Negotiating sensitivity and specificity

A new study by Cox et al highlights this shifting paradigm, along with the important compromises between sensitivity and specificity inherent in any screening test. The authors carried out a post-hoc analysis of data from the ATHENA (Addressing THE Need for Advanced HPV Diagnostics) HPV trial, which took place in the United States. Specifically, they examined 10 different screening approaches—varying combinations of cytology, HPV testing, and HPV genotyping—used among a cohort of 34,354 women aged 30 years and older who were enrolled in ATHENA. The goal was to determine the sensitivity and specificity of each strategy and weigh these factors against the potential harm associated with each screening mode. The number of tests required at baseline and the number of initial colposcopic examinations needed to detect CIN2 or greater served as surrogates for harm.

While false-positive outcomes may lead to unnecessary colposcopic exams and excisional procedures, false negatives can result in undetected disease. How we balance these trade-offs most commonly depends on the prevalence of disease where the screening test is offered. The authors proposed 4 potential screening strategies that may increase sensitivity and reduce needless and potentially harmful colposcopies. Commonalities are the use of HPV testing and genotyping. Two also incorporate cytology in the initial screening, while the other 2 include cytology only as a reflex test for women who are negative for HPV 16 and HPV 18 in the setting of a positive pooled high-risk HPV result. Journal Club members agreed that any implemented strategy must be validated in a prospective fashion.

Other considerations

The slow process—years to decades—from the detection of cervical intraepithelial neoplasia (CIN)2+ or CIN3+ to the development of cancer also influences screening decisions. Since most women with CIN2+ do not have cancer and those with CIN are often identified in subsequent screening rounds, maximizing sensitivity for a surrogate endpoint like CIN2+ may not be critical. In fact, recent guidelines specifically discourage use of CIN2 as a primary screening target. Cox and colleagues used CIN3+ as the endpoint in their study, a decision that lent strength to the study, according to Journal Club members.

Because the study was conducted among women in the United States, participants pointed out that its results were not necessarily generalizable to “real world” clinical practice and follow-up; that the study is most relevant to developed countries with a low cervical cancer incidence and well-established primary care systems. It was also suggested that the results might not even be generalizable to all screening populations within the United States. Perhaps populations with higher rates of cervical cancer might benefit from different screening strategies than those used in populations at lower risk. Or, separate plans might be required for patients who follow recommendations closely and those tending to be less compliant.

While Cox et al propose tactics that could be used in future validation studies, Journal Club members believed that it was necessary to keep an open mind to the potential value of identifying additional proposed biomarkers for disease. For example, detection of E6/E7 mRNA or p16 immunostaining could be tested head-to-head with HPV genotyping. ATHENA used the Cobas HPV test (Roche Molecular Diagnostics, Pleasanton, CA), but strategies involving the other commercially-available FDA-approved HPV test, Hybrid Capture 2, should probably be explored alongside, as well.

As our screening capabilities continue to change, we must also be mindful of cost, an issue to which the authors allude in their conclusion. In an era of increased scrutiny of health care expenses, there is a societal expectation that along with the sensitivity and specificity of screening programs, cost-effectiveness will also be optimized.