Cost-effectiveness of Population-Wide Genomic Screening for Hereditary Breast and Ovarian Cancer in the United States

Key Points Question Is it cost-effective to implement population-wide genomic screening for hereditary breast and ovarian cancer (HBOC)? Findings This decision analytical model study found that genomic screening for HBOC among unselected women may be cost-effective depending on the age distribution of the women screened. Cascade testing of first-degree relatives added a modest improvement in clinical and economic value. Meaning Population-level genomic screening for HBOC targeting women aged 20 to 35 years could be considered in settings in which the outcomes of screening can be evaluated, particularly to avoid a reduction in mammography screening among patients with negative test results.

carriers entered the Markov model in starting state 2, in which they received mammography only according to age-based guidelines for average risk individuals. 8,9 Known carriers could opt for prophylactic RRM or RRSO based on age-based cumulative uptake among BRCA1/2 positive women ( Figure 2 in main text). 10 We modeled a 1-year health state for the year of the procedure, wherein we applied procedure costs and disutilities ( Table 1 in main text). Women who underwent RRM or RRSO then transitioned to post-RRM or post-RRSO procedure health states with reduced risks of cancer incidence. 11 Individuals with one procedure could transition to another 1-year state for a second procedure using a weighted probability of RRM and RRSO, with weighted procedure costs and disutilities; these patients then transitioned to a post-2nd procedure health state where we assumed their breast cancer risk was zero, but the age-based, RRSO-adjusted probability of ovarian cancer remained. 11 We assumed the small number of individuals who receive RRM and/or RRSO and nonetheless go on to develop cancer transition to earlier stage cancers due to continued intensive screening.
All individuals were at risk for breast or ovarian cancer according to age-based cancer incidence among carriers and non-carriers ( Figure 2 in main text). [12][13][14] We assumed women undergoing intensive mammography plus MRI would be diagnosed at an earlier stage, on average, than those undergoing standard mammography based on a study showing there were significantly fewer patients with positive lymph nodes at the time of cancer removal surgery in the MRI-screened group compared to the mammographyonly screened group. 15 We assumed equivalent incidence of earlier versus later stage © 2020 Guzauskas GF et al. JAMA Network Open.
breast or ovarian cancer, however we applied a mortality risk reduction to earlier stage breast cancer health states over the individual's remaining lifetime. 15 In both the earlier and later stage cancer states, we modeled first year-specific treatment costs and utility values. Patients who survived the first year of breast or ovarian cancer then transitioned to post-breast or post-ovarian cancer health states with long-term/continuing treatment costs and utility values.
We also derived age-based estimates of non-BRCA variant (ATM, CHEK2, MSH6, PALB2, RAD51C, TP53) cancer incidence from Lu et al., who conducted whole-exome sequencing and gene-phenotype associations on a sample of 11,416 patients with clinical features of breast cancer, ovarian cancer, or both from 1200 hospitals and clinics across the United States, plus 3988 controls who were referred for genetic testing for noncancer conditions. 13 Odds ratios from Lu et al. of breast and/or ovarian cancer risk were converted to relative risk estimates and applied to the cancer incidence data derived for the noncarrier population. Non-BRCA mutations were modeled in the Markov model as an HBOC variant prevalence-weighted pooled group. Annual agebased cancer incidence among noncarriers was derived from the Surveillance, Epidemiology, and End Results (SEER) Program. 14

Breast Screening Uptake
We based uptake of mammography and MRI on current guidelines. [6][7][8][9] Noncarriers and unknown carriers were assumed to undergo routine mammography according to guidelines for average risk women; starting at age 40, we modeled that 50% chose to receive optional annual mammography, increasing to the recommended 100% from ages 45-54, then all women received recommended biannual mammography from age 55 until death. Known carriers were assumed to undergo routine mammography according to guidelines for increased risk women, which recommend annual mammography alternated with annual breast MRI every 6 months. We assumed 75% of women with known increased risk opted for intensive screening with MRI in addition to mammography. We further assumed the remainder of known increased risk women received mammography only at the recommended frequency.

Cascade Testing Module
We used a decision tree to organize key elements of cascade testing including (a) the probability a newly identified carrier will inform their family members, (b) the number of living first degree female relatives, (c) the probability that informed relatives will opt to undergo testing, (d) carrier/noncarrier status, and (e) testing and/or family history testing The cascade testing module utilized age-based results from the primary screening model, combined with publicly available data on the number of first-degree female relatives of each identified carrier, 18 to calculate the incremental cost and benefits and their impact on the overall model ICER. We estimated age-based incremental outcomes from ages 20 to 100, and estimated a weighted (by number and ages of mother, sisters, and daughters) sum of incremental cost and QALYs over all newly identified family members. These weighted incremental estimates were then "fed back" into the primary population screening model, such that the incremental outcomes from the cascade model are tied to the age of the patient entering the primary model. The process of using the primary model to inform cascade testing incremental outcomes, and then    In one-way sensitivity analysis, one parameter at a time is varied to its low and high value while keeping all other parameters constant. Parameters with the greatest impact on results have the largest bars in the "tornado" diagram and are located on top. eFigure 3. Results of One-Way Sensitivity Analysis