Clinical validation of p16/Ki‐67 dual‐stained cytology triage of HPV‐positive women: Results from the IMPACT trial

Abstract Triage strategies are needed for primary human papillomavirus (HPV)‐based cervical cancer screening to identify women requiring colposcopy/biopsy. We assessed the performance of p16/Ki‐67 dual‐stained (DS) immunocytochemistry to triage HPV‐positive women and compared it to cytology, with or without HPV16/18 genotyping. A prospective observational screening study enrolled 35 263 women aged 25 to 65 years at 32 U.S. sites. Cervical samples had HPV and cytology testing, with colposcopy/biopsy for women with positive tests. Women without cervical intraepithelial neoplasia Grade 2 or worse (≥CIN2) at baseline (n = 3876) were retested after 1 year. In all, 4927 HPV‐positive women with valid DS results were included in this analysis. DS sensitivity for ≥CIN2 and ≥CIN3 at baseline was 91.2% (95% confidence interval [CI]: 86.8%‐94.2%) and 91.9% (95% CI: 86.1%‐95.4%), respectively, in HPV16/18‐positive women and 83.0% (95% CI: 78.4%‐86.8%) and 86.0% (95% CI: 77.5%‐91.6%) in women with 12 “other” genotypes. Using DS alone to triage HPV‐positive women showed significantly higher sensitivity and specificity than HPV16/18 genotyping with cytology triage of 12 “other” genotypes, and substantially higher sensitivity but lower specificity than using cytology alone. The risk of ≥CIN2 was significantly lower in HPV‐positive, DS‐negative women (3.6%; 95% CI: 2.9%‐4.4%), compared to triage‐negative women using HPV16/18 genotyping with cytology for 12 “other” genotypes (7.4%; 95% CI: 6.4%‐8.5%; P < .0001) or cytology alone (7.5%; 95% CI: 6.7%‐8.4%; P < .0001). DS showed better risk stratification than cytology‐based strategies and provided high reassurance against pre‐cancers both at baseline and at 1‐year follow‐up, irrespective of the HPV genotype. DS allows for the safe triage of primary screening HPV‐positive women.


What's new?
Primary screening for human papillomavirus (HPV) requires efficient triage of HPV-positive women to colposcopy and biopsy. In this prospective observational trial in the United States, with 1-year longitudinal follow-up, the authors investigated the performance of p16/Ki-67 dual-stain cytology for the triage of women identified as HPV-positive during primary screening. Compared to HPV16/18 genotyping combined with cytological triage of other HPV genotypes, dual-stain cytology was significantly more sensitive for predicting risk of cervical intraepithelial neoplasia grade 2/3 or worse. The findings indicate that dual-stain cytology is effective for triage of HPV-positive women, either alone or when combined with partial HPV genotyping. screening. 4 Cervical cytology has been used to triage HPV-positive women but because of its low sensitivity for high-grade precursors, cytology-negative women need to be retested at a short interval. 5 HPV16/18 genotyping is also used in some settings for triage due to the elevated risk of high-grade precursors and invasive cancers associated with these genotypes. Triage with HPV16/18 genotyping alone also has limited sensitivity since only approximately 50% of highgrade cervical cancer precursors are associated with these genotypes. 6 To address this limitation, HPV16/18 genotyping has been combined with cytological triage of women with the 12 "other" HPV genotypes. However, the limited sensitivity of cytology means that a relevant proportion of women with the 12 "other" genotypes with a negative cytology may have precancer.  invited to join the IMPACT trial, as previously described in detail. 3 Subjects willing and able to provide written informed consent were eligible unless they were pregnant, had a known history of ablative or excisional cervical therapy within the past 12 months, known history of hysterectomy or current or planned participation in another cervical cancer screening, treatment or vaccination study. Women were referred to colposcopy and biopsy/endocervical curettage within 12 weeks after enrolment if test results showed abnormal cytology (ie, ASC-US or worse), a positive HPV test result or combined unsatisfactory cytology and HPV-negative test results. All study-related costs including costs for cytology and HPV testing, costs for colposcopy visits and biopsy evaluations, as well as costs for treatment performed according to the study protocol were covered by the sponsor of the trial (Roche).
The IMPACT trial consisted of two phases, a baseline (cross-sectional) and a 1-year follow-up phase. Women who met the clinical endpoint (ie, biopsy-confirmed ≥CIN2 [cervical intraepithelial neoplasia Grade 2] after the baseline colposcopy/biopsy visit) exited the study. Women who did not meet the primary endpoint and/or did not undergo treatment at baseline were invited to participate in the follow-up phase of the trial. Subjects included in the follow-up phase underwent an additional round of HPV and cytology testing after 12 months and, analogous to baseline procedures, were referred to colposcopy/biopsy if positive for either of these tests. The flow of the subjects through the baseline and 1-year follow-up phases of the IMPACT trial is shown in Figure S1.

