September 1st, 2013
A TASTE of Registry-Based Randomized Trials
CardioExchange’s Harlan M. Krumholz interviews Ole Fröbert, lead investigator of the TASTE trial, about his study group’s registry-based trial design. The article is published in The New England Journal of Medicine and was presented at the European Society of Cardiology conference in Amsterdam.
Researchers in Sweden conducted a multicenter, prospective, open-label, randomized controlled trial to investigate the effect of thrombus aspiration on hard clinical endpoints in patients with ST-segment-elevation myocardial infarction (STEMI). They used the infrastructure of the population-based registry Swedish Coronary Angiography and Angioplasty Registry (SCAAR) to facilitate patient enrollment and data collection.
In all, 7244 patients with STEMI undergoing percutaneous coronary intervention (PCI) were randomized to undergo manual thrombus aspiration followed by PCI or PCI alone. All-cause mortality at 30 days was similar in the two groups — 2.8% of the thrombus-aspiration group and 3.0% in the PCI-only group — as was the rate of stroke or other neurologic complications at discharge. The thrombus-aspiration group had slightly but nonsignificantly lower rates of hospitalization for recurrent MI at 30 days (0.5% vs. 0.9%; P=0.09) and of stent thrombosis (0.2% vs. 0.5%; P=0.06). Findings were consistent across major prespecified subgroups, including those defined according to pre-PCI thrombus burden and coronary flow. No participants were lost to follow-up.
Krumholz: You have achieved a remarkable study at a low cost. What did you learn from this experience?
Fröbert: First of all that it is doable. The Scandinavian countries have a long tradition of health registries, some going back to the 1950s. More recently, several registries have become web-based, whereby a nurse or physician enters information while examining a patient. For some registries, entering information is compulsory and a prerequisite for administration. Personal identification numbers allow patients in Scandinavia to be tracked, and information on hard endpoints such as death and rehospitalization can be extracted from the same or other national registries. The quality and the fidelity are thus highly reliable.
Adding a randomization module to the existing SCAAR registry was a relatively minor alteration. All information already entered was able to serve as a substitute for case record forms, and all outcome information could be extracted from the registries when needed. Accordingly, administration related to conducting a registry-based RCT is negligible, with no need for dedicated follow-up. We also learned that if the additional work related to randomization is kept at a minimum (as in the TASTE trial), colleagues and hospitals that don’t often take part in clinical trials are eager to participate. Therefore, the spectrum of patients was broader and probably more representative than what conventional clinical trials typically have. With the costs reduced to <1% of the usual costs of an RCT, our concept opens up potentially new ways to conduct clinical trials and improve quality of care in areas with no or limited revenue.
Krumholz: Did you lose anything by taking this approach of building on the registry?
Fröbert: A strength of the registry-based RCT concept is that hard endpoints can be addressed with close to 100% accuracy. Surrogate endpoints such as ST-segment resolution or coronary flow are more susceptible to inaccuracy. Typically, this does not affect one group more than another in an RCT, but underreporting is a possibility. However, because enrollment in a registry-based RCT is simple for the physician, it is relatively easy to recruit enough patients and achieve sufficient power to learn whether the investigated treatment actually reduces mortality.
Krumholz: Is there anything you should have done differently?
Fröbert: The outcome in the TASTE trial was neutral, so additional surrogate endpoints might have been relevant in helping us learn more about the pathophysiology related to thrombus burden, distal embolization, and no reflow. In future trials, we will dedicate resources to adjudication of key surrogate endpoints to get a broader picture.
Krumholz: Are you conducting additional trials in this way?
Fröbert: Yes: Detox-AMI, which is already recruiting participants, will be a 6600-patient registry-based RCT of oxygen therapy versus room air in patients with suspected myocardial infarction. Additional trials on invasive therapies will start sometime soon.
Krumholz: What advice would you give the ACCF and AHA if they wanted to conduct RCTs on top of their registries?
Fröbert: In the U.S. and many other countries, the lack of personal identification numbers creates a challenge. However, given the amount of information being gathered in heart registries today, it seems relatively straightforward to conduct trials using existing registries such as the National Cardiovascular Data Registry, when key variables have accuracy above 90%, as platforms for RCTs. Strengths and weaknesses could be investigated in smaller pilot studies before launching larger trials.
JOIN THE DISCUSSION
What is your expectation about the future of registry-based randomized trials such as TASTE?