PERSPECTIVES OPINION
Registry-based randomized clinical trials—a new clinical trial paradigm Stefan James, Sunil V. Rao and Christopher B. Granger Abstract | Randomized clinical trials provide the foundation of clinical evidence to guide physicians in their selection of treatment options. Importantly, randomization is the only reliable method to control for confounding factors when comparing treatment groups. However, randomized trials have limitations, including the increasingly prohibitive costs of conducting adequately powered studies. Local and national regulatory requirements, delays in approval, and unnecessary trial processes have led to increased costs and decreased efficiency. Another limitation is that clinical trials involve selected patients who are treated according to protocols that might not represent real-world practice. A possible solution is registry-based randomized clinical trials. By including a randomization module in a large inclusive clinical registry with unselected consecutive enrolment, the advantages of a prospective randomized trial can be combined with the strengths of a large-scale all-comers clinical registry. We believe that prospective registry-based randomized clinical trials are a powerful tool for conducting studies efficiently and cost-effectively. James, S. et al. Nat. Rev. Cardiol. advance online publication 17 March 2015; doi:10.1038/nrcadio.2015.33
Introduction
The highest level of scientific clinical evidence stems from prospective, randomized, controlled clinical trials, which control for confounding factors by establishing two groups of patients who have balanced baseline characteristics. When well designed and performed, these trials are the gold standard for comparative studies. Nonrandomized, underpowered, or single-centre randomized trials, unsupported by other independent evidence, are generally not considered to be adequate to support a conclusion of efficacy.1,2 International guideline committees require more than one adequate and well-controlled trial supporting a treatment to recommend that therapy with the highest level of evidence, which reflects the need for independent verification of experimental results.1,2 In medicine, and particularly within the field of cardiology, an increasing number Competing interests S.J. has received institutional research grants from Medtronic, Terumo, and Vascular Solutions for the conduct of the TASTE trial. The other authors declare no competing interests.
of pharmaceutical agents, medical devices, and clinical procedures are being tested in appropriately designed prospective randomized trials. However, only a minority of the recommendations given in international guidelines are based on the highest level of evidence from prospective randomized trials.3 In addition, many trials are limited by the specific recruitment of patients using narrow inclusion criteria and multiple exclusion criteria, thereby limiting the trial’s generalizability to real-world patients seen in practice settings. 4 Drugs and medical devices are, therefore, frequently used outside of their approved indication.5 Furthermore, large-scale randomized clinical trials are complex and expensive to perform, and economic revenue is typically the primary incentive to initiate such trials.6 Funding from government and other sources for large trials to address many of the important clinical strategies, therapies, and medical devices that require evaluation is limited. Consequently, many trials are too small to provide reliable estimates of the risk–benefit balance. Furthermore, the regulatory and ethical requirements for randomized clinical trials, developed
NATURE REVIEWS | CARDIOLOGY
to protect patients and to ensure quality, impede conduct of simple and important trials because of the economic barriers required to overcome them.6,7 A ‘catch 22’ situation might, therefore, exist in which to protect the patients’ interests, evidence of effectiveness of common treatments is never established.8 Ironically, patients are routinely treated in practice with therapies that have not been verified for efficacy and safety, and for which they have not provided consent. Therefore, novel ways to use existing data-collection platforms, such as qu ality-improvement registries, administrative claims, and electronic health records, can provide an opportunity to enhance patient enrolment in large, simple trials. Patients can then be randomly allocated with most of their required baseline medical history already recorded, minimizing the need for additional data collection and onsite monitoring. This concept is the foundation for the design of registry-based randomized trials, which we believe are a powerful tool to increase the efficiency and cost-effectiveness of clinical trials.
