Overdiagnosis of ductal carcinoma in situ by grade and definition in population-based screening: A modeling study.

Aim: To estimate ductal carcinoma in situ (DCIS) overdiagnosis overall and by grade in population-based screening and to determine the variation in overdiagnosis estimates by definition.

Methods: Using a fully validated micro-simulation Markov model for DCIS (SimDCIS), the number, rate, and proportion of DCIS overdiagnoses were estimated overall and by grade. Overdiagnoses comprised excess DCIS cases in the screened versus the unscreened population; overdiagnosis rate equaled the number of DCIS overdiagnoses per 100,000 screened women; and DCIS overdiagnosis proportion equaled overdiagnosed DCIS divided by total diagnosed DCIS in the screened population. Base estimates for overdiagnosed DCIS were from a population perspective (ages 50-100 years) and included screen-detected, clinically detected, or progressed DCIS (i.e., invasive breast cancer with DCIS precursor). Overdiagnosis was also estimated for alternative definitions and perspectives. Univariate and probabilistic sensitivity analyses were performed to estimate uncertainty.

Results: Base definitions yielded an overdiagnosis rate of 38.1 (range, 25.7-58.7) per 100,000 screened women and a proportion of 20 % (range 13 %-30 %). Stratification by grade showed 24 %, 20 %, and 18 % proportion overdiagnosis for grades 1, 2, and 3, respectively. Varying the definition led to overdiagnosis estimates from 18 % to 94 %; these overdiagnosis estimates increased by 36 %-49 % when excluding invasive breast cancer and by 54 %-71 % when including only screen-detected DCIS. Individual perspective estimates were 12 % higher than population perspective estimates.

Conclusion: In biennial screening, approximately 1 in 5 DCIS is overdiagnosed, but with minimal variation between grades. A consensus definition and perspective for overdiagnosis would reduce the observed variation in DCIS overdiagnosis estimates.