The Challenge of Estimating Treatment Effects in Cardiac Surgery

The comparison of the outcomes of different surgical techniques is the foundation of progress in cardiac surgery. Still, estimating treatment effects is a complex methodological challenge.

Randomized trials have low risk of bias but limited external validity. On the other hand, observational studies have a higher risk of bias and confounders but are more representative of clinical practice. While in theory biases may lead to an overestimation of the treatment effect in observational studies and the lower external validity may reduce the treatment effect in randomized trials, it has been reported that in other fields the estimates of treatment effects are qualitatively and quantitively similar between the two. Although no data specific to cardiac surgery exist, we believe that this concordance does not hold true for cardiac surgery studies (and probably for surgical studies in general).

For several decades the cardiac surgery community has relied on observational comparative studies, often due to the perception than randomized trials are too challenging and complex. We have described how only less than 10% of the available evidence on the 10 most commonly performed surgical procedures comes from randomized trials and how the number of published cardiac surgery trials has decreased significantly over the course of the last 20 years.

This attitude has resulted in the fact that many research questions have been unresolved for decades. On the other hand, few committed investigators have been able to successfully complete large trials showing that randomized trials in cardiac surgery are possible and, based on previous points, absolutely necessary. It is the responsibility of the surgical community to work collectively to design and implement the high-quality trials that are needed to provide answers to critical clinical questions.