|Home | About | Journals | Submit | Contact Us | Français|
Results from a prototype real-time PCR assay that separately detected human papillomavirus genotype 16 (HPV16), HPV18, and 12 other carcinogenic HPV genotypes in aggregate (cobas 4800 HPV test) and results from a PCR assay that detects 37 HPV genotypes individually (Linear Array) were compared using a convenience sample of cervical specimens (n = 531). The percentage of total agreement between the two assays was 94.7% (95% confidence interval, 92.5 to 96.5%). The Linear Array test was more likely than cobas 4800 HPV test to test positive for the 12 other carcinogenic HPV genotypes among women without evidence of cervical disease (P = 0.004).
Persistent cervical infections by approximately 15 carcinogenic human papillomavirus (HPV) genotypes cause cervical cancer and its immediate precursor lesions (17, 22, 27). DNA testing for carcinogenic HPV has proven itself to be more sensitive, albeit slightly less specific, for the detection of cervical intraepithelial neoplasia lesion grade 2 (CIN2) and more severe lesions (CIN2+) (1, 10, 11, 15, 18, 19). Carcinogenic HPV DNA testing of cervical samples has been introduced as an adjunct to cervical cytology into primary cervical cancer screening in the United States (20, 26), and the International Agency for Research on Cancer has endorsed its use as a stand-alone option in primary cervical cancer screening (14).
The introduction of carcinogenic HPV DNA testing with cervical cytology as a primary screening modality creates a clinical dilemma—the management of carcinogenic HPV DNA-positive, cytology-negative women (4), who remain at a low but significant risk of CIN2+. One solution for identifying women at risk for CIN2+ in this subgroup is the separate detection of HPV genotype 16 (HPV16) and HPV18, the most carcinogenic HPV genotypes, referring HPV16- or HPV18-positive women immediately to colposcopy and following up with HPV16- and HPV18-negative women after a year (13, 26). While the first generation of clinical HPV tests pooled all carcinogenic HPV genotypes and did not include separate HPV genotyping, at least for HPV16 and HPV18, most of the next generation of clinical tests will offer at least HPV16 and HPV18 genotyping. The utility of full HPV genotyping is less certain (2), as it has been shown that repeatedly testing positive for any carcinogenic HPV is a good proxy for HPV genotype-specific persistence, especially in women aged 30 years and older (7).
One of the next-generation tests, the cobas 4800 HPV test, detects a pool of 12 carcinogenic HPV genotypes in aggregate, with concurrent, separate detection of HPV16 and HPV18. We conducted the first evaluation of the cobas 4800 HPV test and compared the results to previously reported results (8) generated by a now well-validated Linear Array HPV genotyping test (5, 6, 9, 12) and to cervical diagnoses.
We acquired a convenience sample of 552 anonymized residual PreservCyt (PC; Hologic) specimens, after cytologic interpretation and the histologic diagnosis had been rendered, with human subject research approvals as previously described (3, 8). Women with abnormal cytology, including carcinogenic HPV-positive atypical squamous cells of undetermined significance, underwent colposcopy. Women were assigned a worst histologic diagnosis based on the review of the biopsy specimen and loop electrosurgical excision procedure tissue sample and a pathology review as previously described (3, 8). Five sets of 2-ml aliquots were produced from the residual specimens, balanced for the aliquot order, and were stored at −20°C until used; some sets were missing aliquots because of insufficient specimen volume. Of the 552 PC specimens, nine (1.6%) were excluded because women either had unsatisfactory histology or had disease unrelated to cervical neoplasia (e.g., endometrial carcinoma).
In 2006, one set of aliquots was tested for 37 HPV genotypes (genotypes 6, 11, 16, 18, 26, 31, 33, 35, 39, 40, 42, 45, 51 to 56, 58, 59, 61, 62, 64, 66 to 73, 81, 82 subtype [IS39], 82 subtype [W13b], 83, 84, and 89) by using the commercially available Linear Array HPV genotyping test (Roche Molecular Systems) as previously described (5, 8, 12). For Linear Array testing, 285 μl of PC was processed, and the extracted DNA was eluted into 140 μl, of which 50 μl was used in the PCR amplification reaction (0.51% of the 20-ml PC specimen).
In 2009, a second set of aliquots was tested for HPV by the cobas 4800 HPV test (Roche Molecular Systems), masked to the results from Linear Array testing. The cobas 4800 system features fully automated sample preparation combined with real-time PCR technology, plus software that integrates the two components. The cobas 4800 HPV test was designed for the amplification and detection of a broad spectrum of high-risk HPV genotypes, along with the coamplification of the human cellular gene β-globin. Briefly, 94 PC aliquots plus a positive and negative control were processed concurrently. HPV and cellular DNA were released by lysing PC aliquots under denaturing conditions at elevated temperatures in the presence of proteinase K, chaotropic agents, and detergents. Isolation and purification of the released nucleic acid occurred on magnetic beads, followed by elution with a low-salt reagent. PCR amplification and detection occurred in a single tube, where probes with four different reporter dyes track the different targets in the multiplex reaction. Reporter dye 1 tracked the high-risk HPV pool with 12 high-risk HPV targets (HPV31, -33, -35, -39, -45, -51, -52, -56, -58, -59, -66, and -68); dyes 2 and 3 tracked HPV16 and -18, respectively; and dye 4 targeted β-globin to provide a control for cell adequacy, extraction, and amplification. Probe cleavage during the amplification process generated an increase in fluorescence, resulting in a sigmoid growth curve in the ideal case. A proprietary algorithm that models the expected curve was employed to generate a cycle threshold (CT) for determining the presence of HPV or β-globin target. These results were generated with a prototype system where a validated clinical cutoff was not available and thus a conservative cutoff of 45 was used to determine positivity. For the cobas 4800 HPV test, 400 μl of PC was processed, and the extracted DNA was eluted into 150 μl, of which 25 μl was used in the real-time PCR amplification reaction (0.33% of the 20-ml PC specimen).
