A new model based on data from the Breast Cancer Surveillance Consortium (BCSC) suggests that screening-detected overdiagnosis of breast cancer is less common than estimates from studies on excess incidence, but the model also takes into account indolent tumors and produced a more accurate estimation than models that didn’t take into account this aspect.
“There is a distinct lack of consensus of the exact level of overdiagnosis within the current U.S. mammography practice. The uncertainty over the degree of overdiagnosis is a major issue for the formulation of guidelines and policies. We have been able to overcome the limitations of previous studies and come up with an accurate estimate of the amount of overdiagnosis in U.S. mammography practices. In an interview, Marc D. Ryser, PhD, said that approximately one in seven cancers detected by screening among women aged 50-74 will be overdiagnosed. Around one third of all overdiagnosed cancers can be attributed to non-progressive cancers. Ryser is a Duke University, Durham, N.C., expert in statistical and mathematical models in the field of population health science. He presented the results of the model at the 2021 San Antonio Breast Cancer Symposium.
While previous models have provided estimates ranging between 0% and 54%, it is difficult to compare them due to their heterogeneity. Ryser declared that the models differ in their study populations, estimation methods, and definitions of overdiagnosis.
There are two ways to assess the risk of being diagnosed with overdiagnosis. The second is a model-based method. This relies on models of natural history of disease and clinical data to calculate the latency of tumors and utilizes this information to predict overdiagnosis. These models aren’t able to account for cancers that are inactive or are unlikely to cause death during a patient’s lifetime. The foundations of these models may also be unclear. The excess-incidence model however compares the incidence of the unscreened and screened populations. It is based on the assumption that excessive cancers result from overdiagnosis. However, this may be affected by bias.
Ryser’s group employed an approach based on models to overcome these limitations. However they also allowed for aggressive tumors. They made sure that the fundamental assumptions of the model, and took advantage of a contemporary high-quality and high-quality source of data in the BCSC.
The screenings were conducted on 35,986 women who were between 50 and 74. They first conducted screenings in 2000 and then in 2018. To determine the likelihood of being diagnosed with benign tumors, they calculated the risk of non-breast cancer mortality based on age-adjusted mortality risks adjusted for age cohorts. There were 718 cases of breast cancer diagnosed. 3.6 percent of tumors that were detected were indolent (95 percent confidence interval, 0.2%-13.8%). The overdiagnosis rate was 15.3 percent for an annual screening program (95% prediction interval 9.7%-25.2 percent). 6.0 percent of the overdiagnosis rate was expected to be caused by indolent tumors (95% prediction interval, 9.7%-25.2%) that don’t even progress, and 9.3 percent of tumors are growing but not quickly enough to cause the patient to die in their lifetime. An annual screening program had an expected overdiagnosis rate of 14.6% (95 percent PI, 9.4%-23.9%).
Ryser identified specific studies that used the same definitions of overdiagnosis that his group utilized and compared them to the 15.3 percent incidence that his group estimated. Studies that used excess-incidence methods resulted in higher estimates, while modeling studies yielded lower estimates.
The model failed to differentiate between ductal and invasive cancers in situ. It also failed to account for patient race or breast density.
The National Institutes of Health funded the study. Ryser is not able to provide relevant financial information.
This article was originally published on MDedge.com. It is part of the Medscape Professional Network.
Content Source: https://www.medscape.com/viewarticle/964679?src=rss