# Measuring the Contribution of Genomic Predictors to Improving Estimator Precision in Randomized trials

## TL;DR

This paper explores how genomic data, like the MammaPrint signature, can improve the precision of treatment effect estimates in clinical trials, potentially reducing required sample sizes and costs.

## Contribution

The study quantifies the precision gains from combining genomic and clinical data in covariate-adjusted randomized trials.

## Key findings

- Adjusting for clinical factors alone improves estimator precision by 5-6%.
- Adjusting for genomic factors alone provides a 5% precision gain.
- Combining clinical and genomic factors yields an additional 2-3% precision gain.

## Abstract

The use of genomic data in the clinic has not been as widespread as was envisioned when sequencing and genomic analysis became common techniques. An underlying difficulty is the direct assessment of how much additional information genomic data are providing beyond standard clinical measurements. This is hard to quantify in the clinical setting where laboratory tests based on genomic signatures are fairly new and there are not sufficient data collected to determine how valuable these tests have been in practice. Here we focus on the potential precision gain from using the popular MammaPrint genomic signature in a covariate-adjusted, randomized clinical trial. We describe how adjustment of an estimator for the average treatment effect using baseline measurements can improve precision. This precision gain can be translated directly into sample size reduction and corresponding cost savings. We conduct a simulation study using genomic and clinical data gathered for breast cancer patients and find that adjusting for clinical factors alone provides a gain in precision of 5-6%, adjusting for genomic factors alone provides a similar gain (5%), and combining the two yields a 2-3% additional gain over only adjusting for clinical covariates.