The Improving Risk Informed HPV Screening (IRIS) Study: Design and Baseline Characteristics

The recognition that persistent infections with carcinogenic HPV were a necessary cause of almost all cervical cancers (1–3) and its immediate precursor abnormalities (precancer) has led to introduction of human papillomavirus (HPV) testing into cervical cancer screening (4–7). Across primary HPV screening strategies the question is how best to manage HPV-positive individuals, most of whom have transient infections that will not progress to cancer. We cannot yet accurately predict which individuals testing HPV-positive will fail to control their infections and eventually develop precancer and, if left untreated, possibly cancer. Yet, it is not feasible to send all people testing HPV-positive to colposcopy-guided biopsy to search for precancer. Therefore, screening strategies that utilize high-risk HPV (HrHPV) tests require additional triage tests to determine who among those screening HrHPV-positive need to go to colposcopy. Clinical scenarios postcolposcopy and posttreatment also require management decisions that incorporate HrHPV test and triage.

In primary HPV screening, currently recommended triage strategies for individuals testing HPV-positive include HPV16/18 genotyping and/or cytology, approaches with limitations including lower sensitivity of cytology and the reduced utility of HPV16/18 typing in HPV-vaccinated populations (8, 9). New robust and efficient biomarkers are required to triage HPV-positive women. Cytology-based p16/Ki-67 dual stain is reproducible with higher sensitivity and specificity than cytology, resulting in improved risk stratification that reduces colposcopy referral while finding more precancer (10). Because it provides excellent long-term risk stratification compared with cytology, dual stain–negative, HPV-positive women might be released to extended surveillance intervals (11). An automated algorithm using machine learning for cytology can match manual performance and provide a quantitative result, allowing for evaluation of dual stain at higher thresholds that confer a risk similar to high grade squamous intraepithelial lesion (HSIL) cytology (12). Another promising biomarker is HPV and host DNA methylation which is associated with increased disease severity (13–15). The performance of these novel biomarkers in clinical scenarios such as postcolposcopy and posttreatment has not been established.

In existing management guidelines, HPV16 is considered uniquely carcinogenic and HPV18 and HPV45 are associated with higher risks, especially for adenocarcinomas; all warranting immediate colposcopy. Differentiating between other HPV16-related types HPV31, HPV33, HPV35, HPV52, and HPV58 that pose substantially higher risk than the lowest-risk group consisting of HPV39, HPV51, HPV56, HPV59, and HPV68 may provide additional risk stratification (16).

The abundance of new test options demands well-powered head—to–head evaluations in a well-defined screening population to evaluate their performance relative to each other and to identify optimal management strategies. Such evaluations are required for screening and management guidelines that provide specific recommendations on how new technologies should be incorporated in clinical management (6, 8). To determine appropriate recommendations for a given test result, absolute risks of cervical intraepithelial neoplasia grade 3 or worse (CIN3+) are required at multiple follow-up time points (17, 18). Beyond screening, novel biomarkers have the potential to improve management of individuals who have abnormal screening results that do not trigger immediate colposcopy referral, individuals with colposcopy findings that do not require treatment, and individuals previously treated (8).

To address this, we created a biorepository of cervical specimens from a large population to evaluate the performance of biomarkers in cervical cancer screening and management while also allowing for evaluation of the impact of HPV vaccination on screening and management performance. Cervical biobank studies storing residual liquid-based cytology and HPV testing samples are a robust resource for natural history studies and longitudinal analyses of novel biomarkers (19–21). The Improving Risk Informed HPV Screening (IRIS) study includes stored cervical specimens from screening and follow-up linked with relevant cofactors, clinical history, and follow-up information. Specimens from the IRIS cohort will enable well-powered etiologic analyses of host and viral somatic changes, HPV genome variation, association of host ancestry, cervicovaginal microbiome, and HPV cofactors with the progression from HPV infection to precancer. In addition, the IRIS study will allow for the evaluation of biomarkers associated with other gynecologic cancers, including endometrial and ovarian cancers in prospectively collected cervical specimens from a large population.

