Clinical Evaluation of Human Papillomavirus Screening With p16/Ki-67 Dual Stain Triage in a Large Organized Cervical Cancer Screening Program

Key Points Question What are efficient approaches for triage of human papillomavirus–positive women in cervical cancer screening? Findings This cohort study of 3225 women found that p16/Ki-67 dual stain, alone or in combination with human papillomavirus 16/18 genotyping, provides better risk stratification than comparable cytologic-based strategies. Meaning Triage of human papillomavirus–positive women with dual stain may lead to lower referral to undergo colposcopy with similar detection of precancerous lesions compared with cytologic screening, making cervical cancer screening more efficient.


C
ervical cancer screening is shifting from primary Papanicolaou cytologic testing to primary human papillomavirus (HPV) testing worldwide. 1Women with negative results for HPV are reassured of a low risk of cervical cancer for many years and screening intervals can be safely extended compared with primary cytologic screening. 2Although HPV infections are common in the population, most HPV-positive women have transient infections.Three primary screening strategies are currently approved in the United States: Papanicolaou cytologic testing, HPV testing with Papanicolaou cytologic cotesting, and HPV testing with partial genotyping. 3For all screening approaches, triage tests are needed to decide which patients should be referred to undergo colposcopy for diagnostic evaluation. 4,5Per current HPV screening recommendations, HPV16/18-positive women are referred to undergo colposcopy, while women with positive results for other high-risk types of HPV undergo cytologic testing. 6,7Partial genotyping for HPV16 and HPV18 identifies the 2 genotypes with the highest risk of cancer, 8 but cannot distinguish a transient infection from a prevalent precancerous lesion.Triage with Papanicolaou cytologic testing is subjective, and its sensitivity varies widely, requiring retesting of HPV-positive women with negative cytologic results. 9rimary screening and triage strategies should be evaluated together, because the safety and efficiency of a screening approach and subsequent management depends on the combination of component test results.An ideal screening and triage approach should identify as many precancerous lesions as possible, while referring as few women as possible to undergo colposcopy.][12][13][14][15] We conducted a large clinical evaluation study at Kaiser Permanente Northern California (KPNC) evaluating DS together with HPV16/18 genotyping compared with the approved primary HPV screening strategy.We studied the performance of DS cytologic testing for triage of HPV-positive women, focusing on tradeoffs between colposcopy referral and detection of precancerous lesions in clinical practice.

Study Population and Clinical Procedures
The study was conducted within the routine cervical cancer screening program at KPNC.Per KPNC guidelines, women with negative HPV and Papanicolaou cytologic (cotest) results should return for regular screening after 3 years.Women with HPVpositive atypical squamous cells of undetermined significance (ASC-US), low-grade squamous intraepithelial lesions, or more severe cytologic results are referred to undergo immediate colposcopy.Women with positive HPV results and negative cytologic results should undergo repeat cotesting after 1 year and are referred to undergo colposcopy when either HPV or Papanicolaou cytologic results (at a threshold of ASC-US or worse [ASC-US+]) are positive.Women with negative results of a repeat cotest are not referred to undergo colposcopy.Per KPNC recommendations, all women undergoing colposcopy should receive at least 1 biopsy.Pathologists evaluating cervical biopsies routinely had access to patients' full medical records.Clinical management was not based on results of DS or HPV16/18 testing.In total, 3416 HPV-positive women who underwent cotesting at KPNC between September 16 and October 31, 2015, were included in the study.Follow-up for cervical precancer and cancer end points was conducted using electronic medical records through December 31, 2018.We excluded 85 women without evaluable DS results and 106 women without colposcopy when indicated per the algorithms described above (Figure 1).We evaluated the performance of DS and partial genotyping in the following 2 populations: (1) all HPVpositive women who underwent cotesting for screening, triage, surveillance, and management, representing all women currently undergoing cotesting at KPNC (n = 3225); and (2) the subset of HPV-positive women without abnormal cotest results in the preceding 42 months, representing a screening population of women who are identified to be newly HPV-positive (n = 2066).The study was approved by the KPNC Institutional Review Board and was exempted from institutional review at the National Cancer Institute by the Office of Human Subjects Research.Patient consent was waived because deidentified discarded specimens were used in this study.

