Prospective Evaluation over 15 Years of Six Breast Cancer Risk Models

Abstract

Prospective validation of risk models is needed to assess their clinical utility, particularly over the longer term. We evaluated the performance of six commonly used breast cancer risk models (IBIS, BOADICEA, BRCAPRO, BRCAPRO-BCRAT, BCRAT, and iCARE-lit). 15-year risk scores were estimated using lifestyle factors and family history measures from 7608 women in the Melbourne Collaborative Cohort Study who were aged 50–65 years and unaffected at commencement of follow-up two (conducted in 2003–2007), of whom 351 subsequently developed breast cancer. Risk discrimination was assessed using the C-statistic and calibration using the expected/observed number of incident cases across the spectrum of risk by age group (50–54, 55–59, 60–65 years) and family history of breast cancer. C-statistics were higher for BOADICEA (0.59, 95% confidence interval (CI) 0.56–0.62) and IBIS (0.57, 95% CI 0.54–0.61) than the other models (p-difference ≤ 0.04). No model except BOADICEA calibrated well across the spectrum of 15-year risk (p-value < 0.03). The performance of BOADICEA and IBIS was similar across age groups and for women with or without a family history. For middle-aged Australian women, BOADICEA and IBIS had the highest discriminatory accuracy of the six risk models, but apart from BOADICEA, no model was well-calibrated across the risk spectrum.

Variability Among Breast Cancer Risk Classification Models When Applied at the Level of the Individual Woman

“…In our analysis, we found substantial variability in categorizing individuals as high risk for developing invasive breast cancer based on the risk model and the 5-year cutoff used. Thus, women are likely receiving vastly different recommendations depending on which breast cancer risk model is used and which cutoff is used to define “high risk.” While these models perform similarly at the population level, as noted in our study and in prior studies, 3 – 5 , 18 , 33 , 34 our analyses highlight marked disagreement between models in who they identify as “high risk.” For example, more than 20% of women would be classified as high risk of developing invasive breast cancer in the next 5 years by one model but average risk by a second model if using the ≥ 1.67% threshold. We found that if all three models were used, almost half of women would be considered at high 5-year risk by at least one model when using the ≥ 1.67% cutoff.…”

“… 2 These commonly used risk prediction models have similar accuracy and good discrimination when assessed in large populations. 3 – 5 …”

Clinical implications of incorporating genetic and non-genetic risk factors in CanRisk-based breast cancer risk prediction

“…The BOADICEA algorithm embedded in the CanRisk interface has been the subject of numerous validation studies considering different risk populations, model components and time scales, where the model has generally been shown to be well-calibrated with good performance in comparison to other risk models [ 5 , [20] , [21] , [22] , [23] , [24] , [25] ].…”