Already a DIA Member? Sign in. Not a member? Join.

Sign in

Forgot User ID? or Forgot Password?

Not a Member?

Create Account and Join

Menu Back to Poster-Presentations-Details

W-07: A General Framework for Utilizing Real World Data with Clinical Trials





Poster Presenter

      Xiaoyun (Nicole) Li

      • Senior Principal Scientist
      • Merck & Co., Inc.
        United States

Objectives

This manuscript provides a general framework for utilizing real-world data with clinical trials

Method

Real world data (RWD) has the potential to be used alongside clinical trial data for a variety of applications, including (1) Virtual control arms, (2) Augmenting an existing clinical trial arm, and (3) Strategy planning for clinical studies. RWD data are used to compare with clinical trials data. Co-author - Wei Zhou

Results

The first step of utilizing RWD with clinical trial data is to define the question of interest and thus the target population of interest, which may also include inclusion/exclusion criteria. For example, how would overall survival of RWD pembrolizumab patients look if they had characteristics similar to clinical trial pembrolizumab patients? Next, the relevant RWD and clinical trial data are pooled and all baseline characteristics (biomarkers, labs, vitals, etc) that are in both data sources are included. Propensity scores, or the probability of being in a clinical trial, are estimated using the pooled data set and are then used to weight the RWD and/or clinical trial data towards the target population of interest. While a common approach is to use logistic regression to estimate propensity scores, we illustrate the advantage of robust machine learning models, specifically the SuperLearner algorithm. Last, since oncology trials often explicitly state that patients should have a life expectancy of at least 3 months, we propose a novel early death “nomogram” based on external RWD that assigns more weight to RWD patients that are likely to survive beyond 3 months. We then apply our proposed methods for (1) Comparing the overall survival of NSCLC patients treated with pembrolizumab monotherapy to their corresponding clinical trials, (2) Augmenting pembrolizumab clinical trial arms with pembrolizumab RWD, (3) Pembrolizumab RWD as Virtual ‘Control’ Arms, and (4) Use existing RWD for prediction and new study design planning. Overall, while further validation and development is ongoing, these results illustrate practical use cases and analytical considerations for utilizing RWD with clinical trial data.

Conclusion

We demonstrated the use of real world data in clinical trials and use advanced machine learning models and propensity scores methods to utilize real world data in clinical trials. The results show that we can use existing real-world data for predictions and new study design planning.

Be informed and stay engaged.

Don't miss an opportunity - join our mailing list to stay up to date on DIA insights and events.