Propensity‐score‐based priors for Bayesian augmented control design

Abstract

Drug developers are required to demonstrate substantial evidence of effectiveness through the conduct of adequate and well-controlled (A&WC) studies to obtain marketing approval of their medicine. What constitutes A&WC is interpreted as the conduct of randomized controlled trials (RCTs). However, these trials are sometimes unfeasible because of their size, duration, and cost. One way to reduce sample size is to leverage information on the control through a prior. One consideration when forming data-driven prior is the consistency of the external and the current data. It is essential to make this process less susceptible to choosing information that only helps improve the chances toward making an effectiveness claim. For this purpose, propensity score methods are employed for two reasons: (1) it gives the probability of a patient to be in the trial, and (2) it minimizes selection bias by pairing together treatment and control within the trial and control subjects in the external data that are similar in terms of their pretreatment characteristics. Two matching schemes based on propensity scores, estimated through generalized boosted methods, are applied to a real example with the objective of using external data to perform Bayesian augmented control in a trial where the allocation is disproportionate. The simulation results show that the data augmentation process prevents prior and data conflict and improves the precision of the estimator of the average treatment effect.

A Food and Drug Administration analysis of survival outcomes comparing the Adjuvant Paclitaxel and Trastuzumab trial with an external control from historical clinical trials

“…By matching patient-level data, the balance of patients' baseline characteristics can be examined to provide a measure of similarity between patients in the treatment and control arms. 12 One major strength of our analyses is the number of patients pooled. In addition, the analysis provided a unique ability to explore survival outcomes for patients with verylow-risk features (T1mi/T1a/T1b).…”

Combining Real-World and Randomized Control Trial Data Using Data-Adaptive Weighting via the On-Trial Score

“…Lin's method uses an on-trial score, similar to a propensity score, where the outcome of interest is inclusion in the trial to construct a matched set of external standard-of-care patients and weight their likelihood contribution. 12 The on-trial score is estimated as the probability that a patient is in the clinical trial given their baseline covariates using a logistic regression model. Next, optimal pair matching is performed using the on-trial score so that each trial patient receiving the intervention is matched with an external standardof-care patient.…”

“…11 Another recently proposed method uses a modification of the propensity score, called the on-trial score, to create matches between the external standard-of-care patients and the trial standard-of-care patients in order to create a hybrid standard-of-care arm consisting of patients most similar to those in the clinical trial intervention arm. 12 In this paper, we propose a new data-adaptive weighting method that addresses the limitation of assigning a single weight to the entire group of external standard-of-care patients by assigning weights to each individual in the external standard-of-care arm based on sim-ilarity to trial patients using the on-trial score. The use of individualized weights helps to account for the fact that patients included in EHR databases may be more heterogeneous than patients included in clinical trials.…”

Propensity‐score‐based meta‐analytic predictive prior for incorporating real‐world and historical data

“…To address this issue, Han et al 14 proposed a Bayesian hierarchical model that incorporates patient‐level covariates to enhance the efficiency of borrowing. Lin et al 15 extended the power prior approach and refined propensity‐score matching methods by defining individual discounting weights based on pre‐treatment characteristics. More recently, Wang et al 16 proposed a method for propensity score‐integrated power prior (PS‐power prior).…”