Observational studies
Large-scale clinical registries initiated to assess the quality of clinical performance are, at present, successfully collecting data from consecutive patients in many hospitals and health-care organizations; Denmark, Sweden, and the UK have some of the most complete national databases.9 Observational studies from these large-scale registries or from trial databases provide valuable information and enhance clinical guidance for diseases in which prospective randomized trials are missing or difficult to conduct (Box 1). 10 The early detection of stent thrombosis and high mortality with the first generation of drug-eluting stents and the early evaluation of drug-eluting balloons within the SCAAR registry are examples of important large observations.11,12 Retrospective analyses of prospectively collected data are ideally used for descriptive studies to assess treatment patterns and evaluate outcomes associated with therapies applied in clinical practice. These analyses might also provide complementary data to those from randomized trials.13 Ideally, with complete nonselective enrolment, a ADVANCE ONLINE PUBLICATION | 1
© 2015 Macmillan Publishers Limited. All rights reserved
PERSPECTIVES Box 1 | Main features of trials Observational studies Observational studies, drawn from large populations, are complementary to prospective randomized trials ■■ Despite appropriate statistical adjustments, some confounding factors cannot be completely eliminated ■■ The interpretation of observational studies assessing treatment effects must be approached with caution ■■ Results should be considered nondefinitive and hypothesis generating Registry-based randomized clinical trials ■■ Randomly assigning patients in a clinical quality registry combines the features of a prospective randomized trial with a large-scale clinical registry ■■ Registry-based trials are less selective and enable fast enrolment, control of nonenrolled patients, and the possibility of very long-term follow-up ■■ Inexpensive and simple designs are the main strengths of registry-based randomized clinical trials ■■ The clinical registry can be used to identify patients for enrolment, perform randomization, collect baseline variables, and detect end points
large-scale registry will truly reflect the overall patient population, particularly if the registry is integrated into the healthcare system. However, observational studies have limited value for assessing differences in therapeutic options owing to selection bias and confounding factors, both measured (for which adjustment is possible) and unmeasured (for which adjustment is not possible). Observational studies are, therefore, not sufficiently reliable to estimate modest treatment effects, and should not be used for this purpose. For many treatments, including ones that are commonly used, randomized trials are highly unlikely to be performed owing to complex study procedures, small patient populations, lack of funding opportunities for academia, or lack of financial incentives for industry. For these patient subsets, results from observational studies might be the only data available to evaluate safety and efficacy.14 For the same reasons, many prospective randomized trials are not adequately powered for important clinical end points, and surrogate end points that have not been sufficiently validated are often used. An observational study might also provide important data for reliable power calculations that are needed to achieve clinically meaningful results, particularly in low-risk populations and in trials to evaluate low-frequency events.
Registry-based randomized trials
A prospective trial requires several important elements to be successful, including the identification of eligible patients, attainment of patient consent, random assignment of treatment, collection of baseline variables, and the detection and adjudication of clinical end points (Box 2). A clinical registry can be used for some, and ideally most, of these aspects. A registry can be used for collection of baseline variables and to identify eligible patients for the trial, and by using interactive methodology, can also actively propose enrolment in a trial. By incorporating randomization in a clinical quality registry, some of the critical attributes of a prospective randomized trial can be combined with the practical features of a large-scale clinical registry including the main strengths of minimally-selected consecutive enrolment and automated and efficient patient identification and followup (Table 1).15,16 We describe such a prospective randomized trial built on a registry platform as a ‘prospective registry-based randomized clinical trial’ (RRCT). An RRCT can efficiently and effectively be used to assess hard clinical end points in large patient cohorts, and is well suited for open-label evaluation of commonly used therapeutic alternatives in settings with existing registries. A simple trial design with open-label randomization, limited monitoring of trial processes, data focused on key outcomes, and no centralized event adjudication is inexpensive, but has limitations and is not suitable for all types of trials. For drugs or medical devices that require comprehensive safety reporting or intense pharmacokinetic or pharmacodynamic modelling, or trials that require strictly defined end points, a traditional trial methodology including blinding of treatment alternatives, dedicated follow-up, systematic monitoring, and formal adjudication is needed to ensure patient safety and sufficient quality of end point reporting. However, the linking of numerous functionalities to a clinical registry might still be possible. To date, RRCTs are being used to evaluate treatments, strategies, and devices or acute-phase pharmacological agents, but no strict limit exists as to what therapy can be evaluated with the registry link, as long as patient safety is assured and existing regulations are followed. RRCTs with efficient and streamlined trial conduct might also be used for the evaluation of pharmaceutical agents for new indications. Depending on the trial hypotheses, outcome variables, and specific
2 | ADVANCE ONLINE PUBLICATION
aims, the registry can be used for most of the typical trial functionalities or for collection of baseline variables only. Benefits of the linkage to a registry might include the capacity to identify patients, to enrol larger proportions of patients, and to conduct long-term follow-up at low cost. Another important strength of an RRCT is the capacity to describe and follow up the complete reference population, that is, all patients who are eligible but nonrandomized, as well as noneligible individuals.