Kaiser Permanente Northern California (KPNC) is a large integrated healthcare delivery system providing cervical cancer screening to individuals age 25 and older with concurrent HPV and Pap testing (“cotesting”) every 3 years. The KPNC Regional Laboratory processes almost 300,000 pairs of cytology and HPV tests each year. The representative collection of baseline specimens occurred between 2016 and 2018 from a consecutive sample of individuals age 25 to 65 undergoing cotesting at KPNC. After baseline discard specimens were collected, members were mailed an opt-out consent letter informing about the study and providing the opportunity to opt-out of participation. Women could respond by phoning a toll-free number or returning a prestamped, preaddressed letter. Specimens for women who chose to opt out or had returned mail were destroyed and their specimen test results and personal health information are not used and removed from the study dataset.

Specimens were selected according to cotest result, with an objective of sampling almost all HPV-positive individuals and a random sample of individuals testing negative for both HPV and cytology, to allow weighting results back to the full KPNC population (Table 1). The time periods of collection varied by sampling strata: HPV-positive, HPV-negative/cytology-positive, and HPV-negative/cytology-negative (referred to as “cotest negative”). As presented in Table 1, HPV-positive and HPV-negative/Pap-positive specimens were collected during the entire period (April 2016–March 2018 and January 2016–August 2018) while cotest-negative specimens were sampled early in the study to maximize follow-up time during the study period (March 2016–January 2017).

The details of specimen collection, processing, aliquoting, shipping, and storage are described separately (22). Briefly, Surepath (Becton Dickinson) and standard transport medium (STM; Qiagen) specimens collected for cytology and HPV testing were stabilized as needed and transferred into storage tubes as previously described (21). Immediate cryopreservation was not possible in the clinical setting, but specimens were frozen as soon as possible, allowing for compatibility with future analysis of RNA and DNA protein that would allow a range of assays from RNA expression, methylation, and microbiome. They were then shipped to the NCI Biorepository for long-term storage. From a stratified random sample of HrHPV-negative and -positive specimens, a total of 48,794 liquid-based cytology slides were created and stained with p16/ki-67 dual stain between June 2016 and March 2018 (22).

Follow-up HPV and cytology specimens for IRIS individuals were collected through November 2020. Deidentified HPV, cytology, and histopathology results as well as demographic and cofactor information for individuals were obtained from electronic medical records (EMR) for baseline and will continue to be collected for an additional 5 years or more. The IRIS study is an NCI-funded public resource. Study data will be made available for research purposes in accordance with institutional review boards upon request. The number of unique individuals in the KPNC screening program with cotest results during the study period was also collected. The study was approved by ethical review boards at Kaiser Permanente and the NCI.

The number of follow-up cotest specimens collected as of November 2020 were summarized by time of baseline collection and baseline cotest result. Each individual's status in the cervical cancer screening and management process was classified hierarchically using their clinical history prior to the baseline cotest. Those with no previous information available were categorized as an “unknown clinical history.” Those with any history of excisional treatment were considered “previously treated.” Individuals with a record of colposcopy prior to baseline were categorized as “postcolposcopy” and further classified as “recent” or “distant” depending on whether the colposcopy was conducted within 42 months prior to baseline cotest. Those whose previous screening result was abnormal [HPV-positive or cytology low-grade squamous intraepithelial lesion (LSIL) or worse] were categorized as “previous screen abnormal” while those whose previous screen was normal [HPV-negative and cytology atypical squamous cells of undetermined significance (ASC-US) or negative) were categorized as “routine screening.” Based upon the 3-year screening intervals at KPNC, routine screening was divided by time of screening visit into “early” (<=30 months), “timely” (over 30 and less than or equal to 42 months), or “late” (>42 months). This clinical history among women participating in the IRIS study was compared with the entire KPNC screening population. Baseline cotest results and follow-up activities were summarized by categories of history of previous screening and treatment. Follow-up activities included repeat cotests, colposcopy visits, and histology results from biopsy and excisional procedures. The current membership in status was also summarized.

The planned number of individuals enrolled in IRIS was 50,000 HPV-positive individuals, 10,000 HPV-negative/cytology-positive individuals, and 10,000 women negative for both HPV and cytology with the expectation that approximately 2,500, 50, and 5 precancers (CIN3+) would be detected in each group, respectively, most of which would be diagnosed within 3 to 4 years after baseline.

