Economic impact of multigene panel testing for hereditary breast and ovarian cancer

This retrospective cohort study used Myriad Genetic Laboratories, Inc.’s (Myriad) test result database linked to administrative claims data from the Optum Research Database (ORD). Using Optum’s De-identified Linking Software, a unique and de-identified value was created for each necessary individual attribute within the two databases and the resulting de-identified keys were used to establish the de-identified linkages. This parallel process allowed for de-identified linking of individuals from the two databases with no exchange of Protected Health Information. The ORD is a nationally representative, de-identified research database that contains included enrollment information and medical and pharmacy claims. Medical claims included diagnosis and procedure codes from the International Classification of Diseases, 9th and 10th Revisions, Clinical Modification (ICD-9-CM, ICD-10-CM); Current Procedural Terminology, Version 4 (CPT-4) procedure codes; HealthCare Common Procedure Coding System codes; revenue codes; and site of service codes. Outpatient pharmacy claims, available for approximately 38% of study patients, included National Drug Codes for dispensed medications. De-identified data were used to support this analysis; therefore, pursuant to the Health Insurance Portability and Accountability Act, an Institutional Review Board approval or waiver of authorization was not required.

The patient sample consisted of a majority commercial and a small number of Medicare Advantage enrollees included in both Myriad’s test result database and the ORD. Patients must have had ≥1 claim for a panel test (CPT codes 81211, 81213 and 81479 on the same day) or ≥1 claim for a SS test (CPT codes 81211 or 81213) between 1 January 2014 and 31 December 2016, with the name Myriad on the same claim as the specified CPT codes. The date of the first test claim was considered the index date. Test type identified in the ORD was confirmed in the linked Myriad test result database. Patients were also required to have continuous enrollment in the health plan with medical benefits for ≥6 months prior to (baseline period) and ≥12 months including and following (follow-up period) the index date and be ≥18 years of age as of the index date. A follow-up period of 12 months was required to adequately capture healthcare utilization and costs. Patients with a prior claim for HBOC genetic testing, a prior nondiagnostic claim with a cancer diagnosis during the baseline period or a claim indicating a personal history of cancer during the study period were excluded. Exclusion of patients with a personal history of cancer at the time of testing removed the effect of treatment for a cancer diagnosed prior to genetic testing on healthcare utilization and costs.

The SS test identified an elevated risk of HBOC, hereditary pancreatic cancer, prostate cancer and melanoma due to variants in BRCA1 and BRCA2. The panel test identified an elevated risk for eight hereditary cancers (breast, ovarian, colorectal, endometrial, melanoma, pancreatic, gastric and prostate) due to variants in the following genes: BRCA1, BRCA2, ATM, CDH1, CHEK2, NBN, PALB2, PTEN, STK11, TP53, BRIP1, RAD51C, RAD51D, MSH2, MLH1, MSH6, PMS2, EPCAM, APC, BARD1, BMPR1A, CDK4, CDKN2A, MUTYH and SMAD4. Variant classification was based on joint guidelines from the American College of Medical Genetics and Genomics and the Association for Molecular Pathology [27], as previously described [28,29]. Patients were assigned to groups based on the test received (panel or SS) and test result: positive (identification of ≥1 pathogenic variant [laboratory classification of deleterious or suspected deleterious] or a variant with a special interpretation finding noting possible increased cancer risks), VUS (identification of ≥1 variant without sufficient evidence of increased cancer risk and no pathogenic variants), negative (no identification of pathogenic variants or VUS). This resulted in six groups: panel–positive, panel–VUS, panel–negative, SS–positive, SS–VUS, SS–negative. The positive group included individuals with variants reported as ‘Special Interpretation,’ such as monoallelic MUTYH pathogenic variants and APC I1307K. This reflects the reporting of these findings as potentially actionable based on personal and family cancer history and/or ancestry.

All variables were measured over the 6-month baseline period through the 12-month follow-up period.

All comparisons were analyzed descriptively. Numbers and percentages were provided for dichotomous and polychotomous variables and means and standard deviations were provided for continuous variables. Chi-square tests and unpaired t-tests were used to compare categorical and continuous variables, respectively. Three generalized linear models with gamma distributions and log link (using a stepwise selection methodology) were examined to account for confounding of the relationship between the test group and the number of all-cause ambulatory visits, costs associated with all-cause ambulatory visits and total medical costs. The models were adjusted for test result, age category (18–34, 35–44, 45–54, 55–64 and ≥65 years), region, baseline Charlson co-morbidity score, varying baseline AHRQ co-morbidities, baseline count of ambulatory visits, and other baseline cost and utilization measures depending on which variables were indicated by stepwise selection for model inclusion.

