AJR: USPSTF underestimated breast cancer screening benefits
The key distinction of the current study by R. Edward Hendrick, PhD, of the University of Colorado-Denver, and Mark A. Helvie, MD, of the University of Michigan in Ann Arbor, was that it estimated the number needed to screen (NNS) to prevent one breast cancer death compared with the number needed to invite (NNI) to a screening trial to prevent one breast cancer death. Screening recommendations from the USPSTF were based in part on an analysis of NNI from randomized controlled trials (RCTs), but can paint a much different picture than an analysis of NNS.
“This difference has two main contributors,” wrote the authors. “NNI based on RCTs does not account for nonattendance among women invited to screening or for crossover of uninvited control group women who attend screening, whereas NNS compares women who are screened to women who are not.”
The other issue, according to Hendrick and Helvie, is that RCTs of screening mammography are based on screening that occurred between 1963 and 1990, with the exception of the Age trial in the mid-1990s.
“These data do not reflect current mammography technology, nor do they reflect current screening practice or interpretation skills. Just as it would be inappropriate to judge the current performance of many other technologies (e.g., computers and computing skills) on the basis of their capabilities in the 1970s and 1980s, there are important limitations to assessing modern screening mammography on the basis of RCT data that are decades old.”
Hendrick and Helvie turned to modeling results from the Cancer Intervention and Surveillance Modeling Network (CISNET), which includes U.S. screening mammography data from 1996 to 2007, to estimate NNS and NNS per life-year gained (NNS/LYG) with annual or biennial screening mammography, and then compared the results with NNI values reported by USPSTF.
Results showed that for women aged 40 to 49 undergoing annual screening mammography, the estimated NNS is 746, well below the NNI of 1,904 estimated by the USPSTF’s analysis of RCT data. Older age groups also saw lower estimates of NNS compared with NNI, and overall, annual screening of women between 40 and 84 years yields an NNS of 84 and a NNS/LYG of 5.3. Biennial screening for women ages 50 to 74 pushes the NNS to 144 and yields an NNS/LYG of 9.1.
The authors noted that the USPSTF emphasized the NNS to save a life as a metric, but stated no numeric threshold for appropriateness. It recommends biennial screening for women between 50 and 59 years, in whom the NNI was estimated to be 1,339, but recommended against routine screening for younger women, in whom NNI was estimated to be 1,904, suggesting that a NNI of 1,339 should be judged as appropriate to recommend routine screening.
“CISNET-based estimates of NNS in this article show NNS values less than 1,339 for all age decades, including 40–49 years. Our estimates of NNS, which range from 233 to 746 for annual screening in each age decade from 40 to 79 years and 1,316 for the five-year interval from 80 to 84 years, suggest that annual screening from 40–84 years is reasonable on the basis of the NNI criteria of 1,339.”
Despite the differences between NNS and NNI, the two values have been considered essentially equivalent in both academic publications and the mainstream media, leading to a skewed view of the value of screening.
“Harms of mammography have been incorrectly estimated on the basis of RCT-derived NNI rather than NNS, negatively influencing the benefit-to-harm ratio of screening mammography, even though potential harms (e.g., pain from compression, callback, biopsy of negative findings, and radiation exposure) accrue only to women who actually undergo mammography, not to women who receive an invitation to attend,” wrote the authors.