This sample size was based upon calculations for statistical power to evaluate novel biomarkers for primary screening and triage of HPV-positive results (23). The required precision depended upon the anticipated performance characteristics of individual risk markers or combined strategies. The precision of absolute risk estimates would be high overall, for example for evaluation of primary screening, a 3-year CIN3 risk of 5.5%, 1.4%, and 0.45% would have a SE of 0.22%, 0.16%, and 0.06%, respectively. For triage, an immediate CIN3 risk of 19.7% and 1.4% would have a SE of 0.7% and 0.11%, respectively.

Between January 2016 and August 2018, 897,680 KPNC individuals age 25 and older had a cotest at KPNC (Table 1). Baseline specimens from individuals with cotest negative, HPV-positive/cytology-positive, and HPV-positive results were collected at different, overlapping time periods. During these time periods, 305,083 individuals had a cotest across the three strata (Table 1). Overall, the IRIS study collected baseline specimens for 81,348 individuals– representing 26.7% of the 305,083 individuals with a cotest during the sampling time periods. The proportion of the individuals in the sampling strata with a specimen collected varied by cotest result: 5.9% of women testing cotest-negative, 80.3% of women testing HPV-negative/Pap-positive, and 90.8% of women testing HPV-positive were sampled, respectively. Opt-out requests were received for 3,428 (4.2%) individuals and those with HPV-positive baseline results were less likely to opt out than other sampling groups (3.8% vs. 5.3% HPV-negative/Pap-positive, and 5.0% HPV-positive; P < 0.001). The final analytic cohort included 77,920 individuals.

Demographic characteristics of individuals are presented in Table 2. Those with unknown screening history were younger and therefore less likely to have been a KPNC member for 5 years or more (mean age of individuals with 5 or more years membership was 42.2 years). Individuals with previous abnormal screen were also more likely to be younger (likely because they had not had previous colposcopy or treatment).

The IRIS cohort consisted of women undergoing cotesting for a variety of indications (Table 3): about one quarter (27.4%) had no previous clinical history at KPNC. 39.2% were routine screening visits, 6.2% had a previous abnormal screening test, 22.0% had a previous colposcopy, and 5.3% had been previously treated. Most routine screening visits were conducted around 3 years (30–42 months) after their previous screen. IRIS individuals with previous colposcopy were more likely to have had the procedure in the past 42 months versus later (14.4% vs. 7.6% of total cohort). According to the study design, individuals selected for IRIS were more likely to have previous treatment, colposcopy, and abnormal screening compared with the overall population. By study design, individuals with abnormal test results were more likely to be selected for IRIS and therefore they were less likely to be in routine screening than the overall population with a cotest at KPNC (39.2% vs. 58.4%). The 30,520 individuals in routine screening represent approximately 58.4% of the full routine screening population.

Baseline cotest results varied by clinical history with negative baseline screening results more common among women in routine screening versus previous colposcopy or treatment (Table 3). Individuals with a high-grade baseline cytology result were more likely to have no prior screening record or previous treatment. Due to the sampling scheme, IRIS oversampled individuals with HPV-negative/HSIL or worse cytology results and they were more likely to have been previously treated.

Follow-up activities stratified by category of previous clinical history are presented in Table 4. As of February 2020, 75.3% of IRIS individuals were still active KPNC individuals. Individuals with unknown clinical history were less likely to be active members (63.9%), compared with those who had previous treatment (83.3%) or colposcopy (80.5%). As of November 2020, most individuals (83.7%) had either a follow-up cotest and/or a colposcopy. Specifically, 75.9% had a follow-up cotest (99.0% of which were collected for IRIS) and 55.7% had colposcopy either at baseline or during follow-up. These procedures were more common when the individual had previous abnormal screening results, colposcopy, or treatment. Through February 2021, 5,563 cervical intraepithelial neoplasia grade 2 (CIN2), 2,756 CIN3 (including 219 adenocarcinoma in situ), and 146 cervical cancers (including 69 adenocarcinoma) were identified among IRIS individuals so far, with passive follow-up continuing. Among individuals with CIN2+ histologic diagnoses, CIN2 was more common among individuals undergoing timely or early routine screening (226 of 303, 74.6%) and (835 of 1,095, 76.3%) while individuals with previous treatment had a greater proportion of CIN3 (346 of 806, 42.9%) and cancer (27 of 807, 3.3%) diagnoses.

