Validation study of the lambda model for predicting the BRCA1 or BRCA2 mutation carrier status of North American Ashkenazi Jewish women

Abstract

LAMBDA is a model that estimates the probability an Ashkenazi Jewish (AJ) woman carries an ancestral BRCA1 or BRCA2 mutation from her personal and family cancer history. LAMBDA is relevant to clinical practice, and its implementation does not require a computer. It was developed principally from Australian and UK data. We conducted a validation study using 1286 North American AJ women tested for the mutations 185delAG and 5382insC in BRCA1 and 6174delT in BRCA2. Most had a personal or family history of breast cancer. We observed 197 carriers. The area under the receiver operator characteristic (ROC) curve (a measure of ranking) was 0.79 [95% confidence interval (CI) = 0.77-0.81], similar to that for the model-generating data (0.78; 95% CI = 0.75-0.82). LAMBDA predicted 232 carriers (18% more than observed; p = 0.002) and was overdispersed (p = 0.009). The Bayesian computer program BRCAPRO gave a similar area under the ROC curve (0.78; 95% CI = 0.76-0.80), but predicted 367 carriers (86% more than observed; p < 0.0001), and was substantially overdispersed (p < 0.0001). Therefore, LAMBDA is comparable to BRCAPRO for ranking AJ women according to their probability of being a BRCA1 or BRCA2 mutation carrier and is more accurate than brcapro which substantially overpredicts carriers in this population.

Prediction ofBRCAMutations Using the BRCAPRO Model in Clinic-Based African American, Hispanic, and Other Minority Families in the United States

“…The AUC was 0.75 for all minority groups combined, which is within the range of those reported in non-Hispanic white populations (0.71 to 0.83). 9,10,13,14,16,[32][33][34] Of the minority groups examined, BRCAPRO performed the best in Hispanics, as indicated by the highest AUC (0.83) and the smallest Brier score (0.089), in which the AUC was higher than in a previous observation in Hispanics (0.77). 18 The model had the worst performance in African Americans (AUC, 0.68; Brier score, 0.208), in which the AUC was lower than our previous observation in African Americans (0.77).…”

“…Interestingly, this pattern was also observed for the Breast and Ovarian Analysis of Disease Incidence and Carrier Estimation Algorithm (BOADICEA) model, which allows for polygenic locus effects, 10 and for the Log Odds of the Probability of Carrying an Ancestral Mutation in BRCA1 or BRCA2 for a Defined Personal and Family Cancer History in an Ashkenazi Jewish Woman (LAMBDA) model, which was developed empirically. 34 Reasons for this pattern are unclear and may be related to families that have a single breast cancer diagnosis and limited family structure, familial clustering caused by environmental factors, heterogeneous penetrance caused by genetic/environmental modifiers and mutation spectrum, other low penetrance genes, and imperfect sensitivity of molecular methodology. We showed that lack of information in families that have a single breast cancer diagnosis is an important cause of this pattern.…”

Predicting the likelihood of carrying a BRCA1 or BRCA2 mutation: validation of BOADICEA, BRCAPRO, IBIS, Myriad and the Manchester scoring system using data from UK genetics clinics

“…The area under the ROC curve statistics were reasonably similar across models (0.72–0.77) but BOADICEA performed best. Previous validation studies have also found that BRCAPRO, BOADICEA, Myriad, and the Manchester scoring system discriminate reasonably well between carriers and non-carriers 6 11 15 17 – 19…”

“…We found that all models were overdispersed, suggesting that some predictions may be too extreme11—that is, either due to low predictions in carrier individuals, or high predictions in non-carriers. We found that this was mainly driven by observations in the low predicted carrier probability category, and may be explained by the reasons outlined above.…”

“…To evaluate whether the predicted probabilities fit the observed mutation status at the individual level (accuracy) we also carried out tests that assess the probabilities given by the models for systematic overprediction or underprediction (underdispersion or overdispersion) as described previously 11 12. The degree of overprediction or underprediction was estimated using logistic regression by treating the observed mutation status as the dependent variable and the log-odds of the probability of detecting a mutation given by the model as the independent variable.…”