Background: Novel metastatic cancer therapies have improved overall survival (OS), yet real-world data suggest this is driven by increases in patient groups that already had the highest OS. Although a better understanding of the impact of new therapies on health disparities is needed, individual patient data (IPD) from randomized controlled trials (RCTs) to enable such analyses are rarely available and, hence, a framework for quantifying disparities based on aggregate data is needed.
Objectives: To quantify the impact of novel cancer therapies on disparities in OS based on aggregate data.
Methods: Nine (N=9) recent Food and Drug Administration-approved therapies for metastatic breast cancer, colorectal cancer, or non-small cell lung cancer, were identified. Inclusion was based on the availability of published Kaplan-Meier (KM) data for OS from phase III RCTs, with at least 100 patients in each arm, and OS reaching ≤30%. IPD were replicated and pooled across all cancers for the novel therapies and comparators separately and weighted based on sample size, ensuring equal contribution in the analyses. Censoring was handled by fitting parametric survival distributions. For the pooled interventions and comparators separately, OS disparity was defined as the absolute difference in median OS between the highest and lowest group, with stratification based on either 2 or 5 groups. A linear regression model was fitted to OS for the 5 groups, with the regression coefficient considered as the Inequality Gradient (IG). The impact of the novel therapies was defined as the change in OS disparity compared to the comparator therapies. A probabilistic analysis was performed to account for parameter uncertainty and a scenario analysis was conducted using mixture distributions to explore structural uncertainty in the survival modelling.
Results: Using the 2-level stratification, disparities in OS increased by 8.5 (95%-confidence interval: 5.6; 11.5) months or 36.5% from 23.3 to 31.8 months. Considering 5 groups, OS disparities increased by 22.8 (14.2; 31.7) months or 39.7% from 57.8 to 80.5 months. The IG increased by 5.2 (3.3; 7.2) or 39.2% from 13.3 to 18.5. Conservative survival extrapolations using a mixture model yielded similar results in the scenario analysis, but the 5-group scenario results were not significant due to increased parameter uncertainty.
Conclusions: Novel metastatic cancer therapies increase OS disparities relative to their RCT comparators. The proposed framework can be used for other outcomes commonly reported in KM curves, such as progression-free survival. Although it cannot replace equity-informed analyses using IPD, for example on socioeconomic factors, it can provide important insights in the absence of such data.
