237P Overall survival (OS) estimation using target trial emulation of major randomized controlled trials (RCTs) with the ESME metastatic breast cancer (MBC) real-world data (RWD)

Trial emulation applies RCT methodology to observational data, allowing estimation of long-term OS by mimicking existing RCTs. We emulated eight major RCTs in MBC to validate this approach using the ESME-MBC data. We selected the emulated RCTs based on feasibility regarding power and key confounders among 13 pre-selected pivotal RCTs since 2000. Emulation was combined with appropriate adjustment using Stabilized Inverse Probability of Treatment Weight as the reference method, and alternative methods as sensitivity analyses. The RWD and RCT results were compared using three predefined agreement metrics. Table: 237POS estimates and agreement metricsEmulation sample sizeHazard Ratio [95% Confidence Interval]AgreementRCTAdjusted RWDSAEASDPALOMA-3 (palbociclib)1, 1390.81 [0.64-1.03]0.81 [0.69-0.95]✓✓PALOMA-2 (palbociclib)2, 4130.96 [0.78-1.18]0.81 [0.72-0.92]✓✓MONALEESA-2 (ribociclib)1, 0030.76 [0.63-0.93]0.55 [0.39-0.79]✓✓CLEOPATRA(pertuzumab)1, 0620.68 [0.56-0.84]0.44 [0.36-0.55]✓EMILIA (T-DM1)8360.68 [0.55-0.85]0.64 [0.52-0.77]✓✓✓RIBBON-1 (bevacizumab4, 7031.11 [0.86-1.43]1.16 [1.07-1.25]✓✓AVADO (bevacizumab)8731.03 [0.70-1.33]0.96 [0.77-1.20]✓✓✓BOLERO-2 (everolimus)9040.89 [0.73-1.10]1.25 [1.06-1.47]SA, Full statistical significance agreement = when RWD and RCT HRs and CIs are on the same side of the null; EA, Estimate agreement = when the RWD HR falls within the 95% CI of the RCT; SD, Standardized difference agreement = difference in effect size between RCT and RWD relative to their standard deviation, achieved when the null hypothesis of no difference is not rejected, i.e. |z| ≤ 1.96. Open table in a new tab SA, Full statistical significance agreement = when RWD and RCT HRs and CIs are on the same side of the null; EA, Estimate agreement = when the RWD HR falls within the 95% CI of the RCT; SD, Standardized difference agreement = difference in effect size between RCT and RWD relative to their standard deviation, achieved when the null hypothesis of no difference is not rejected, i.e. |z| ≤ 1.96. These 8 emulations provide important insight into the real-world efficacy of major MBC drugs released in the past 15 years. Discrepancies may arise from residual confounding and drastic changes in prescription practices following drug approval. This approach provides opportunities to tackle challenges in demonstrating comparative effectiveness by applying external control arms.