Clinical Routine HPV Testing and Cytologic Screening
Hybrid capture 2 (HC2 High-Risk HPV DNA Test; Qiagen Inc) was conducted on specimen transport medium specimens per the manufacturer's instructions.Papanicolaou cytologic samples were collected in SurePath (Becton Dickinson) fixative.SurePath slides were prepared, stained, and processed on the BD FocalPoint Slide Profiler (Becton Dickinson).Medical history, HPV test results, and FocalPoint results were transmitted to the cytotechnologists reviewing slides on guided screening microscopes.All slides from HPV-positive women were evaluated by guided screening-assisted screening and full manual review.All abnormal slides were sent for pathology review.In addition, all negative Papanicolaou cytologic slides from HPV-positive women were rescreened manually.

DS and cobas 4800 Testing
SurePath tubes containing the residual pellet after preparation of routine Papanicolaou cytologic slides from HPV-positive women with final Papanicolaou cytologic results were removed consecutively from storage.SurePath slides were prepared by an experienced laboratory assistant.The prepared slides underwent staining by a histology technician who successfully completed CINtec PLUS (Roche Diagnositics) DS training, using the CINtec PLUS kit following the manufacturer's instructions.The DS cytologic slides were evaluated by cytologists after a 2-day training and certification to evaluate p16/Ki-67 DS cytologic slides.The cytotechnologists marked at least 1 positive cell on any DS-positive slides and documented their results electronically.The pathologists reviewed all slides and documented their results electronically.Slides with at least 1 DS-positive cell were considered positive.
The residual matching specimen transport medium tubes were removed from frozen storage and processed on the cobas 4800 instrument (Roche) by licensed and trained personnel per the manufacturer's protocol.Results from cobas testing are either HPV16/18 positive, other 12 high-risk HPV positive, or HPV negative.

Statistical Analysis
We evaluated the distribution of HPV16/18 and other high-risk HPV types (high-risk HPV12), and positive DS results by baseline cytologic findings and poorest histologic test results.Cytologic findings were classified by the Bethesda System 16 : negative for intraepithelial lesions or malignant neoplasm (NILM); ASC-US; low-grade squamous intraepithelial lesions; atypical squamous cells, cannot rule out high-grade squamous intraepithelial lesion; and high-grade squamous intraepithelial lesions.Final diagnosis was established by histopathologic findings classified according to the cervical intraepithelial neoplasia (CIN) scale: no indication for biopsy, normal, other (including atypia), CIN grade 1 (CIN1), CIN grade 2 (CIN2), CIN grade 3 (CIN3) and adenocarci-noma in situ, or cancer.Women who did not have an indication for colposcopy ("no indication for biopsy," eg, because of a negative results on a repeat cotest) were considered free of disease.The primary end point included CIN3, adenocarcinoma in situ, and cancer (CIN3+).The secondary end point included CIN2, CIN3, adenocarcinoma in situ, and cancer (CIN2+).We calculated sensitivity, specificity, positive predictive value, and negative predictive value for detection of CIN2+ and CIN3+ for primary HPV screening with partial genotyping and DS triage, and for DS triage in HPV-positive women without partial genotyping.The strategy currently approved by the US Food and Drug Administration (FDA) of primary HPV screening with partial genotyping and cytologic triage was the reference for the comparisons.
Differences in positivity, sensitivity, and specificity were evaluated using an exact McNemar χ 2 and differences in predictive values were evaluated using the method developed by Leisenring et al 17 using the R package, DTComPair (R Foundation for Statistical Computing).In a supplemental analysis, we stratified clinical performance measures by age groups (<29, 30-34, 35-44, and ≥45 years).We evaluated absolute risks of CIN2+ and CIN3+ in strata by cytologic testing, HPV16/18 genotyping, and DS.Exact binomial 95% CIs were calculated for proportions.We calculated the absolute risk of CIN3+for the currently approved primary HPV screening strategy with partial genotyping and cytologic triage, for DS in combination with partial genotyping, and for DS alone without partial genotyping (eFigure 1 in the Supplement).For each test result, we plotted the absolute risk of CIN3+ in relation to internal risk benchmarks derived from this study population: the risk of CIN3+ during the observation period in women with positive HPV test results and ASC-US (5.7% [53 of