Low-cost, streamlined design
Conducting prospective randomized trials can be tedious and expensive. Conse quently, owing to both cost and the regulatory burden, many new and existing interventions are not evaluated, and the trials that are conducted are smaller and less informative than they might otherwise be. The RRCT design solves some of these cost issues. A large reduction in cost can be achieved by using existing hospital and registry networks. The TASTE trial17,18 involved 7,400 patients and was performed and completed with >90% cost saving compared with similar randomized trials with a conventional design. Internet-based platforms, current systems for patient identification, baseline and study data collection, and event detection can all be used from existing databases rather than arranging and rebuilding electronic records and other infrastructures for every new trial. Other cost-saving aspects of an RRCT include the limited number of trial-specific patient visits and fewer problems with patient retention. Furthermore, cost can be reduced with the use of existing collaboration networks for site selection, web-based start-up meetings, and focused risk-based monitoring of critical variables, excluding only those variables that do not affect the main outcome of the trial. However, a RRCT should still be based on appropriate power calculations to enrol enough patients and acquire sufficient outcome events to demonstrate effectiveness. The link to a registry Box 2 | Collection of outcome variables ■■ Detection of outcomes can be performed by a registry or a separate method such as telephone calls or outpatient visits ■■ Short-term and long-term all-cause mortality is a reliable hard end point that does not require definition and has low susceptibility to ascertainment bias ■■ Outcome variables can be adjudicated if subject to uncertainty
www.nature.com/nrcardio © 2015 Macmillan Publishers Limited. All rights reserved
PERSPECTIVES can facilitate an accurate estimation of event rates, and the event rate of the enrolled and nonenrolled patients can be followed up to place the trial population in the context of an unselected population.
Table 1 | Advantages and disadvantages of different clinical trial designs Clinical trial design
Advantages
Disadvantages
Registry studies and observational studies
Ideal for description of standards Unselected patient populations (generalizable cohorts) Large number of events allows for the identification of rare events Inexpensive
Data quality is variable and questionable Cannot be used for comparative outcomes research Confounding factors cannot be adjusted for, despite advanced statistical models
Randomized clinical trials
Well-designed studies with adequate power (gold-standard clinical design) Removes confounding factors
Highly selected populations owing to specific inclusion and exclusion criteria Often performed at specialized study centres Often include surrogate end points Requires long time to plan and complete Expensive Often sponsored by industry (only studies with economic interest will be performed)
Registry-based randomized clinical trials
Randomization removes potential confounding Less-selected patient populations Large number of events allows for the identification of rare events Simple design Inexpensive
Data quality might be variable and questionable Variables might not be well-defined Limited possibility for collection of detailed safety reporting, biospecimens, and pharmacokinetics or pharmacodynamic indices
The TASTE experience
The RRCT concept was first implemented within the SWEDEHEART registry, 9 in which manual thrombus aspiration was prospectively evaluated as an adjunctive treatment to primary percutaneous coronary intervention (PCI) for acute myocardial infarction, with mortality as the primary end point.17 The TASTE trial18 was a multicentre, prospective, randomized, controlled, open-label clinical trial involving patients with ST-segment elevation myocardial infarction (STEMI) undergoing PCI. The national comprehensive SWEDEHEART registry 9,19 was used for all aspects of the TASTE trial. The registry contains data for all consecutive patients from all coronary intervention centres in Sweden and Iceland. SWEDEHEART is funded solely by national health authorities and provides immediate and continuous feedback on processes and quality-of-care measures. Baseline and procedural data are entered into the registry directly online. Validation of source data against electronic health records has an overall agreement of 95%.9,19 The online registry was used to identify patients suitable for inclusion in TASTE by confirming that the patients were >18 years of age, that PCI was planned because of STEMI, and that they had not previously been included in the trial. The treating