Didem Egemen

Primary human papillomavirus testing vs cotesting: clinical outcomes in populations with different disease prevalence

Abstract Implementation of primary human papillomavirus (HPV) testing has been slow in the United States perhaps because of concerns of decreased sensitivity compared with concurrent HPV and cytology testing (“cotesting”). We used the National Breast and Cervical Cancer Early Detection Program and the Kaiser Permanente of Northern California cohort to quantify potential trade-offs with primary HPV compared with cotesting in 4 US populations with differing precancer or cancer prevalence. In all settings, cotesting required more lab tests and more colposcopies compared with primary HPV testing. Additional cervical intraepithelial neoplasia grade 3 or cancer immediately detected from cotesting vs primary HPV decreased with decreasing population-average cervical intraepithelial neoplasia grade 3 or cancer prevalence from 71 per 100 000 screened among never or rarely screened individuals in the National Breast and Cervical Cancer Early Detection Program (prevalence = 1212 per 100 000) to 4 per 100 000 screened among individuals with prior HPV-negative results in Kaiser Permanente of Northern California (prevalence = 86 per 100 000). These data suggest that cotesting confer an unfavorable benefit-to-harm ratio over primary HPV testing.

Artificial intelligence–based image analysis in clinical testing: lessons from cervical cancer screening

Abstract Novel screening and diagnostic tests based on artificial intelligence (AI) image recognition algorithms are proliferating. Some initial reports claim outstanding accuracy followed by disappointing lack of confirmation, including our own early work on cervical screening. This is a presentation of lessons learned, organized as a conceptual step-by-step approach to bridge the gap between the creation of an AI algorithm and clinical efficacy. The first fundamental principle is specifying rigorously what the algorithm is designed to identify and what the test is intended to measure (eg, screening, diagnostic, or prognostic). Second, designing the AI algorithm to minimize the most clinically important errors. For example, many equivocal cervical images cannot yet be labeled because the borderline between cases and controls is blurred. To avoid a misclassified case-control dichotomy, we have isolated the equivocal cases and formally included an intermediate, indeterminate class (severity order of classes: case>indeterminate>control). The third principle is evaluating AI algorithms like any other test, using clinical epidemiologic criteria. Repeatability of the algorithm at the borderline, for indeterminate images, has proven extremely informative. Distinguishing between internal and external validation is also essential. Linking the AI algorithm results to clinical risk estimation is the fourth principle. Absolute risk (not relative) is the critical metric for translating a test result into clinical use. Finally, generating risk-based guidelines for clinical use that match local resources and priorities is the last principle in our approach. We are particularly interested in applications to lower-resource settings to address health disparities. We note that similar principles apply to other domains of AI-based image analysis for medical diagnostic testing.

Long-Term Prospective Cohort Study of Cervical Cancer Screening Using Triage of Women Who Are Human Papillomavirus–Positive With Dual Stain and Human Papillomavirus Genotyping

PURPOSE Primary human papillomavirus (HPV) testing has the best tradeoff of benefits and harms for cervical screening but requires triage to determine management among HPV positives. We conducted a prospective observational study to evaluate triage of women who are HPV-positive using dual stain (DS) and HPV genotyping. MATERIALS AND METHODS We included 9,645 consecutive women who are HPV-positive undergoing cervical screening in two periods between 2015 and 2017 in the organized cervical screening program at Kaiser Permanente Northern California. Absolute risk and clinical performance of DS and cytology for detection of cervical intraepithelial neoplasia grade 3 and greater (CIN3+) were estimated overall and by HPV genotype and by age. Cumulative absolute risk of CIN3+ was modeled over 5 years using a prevalence-incidence mixture model, which allows estimating risk accounting for differences in disease ascertainment, surveillance intervals, and compliance. RESULTS The baseline risk of CIN3+ was 9.4% and 0.8% for women testing positive and negative for DS, respectively, and 6.9% and 2.0% for women testing positive and negative for cytology, respectively. Sensitivity, specificity, and predictive values for CIN3+ detection were better for DS compared with cytology over 5 years ( P < .001 for all comparisons). Risk in women with HPV16-positive/negative for intraepithelial lesion or malignancy was substantially higher than the risk in women with HPV16-positive/DS-negative (7.5% v 2.9%, P < .001). DS had better triage performance compared with cytology in all age groups and in women positive for HPV types other than HPV16 or HPV18. CONCLUSION Long-term reassurance of low risk among DS negatives suggests that DS detects molecular changes earlier in the carcinogenic pathway than cytology. DS has better risk stratification than cytology overall, within HPV risk strata, and across all screening age groups and is a better option for triage of vaccinated populations.

Initial evaluation of a new cervical screening strategy combining human papillomavirus genotyping and automated visual evaluation: the Human Papillomavirus–Automated Visual Evaluation Consortium

Abstract The HPV-Automated Visual Evaluation Consortium is validating a cervical screening strategy enabling accurate cervical screening in resource-limited settings. A rapid, low-cost human papillomavirus (HPV) assay permits sensitive HPV testing of self-collected vaginal specimens; HPV-negative women are reassured. Triage of positive participants combines HPV genotyping (4 groups in order of cancer risk) and visual inspection assisted by automated cervical visual evaluation that classifies cervical appearance as severe, indeterminate, or normal. Together, the combination predicts which women have precancer, permitting targeted management to those most needing treatment. We analyzed CIN3+ yield for each HPV-Automated Visual Evaluation risk level (HPV genotype crossed by automated cervical visual evaluation classification) from 9 clinical sites (Brazil, Cambodia, Dominican Republic, El Salvador, Eswatini, Honduras, Malawi, Nigeria, and Tanzania). Data from 1832 HPV-positive participants confirmed that HPV genotype and automated cervical visual evaluation classification strongly and independently predict risk of histologic CIN3+. The combination of these low-cost tests provided excellent risk stratification, warranting pre-implementation demonstration projects.