A total of 11,040 patients met inclusion and exclusion criteria; 6359 in the panel group and 4681 in the SS group (Figure 1). Among the panel group, 5.9% tested positive, 60.3% negative and 33.8% VUS. Among the SS group, 2.4% tested positive, 95.2% negative and 2.4% VUS. In the panel group, pathogenic variants were most common in MUTYH (31.3%), BRCA2 (10.4%), APC (10.2%), BRCA1 (8.6%), CHEK2 (8.6%), ATM (7.5%) and PALB2 (5.1%). Among patients who tested positive in the SS group, 51.3% had a BRCA1 variant and 48.7% had a BRCA2 variant. Most study patients were female (98.1% panel, 99.0% SS; p = 0.041) and commercially insured (99.3% panel, 98.9% SS; p < 0.001) (data not shown). Mean age was slightly lower (43.2 vs 43.7 years; p = 0.041) as was the Charlson comorbidity score (0.19 vs 0.22; p = 0.041) among the panel group compared with the SS group. Slightly more patients in the panel group versus the SS group had pharmacy coverage through the health plan (39.6 vs 36.5%; p < 0.001).

Baseline healthcare utilization was similar between the panel and SS groups and across all test results with the exception of emergency room visits among VUS patients. Patients with a VUS result on the panel test had more emergency room visits compared with VUS patients who received the SS test (0.6 vs 0.3 mean visits; p = 0.041).

Overall, there were no significant differences in follow-up all-cause healthcare utilization among patients who underwent panel compared with SS testing (Table 1). Almost all patients had an all-cause ambulatory visit during the 12-month follow-up period, averaging 1.4 to almost two visits per month and approximately 32% had an emergency room visit. Ambulatory visit counts did not differ between the panel versus the SS groups either before (16.5 vs 16.7; p = 0.631) or after adjustment (17.3 vs 17.3, p = 0.809). When test results were considered in unadjusted analyses, a higher proportion of patients who were positive experienced an inpatient stay in the SS versus panel group (18.6 vs 8.3%; p = 0.002), with a higher mean number of hospitalizations (0.23 vs 0.07; p = 0.021).

Risk-reduction/early detection-specific healthcare utilization among patients who had panel testing (n = 6359) was lower compared with those who had SS testing (n = 4681; 69.5 vs 66.2%; p < 0.001; Table 1). Imaging was the most commonly utilized intervention regardless of test type or result, with approximately 63% of patients undergoing at least one imaging study. Significantly more patients in the SS group underwent imaging compared with the panel group (65.0 vs 61.6%; p < 0.001). In addition, risk-reducing surgery was more common within the SS group compared with the panel group (3.7 vs 2.4%; p < 0.001).

There were no significant differences in the overall risk-reduction/early-detection-specific utilization among patients who tested positive in the panel (n = 374) versus SS group (n = 113; 73.5 vs 80.5%; p = 0.131; Table 1). Among patients who tested positive, those in the SS group had a higher percentage of claims for both risk-reducing surgery (19.5 vs 6.2%; p < 0.001) and CA-125 monitoring (32.7 vs 13.1%; p < 0.001) compared with patients in the panel group.

Among patients who tested negative (n = 8292), overall risk-reduction/early detection-specific utilization was lower in the panel group compared with the SS group (64.3 vs 69.2%; p < 0.001; Table 1). This included increased utilization of risk-reducing surgery (3.3 vs 2.0%; p < 0.001) and imaging (64.8 vs 59.7%; p < 0.001) within the SS group.

In the VUS group, there was no difference in the overall utilization of risk-reduction/early detection-specific interventions for breast or ovarian cancer among the panel (n = 2149) or SS groups (n = 112; 68.3 vs 67.9%; p = 0.928; Table 1). In addition, there were no significant differences in any category of risk-reduction/early-detection interventions.

When follow-up all-cause healthcare utilization was compared across test results, positive patients receiving either the panel or SS tests had a significantly higher count of ambulatory visits compared with patients who tested negative or VUS (p ≤ 0.007 for all comparisons). Additionally, among patients who received the SS test, both the percentage of patients who had an inpatient stay and the mean number of inpatient stays were significantly higher among those who tested positive compared with those who tested negative or VUS (p ≤ 0.005 for all comparisons). There were no significant differences in follow-up all-cause healthcare utilization among patients who tested negative compared with patients who tested VUS in either the panel or SS test. Patients who tested positive (n = 487) also received more risk-reduction/early detection-specific interventions compared with patients with other test results (negative or VUS), regardless of test type.