clinician had to confirm only a few inclusion criteria, such as correspondence between an electrocardiogram finding and stenosed artery, and obtain verbal consent from the patient. The ethics committee requested a written confirmation of the consent from the patient as soon as possible after the procedure. Patients were randomly assigned treatment directly in the registry clinical report form, and no additional studyrelated activities for staff or patients were included in the study. All end points were automatically collected from national registries. No study-specific monitoring of data or event adjudication was performed during the trial. Consequently, all-cause mortality collected from the national population registry was chosen as the primary end point of the TASTE trial. Secondary end points such as stent thrombosis, repeat myocardial infarction, stroke, and readmission to hospital for heart failure were collected
from the SCAAR/SWEDEHEART registry and from the mandatory national hospital admission registry. In total, 7,244 patients were randomly assigned in the study.20 During 2 years and 9 months of enrolment, 61% of all patients who presented with STEMI and were referred for PCI, and 82% of all patients who were potentially eligible for enrolment in Sweden and Iceland were randomly assigned to receive either thrombus aspiration as an adjunct to PCI, or PCI alone (Figure 1). Patients who were not randomly assigned included those unable to give informed verbal consent and, therefore, did not fulfil the inclusion criteria. No patients were lost to follow-up for the primary end point owing to automated, personalized identification number tracking. The trial showed no significant difference in the primary end point of all-cause mortality at 30 days or at 1 year.18 While the cost of such a trial is subsidized by the existing registry and willingness of investigators to participate for minimal monetary compensation, the additional cost involved in establishing and administering the SCAAR/SWEDEHEART registries was ~US$400,000, compared with tens of millions of dollars for a study of equivalent size using a traditional industry‑funded trial model.18
Other registry-based trials
A number of stent trials performed by investigators in the Danish research network (the SORT OUT trials) have involved the
NATURE REVIEWS | CARDIOLOGY
random allocation of patients using a postal or interactive voice randomization system. In the SORT OUT studies, data was collected in a specific clinical report form, but national registries were used to detect clinical events such as death, myocardial infarction, and revascularization.21,22 Detected events were thereafter adjudicated with standard methodology. This trial design has two important advantages: the ability to describe baseline demographics and clinical outcomes in all patients treated during the study period, and not only those included in the randomized controlled trial; and systematic end-point detection through registries without additional patient visits. A registry-based trial methodology was also used in the SAFE‑PCI for Women trial in the USA, which incorporated a randomized trial into the existing cardiovascular research infrastructure of the NIH National Cardiovascular Data Registry™ CathPCI Registry.23 SAFE‑PCI for Women24,25 was a prospective, randomized trial to compare radial and femoral artery access in women undergoing PCI. The primary efficacy end point was a composite of bleeding or vascular complication requiring intervention. The trial design had two major advantages over traditional methodologies: using a clinical registry for identification of trial sites and investigators, and providing trial data. The registry enabled the identification of operators and sites that could include patients whose risk of femoral bleeding or vascular complications was balanced by a risk ADVANCE ONLINE PUBLICATION | 3
© 2015 Macmillan Publishers Limited. All rights reserved
PERSPECTIVES 14,000 12,000
All primary PCIs Randomly allocated in trial
Patients
10,000 8,000 6,000 4,000 n = 7,244 61% of all patients 82% of all eligible patients
2,000
Ju n J e Se Au uly pt gu e s O mb t No cto er Devember ce be Ja mb r Fe nuaer br r y u M ar y ar Apch r M il a Ju y ne Se Au July pt gu em st NoOct ber o Devember ce be Ja mb r Fe nuaer br r y u M ar y ar Apch r M il Ju ay ne Se A Jul u pt gu y e s O mb t No cto er Devember ce be Ja mb r Fe nuaer br r y u M ar y ar ch
0
2010
2011
Date
2012
2013
Nature Reviews | Cardiology Figure 1 | Inclusion rate in the TASTE trial.18 All primary PCI procedures in Sweden between June 2010 and March 2013 are displayed in blue, and patients enrolled in the TASTE trial in red, during the enrolment period of 2.5 years. Abbreviations: PCI, percutaneous coronary intervention.
