Directly Comparing Randomized and Real-World Controls in Adjuvant Breast Cancer


Boyd M1, Walker B2, Silver M3, Shi J4, Bulsara P4, Robert N1
1Ontada, Irving, TX, USA, 2Tulane University, New Orleans, LA, USA, 3Ontada, SALT LAKE CTY, UT, USA, 4Ontada, The Woodlands, TX, USA


To directly compare contemporaneous real-world and randomized controlled trial control arms to assess the viability of real-world controls in early breast cancer.


The patients were women with HER2-negative high-risk breast cancer in the adjuvant setting treated with docetaxel, doxorubicin, and cyclophosphamide for six cycles between May 29, 2007 and November 21, 2013, mirroring the control arm and study timeframe from our reference trial Anthracyclines in Early Breast Cancer: The ABC Trials. Data were sourced from the iKnowMed electronic medical record system which included both real-world and trial patients allowing for their direct comparison. We used the 1:1 greedy nearest neighbor matching method based on propensity scores with a caliper of 0.1 to balance the sample by age, body mass index, race, weeks since diagnoses, stage, ECOG status, hormone receptor status, number of positive nodes, smoking history, and region and network of care. This resulted in a final sample of 306 patients (i.e., 153 real-world and 153 trial). Our primary endpoint was disease-free-survival estimated using the Kaplan-Meier method with 95% confidence intervals.


The matching process resulted in balance between arms along available covariates using a standardized difference threshold of 0.2. The Kaplan-Meier curves and the distribution of disease-free survival did not suggest significant differences between real-world and trial arms (median not reached in real-world or trial arm, log-rank p-value 0.98).


While this study has several limitations (e.g., use of structured data only, indirect information on disease-free intervals, and disease-free survival as a proxy outcome), owing in part to the ability to appropriately balance the sample on observed covariates, these results suggest that real-world data could have served as a synthetic control in this setting.

Conference/Value in Health Info

2021-05, ISPOR 2021, Montreal, Canada

Value in Health, Volume 24, Issue 5, S1 (May 2021)




Clinical Outcomes, Health Technology Assessment, Organizational Practices, Real World Data & Information Systems

Topic Subcategory

Best Research Practices, Clinical Outcomes Assessment, Decision & Deliberative Processes, Health & Insurance Records Systems



Your browser is out-of-date

ISPOR recommends that you update your browser for more security, speed and the best experience on Update my browser now