Early Detection and Diagnosis
Evaluation of a new p16INK4A ELISA test and a high-risk HPV DNA test for cervical cancer screening: Results from proof-of-concept study
p16INK4a, a cell cycle regulation protein, accumulates in abnormal epithelial cells infected with high-risk human papilloma virus (HPV). In immunostaining studies, p16INK4a has shown potential as a marker of high grade cervical intraepithelial neoplasia (CIN) and invasive cervical cancer. To evaluate its potential use in cervical cancer screening, we conducted a feasibility study to compare the performance of a new enzyme linked immunosorbant assay (ELISA) for p16INK4a (mtm laboratories, Heidelberg, Germany) to that of the Hybrid capture 2™ (hc2) test for high-risk HPV DNA for the detection of CIN3. Three hundred and nineteen women were referred from Western Washington Planned Parenthood clinics for colposcopy examination and cervical biopsy because of abnormal Pap test results. Cervical samples were obtained from study participants for p16INK4a ELISA, liquid-based cytology and hc2. The order (first and second) for obtaining samples for cervical cytology and p16INK4a ELISA changed with every other subject. Concentrations of p16INK4a protein were higher when the sample was taken before the cytology. The sensitivity of p16INK4a ELISA (concentration ≥ 8 units/ml) taken as first sample was 90.0% for CIN3, and the sensitivity of HC2 taken as a second sample was 85%. In the same group, the specificity of p16INK4a ELISA (46.9%) was slightly better than hc2 (35.4%) Results from this proof-of-concept study suggest that p16INK4a ELISA has a similar sensitivity and slightly better specificity for CIN3 compared to hc2. These findings support proceeding with a larger study with samples from a population of women presenting for routine cytology screening. © 2007 Wiley-Liss, Inc.
The Pap smear has been an effective screening tool for cervical cancer, but it is limited with respect to its sensitivity, cost and requirement of significant infrastructure and technology to be effective. Human papillomavirus (HPV) testing has been proposed as a possible screening tool in resource poor settings where Pap screening is not feasible; however testing for high-risk HPV deoxyribonucleic acid (DNA) is limited by its poor specificity.1 The search for specific biomarkers of HPV infected precancerous cells has yielded a number of possible markers of disease, including p16INK4a.
p16INK4a is a protein involved in cell cycle regulation and accumulates in abnormal epithelial cells infected with high-risk HPV. p16INK4a is a cyclin dependent kinase (CDK) inhibitor that regulates the activity of CDK4 and CDK6. It is inactivated in many cancers by genetic deletion or hypermethylation.2 In non-HPV associated tumors, this inactivation leads to increased CDK activity and inactivation of the retinoblastoma protein (pRb), resulting in disruption of the cell cycle and increased cell proliferation. However, in HPV-associated tumors, inactivation of pRb by high-risk HPV oncoprotein E7 leads to cell proliferation and markedly increased levels of p16INK4a. Staining of histology and cytology specimens for p16INK4a has shown that detection of antibodies to the protein is a sensitive and specific marker of high-grade cervical intraepithelial neoplasia (CIN) and invasive cervical cancer.2, 3, 4, 5 In a number of studies, p16INK4a staining of biopsy tissue was seen in over 90% of cases with high-grade CIN and invasive cervical cancer, being generally absent from the normal tissue and staining only zonal or patchy areas of the low-grade CIN cases.2, 6, 7 Findings from these immunostaining studies suggest that compared to high-risk HPV DNA, which is detected in the majority of both high- and low-grade lesions,8 p16INK4a might be a better marker for identifying those lesions most likely to progress.9
We evaluated a new cervical cancer screening test that is based on a protocol that includes lysis of the cervical sample and subsequent measurement of solubilized p16INK4a using an enzyme-linked immunosorbant assay (ELISA) test for p16INK4a (mtm laboratories, Heidelberg, Germany). Results were compared to HPV testing by Hybrid Capture 2 (hc2, Digene, Gaithersburg, MD). An unpublished pilot study from a German dysplasia clinic suggested that very high levels of p16INK4a protein was detected in virtually all lysed cervical specimens of women who had biopsy proven CIN (Wentzensen, unpublished data). We present results from a feasibility study evaluating p16INK4a protein levels in lysed cervical samples that was performed in an outpatient clinic setting in the United States. A protein-based assay with high sensitivity and specificity has the potential to provide a low cost and a low technology alternative to Pap screening. Moreover, this technology could be adapted to produce a point-of-care test, which could be considered a major advancement for cervical cancer screening programs worldwide.