of procedural failure with radial access, or those at risk of vascular complications from both radial and femoral approaches. Patient demographics, medical history, concomitant medications, procedural details, and index hospitalization clinical outcomes are routinely coded into the CathPCI registry data collection form using standardized data elements and electronically captured without the involvement of the study site coordinator. For the SAFE‑PCI for Women trial itself, extra information specific to vascular access and bleeding definitions that were not in the CathPCI Registry was obtained using additional electronic case report form pages.24 Overall, this approach reduced the site coordinator workload by ~65% per patient, compared with the traditional study clinical report forms.25 The resulting SAFE‑PCI database was designed to be compliant with regulatory requirements for investigational new drug or device exemption studies. The trial was successful, but the cost savings was far less than that in the TASTE trial, owing to the lack of full integration of this registry into clinical care and necessary redundancies. The DETO2X‑AMI trial26 is the first RRCT to assess the efficacy of a pharmacological strategy using the SWEDEHEART registry as a basis for randomization, baseline variable collection, and primary outcome detection. Investigators in this trial are evaluating whether supplemental oxygen, as compared with normal room air, can reduce all-cause mortality in 6,600 patients with suspected or confirmed myocardial infarction. Patient
randomization is performed in ambulances, emergency departments, coronary care units, and catheterization laboratories, by means of an easily accessible online randomization module in which a 10‑digit patient personal identification number is entered. Inclusion criteria, such as symptoms indicating myocardial infarction, oxygen saturation >90%, age >30 years, and clinically significant electrocardiogram changes or elevated troponin levels, and exclusion criteria such as cardiac arrest before inclusion, need for ambulatory continuous oxygen therapy, and inability to consent, need to be confirmed before a patient is allocated to one of the treatment alternatives. The randomization module automatically connects with the SWEDEHEART registry for collection of name and sex, and all data are automatically transferred to the registry. All relevant trial data are obtained directly from the registry with no other documentation needed. Mortality data obtained outside the hospital is collected by merging with the Swedish Population Registry,27 which includes information of the vital status of all Swedish citizens. Merging with additional public registries, including the National Patient Registry, which contains information on hospital discharge, allows for the collection of other clinical end point data. However, a trial design that uses national registries as the only basis for follow-up is limited by the lack of formal central adjudication of clinical events. The DETO2X‑AMI trial26 also has a potential for ascertainment bias, owing to its open-label design. Consequently,
4 | ADVANCE ONLINE PUBLICATION
1‑year all-cause mortality from the Swedish Population Registry was chosen as the primary end point. The estimated completion of enrolment in this study is early 2016. Finally, the VALIDATE-SWEDEHEART trial28 is an open-label, multicentre, prospective, RRCT involving patients with myocardial infarction and undergoing PCI. The trial aim is to determine whether bivalirudin is superior to heparin alone in reducing death, myocardial infarction, and major bleeding events in patients with STEMI or NSTEMI, pretreated with ticagrelor or prasugrel, and undergoing PCI. Investigators will enrol 3,000 patients with STEMI and 3,000 patients with NSTEMI undergoing PCI. VALIDATESWEDEHEART will use a hybrid RRCT design, whereby randomization is performed and baseline data are collected from the SWEDEHEART registry.28 Clinical end point data are collected through national registries and phone calls after 7 days and 180 days, and adjudicated by a blinded central end-point committee. Long-term outcomes will also be accumulated from later events recorded in the registry or in public databases. The estimated completion of enrolment is early 2016. RRCTs are currently promoted by the National Institutes of Health6 and the Patient-Centered Outcomes Research Institute in the USA as priorities to address public-health needs.6,29
Conclusions
Study randomization controls for confounding baseline differences and is an important feature for studies designed to compare treatment effect. By including a randomization module in a large, all-inclusive clinical registry with unselected consecutive enrolment, some of the most important features of a prospective randomized trial can be combined with the inclusiveness and efficiencies of a large-scale all-comers clinical registry. Such trials can be described as a prospective RRCT and are a powerful and highly cost-effective tool to establish clinical evidence in many areas of medicine that might not otherwise be appropriately evaluated. The integration of clinical trials and real-world practice is critical to determine which treatments will improve public health efficiently. Uppsala Clinical Research Center, Uppsala University Hospital, Dag Hammarskjölds väg 14B, 75237 Uppsala, Sweden (S.J.). The Duke Clinical Research Institute, Duke University Medical Center, 2301 Erwin Road, Durham, NC 27710, USA (S.V.R., C.B.G.). Correspondence to: S.J. [email protected]
www.nature.com/nrcardio © 2015 Macmillan Publishers Limited. All rights reserved
PERSPECTIVES 1.