HPV16, -18, -31, -33, -35, -39, -45, -51, -52, -56, -58, -59, -66, and -68 were classified as carcinogenic HPV genotypes. HPV testing results from Linear Array and cobas 4800 HPV tests were also categorized hierarchically according to cancer risk (HPV risk category) as follows: HPV16 > HPV18 > carcinogenic HPV types other than HPV16 and HPV18 > negative for carcinogenic HPV.
Seven (1.3%) of the 543 eligible specimens did not have an aliquot for Linear Array testing: three were judged invalid because of the absence of a β-globin signal, and four aliquots were missing. Of the 536 specimens with Linear Array results, two (0.4%) did not have an aliquot for testing by the cobas 4800 HPV test, and three (0.6%) were negative for β-globin by the cobas 4800 HPV test and were excluded from this analysis. This analysis was restricted to women with satisfactory Linear Array and cobas 4800 HPV test results and a final diagnosis which was categorized as normal (which included women with no biopsy specimens taken based on a colposcopic impression of normality) or histologic diagnoses of CIN1, CIN2, or CIN3, or more severe (CIN3+) (n = 531). The hierarchical HPV results for Linear Array and cobas 4800 HPV tests were compared using contingency tables. Kappa value, percentage of total agreement, and percentage of positive agreement were calculated. Differences in HPV categorization and percentage of positive agreement were tested for statistical significance (P < 0.05) using an exact symmetry or McNemar's χ2 test, respectively.
The comparison of the results of the Linear Array and cobas 4800 HPV tests, according to the cancer risk categories, is shown in Table Table1.1. The percentage of total agreement was 94.7% (95% confidence interval [95% CI], 92.5 to 96.5%), and the kappa value was 0.92 (95% CI, 0.89 to 0.95). There was a significant difference in the categorization of the pooled 12 carcinogenic HPV genotype results (P = 0.003), primarily due to the tendency for the Linear Array test to test positive and the cobas 4800 HPV test to test negative for carcinogenic HPV genotypes other than HPV16 and HPV18. When the results of the tests for all 14 carcinogenic HPV genotypes were pooled, there was 95.5% (95% CI, 93.3 to 97.1%) total agreement and a kappa value of 0.91 (95% CI, 0.87 to 0.94) between the two tests. The Linear Array test was more likely to test positive for any of the 14 carcinogenic HPV genotypes (P = 0.002) than was the cobas 4800 HPV test.
In Table Table2,2, we compared the HPV tests for the detection of HPV16, HPV18, any carcinogenic HPV genotypes other than HPV16 and HPV18, or any carcinogenic HPV genotypes for all women and stratified the results based on the final-diagnosis category. The agreement between the two HPV tests was in general very good to excellent, with ≥90% total agreement, ≥78% positive agreement, and kappa values of 0.81 or greater.
There were no statistically significant differences in the detection of HPV16 or HPV18 overall or by diagnostic category. There were no statistically significant differences in the detection of the pool of carcinogenic HPV genotypes other than HPV16 and HPV18, except that the Linear Array test was more likely than the cobas 4800 HPV test to test positive for carcinogenic HPV genotypes other than HPV16 and HPV18 among women who were without evident cervical histopathology (P = 0.004). Consequently, the Linear Array test was more likely than the cobas 4800 HPV test to test positive for any carcinogenic HPV genotype among women who were without evident disease (P < 0.001) and also for all women (P = 0.002).
We could not assess the exact number of carcinogenic HPV genotypes present, because the cobas 4800 HPV test detects the non-HPV16/18 carcinogenic HPV genotypes as a pool. However, we combined the three results for the cobas 4800 HPV test for the minimum number of HPV genotypes present (1, 2, or 3) (e.g., if positive for HPV16 and for the pool of 12 carcinogenic HPV genotypes, a “2” was assigned), and we did the same for the Linear Array test results. Among women who tested carcinogenic HPV positive by the cobas 4800 HPV test, 18.5% had at least two carcinogenic HPV genotypes, compared to 16.7% of those who tested carcinogenic HPV positive by the Linear Array test. There was no difference in the minimum numbers of carcinogenic HPV genotypes detected by the two assays among women who tested positive for carcinogenic HPV by both assays (P = 0.8).
In summary, we found very good agreement between the well-validated Linear Array HPV genotyping assay (5, 6, 8, 12, 23, 25) and the new prototype clinical assay, the cobas 4800 HPV test, which offers partial HPV genotyping for HPV16 and HPV18. We noted one discrepancy, that more women without evident cervical histopathology tested positive by the Linear Array test and negative by the cobas 4800 HPV test for carcinogenic HPV excluding HPV16 and HPV18. Prospective data for this convenience sample were not available to assess the subsequent risk of CIN2+ for Linear Array-positive/cobas 4800 HPV test-negative women without evident cervical histopathology. Notably, only 3 of the 14 (21%) women without evident cervical histopathology who were Linear Array-positive/cobas 4800 HPV test-negative for carcinogenic HPV excluding HPV16 and HPV18 were positive for E6/E7 mRNA (3).
More generally, sufficiently powered, formal evaluations against rigorously defined endpoints of CIN3+ with receiver-operator curve analyses are needed to establish the optimal cut point for clinical performance by the cobas 4800 HPV test (16, 24), as has been done for other HPV assays (21). And the reproducibility and reliability of the cobas 4800 HPV test should be established before it can be widely used for cervical cancer screening (16, 24).
This work was supported in part by the intramural research program of the NIH/NCI.
Published ahead of print on 12 August 2009.