Association of drug industry payments to oncologists and use of guideline-preferred cancer treatments.

Abstract

11059 Background: Payments from the drug industry to oncologists are common and influence physicians’ prescribing decisions. However, whether payments affect the likelihood that patients receive optimal cancer treatment has not been assessed. Methods: This was a population-based cohort study. We used fee-for-service Medicare claims to identify patients with incident cancer diagnosis (new cancer diagnosis code after >= 1 year with no cancer diagnosis codes) from 2014-2020. We used National Comprehensive Cancer Network (NCCN) Guidelines to identify treatment scenarios (e.g., metastatic melanoma, stage III colon adjuvant therapy) with variation in the clinical benefit among recommended treatment options. The patient cohort corresponding to each scenario was identified using claims. We used NCCN Guidelines to identify the “optimal” treatment for each scenario (defined as the designated “preferred” treatment, with NCCN Evidence Blocks scores used for tiebreaks) at each time point across the study period. The primary outcome was whether a patient received the treatment that was, as of their diagnosis date, the optimal treatment for their cancer (vs. any non-optimal treatment). The primary exposure[s] were whether a patient’s medical oncologist received personal (“general”) payments (source: linked Open Payments data) related to 1) optimal treatment or 2) any non-optimal treatment, during the year prior to the patient’s cancer diagnosis. We fit Poisson-family generalized estimating equations, with physician-level clustering and adjustment for scenario, year, and patient characteristics, to estimate the relative risk (RR) of optimal treatment. Results: There were 14 clinical scenarios comprising 15,835 patients. There were 3,429 (21.7%) patients whose oncologist received payment for the optimal treatment, 4,891 (30.9%) for non-optimal payments, and 1,993 (12.6%) to both payment types. Oncologist receipt of payment for optimal treatment was associated with a greater likelihood (RR=1.07, 95%CI:1.01-1.13), and payment for non-optimal treatment was associated with lower likelihood (RR= 0.94, 95%CI:0.89-0.98) of patients receiving optimal treatment. Associations became stronger as payment dollar value increased; for optimal payments ≥$10,000, RR=1.26 (95%CI:1.02-1.57), and for non-optimal payments ≥$10,000, RR=0.70 (95%CI: 0.49-1.00). An estimated 41.6% of unexposed patients received optimal treatment; versus 52.6%, 29.3%, and 33.8% among patients exposed to ≥$10,000 of payments for optimal treatment only, ≥$10,000 non-optimal payments only, and ≥$10,000 of both payment types, respectively. Conclusions: Industry payments related to the optimal cancer treatment were associated with increased use of the optimal treatment. However, payments related to non-optimal treatments were more common and were associated with decreased use of optimal treatment.