The Randomized Registry Trial

Because healthcare in the United States has become more value-based, patients, providers, and other stakeholders urgently need more high-quality clinical evidence. 1 Such evidence can inform clinical decisions, guide healthcare policies, evaluate care quality, and, ultimately, provide and assess value. Randomized clinical trials (RCTs) represent a criterion standard for evidence because of their ability to remove systematic biases and enhance insights into causality. Their complexity, need for scale, difficulties in patient and provider accrual, and immense costs, however, hamper their performance and completion. To overcome some of the challenges facing RCTs, trialists have recently begun to seek other ways to attain evidence. One such strategy is to leverage the abundance of clinical registries as data platforms for clinical trials. The registry-based randomized trial, or randomized registry trial (RRT), aims to approach the statistical rigor of randomization, whereas expediting and enhancing patient enrollment, minimizing cost, and addressing the generalizability of findings (Table 1). 2–5 TABLE 1: Comparison of Study DesignsThe first published RRT was conducted in Sweden. 3 In less than 3 years, 7244 patients across 3 countries in more than 30 sites were randomized. The trial was estimated to have costed only $50 per patient. 2,3,5 The first RRT in the United States was the Study of Access Site for Enhancement of Percutaneous Coronary Intervention (SAFE-PCI) for Women Trial, 2,6 conducted with the National Institutes of Health National Cardiovascular Data Registry CathPCI Registry platform. 7 In this pragmatic, open-label trial, women were randomized to percutaneous coronary intervention via radial versus femoral access to compare postprocedure bleeding. Using the registry, high-volume hospitals with physicians proficient at both radial and femoral access were identified and recruited. The trial reached 99.2% accrual in only 22 months. 6 Registry data were copied into each participant's electronic trial case report and supplemented with trial-specific data by each site's trial coordinator, reducing the coordinator's workload by approximately 65%. 6 Registry abstractors and site trial coordinators collaborated closely. The registry abstractors accelerated their workflow to ensure timely collection of data for patient safety monitoring. An independent Data and Safety Monitoring Committee reviewed data on a quarterly basis, and all participating sites received approval by their respective institutional review boards. Significant investments have been made in the development of and participation in registries—and their use has expanded. 8 In surgery, registries focused on quality improvement, such as the American College of Surgeons’ National Surgical Quality Improvement Program (ACS NSQIP), have been developed to evaluate and improve surgical care and outcomes. In addition, research using these registry data has made considerable contributions to clinical practice and health policy. Finally, registries are now playing an increasingly important role in value-based healthcare. The Quality Payment Program of the Medicare Access and CHIP Reauthorization Act relies on Qualified Clinical Data Registries to track the quality of care provided and adjust reimbursement. 9 Recently, the ACS NSQIP and the Americas Hepato-Pancreato-Biliary Association (AHPBA) have designed an RRT for surgical care. As one of the first RRTs in surgery, it aims to test a quality improvement intervention using the ACS NSQIP. Patients undergoing pancreatoduodenectomy at participating centers are randomized to receive 1 of 2 widely available antibiotics as surgical antimicrobial prophylaxis, and then the ACS NSQIP tracks their outcomes to determine which regimen reduces surgical site infections more effectively. 10 This relevant yet unanswered clinical question has garnered broad engagement by the surgical community who performs these procedures. Importantly, there is equipoise. In setting up the trial, we have learned a number of important lessons using this novel trial design that are worth sharing. First, in contrast to the typical RCT, whereby the design fits the question, the RRT requires that the question fit the registry. The RRT design is not meant to replace traditional RCTs. Instead, the RRT offers an alternative strategy to answer questions that would otherwise remain unanswered because of time and money constraints. RRTs are fundamentally pragmatic, and most appropriate for testing hypotheses involving pharmaceuticals, devices, and treatments widely accepted in real-world settings but variable in their use. 4,5 Asking the right question is therefore paramount. Is a registry available that can answer the proposed question? If so, does the registry act as an adjunct to a traditional RCT or is it the complete infrastructure? The ACS NSQIP-AHPBA trial asked a clinically relevant question that can be answered wholly within the ACS NSQIP framework. Second, the primary purpose of the registry should remain undisturbed by the methods and conduct of the RRT. The primary purpose of the registry cannot be compromised to benefit the trial. For instance, we could not allow only patients undergoing pancreatoduodenectomy who participate in the trial to be abstracted. Otherwise, hospital benchmarking results could be affected. Trial designs must be thoughtfully tailored to the capabilities and scope of the target registry. Third, high data quality and rigorous statistics are vital. Accuracy of the findings is heavily dependent on data quality. Validation procedures and data audits are routine in high-caliber data registries to ensure data integrity—these will be integral to successful RRTs. Randomization and analytic strategies used in RCTs remain applicable to RRTs. Interim analyses should be planned accordingly. For example, a futility analysis is incorporated into the ACS NSQIP-AHPBA trial. As in all RCTs, strong statistics are important. Fourth, because most registries rely on data abstractors to collect data, we must balance the workload and workflow of an RRT with those of the registry. This includes deciding whether or not additional data are needed, and what the added collection burden will be. Future advances in automated data abstraction may help mitigate these concerns and further promote the RRT as a viable trial paradigm. Unlike the SAFE-PCI for Women Trial, 6,7 the current trial requires minimal additional data collection beyond what is routinely collected in the ACS NSQIP. In fact, the ACS NSQIP currently collects all primary and secondary outcomes, and only two additional data points are needed: whether the patient is enrolled in the trial and the reason for trial violations, if any. Fifth, a strong ethos of collaboration is integral to any successful RRT. Mindfulness of each stakeholder's needs and limitations cannot be emphasized enough. Key individuals that understand both trial methodology and registry processes should be included in the primary working group during the course of the trial. In our trial, representatives of the ACS NSQIP work closely with the AHPBA Clinical Trials Committee. Data sharing agreements have been discussed. An open dialogue with institutional review boards is also important given the novelty of the trial design. 5 Despite challenges, RRTs can be a viable substrate to enhance professional collaboration by teaming with societies to develop and conduct clinical trials. In conclusion, the RRT is a promising design that may help to address the contemporary challenges facing RCTs. Numerous high-quality clinical registries dedicated to surgery are presently available across many surgical specialties and each could be poised to become platforms for RRTs. For suitable questions, trialists now have the opportunity to consider RRTs as an efficient alternative to the traditional RCT. The ACS NSQIP-AHPBA collaboration is one of the first formal explorations of an RRT in surgery and may serve as proof of concept. Because significant time, effort, and money have been invested into registries, leveraging this infrastructure to conduct randomized trials may satisfy our urgent needs for high-quality, speedy evidence. Acknowledgments Dr. Liu's position is supported by a research fellowship from the Department of Surgery, University of Chicago Hospitals, under the auspices of Dr. Jeffrey B. Matthews, and by the American College of Surgeons Clinical Scholars in Residence Program.