Steg, P. G., James, S. K. & Gersh, B. J. 2012 ESC STEMI guidelines and reperfusion therapy: evidence-based recommendations, ensuring optimal patient management. Heart 99, 1156–1157 (2013). 2. O’Gara, P. T. et al. 2013 ACCF/AHA guideline for the management of ST-elevation myocardial infarction: a report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines. Circulation 127, e362–e425 (2013). 3. Tricoci, P., Allen, J. M., Kramer, J. M., Califf, R. M. & Smith, S. C. Jr. Scientific evidence underlying the ACC/AHA clinical practice guidelines. JAMA 301, 831–841 (2009). 4. Wiviott, S. D. et al. Prasugrel versus clopidogrel in patients with acute coronary syndromes. N. Eng. J. Med. 357, 2001–2015 (2007). 5. Carlsson, J. et al. Outcome of drug-eluting versus bare-metal stenting used according to on- and off-label criteria. J. Am. Coll. Cardiol. 53, 1389–1398 (2009). 6. Reith, C. et al. Randomized clinical trials— removing unnecessary obstacles. N. Eng. J. Med. 369, 1061–1065 (2013). 7. Duley, L. et al. Specific barriers to the conduct of randomized trials. Clin. Trials 5, 40–48 (2008). 8. Eapen, Z. J., Lauer, M. S. & Temple, R. J. The imperative of overcoming barriers to the conduct of large, simple trials. JAMA 311, 1397–1398 (2014). 9. Jernberg, T. et al. The Swedish web-system for enhancement and development of evidencebased care in heart disease evaluated according to recommended therapies (SWEDEHEART). Heart 96, 1617–1621 (2010). 10. James, S. K. et al. Long-term safety and efficacy of drug-eluting versus bare-metal stents in Sweden. N. Eng. J. Med. 360, 1933–1945 (2009).
11. Lagerqvist, B. et al. Long-term outcomes with drug-eluting stents versus bare-metal stents in Sweden. N. Eng. J. Med. 356, 1009–1019 (2007). 12. Bondesson, P. et al. Comparison of two drugeluting balloons: a report from the SCAAR registry. EuroIntervention 8, 444–449 (2012). 13. Graham, D. J. et al. Cardiovascular, bleeding, and mortality risks in elderly Medicare patients treated with dabigatran or warfarin for nonvalvular atrial fibrillation. Circulation 131, 157–164 (2015). 14. Benson, K. & Hartz, A. J. A comparison of observational studies and randomized, controlled trials. N. Eng. J. Med. 342, 1878–1886 (2000). 15. James, S., Frobert, O. & Lagerqvist, B. Cardiovascular registries: a novel platform for randomised clinical trials. Heart 98, 1329–1331 (2012). 16. Lauer, M. S. & D’Agostino, R. B. Sr. The randomized registry trial—the next disruptive technology in clinical research? N. Eng. J. Med. 369, 1579–1581 (2013). 17. Fröbert, O. et al. Thrombus Aspiration in STElevation myocardial infarction in Scandinavia (TASTE trial): a multicenter, prospective, randomized, controlled clinical registry trial based on the Swedish angiography and angioplasty registry (SCAAR) platform. Study design and rationale. Am. Heart J. 160, 1042–1048 (2010). 18. Lagerqvist, B. et al. Outcomes 1 year after thrombus aspiration for myocardial infarction. N. Eng. J. Med. 371, 1111–1120 (2014). 19. Harnek, J. et al. The 2011 outcome from the Swedish Health Care Registry on Heart Disease (SWEDEHEART). Scand. Cardiovas. J. 47 (Suppl.), 1–10 (2013). 20. Fröbert, O. et al. Thrombus aspiration during ST‑segment elevation myocardial infarction. N. Eng. J. Med. 369, 1587–1597 (2013).