With over 77,000 individuals included and 2,500+ precancers (CIN3) detected in the first 2 to 4 years of follow-up, the IRIS study is well-positioned to rapidly evaluate novel biomarkers in a variety of clinical scenarios to inform the changing landscape of cervical screening and management. The IRIS study will also permit studies of HPV carcinogenesis. To date, most of the women in the IRIS cohort remain KPNC members and follow-up information will be obtained from EMR records for at least 5 more years to identify outcomes as they occur over time during follow-up clinical care.

KPNC has a very large screening program that tested almost 900,000 members during a 20-month period between 2016 and 2018. By embedding the study within the regional laboratory of a large integrated health delivery system, baseline and follow-up discard specimens were systematically identified, retrieved, and processed, and stored without disrupting regular lab processes. Access to EMRs allowing for ascertainment of deidentified relevant clinical history, test results, as well as current and long-term outcomes creates additional efficiency for this research. Within the first 2 to 4 years of follow-up, our target numbers of precancer (CIN3) cases were reached for individuals with HPV-positive results at baseline (n = 2,701) and HPV-negative/Pap-positive results at baseline (n = 53). In the lowest-risk sampling stratum (HPV-negative/Pap-negative result at baseline), two of the anticipated five CIN3 cases had been detected. Case ascertainment is continuing in subsequent screening and management rounds. Ascertainment of outcomes among cotest-negatives is expected to take longer compared with HPV-positives, since clinical management is based on these results.

Because exact sampling fractions are available, analyses from the IRIS study can be weighted back to the entire KPNC screening population. The membership represents a well-screened population; their demographic characteristics align with those of the U.S. census-enumerated population in the Bay Area Metropolitan Statistical Area, with the exception of lowest and highest incomes (24). While the KPNC screening populations represents a high-resource setting, the risk estimates calculated within strata of cytology and HPV result from KPNC have shown to be portable to some other U.S. settings (18).

In the United States, the management of abnormal cervical cancer screening results are guided by the American Society for Colposcopy and Cervical Pathology (ASCCP) Consensus Guidelines. The 2019 ASCCP Risk-Based Consensus Management Guidelines were redesigned to current test results and history of cervical cancer screening tests and treatments for precancer. While KPNC represents an optimal scenario with regard to having prior history available, widely applicable estimates can be generated by collapsing across subsets of the population with and without known prior history. Importantly, the framework utilizes clinical history and risk calculations given test results to incorporate new technologies without the need for frequent interim guidance or full consensus conferences. Outside of IRIS, the full KPNC screening population is a unique large screening cohort that provides data for analyses pertaining to HPV testing, cytology, and other cofactors (17, 25–31).The KPNC screening population includes 1.5 million individuals aged 25 to 65 years routinely screened and provides immediate and 5-year risks of CIN3+ for combinations of current test results paired with history of screening test and colposcopy/biopsy result (17). Because the IRIS cohort is a subcohort within this larger effort, the results from IRIS will be directly applicable to the risk estimates that underlie current screening and management guidelines in the United States and will allow rapid update of guidelines when new technologies receive regulatory approval. As new biomarkers are developed, the large IRIS study is well poised to evaluate assays and provide risk calculations for many clinical scenarios, allowing direct head—to–head comparisons of screening, triage, and management tests which can accelerate moving new promising assays into clinical practice.

While the FDA continues to evaluate, approve, and regulate new technologies for management of abnormal cervical cancer screening results, the FDA frequently refers to clinical guidelines such as the 2019 ASCCP Consensus Guidelines, and its continuation, the Enduring Guidelines effort, to provide specific recommendations on how new technologies should be incorporated in patient management. Similarly, international guidelines are using risk-based approaches to evaluate new technologies and make practice recommendations (32). IRIS allows for evaluation of both common testing scenarios such as routine screening and unknown screening history as well as unique clinical scenarios such as postcolposcopy, posttreatment, and previous abnormal screening result in the absence of colposcopy. Using a risk-based approach, follow-up data from IRIS can be used to evaluate novel biomarkers to inform future clinical practice guidelines.

