Date of Award

5-1-2017

Degree Type

Dissertation

Degree Name

Doctor of Philosophy

Field of Study

Biostatistics

Department

School of Public Health

First Advisor

Sumihiro Suzuki

Second Advisor

Swati Biswas

Third Advisor

Karan P. Singh

Abstract

Identifying women at high risks of carrying the breast cancer susceptibility genes is crucial for providing timely surveillance and necessary health management interventions. BRCAPRO is one of the most widely used statistical models for breast cancer risk prediction in genetic counseling. It provides carrier probabilities of BRCA1/2 mutations and calculates the risks of developing breast and ovarian cancers. This calculation requires extensive personal and family history information, which makes it difficult to use in primary care where a wider population could be reached. Thus, we developed a two-stage approach for the genetic risk prediction of BRCA1/2 mutation. In the first stage, limited information on the counselee and her family history of cancer are used in simplified versions of BRCAPRO. If the risk at this stage is found to be high, the full BRCAPRO model utilizing the complete family history is implemented in the second stage. We aimed to balance the tradeoff between the amount of information used and the accuracy of the predictions. We explored several first stage tools. BRCAPROLYTE uses information on the affected relatives up to the second degree only. BRCAPROLYTE-Plus additionally includes unaffected relatives by imputing their ages. BRCAPROLYTE-Simple eliminates the need to collect information on the numbers and types of unaffected relatives and imputes them and their ages instead. The study cohorts include 1,917 families mostly at high risk from the Cancer Genetics Network, 796 high-risk families from MD Anderson Cancer Center, and 1,344 population-based families from Newton-Wellesley Hospital. To evaluate the models, we used sensitivity, specificity, area under the curve, and observed versus expected number of carriers. We also considered clinical criteria of number of referrals made by each model. We found the proposed two-stage approach (with BRCAPROLYTE, BRCAPROLYTE-Plus, and BRCAPROLYTE-Simple at the first stage) has very limited loss of discrimination and comparable calibration with BRCAPRO. It identifies a similar number of carriers without requiring a full family history evaluation on all probands. Thus, our two-stage approach allows for practical large-scale genetic risk assessment in primary care.

Comments

Atienza, Philamer M., Adaptation of the Genetic Risk Prediction Model BRCAPRO for Primary Care Settings. Doctor of Philosophy (Biostatistics), May 20, 2017, 100 pp., 12 tables, 8 illustrations, bibliography, 94 titles.

Share

COinS