OBJECTIVES: Payers might be more interested in real-world effectiveness rather than efficacy of novel treatments. We aimed to assess the use of real-world evidence along with efficacy data reported in a clinical trial setting (CTS) to estimate effectiveness in a real-world setting (RWS).

METHODS: Two RCTs in non-small cell lung cancer (NSCLC) and renal cell carcinoma (RCC) were selected. First, patient-level data (PLD) was extracted from reported Kaplan-Meier curves for overall survival (OS). Cox proportional hazard model and parametric survival models were fitted on PLD. Second, based on reported treatments in respective control arms, similar patient populations were identified from a German sickness fund dataset. Parametric survival models were fitted over the retrieved PLD. This curve was considered to represent the effectiveness of control treatments in RWS. Third, estimated HRs were applied to RWS to estimate OS in active treatment arms. Finally, we estimated mean OS in both CTS and RWS.

RESULTS: In NSCLC, estimated mean OS in the CTS was 40.7 months (95% confidence interval (CI): 25.0-61.1) and 24.8 (95% CI: 17.2-35.6) for the active and control arms, respectively. In the RWS, estimated mean OS was 15.4 (95% CI: 12.2-19.8) for the control arm, and based on a HR of 0.61 (95%CI: 0.42-0.90), the mean OS for the active treatment was 23.3 (95%CI: 16.8-32.2). In RCC, estimated mean OS in the CTS was 27.8 (95%CI: 17.5-44.8) and 23.8 (95% CI: 17.1-33.8) for the active and control arms, respectively. In the RWS, estimated mean OS was 31.4 (95% CI: 23.1-43.4) for the control arm and based on the HR of 0.72 (95%CI: 0.55-0.95), the mean OS was 41.8 (95%CI: 32.9-52.5) for the active arm.

CONCLUSIONS: If treatment effect is expected to be comparable in both CTS and RWS, this methodology may be used to estimate the effectiveness of novel treatments.