| Test methods
Women had one cervical sample collected into a liquid-based cytology vial (PreservCyt; Hologic Inc, Marlborough, MA) using either spatula/ brush or broom-type collection devices (approximately half of the cohort per device). Specimens were shipped to 1 of 4 central laboratories in the United States participating as clinical laboratory study sites for the trial, and all laboratory testing was performed by these four laboratories.
HPV testing using both the cobas 4800 HPV Test and cobas HPV for use on the cobas 6800/8800 Systems (cobas 6800/8800 HPV test; Roche Molecular Systems, Inc, Pleasanton, CA) and cytology testing using the ThinPrep Pap Test (Hologic, Inc) were performed on all women enrolled into the IMPACT trial, according to the respective manufacturer's instructions. The use of both cobas 4800 and 6800/8800 HPV tests (each of them providing separate results for HPV16, HPV18 and the 12 "other" HPV types as a group) on every women allowed for the assessment of the performance of the highthroughput cobas 6800/8800 HPV test compared to the cobas 4800 HPV test in primary HPV screening, co-testing with cytology and ASC-US triage, as described in more detail recently. 3

| Clinical endpoints
Clinical endpoints for the study were biopsy-confirmed ≥CIN2 (ie, purposes, all tissue specimens were subjected to a central pathology review (CPR) as previously described in detail. 3 CPR results on H&E with p16-stained slides added to the review per Lower Anogenital Squamous Terminology (LAST) criteria (but without using HPV16/18positive ASC-US as an inclusion criterion) were used as the primary reference diagnoses for the trial. 18

| Study objectives and statistical methods
Co-primary objectives for the IMPACT trial were (a) to evaluate the performance of DS for identification of ≥CIN2/≥CIN3 when used to triage HPV-positive women, stratified by HPV16/18 vs 12 "other" There were no missing data for DS results, and unknown CPR reference diagnoses were not imputed. In disposition tables, the number of cases with unsatisfactory DS results is shown, and distributions for CPR results are shown for cases with both satisfactory and unsatisfactory DS results to enable assessment of potential bias.
A target sample size of 3500 HPV-positive women was set in order for 95% CIs for 1-NPV for co-primary objectives to span $3.2%. The obtained sample size of 5250 HPV-positive women resulted in precision greater than planned.  Table S1. Figure 1 shows the analysis population of 4927 women with valid DS results and histologic endpoints at baseline as well as cumulative year-1 follow-up numbers.

| DS positivity by cytology and biopsy results
Within the cobas 6800/8800 HPV-positive study population with valid DS results, 536 women with ≥CIN2 were diagnosed at baseline and 632 women were diagnosed with ≥CIN2 cumulatively at baseline and/or year-1 ( Figure 1). Figure Table 2).
In 12 "other" HPV-positive women, sensitivity of DS for ≥CIN2 and ≥CIN3 at baseline was 83.0% and 86.0%, respectively, significantly higher as compared to the respective sensitivity estimates of cytology: 58.8% for ≥CIN2 and 66.7% for ≥CIN3 (Table 2). DS showed lower specificity but similar to slightly higher PPV for ≥CIN2 as compared to cytology in the triage of 12 "other" HPV-positive women.

| Risk of high-grade CIN in HPV-positive women with positive or negative triage test results
The risk of ≥CIN2 and ≥CIN3 among HPV-positive women for the various triage strategies using DS or cytology, either combined with HPV16/18 genotyping or alone, is provided in Table 3 and graphically presented in Figure 2 for ≥CIN3. Results for cobas 4800 HPV-positive women are provided in Table S6 and Figure S3. HPV-positive women with negative DS test results showed a very low cumulative 1-year risk for disease (1-NPV for ≥CIN3: 1.4%), significantly lower than the respective risks when using cytology with HPV16/18 genotyping (2.3%; P = .0181), or cytology alone (3.1%; P < .0001) (     Our study has several strengths and limitations. Strengths include that IMPACT was a large prospective study that enrolled women in 32 clinical centers and assessed DS test performance in 4 central laboratories that also performed the cytology and HPV testing. Disease ascertainment was maximized by referring all women who tested positive with either of two HPV tests or who had an abnormal cytology to colposcopy. Women who fulfilled the initial colposcopy referral criteria were followed-up at 1 year, and those who were either HPV or cytology positive at 1 year were referred for another colposcopy.
To eliminate potential study bias, colposcopy was performed blinded to all test results and a nontargeted biopsy was collected when no lesion was identified at colposcopy. A CPR was performed on both H&E and H&E + p16-stained biopsy specimens. Limitations include the fact that the study follow-up was limited to 1 year and therefore the assessment of the negative disease prediction of a negative DS for a period longer than 1 year cannot be made.
In conclusion, the results of the IMPACT trial demonstrate that DS is safe and effective for the triage of HPV-positive women identified during primary HPV screening. DS alone or in combination with HPV16/18 genotyping offers an alternative to current triage strategies which are based on cytology, either alone or combined with HPV16/18 genotyping. DS-based triage provides consistently higher sensitivity than cytology-based triage, providing better reassurance against ≥CIN2 and ≥CIN3. Using DS alone as the triage reduces the complexity of triage strategies for HPV-positive women.

ACKNOWLEDGMENTS
The IMPACT trial was performed as a registration study, sponsored by Roche. The sponsor was responsible for the design and conduct of the study; collection, management, analysis and interpretation of the data; preparation, review and approval of the manuscript; and decision to submit the manuscript for publication. Drs. Wright and Stoler were paid for their work of members of the central pathology panel.
They were not paid to write this manuscript. All authors had full access to the full data and accept responsibility to submit for publication.

CONFLICT OF INTEREST
Drs. Stoler and Wright are consultants to Roche, BD Life Sciences, Inovio, and QSquared Solutions. They are speakers for Roche and BD Life Sciences. The other authors are employees of Roche.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.

ETHICS STATEMENT
All patients enrolled into the trial provided their informed consent before any study procedures. The study protocol and all amendments were approved by an Institutional Review Board (IRB). The trial was Helsinki. This is an observational, noninterventional diagnostic study and in alignment with FDAAA2007, World Health Organization (WHO) and ICMJE was not registered with ClinicalTrials.gov.