In Reply Romeo and colleagues note that randomized clinical trials (RCTs) to determine the clinical efficacy of mechanical circulatory support especially percutaneous ventricular assist devices (PVAD) are not justified owing to the challenges in conducting RCTs in critically ill patients and the small expected treatment benefit of mechanical circulatory support devices.
We disagree with that assertion. Several trials of interventions in critically ill patients have been successful in improving patient outcomes and have substantially changed clinical practice.1-3 In contrast, comparative effectiveness studies based on observational data, including ours4 and that by Inohara et al,5 should be interpreted with caution, as there is potential for confounding by indication (selection bias) despite the use of robust statistical methods such as propensity score analysis. The main purpose of our study was to describe the exponential growth in usage of PVADs, an expensive new technology, even though its effectiveness on improving clinical end points has not been conclusively proven in RCTs. Our hope is that these findings spark a debate on the process of device regulation in the United States, which is in need of reform as also noted by the Institute of Medicine.6 We do not believe that patient interests are served when high-risk medical devices receive regulatory approval based on evidence of substantial equivalence to a predicate device using the 510(k) pathway rather than hard data obtained from well-conducted RCTs to establish device efficacy and safety. Moreover, once a device becomes commercially available, it becomes even more difficult to conduct RCTs because patients and physicians have ready access to the device outside a clinical trial. The case of the intra-aortic balloon pump (IABP), which was incorporated into clinical practice without clear outcome data, serves as an example. It is possible that the sickest patients (and potentially the ones most likely to be benefit) were not randomized in the IABP SHOCK II trial7 owing to ready availability of IABP outside clinical trials, which may explain why the observed mortality was lower than expected. Such concerns can limit the findings from RCTs such as the IABP SHOCK II trial to influence clinical practice.