DS Positivity by Cytologic and Histologic Results
Of 3225 women included in this study, 3 had cancer, 233 had CIN3 or adenocarcinoma in situ, 236 had CIN2, 2056 had his-tologic findings less severe than CIN2, and 697 did not have an indication for colposcopy (Table 1).Positive DS results increased from 36.1% (476 of 1318) in women with NILM cytologic findings to 96.0% (96 of 100) in women with high-grade squamous intraepithelial lesions, and from 32.9% (229 of 697) in women without biopsy to 88.4% (206 of 233) in women with CIN3 or adenocarcinoma in situ.

Cytologic and DS Triage for Detecting Precancerous Lesions
Compared with the currently approved primary HPV screening strategy with partial genotyping and cytologic triage, a combination of DS with HPV16/18 genotyping had significantly lower positivity than cytologic testing alone (1818 [56.4%; 95% CI, 54.6%-58.1%]vs 2128 [66.0%; 95% CI, 64.3%-67.6%];P < .001),while the sensitivity was equal for both strategies (Table 2).Although DS alone had a slightly reduced sensitivity compared with the combined strategies, the sensitivity was higher than that with cytologic testing alone and the specificity was significantly higher compared with all other strategies.These results were consistent across all age groups (eTable 1 in the Supplement).Similar patterns were observed for detection of CIN2+ (Table 2).For triage of HPV-positive women with partial genotyping, DS showed better risk stratification for CIN3+ than did Papanicolaou cytologic testing, with women with positive DS results having a higher risk than women with positive Papanicolaou test results for CIN3+ (218 of 1818 [12.0%; 95% CI, 10.5%-13.5%]vs 219 of 2128 [10.3%; 95% CI, 9.0%-11.6%];P = .005)(Table 3).

Cervical Precancerous Lesion Risk in Combinations of DS, Positive HPV16/18 Results, and Cytologic Testing
Figure 2 shows the risk of CIN3+ among HPV-positive women for the currently approved primary HPV screening strategy with partial genotyping and cytologic triage, for DS in combination with partial genotyping, and for DS alone without partial genotyping in relation to clinical management thresholds.The strategy of partial genotyping with DS yielded better risk stratification compared with the current standard: more women had a very low risk (1407 [43.6%] vs 1097 [34.0%]), and fewer women would be referred to undergo colposcopy immediately (1610 [49.9%] vs 2128 [66.0%]).More important, while women with NILM cytologic findings and positive HPV16/18 results had a risk high enough for colposcopy referral, the risk in women with negative DS results and positive HPV16/18 results was clearly below the threshold for colposcopy referral.Women with negative DS results and negative HPV16/18 results had the lowest risk of all combinations.The strategy of DS alone without partial genotyping provided good risk stratification, with half the women under the threshold for return testing at 1 year and the other half clearly above the threshold for referral to undergo colposcopy.The addition of partial genotyping to DS did not increase the percentage of women above the threshold for referral to undergo colposcopy.The individual risk estimates for all combinations of cytologic test-ing, HPV16/18 genotyping, and DS results are summarized in eTable 2 in the Supplement.

Detection in Women Without Previous Abnormal Cotest Results
Among the 2066 women without previous abnormal cotest results, the results with respect to positive DS and HPV16/18 positivity (eTable 3 in the Supplement), and the performance of these assays in comparison with the approved primary HPV screening strategy (eTable 4 in the Supplement) were similar to the overall population.Risks in strata of cytologic testing, HPV genotyping, and DS were largely similar to risks observed in the overall population (eTable 5 in the Supplement).Compared with the overall population of HPVpositive women, we observed very similar risk stratification of the approved strategy and the DS strategies in relation to clinical management thresholds, with one notable difference being the lower risk observed in women with negative DS results and positive HPV16/18 results, which was below the 1-year return threshold (eFigure 2 in the Supplement).