Baseline healthcare costs were similar between the panel and SS groups (US$4120 vs US$3975; p = 0.458). Total unadjusted mean annual follow-up healthcare costs were similar between the panel and SS groups (US$14,425 vs US$14,384, p = 0.942; Figure 2). When test results were considered, there were no significant differences in total healthcare costs for patients in the panel and SS groups who tested negative or VUS.

Mean annual follow-up healthcare costs were the highest among patients who tested positive, regardless of test type (Figure 2). Unadjusted total and medical costs were significantly higher among patients who tested positive in the SS test compared with those in the panel group (overall costs: US$28,401 vs US$20,338, p = 0.032; medical costs: US$27,469 vs US$19,092; p = 0.023). Ambulatory costs accounted for more than half of the total healthcare costs and were higher in positive patients in the SS group compared with those in the panel group (US$17,341 vs US$11,347; p = 0.040). After adjustment for confounding variables, costs were not significantly different between the test groups (adjusted medical costs: US$13,546 SS vs US$13,557 panel, p = 0.978; adjusted ambulatory costs: US$8064 SS vs US$7765 panel; p = 0.338).

Mean annual risk-reduction/early detection-specific healthcare costs were lower among the panel group compared with the SS group in the overall study population in unadjusted analyses (US$763 vs $1044; p = 0.028; Figure 3). Costs were higher among positive patients in the SS group compared with the panel group (US$6052 vs $2267; p = 0.014) due to significantly higher costs for breast cancer surgical procedures (US$4362 vs US$1453; p = 0.041; data not shown). While costs for imaging were similar among both groups (Figure 3), costs for breast cancer-specific imaging were significantly lower among the panel group compared with the SS group (US$320 vs US$464; p = 0.048). Among patients who tested negative, those in the SS group had higher overall mean annual risk-reduction/early detection-specific costs compared with the panel group (US$926 vs US$642; p = 0.014; Figure 3). This was driven by significantly higher costs for breast cancer-specific surgical procedures within the SS group (US$348 vs US$188; p = 0.022). Patients who tested VUS had similar mean risk-reduction/early detection-specific costs regardless of test type (US$718 panel and US$679 SS; p = 0.890).

Among VUS patients in the panel and SS groups, mean all-cause costs were US$14,404 versus US$20,607 (p = 0.361), respectively. When comparing annual costs across test results, follow-up mean estimates for total healthcare costs were similar for patients who tested VUS or negative within the panel group (US$14,404 [SD US$34,920] vs US$13,860 [SD US$26,113]; no statistical testing), and costs were higher among patients who tested VUS compared with negative in the SS group (US$20,607 [SD US$71,191] vs US$13,872 [SD US$26,577]; no statistical testing). Follow-up mean total costs were higher among patients who tested positive in both the panel and SS groups compared with those who tested negative (p < 0.001 for both groups). In the panel group, patients who tested positive had significantly higher total costs (p = 0.002), ambulatory costs (p = 0.020), inpatient costs (p = 0.041) and other medical costs (p = 0.013) than patients who tested VUS. There were no significant differences in follow-up healthcare costs among those who tested VUS and those who tested negative in either the panel or SS groups. Patients who tested positive had the highest risk-reduction/early detection-specific costs regardless of test type.

Results of this study should be interpreted in light of the inherent limitations associated with claims data analyses. Claims data are subject to coding errors and may be incomplete. There may be differences between the underlying risk (i.e., severity of family cancer history) in the panel and SS groups; however, we could not adjust for family cancer history because the available codes lacked the specificity needed. By restricting our analysis to unaffected individuals, bias between the cohorts should be minimized as a family cancer history would be required to meet HBOC testing criteria regardless of whether SS or panel testing was ordered. Healthcare received within a clinical trial may not generate claims for some or all services. Statistical tests were not adjusted for multiple comparisons; therefore, the probability of replicating all statistically significant differences observed in this study is decreased. Due to coding variations, the receipt of some risk-reduction/early detection-specific interventions may not have been captured. Evaluation of drugs available through pharmacy benefits was limited to the 38% who had pharmacy benefits through their health plan. Healthcare costs were adjusted for inflation over the study period using the CPI, which may have overstated the rate of medical price inflation [39]. However, all groups were adjusted using CPI, so comparisons between groups were not affected. Lastly, the study population was from a managed care population and conclusions may not be generalizable to other populations, including the uninsured and those outside the United States. Given the limitations listed above, future research that adjusts for potential confounders may be useful.