Material and methods
This feasibility study was designed to enroll 400 women with abnormal cervical cytology who meet the criteria for colposcopy and biopsy referral, assuming an 8% prevalence of biopsy-confirmed CIN3. The actual prevalence was 11%, so enrollment stopped at 319 women. Women who had abnormal Pap screening at Western Washington Planned Parenthood clinics were referred for colposcopy and approached to partricipate in the study. Eligible women were between 18 and 50 years old, had an intact cervix with no history of hysterectomy or other surgeries on the cervix, were not immunocompromised and had had at least 1 male sex partner. Cervical samples were obtained from study participants for p16INK4a ELISA, cytology and high-risk HPV DNA testing by hc2 prior to colposcopy and biopsy. The Institutional Review Board at the University of Washington approved all study procedures prior to recruitment.
Ectocervical and endocervical samples for p16INK4a ELISA were obtained from the cervix with spatula and cytobrush, respectively, and placed in 5 ml of mtm sample lysis buffer medium (mtm laboratories, Heidelberg, Germany). Cervical specimens for liquid-based cytology were similarly obtained with a spatula and cytobrush and placed in PreservCyt® (Cytyc Corporation, Marlborough, MA). Residual PreservCyt liquid was tested by Hybrid Capture 2™ (hc2) (Digene, Gaithersberg, MD) for high-risk HPV DNA according to the manufacturer's specifications. The high-risk probe (Probe B) that targets oncogenic HPV types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59 and 68 was used. The order for obtaining samples (first or second) for cervical cytology (with hc2 testing of residual PreservCyt® liquid) and for p16INK4a ELISA changed with every other subject based on an assigned study number. During the first half of the study, the p16INK4a cytobrush sample was discarded after swirling and tapping the brush 10 times in the sample lysis medium. For the second half of the study, the cytobrush was retained in the vial containing the sample lysis medium until laboratory processing. The change in collection method was made when comparison with preliminary data from a German study revealed higher concentrations of p16INK4a. The German protocol included leaving the cytobrush in the vial after sampling (Wentzensen, unpublished data).
All study participants underwent colposcopically-directed cervical biopsy with possible endocervical curettage by a study gynecologist. A biopsy at 12 o'clock was performed if the colposcopic exam was satisfactory and no lesions were seen. A digital photograph was taken at the time of colposcopy and later used as reference to measure the size of the lesions. Lesion size was estimated using Cooper Surgical Cerveillance scope digital colposcopy system Model CS2000 (Cooper Surgical, Trumbull, CT) We examined the first 100 subjects who had colposcopy images that were suitable for measuring lesion size. All areas of acetowhite were included in the lesion size measurement regardless of severity by colposcopic appearance.
Residual PreservCyt® cytology specimens from all subjects were tested for high-risk HPV DNA by Hybrid Capture 2 (hc2) (Digene, Gaithersburg MD) according to the manufacturer's specifications. When the volume of the remaining specimen did not meet criteria for sample adequacy according to the manufacturer's instructions, the specimen was centrifuged and tested. If the result returned negative, then the test result was reported as “insufficient”. In our statistical analyses, samples that were insufficient for hc2 testing were combined with those that tested negative by hc2.
Colorimetric sandwich-ELISA analyses were performed using a p16INK4a specific monoclonal antibody for coating of the solid phase (microtiter plate) and a second p16INK4a-specific HRP-labeled monoclonal antibody for detection. Quantitation was achieved by construction of a calibration curve using known concentrations of p16INK4a standards. The concentration of test samples was delineated by interpolation from the standard curve and stated as arbitrary units (units/ml). Study samples were measured in duplicates, each with a volume of 100 μl.
Two study pathologists read all cervical biopsies, and diagnoses were classified as normal, atypia, CIN1, CIN2 or CIN3. Histology was coded as “atypia” when abnormalities of maturation or nuclear changes were seen, but the changes did not meet criteria for squamous or glandular dysplasia. None of the women had adenocarcinoma in situ or invasive cervical cancer. CIN3 was the main endpoint for the analysis of sensitivity and specificity.