Efficiency of Primary HPV Screening With DS Compared With Cytologic Triage
The currently approved HPV screening strategy had the highest immediate referrals to undergo colposcopy (2128 [66.0%]) and overall referrals to undergo colposcopy (2677 [83.0%]).This strategy required 9.7 colposcopies to detect CIN3+ findings at immediate referral.All DS strategies required substantially fewer colposcopies, from 1818 for the combined partial genotyping and DS strategy in which all women with positive  a Women with HPV-negative LSIL or HPV-negative HSIL are not included.The number of women who underwent colposcopy after rescreening is estimated to be about 50% of women with an HPV-positive result at their repeat cotest.The risk of CIN3+ for combinations of cytologic testing with HPV16/18 genotyping, DS with HPV16/18 genotyping, and DS alone is plotted on the y-axis, with number and percentage of women for specific test combinations indicated.The dotted line corresponds to the 1-year return threshold (HPV-positive negative for intraepithelial lesions or malignant neoplasm [NILM], 2.3%), and the dashed line corresponds to the colposcopy referral threshold (HPV-positive atypical squamous cells of undetermined significance or worse [ASC-US+], 5.2%).FDA, indicates US Food and Drug Administration.

Discussion
Cervical cancer screening is transitioning from cytologic testing to primary HPV testing.Although a negative HPV test result provides reassurance against prevalent precancerous lesions or cancer, most HPV-positive women have transient infections that are not associated with cervical precancerous lesions, highlighting the need for additional triage tests.The currently approved algorithm for primary HPV screening in the United States includes partial genotyping and cytologic testing for triage of HPV-positive women. 6,7Within KPNC, this strategy would lead to two-thirds of HPV-positive women being referred to undergo colposcopy immediately, and more than 80% total of HPV-positive women being referred to undergo colposcopy immediately after 1 year.The strategy creates a substantial burden for women and has implications for health care infrastructure and cost, because most referrals to under colposcopy do not lead to detection of precancerous lesions.
A previous study reported that DS has higher sensitivity and specificity compared with cytologic testing for triage of HPVpositive women. 13In the current study, we evaluated the programmatic clinical performance of DS in conjunction with partial genotyping, allowing a direct comparison of a novel DS and HPV16/18 algorithm with the currently approved approach for triage of HPV-positive women.Using a risk-centered approach with internal benchmarks based on current management, 18 we found that performing DS would accurately identify most women at very low risk of precancerous lesions (negative HPV16/18 results and negative DS results) who may safely undergo retesting at extended intervals, a small group of women with somewhat elevated risk that is not high enough for colposcopy referral (positive HPV16/18 results and negative DS results), and the remaining women whose risk is well above the threshold for referral to undergo colposcopy (all women with positive DS results with highest risk among the women with positive DS results and positive HPV16/18 results).Women with normal cytologic test results, but positive HPV16/18 results, have a risk above the threshold for referral to undergo colposcopy, much higher than the risk of women with negative DS results and positive HPV16/18 results.Dual stain alone, without genotyping, provides very similar risk stratification, indicating that DS is also effective for triage of HPV screening tests that do not provide genotyping.Furthermore, the risk stratification of DS among HPV16/18-negative women is better compared with Papanicolaou cytologic testing, suggesting that DS is a good triage option for vaccinated populations that have reduced prevalence of HPV16/18 infections.
We evaluated the approved primary HPV screening strategy and novel DS-based triage strategies with respect to colposcopy referral and disease detection.We found that DSbased strategies were more efficient (indicated by lower numbers of colposcopies needed per case of CIN3+ detected) compared with cytologic testing-based strategies.Our data show the poor efficiency of 1-year repeat testing in currently approved strategies, with large numbers of colposcopies after repeat testing yielding only very few cases of CIN3+.Compared with a combined DS and HPV16/18 genotyping approach, the few additional cases of CIN3+ that are detected by these strategies involving 1-year repeat testing have negative HPV16/18 results and negative DS results at baseline, suggesting that they are small, low-risk precancerous lesions that will either resolve or be safely detected at the next 3-year screening visit.
We also found that primary HPV with DS cytologic testing can supplant a high-quality cotesting program and provide sensitive detection of cervical precancerous lesions while referring fewer women to undergo colposcopy.Dual stain cytologic testing showed improved performance compared with Papanicolaou cytologic testing in both the subset of women undergoing routine screening and the full population including women undergoing cotests for management of abnormal screening results, after colposcopy, or after treatment.Longterm follow-up from a previous study shows that women with negative DS results have a low risk of cervical precancerous lesions over 5 years. 12These data, combined with the low risk estimates in the current study for women with negative DS results, support the finding that retesting intervals in women with negative DS results can be safely extended to 3 years.