NATURE REVIEWS | CARDIOLOGY
21. Thuesen, L. et al. Event detection using population-based health care databases in randomized clinical trials: a novel research tool in interventional cardiology. Clin. Epidemiol. 19, 357–361 (2013). 22. Devereaux, P. J. & Yusuf, S. When it comes to trials, do we get what we pay for? N. Eng. J. Med. 369, 1962–1963 (2013). 23. Thuesen, L. et al. Event detection using population-based health care databases in randomized clinical trials: a novel research tool in interventional cardiology. Clin. Epidemiol. 5, 357–361 (2013). 24. Moussa, I. et al. The NCDR CathPCI Registry: a US national perspective on care and outcomes for percutaneous coronary intervention. Heart 99, 297–303 (2013). 25. Hess, C. N. et al. Embedding a randomized clinical trial into an ongoing registry infrastructure: unique opportunities for efficiency in design of the Study of Access site For Enhancement of Percutaneous Coronary Intervention for Women (SAFE-PCI for Women). Am.Heart J. 166, 421–428 (2013). 26. Hofmann, R. et al. DETermination of the role of Oxygen in suspected Acute Myocardial Infarction trial. Am. Heart J. 167, 322–328 (2014). 27. Swedish Population Registry. Swedish Tax Agency [online], http://www.skatteverket.se (2015). 28. US National Library of Medicine. ClinicalTrials.gov [online], https://clinicaltrials.gov/ct2/show/ NCT02311231 (2015). 29. Patient-Centered Outcomes Research Institute. Research We Support [online], http://www. pcori.org/content/research-we-support (2014). Author contributions All the authors researched data for the article, substantially contributed to discussion of content, wrote, and reviewed and edited the manuscript before submission.
ADVANCE ONLINE PUBLICATION | 5 © 2015 Macmillan Publishers Limited. All rights reserved
Registry-based randomized clinical trials—a new clinical trial paradigm Stefan James, Sunil V. Rao and Christopher B. Granger Abstract | Randomized clinical trials provide the foundation of clinical evidence to guide physicians in their selection of treatment options. Importantly, randomization is the only reliable method to control for confounding factors when comparing treatment groups. However, randomized trials have limitations, including the increasingly prohibitive costs of conducting adequately powered studies. Local and national regulatory requirements, delays in approval, and unnecessary trial processes have led to increased costs and decreased efficiency. Another limitation is that clinical trials involve selected patients who are treated according to protocols that might not represent real-world practice. A possible solution is registry-based randomized clinical trials. By including a randomization module in a large inclusive clinical registry with unselected consecutive enrolment, the advantages of a prospective randomized trial can be combined with the strengths of a large-scale all-comers clinical registry. We believe that prospective registry-based randomized clinical trials are a powerful tool for conducting studies efficiently and cost-effectively. James, S. et al. Nat. Rev. Cardiol. advance online publication 17 March 2015; doi:10.1038/nrcadio.2015.33
Introduction
The highest level of scientific clinical evidence stems from prospective, randomized, controlled clinical trials, which control for confounding factors by establishing two groups of patients who have balanced baseline characteristics. When well designed and performed, these trials are the gold standard for comparative studies. Nonrandomized, underpowered, or single-centre randomized trials, unsupported by other independent evidence, are generally not considered to be adequate to support a conclusion of efficacy.1,2 International guideline committees require more than one adequate and well-controlled trial supporting a treatment to recommend that therapy with the highest level of evidence, which reflects the need for independent verification of experimental results.1,2 In medicine, and particularly within the field of cardiology, an increasing number Competing interests S.J. has received institutional research grants from Medtronic, Terumo, and Vascular Solutions for the conduct of the TASTE trial. The other authors declare no competing interests.