Future analyses will consider the performance of primary screening and triage markers among vaccinated individuals, using the KPNC vaccination registry, and simulated vaccinated cohorts based on models that include HPV typing. Future etiologic studies will evaluate HPV and host somatic changes, epigenetics, and the cervicovaginal microbiome among other markers (13, 33–37). Prior to IRIS, the HPV Persistence and Progression (PaP) Cohort was a very large “convenience sample” created between 2007 and 2010 by storing residual STM specimens from among approximately 55,000 mainly HPV-positive women who were cotested or triaged at KPNC. Three thousand, four hundred and ninety-nine women from the PaP Cohort were also enrolled in IRIS, allowing for even long-term natural history studies.

The IRIS study collected specimens and data from individuals managed according to KPNC clinical practice guidelines. Therefore, management and disease ascertainment were based on screening and management test results. For example, screen-positive individuals are more likely to have repeat sampling than screen-negative individuals and risk analyses require sophisticated statistical methodologies such as prevalence-incidence mixture models (18, 38–40). Furthermore, all assays and natural history studies must be conducted from one of the two clinical collection mediums, STM and SurePath, used at KPNC and collected in IRIS. Although clinical follow-up data is available for 83.7% of the IRIS cohort, 24.7% were no longer KPNC members, preventing the acquisition of long-term outcomes for those individuals.

Beyond cervical endpoints, the IRIS study also is well-placed to study markers for endometrial and gynecologic cancers because cells from the genital tract, including the fallopian tubes and the endometrium shed and can be detected in cervical samples (41, 42). Given the large population of individuals with repeat cervical samples, the IRIS study is amenable to evaluate biomarkers for gynecologic cancers such as somatic mutations and methylation for carcinogenesis and early detection studies. In summary, IRIS is a large, unique cohort that allows for natural history studies and head—to–head evaluation of established assays and new biomarkers for cervical cancer screening and management.

T. Raine-Bennett reports other support from NCI of NIH during the conduct of the study. No disclosures were reported by the other authors.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

J.C. Gage: Conceptualization, formal analysis, supervision, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. T. Raine-Bennett: Conceptualization, data curation, supervision, methodology, project administration, writing–review and editing. M. Schiffman: Conceptualization, formal analysis, investigation, visualization, methodology, writing–review and editing. M.A. Clarke: Formal analysis, validation, investigation, writing–review and editing. L.C. Cheung: Data curation, software, formal analysis, validation, writing–review and editing. N.E. Poitras: Conceptualization, data curation, software, formal analysis, supervision, validation, investigation, methodology, writing–review and editing. N.E. Varnado: Data curation, supervision, investigation, visualization, methodology, project administration. H.A. Katki: Conceptualization, data curation, formal analysis, validation, investigation, visualization, writing–review and editing. P.E. Castle: Conceptualization, data curation, software, investigation, visualization, writing–review and editing. B. Befano: Data curation, software, formal analysis, supervision, validation, investigation, visualization, methodology, writing–review and editing. M. Chandra: Data curation, software, validation, investigation. G. Rydzak: Data curation, software, formal analysis, validation, writing–review and editing. T. Lorey: Conceptualization, resources, supervision, validation, project administration, writing–review and editing. N. Wentzensen: Conceptualization, data curation, formal analysis, supervision, funding acquisition, validation, investigation, methodology, project administration, writing–review and editing.

The authors acknowledge the extensive efforts of laboratory study staff, including Stephanie Phelps, Oscar Lee, Kennedy Hardemion, Howard Moffett, and Chinar Sheth. This paper is dedicated to the memory of Barbara Fetterman, who coordinated the KP Regional Lab through 2017 and coordinated the KP-NCI collaborations. Her long-term determination, tenacity, and commitment were critical for success of these projects. Research reported in this publication was supported by the Intramural research award of the NIH. This publication was supported by grant number T32CA09168 from the NIH.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.