The sensitivity of the p16INK4a ELISA and hc2 tests for biopsy proven CIN3 was based on the percentage of women with CIN3 who tested positive. Specificity was estimated by the percentage of women with ≤CIN1 who tested negative. Although this estimate of specificity is not analogous to estimates derived from screening populations where the majority of women have normal Pap test results, it does allow for a valid comparison of the relative specificity of the p16INK4a ELISA and hc2 tests. For hc2, samples with concentration ≥1 RLU were considered test positive, as recommended by the manufacturer. For p16INK4a ELISA, results were evaluated at a number of different cut-off levels used to define test positivity, ranging from ≥4 to ≥18 units/ml. All samples with concentrations at or exceeding the specified cut-point were considered to be positive in analyses of sensitivity and specificity. Ninety five percent confidence intervals for sensitivity and specificity were estimated using Clopper–Pearson exact methods.10 All tests were performed without the knowledge of other laboratory and clinical test results. χ2-tests were used to evaluate differences in proportions.
The association between p16INK4a concentration and lesion size was examined using a linear regression model that adjusted for histologic diagnosis of the lesion. The variables for lesion size (measured in mm2) and p16INK4a concentration (measured in units/ml) were both logarithmically transformed to eliminate statistical skew and allow for normally distributed data. Histologic diagnosis was coded as dummy variables representing negative, atypia, CIN1, CIN2 and CIN3. Analysis of variance (ANOVA) was also performed to determine if the size of a lesion was associated with p16INK4a concentration. Lesion size was evaluated as a continuous variable, and p16INK4a concentration was evaluated as a series of binary variables using a number of different cut-off levels for p16INK4a positive results, ranging from 6 to 15 units/ml.
Study subjects had a median age of 23 years (range of 18–44 years). Participants reported a lifetime median of 8 sex partners (range 1–250). Overall, 238 of 319 subjects (74.6%) tested positive for high-risk HPV DNA by hc2 (Table I). Based on the colposcopically-directed biopsy sample, 36 women (11.3%) had CIN3, 43 (13.5%) had CIN2, 238 (74.6%) had CIN1 or less, and 2 (0.6%) had biopsy samples that were inadequate for evaluation.
Table I. Demographic and Clinical Characteristics of the 319 Female Participants
| 18–19 yr||32 (10.0)|
| 20–24 yr||158 (49.5)|
| 25–29 yr||96 (30.1)|
| 30–34 yr||21 (6.6)|
| 35–50 yr||12 (3.8)|
| American indian||7 (2.2)|
| Asian||24 (7.5)|
| Black||12 (3.8)|
| Native Hawaiian||2 (0.6)|
| White||243 (76.2)|
| More than one race||23 (7.2)|
| Don't know||8 (2.5)|
| Hispanic||22 (7.0)|
| Not Hispanic||286 (90.8)|
| Don't know/declined to answer||7 (2.2)|
|hc2™ result for high-risk HPV DNA|
| Positive||238 (74.6)|
| Negative||57 (17.9)|
| Insufficient2||24 (7.5)|
|Current use of contraception|
| Hormonal method||239 (74.9)|
| Barrier method only||67 (21.0)|
| Other methods||13 (4.1)|
| 0 pregnancies||263 (83.2)|
| 1 pregnancy||33 (10.4)|
| 2 pregnancies||17 (5.4)|
| 3 pregnancies||3 (1.0)|
|Referral Pap test results|
| ASC-US,HPV positive||166 (52.0)|
| ASC-H||18 (5.6)|
| LSIL||129 (40.4)|
| HSIL||6 (1.9)|
When the p16INK4a sample was taken first, the percentage positive among women with CIN3 was high within the concentration range of 4–10 units/ml (Table II). High positivity for CIN3 with a reasonably low number of positive test results among women with ≤CIN1 was seen at a p16INK4a cut-off in the range of 8–12 units/ml. At each cut-off level, the percentage p16INK4a positive with CIN3 and ≤CIN1 was higher when the samples were obtained first rather than second, whereas for CIN2, the p16INK4a results did not show a clear pattern of association with sample collection order (Table II). There was a trend towards more insufficient hc2 results when the cytology sample was obtained second. When the PreservCyt® cytology sample was collected first, 10 of 158 samples (6.3%) were insufficient for hc2 and when it was collected second, 14 of 161 samples (8.7%) were insufficient for hc2 (p > 05).
