Doubly-robust evaluation of high-dimensional surrogate markers

TL;DR

This paper introduces a new nonparametric method for evaluating high-dimensional surrogate markers in observational studies, leveraging causal inference tools and machine learning to improve efficiency and applicability.

Contribution

It develops a doubly-robust, nonparametric approach for assessing surrogate markers in high-dimensional and observational settings, extending existing methods.

Findings

Method performs well in simulations.

Allows for high-dimensional surrogate evaluation.

Applicable to observational study data.

Abstract

When evaluating the effectiveness of a treatment, policy, or intervention, the desired measure of effectiveness may be expensive to collect, not routinely available, or may take a long time to occur. In these cases, it is sometimes possible to identify a surrogate outcome that can more easily/quickly/cheaply capture the effect of interest. Theory and methods for evaluating the strength of surrogate markers have been well studied in the context of a single surrogate marker measured in the course of a randomized clinical study. However, methods are lacking for quantifying the utility of surrogate markers when the dimension of the surrogate grows and/or when study data are observational. We propose an efficient nonparametric method for evaluating high-dimensional surrogate markers in studies where the treatment need not be randomized. Our approach draws on a connection between quantifying the utility of a surrogate marker and the most fundamental tools of causal inference -- namely, methods for estimating the average treatment effect. We show that recently developed methods for incorporating machine learning methods into the estimation of average treatment effects can be used for evaluating surrogate markers. This allows us to derive limiting asymptotic distributions for key quantities, and we demonstrate their good performance in simulation.