Fig. 2: Bias plots showing unmeasured confounding for comparisons between the 1 L pralsetinib trial cohort and 1 L pembrolizumab cohort, and 1 L pralsetinib trial cohort and 1 L pembrolizumab with chemotherapy cohort. | Nature Communications

Fig. 2: Bias plots showing unmeasured confounding for comparisons between the 1 L pralsetinib trial cohort and 1 L pembrolizumab cohort, and 1 L pralsetinib trial cohort and 1 L pembrolizumab with chemotherapy cohort.

From: Addressing challenges with real-world synthetic control arms to demonstrate the comparative effectiveness of Pralsetinib in non-small cell lung cancer

Fig. 2

A Bias plots for unmeasured confounding corresponds to the comparison with 1 L pembrolizumab comparison (HR 0.38, 95% CI 0.21–0.67), B corresponds to the comparison involving 1 L pembrolizumab with chemotherapy comparison (HR 0.37, 95% CI 0.21–0.64). These graphs plot unconfounded treatment effect estimates as risk ratios (ARR adjusted risk ratio) after adjusting for a hypothetical unmeasured binary confounder over a range of confounder-exposure and confounder-outcome associations on the risk ratio scale. The colors map the strength of an unmeasured confounder (x and y axes) to the robustness of this study’s conclusions (color gradient). The worst-case strengths of measured baseline confounders are shown using HRs from the multiple imputation resulting from QBA for missing data assumptions.

Back to article page