPV and p16 would reduce colposcopy assessment by 73.1% (125 HPV-negative and 82 HPV-positive or p16-negative of 283 total patients referred to colposcopy) and would increase the PPV for > CIN2 of a call for colposcopy from 10.6% (based on cytology only) to 30.5%.
The study design was aimed at investigating the role of p16 in improving the specificity of HPV testing for high-grade dysplasia. Thus, p16 testing was obviously limited to patients with a positive PCR testing for 16, 18, 31, 33, 35, 45, 52 and 58 high-risk HPV types. Had we tested for a wider range of HPV types, the prevalence of HPV- and p16-positive cases in the overall series might have been greater.
Cost analysis based on current expenditures showed that triage by HPV or HPV and p16, while reducing the cost related to colposcopies and biopsies by 41% or 67.1%, would cause a 31% or a 32% increase in overall costs and a 46% or a 68% increase in cost per > CIN2 lesion detected, compared with current assessment protocol. Of course, the present cost analysis is valid only for the Italian scenario, where colposcopy is quite cheap compared with molecular testing, and not for other settings, where the cost of single procedures may be different.
Our data show that triage by HPV and p16 improves considerably the PPV of diagnostic assessment but decreases > CIN2 detection rate by 22% and is associated with substantially higher costs. Although missing a minority of CIN2 (no CIN3 were missed) lesions, most of which are going to regress, might not have a major negative impact on screening efficacy, this may be acceptable if associated with a major cost sparing, but not if it causes a cost increase. Unfortunately, the present cost of molecular methodologies is too high to allow a favorable cost balance. Nevertheless, if the cost of molecular testing, which has been decreasing in recent years, would further decrease with technological progress, triage to colposcopy of patients with cytologic abnormalities associated with a moderate PPV for > CIN2 by using HPV and p16 testing may become a cost-effective procedure in the future.