In their new cervical cancer screening draft recommendation statement, the U.S. Preventive Services Task Force recommends screening for cervical cancer every 3 years with cervical cytology alone in women aged 21-29 years, and either continuing 3-year cytology screening or 5-year high-risk human papillomavirus (HPV) screening in women aged 30-65 years.1 They arrived at the conclusion primarily based on an analysis of benefits and harms of each of the currently used screening modalities. Cytology plus HPV cotesting was considered but rejected based on the benefits versus harms calculations. However, some of the assumptions in their modeling are wrong and this has led them to faulty conclusions.
The USPSTF correctly identified CIN-2/3 and CIN-3+ detected and cervical cancer cases and deaths prevented as benefits. But they used the number of colposcopies and the number of tests conducted as their proxy for harms, primarily because these were more easily measured. They identified other harms, such as greater psychological distress related to being told of a positive HPV result (compared with being told of an abnormal cytology or cotest result), but these harms were not numerical and so they could not be incorporated into their modeling. They also did not measure the psychological distress associated with an extended screening interval or HPV-only screening.
The USPSTF also primarily relied on evidence from seven large, randomly controlled trials of primary screening, mostly conducted in Europe.3 Of the seven trials, two of them (and ) used HPV testing by a polymerase chain reaction methodology that is not approved by the U.S. Food and Drug Administration, not commercially available in the United States, and has different sensitivity and specificity does than FDA-approved tests.
Most of the studies also were done with conventional cytology that is, for the most part, no longer used in the United States. The age range in some of the studies was also very narrow (SWEDESCREEN evaluated women aged 32-38 years only) while extending to very young ages in others (evaluated women as young as age 20 years). It may not be possible to extrapolate from these results to the U.S. experience. On the other hand, the USPSTF elected not to use data from very large retrospective U.S. trials, such as the data from more than 1 million women from Kaiser Permanente Northern California, which helped form the foundation of current guidelines.4
Although screening has dramatically reduced the incidence of cervical cancer in the United States, most screening in this country is opportunistic and lacks population-based registries or regular invitations to screening. The USPSTF notes that a sizable proportion of the U.S.-based female population is not routinely screened. In fact, 11.4% of 21- to 65-year-old women have not been screened in the previous 5 years and the percentages are higher among minority and disadvantaged populations. Although the numbers seem relatively small, more than half of women with cervical cancer are found in this population. These women also are at greater risk of being lost to follow-up. Unlike the patients in the European trials, there is no current mechanism in the United States to assure that patients will return in 5 years. In fact, the USPSTF notes this limitation in their evidence, but doesn’t provide a remedy. These women may not get rescreened for many years after the 5-year interval has passed, and there is no evidence that this interval is safe.
Finally, the USPSTF recommendations for HPV primary screening relied in whole or in part on cytology triage of positive results. There is no U.S. experience for the use of cytology within this context, and many experts have argued that the test performance of cytology in a triage setting would be different than in a screening setting and would further be different in a vaccinated population. Since the modeling on which these draft recommendations are based did not account for these changed performance characteristics, their assumptions must be wrong.
Five years have not yet passed since publication of the last set of guidelines, and we have no idea how this extended screening interval functions in our opportunistic screening system. Stated simply, these draft recommendations have gone too far and too fast and should be reconsidered.