Table II. p16INK4aConcentration Compared to Histology
|p16INK4a collected first||n = 20||n = 26||n = 113|
| cutoff: 4 units/ml||19 (95.0)||23 (88.5)||86 (76.1)|
| cutoff: 6 units/ml||19 (95.0)||20 (76.9)||72 (63.7)|
| cutoff: 8 units/ml||18 (90.0)||20 (76.9)||60 (53.1)|
| cutoff: 10 units/ml||17 (85.0)||17 (65.4)||49 (43.4)|
| cutoff: 12 units/ml||16 (80.0)||14 (53.9)||43 (38.1)|
| cutoff: 15 units/ml||12 (60.0)||10 (38.5)||29 (25.7)|
| cutoff: 18 units/ml||11 (55.0)||10 (38.5)||22 (19.5)|
|p16INK4a collected second||n = 15||n = 17||n = 125|
| cutoff: 4 units/ml||12 (75.0)||16 (94.1)||87 (69.6)|
| cutoff: 6 units/ml||12 (75.0)||14 (82.4)||64 (51.2)|
| cutoff: 8 units/ml||10 (62.5)||13 (76.5)||42 (33.6)|
| cutoff: 10 units/ml||10 (62.5)||10 (58.8)||35 (28)|
| cutoff: 12 units/ml||7 (43.8)||9 (52.9)||29 (23.2)|
| cutoff: 15 units/ml||5 (31.3)||9 (52.9)||20 (16.0)|
| cutoff: 18 units/ml||4 (25.0)||6 (35.3)||15 (12.0)|
Using ≥8 units/ml as a cut off, 163 women had a positive p16INK4a ELISA compared to 236 women with a positive hc2. The sensitivity of p16INK4a ELISA (concentration of ≥ 8 units/ml) for CIN3 was higher when the sample was collected first rather than second (p = 0.05). The sensitivity of p16INK4a ELISA for CIN3 was comparable to that of hc2 (p = 0.48), regardless of the order of sample collection (Table III). Relative to the hc2 test, the p16INK4a ELISA test was more specific with fewer false positive screening results among women with ≤CIN1 (p < 0.0001). The difference in specificities of the 2 tests was more pronounced when the Pap sample used for hc2 testing was collected before (p < 0.0001) rather than after (p = 0.058) the p16INK4a sample (Table III).
Table III. Estimates of sensitivity and specificity for the p16INK4a ELSIA test (cut-off of 8 units/ml) and the hc2 test by order of specimen collection
|p16INK4aspecimen collected first||n = 20||n = 26||n = 113|
| p16INK4a||90.0 (68.3, 98.8)||76.9 (56.4 91.0)||46.9 (37.5, 56.5)|
| hc2™3||85.0 (62.1, 96.8)||88.5 (69.8, 97.6)||35.4 (26.6, 45.0)|
|hc2™specimen collected first||n = 16||n = 17||n = 125|
| p16INK4a||62.5 (35.4, 84.8)||76.5 (50.1 93.2)||66.4 (57.4, 74.6)|
| hc2™3||81.3 (54.4, 96.0)||100.0 (82.4, 100.0)4||25.6 (18.2, 34.2)|
Although the numbers were small, it appeared that leaving the cytobrush in the collection vial improved the sensitivity of the p16INK4a test, especially when the sample was collected following collection of the cytology sample (data not shown).
We evaluated whether lesion size (any acetowhite changes) was related to the concentration of p16INK4a in 100 subjects. We found that CIN3 lesions were significantly larger than less severe lesions. Mean CIN3 lesion size was 85.5 mm2 and for lesions showing CIN2, CIN1 or atypia mean size was 51.6 mm2. With a cut-off of ≥8 units/ml there was no significant difference in lesion size between women with positive p16INK4a and women with negative p16INK4a results (Table IV). However, when a linear regression model was used to evaluate the relationship between lesion size and p16INK4a results, there appeared to be a correlation between increasing p16INK4a level and lesion size even after adjusting for lesion grade (p = 0.02).
Table IV. P16INK4A Concentration Compared to Lesion Size in 100 Women
|6 units/ml||Negative||37||29.4||16|| |
|8 units/ml||Negative||55||38.9||16|| |
|10 units/ml||Negative||63||35.8||14.9|| |
|12 units/ml||Negative||71||38.2||17.0|| |
|15 units/ml||Negatine||82||37.2||17.5|| |
Results from this feasibility study of a new p16 INK4a ELISA test suggest that it might be useful in cervical cancer screening. Previous reports have shown good sensitivity and specificity of immunocytochemical staining of cervical samples, both histology and cytology.2, 3 Some authors have proposed that p16INK4a staining could be used to triage women with low-grade cytology results and predict who is more likely to have CIN3 or worse.4, 5 HPV transformed cells with high levels of p16INK4a can be identified with a protein-based assay. Thus, a biochemical test for p16INK4a might be better than a test for high-risk HPV DNA at identifying the subset of HPV-infected women with significant clinical disease.