Strengths and Limitations
Our study has several strengths.We evaluated a large population with uniform and well-organized screening and management procedures, good disease ascertainment, and limited loss to follow-up.Our large population provides precise risk estimates for various combinations of test results.Furthermore, the risk levels from KPNC for different combinations of cytologic testing and HPV testing were the basis of current US screening and management guidelines, 2,19 allowing us to directly compare DS results with established management thresholds.All study procedures were implemented at the KPNC regional laboratory and performed in parallel with routine clinical operations, showing the feasibility of an integrated HPV screening and DS triage program in a clinical laboratory.A previous study showed the high reproducibility of DS when read by experienced cytotechnologists. 20Currently, approaches for automated evaluation of DS slides are being developed based on digital imaging; these approaches may further improve reproducibility, accuracy, and workflow of DS cytologic testing. 21imitations of our study should be noted.Papanicolaou cytologic testing at KPNC may be more sensitive than Papanicolaou cytologic testing performed at other health facilities.Thus, the specific performance comparisons between DS and Papanicolaou may differ in other settings.In our observational study, clinical management was based on HPV cytologic cotesting, with differential follow-up of women with positive HPV test results and positive cytologic results and women with positive HPV test results and negative cytologic results (repeat testing after 1 year).However, we had high completion rates of follow-up procedures.Furthermore, management guidelines were not always followed exactly, reflecting real-life practice in routine cervical cancer screening programs rather than in a clinical trial.Follow-up is ongoing; it will be important to determine how long a negative DS test result in combination with HPV16/18 genotyping is associated with a low risk of cervical precancerous lesions.

Conclusions
We found that, for primary HPV screening, DS has both higher sensitivity and specificity compared with cytologic testing for triage of HPV-positive women.Because of the greater reassurance of negative DS results, screening intervals can be ex-tended compared with the screening intervals after negative cytologic results.Dual stain reduces unnecessary colposcopy referral and unnecessary cervical biopsies, and may reduce unnecessary treatment compared with Papanicolaou cytologic testing.Our estimates of sensitivity, absolute risk, and colposcopy referral for various triage strategies can guide implementation of primary HPV screening.

Table 1 .
HPV16/18 and DS Positivity by Histologic Test Results and Cytologic Test Results Among 3225 HPV-Positive Women a Abbreviations: AIS, adenocarcinoma in situ; ASC-H, atypical squamous cells, cannot rule out high-grade squamous intraepithelial lesion; ASC-US, atypical squamous cells of undetermined significance; CIN, cervical intraepithelial neoplasia (grades 1-3); DS, dual stain cytologic test; HSIL, high-grade squamous intraepithelial lesions; HPV, human papillomavirus; HPV16/18+, positive for either HPV16 or HPV18; HR HPV12+, positive for at least 1 of the other high-risk HPV types; LSIL, low-grade squamous intraepithelial lesions; NILM, negative for intraepithelial lesion or malignant neoplasm; No Biopsy, No indication for colposcopy/biopsy.a The percentage in the first row in each cytology category gives the total across all cytology groups; the other percentages are within the cytology category.