OBJECTIVES

Value-based models are intended to support appropriate use of one procedure/test over another. Cost-ineffective care may nevertheless occur, due to lack of information or to financial incentives. Guidelines state that the 21-gene assay (Oncotype DX^®, Genomic Health, Inc.) is the preferred multi-gene assay (MGA) to predict chemotherapy benefit in early stage breast cancer (ESBC). We sought to compute the “decrementally cost-effective” savings (Nelson Ann Intern Med 2009) that might justify substituting another MGA, PAM50-ROR (Prosigna^®, NanoString).

METHODS

We conducted a decision analysis using Markov Chain Monte Carlo simulation to estimate posterior distributions in 100,000 women with ESBC. Risk estimates of PAM50-ROR to predict above/below the 21-gene result threshold for chemotherapy benefit (RS=25; Sparano NEJM 2018) were extracted from three published discordance studies. Primary endpoints were: (1) number of women tested with PAM50-ROR recommended adjuvant chemotherapy discordant with 21-gene results (discordant-aCT) and (2) QALY loss from suboptimal therapy choice due to discordance. Incidence of adverse events with chemotherapy, risk of distant recurrence, quality of life effects, and time discounting were based on published estimates (Chandler JCO 2018).

RESULTS

The number of women who were discordant-aCT was 23,853 (IQR 16,044 – 26,870) — almost 1 in 5 women tested. The total QALY loss associated with PAM50-ROR versus 21-gene was 24,593 years (IQR 12,265 – 27,161). To be decrementally cost-effective at the $100K threshold, PAM50-ROR would have to save at least $20,000 per patient.

CONCLUSIONS

Use of appropriate therapy guided by the 21-gene assay reduces distant recurrence incidence and saves costs versus no MGA testing (Hochheiser JCER 2019). Discordant therapy choice resulting from PAM50-ROR use results in QALY losses. PAM50-ROR cannot be decrementally cost-effective, because the required savings is 4× higher than the 21-gene assay price. This case study illustrates that value-based payment models need safeguards to avoid potentially suboptimal substitution of preferred interventions.