Propensity score methods for merging observational and experimental datasets

Treatment effect estimation using the propensity score in clinical trials with historical control

BACKGROUND

Clinical trials assessing new treatment effects require a control group to compare the pure treatment effects. However, in clinical trials on regenerative medicine, rare diseases, and intractable diseases, it may be ethically difficult to assign participants to the control group. In recent years, the use of historical control data has attracted attention as a method for supplementing the number of participants in the control group. When combining historical control data with new randomized controlled trial (RCT) data, the assessment of heterogeneity using outcome data is not sufficient. Therefore, several statistical methods that consider participant outcomes and baseline characteristics, including the propensity score (PS) method have been proposed.

METHODS

We propose a new method considering "information on whether the data are RCT data or not" in the PS model when combining the RCT and historical control data. The performance of the proposed method in estimating the treatment effect is evaluated using simulation data.

RESULTS

When the distribution of covariates is similar between the RCT and historical control data, not much difference in performance is found between the proposed and conventional methods to estimate the treatment effect. On the other hand, when the distribution of covariates is not similar between the two kinds of data, the proposed method shows higher performance.

CONCLUSIONS

Even when it is not known whether RCT and historical control data are similar, the proposed PS model is useful to estimate the treatment effect appropriately in RCTs using historical control data.

Combining observational and experimental datasets using shrinkage estimators

We consider the problem of combining data from observational and experimental sources to draw causal conclusions. To derive combined estimators with desirable properties, we extend results from the Stein shrinkage literature. Our contributions are threefold. First, we propose a generic procedure for deriving shrinkage estimators in this setting, making use of a generalized unbiased risk estimate. Second, we develop two new estimators, prove finite sample conditions under which they have lower risk than an estimator using only experimental data, and show that each achieves a notion of asymptotic optimality. Third, we draw connections between our approach and results in sensitivity analysis, including proposing a method for evaluating the feasibility of our estimators.

Methods for Integrating Trials and Non-experimental Data to Examine Treatment Effect Heterogeneity

Estimating treatment effects conditional on observed covariates can improve the ability to tailor treatments to particular individuals. Doing so effectively requires dealing with potential confounding, and also enough data to adequately estimate effect moderation. A recent influx of work has looked into estimating treatment effect heterogeneity using data from multiple randomized controlled trials and/or observational datasets. With many new methods available for assessing treatment effect heterogeneity using multiple studies, it is important to understand which methods are best used in which setting, how the methods compare to one another, and what needs to be done to continue progress in this field. This paper reviews these methods broken down by data setting: aggregate-level data, federated learning, and individual participant-level data. We define the conditional average treatment effect and discuss differences between parametric and nonparametric estimators, and we list key assumptions, both those that are required within a single study and those that are necessary for data combination. After describing existing approaches, we compare and contrast them and reveal open areas for future research. This review demonstrates that there are many possible approaches for estimating treatment effect heterogeneity through the combination of datasets, but that there is substantial work to be done to compare these methods through case studies and simulations, extend them to different settings, and refine them to account for various challenges present in real data.