It is important to consider sources of contamination from other cell types for a protein marker to be clinically useful. Degenerating endometrial cells have shown some staining for p16INK4a.11 Such cells might pose a source for false positive tests. In another study, Trichomonas vaginalis showed reactivity with some p16INK4a antibodies.12 However, the specificity of the antibody used in that study was different from the specificity of the antibody used for the p16INK4a ELISA in this study. In addition, some authors have suggested that p16INK4a might not be as useful for detection of glandular lesions (adenocarcinoma in situ or adenocarcinoma). Although glandular dysplasia and adenocarcinoma both can stain strongly for p16INK4a, staining also occurs with benign cells showing tubal endometriod metaplasia.11, 13 Since sporadic p16INK4a positivity has been noted in endometriod metaplasia and in some columnar cells, this may be another potential source for higher levels of p16INK4a protein in lysed cervical specimens.
We found that the best trade-off between high sensitivity and high specificity for the p16INK4a ELISA was seen when the “cut-off” concentration for a positive test was between 6–12 units/ml. Our analysis indicated that ≥8 units/ml was a reasonable choice. Although 1 additional case of CIN3 would have been detected if the cut-off level was lowered from ≥8 to ≥6 units/ml, thereby raising the estimated sensitivity from 90 to 95%, specificity at this cut-off level dropped, resulting in a 20% increase in the number of false positive test results.
Sampling order was important when another cervical sample was taken together with the p16INK4a ELISA sample. Collection of the p16INK4a ELISA sample first resulted in significantly higher estimated sensitivity compared to collection of the p16INK4a sample second (after the liquid-based cytology sample).
The sensitivity of hc2 for CIN3 in our study (85% as second and 81% as first sample) was lower than reported for many but not all previous studies.14 If we had excluded observations with indeterminate hc2 results (i.e., samples that were insufficient for hc2 testing) from our analysis, as has been done in some previous studies, the estimated sensitivity of hc2 for CIN3 would increase to 94%. Estimates of sensitivity for a cervical cancer screening test also depend on whether all or only subsets of women who are referred for colposcopy undergo standardized biopsy. When all women referred for colposcopy undergo biopsy versus only a subset, fewer CIN3 or AIS lesions, particularly small lesions and lesions high in the canal or deeper in a glandular crypt, will be missed. These small lesions are difficult to sample with a swab, broom or brush for testing by hc2, cytology, or p16INK4a. Therefore, the sensitivity of a cervical screening test will be lower in studies that require biopsy of all women referred for colposcopy compared to those that do not require biopsy. Finally, because no cervical cancer screening test is 100% sensitive, comparisons of 2 highly sensitive tests in a single study, such as 2 high risk HPV DNA tests or 1 high risk HPV DNA test and 1 p16INK4a test, will cause both tests to appear to be less sensitive than if they were compared separately to a less sensitive test, like cervical cytology, or if each test was evaluated individually. For example, cervical cytology appeared to be much more sensitive for ≥CIN3 before its performance was compared in the same study to that of high risk HPV DNA testing.
An important limitation of this study was that the women enrolled were not from a population undergoing routine cervical cancer screening, but rather they were from a population sample that was enriched for CIN. However, it is important to demonstrate feasibility for detection of endpoints such as CIN3 prior to conducting a large-scale, population-based screening study. While our estimates of sensitivity should not differ substantially from those derived from other populations of young women undergoing cervical cancer screening, our estimates of specificity are undoubtedly lower. Although our absolute measures of specificity are not meaningful in the context of a screening population, relative differences in the specificity estimates for the p16INK4a and hc2 tests are valid measures of what is likely to be observed in a screening population.
The p16INK4a ELISA assay shows promise as a potential screening test for CIN3 with an estimated sensitivity in the range of 80–95%. This level of sensitivity is comparable to what has been observed for hc2 testing. Moreover, relative to the hc2 test, the p16INK4a ELISA test was more specific, resulting in fewer false positive test results. Findings from this feasibility study support proceeding with larger screening studies designed to collect samples for p16INK4a ELISA at the time of routine cytology screening.
If future studies demonstrate high accuracy and reliability, the p16INK4a ELISA test might be especially useful in developing nations as a point-of-care screening tool for CIN3.
mtm-laboratories provided protocols, sample lysis medium for p16INK4a and performed ELISA test in their laboratory. The authors certify that they have not entered into any agreement that could interfere with their access to the data on the research, nor upon their ability to analyze the data independently, to prepare manuscripts and to publish them. Authors R.R., A.R., M.H. and M.V. K-D. have declared a financial interest in mtm-laboratories; these authors are listed as inventors on the patent applications, and R.R. and M.V.K-D. have shareholder interest in the company.