of pharmaceutical agents, medical devices, and clinical procedures are being tested in appropriately designed prospective randomized trials. However, only a minority of the recommendations given in international guidelines are based on the highest level of evidence from prospective randomized trials.3 In addition, many trials are limited by the specific recruitment of patients using narrow inclusion criteria and multiple exclusion criteria, thereby limiting the trial’s generalizability to real-world patients seen in practice settings. 4 Drugs and medical devices are, therefore, frequently used outside of their approved indication.5 Furthermore, large-scale randomized clinical trials are complex and expensive to perform, and economic revenue is typically the primary incentive to initiate such trials.6 Funding from government and other sources for large trials to address many of the important clinical strategies, therapies, and medical devices that require evaluation is limited. Consequently, many trials are too small to provide reliable estimates of the risk–benefit balance. Furthermore, the regulatory and ethical requirements for randomized clinical trials, developed
NATURE REVIEWS | CARDIOLOGY
to protect patients and to ensure quality, impede conduct of simple and important trials because of the economic barriers required to overcome them.6,7 A ‘catch 22’ situation might, therefore, exist in which to protect the patients’ interests, evidence of effectiveness of common treatments is never established.8 Ironically, patients are routinely treated in practice with therapies that have not been verified for efficacy and safety, and for which they have not provided consent. Therefore, novel ways to use existing data-collection platforms, such as qu ality-improvement registries, administrative claims, and electronic health records, can provide an opportunity to enhance patient enrolment in large, simple trials. Patients can then be randomly allocated with most of their required baseline medical history already recorded, minimizing the need for additional data collection and onsite monitoring. This concept is the foundation for the design of registry-based randomized trials, which we believe are a powerful tool to increase the efficiency and cost-effectiveness of clinical trials.
Observational studies
Large-scale clinical registries initiated to assess the quality of clinical performance are, at present, successfully collecting data from consecutive patients in many hospitals and health-care organizations; Denmark, Sweden, and the UK have some of the most complete national databases.9 Observational studies from these large-scale registries or from trial databases provide valuable information and enhance clinical guidance for diseases in which prospective randomized trials are missing or difficult to conduct (Box 1). 10 The early detection of stent thrombosis and high mortality with the first generation of drug-eluting stents and the early evaluation of drug-eluting balloons within the SCAAR registry are examples of important large observations.11,12 Retrospective analyses of prospectively collected data are ideally used for descriptive studies to assess treatment patterns and evaluate outcomes associated with therapies applied in clinical practice. These analyses might also provide complementary data to those from randomized trials.13 Ideally, with complete nonselective enrolment, a ADVANCE ONLINE PUBLICATION | 1
© 2015 Macmillan Publishers Limited. All rights reserved
PERSPECTIVES Box 1 | Main features of trials Observational studies Observational studies, drawn from large populations, are complementary to prospective randomized trials ■■ Despite appropriate statistical adjustments, some confounding factors cannot be completely eliminated ■■ The interpretation of observational studies assessing treatment effects must be approached with caution ■■ Results should be considered nondefinitive and hypothesis generating Registry-based randomized clinical trials ■■ Randomly assigning patients in a clinical quality registry combines the features of a prospective randomized trial with a large-scale clinical registry ■■ Registry-based trials are less selective and enable fast enrolment, control of nonenrolled patients, and the possibility of very long-term follow-up ■■ Inexpensive and simple designs are the main strengths of registry-based randomized clinical trials ■■ The clinical registry can be used to identify patients for enrolment, perform randomization, collect baseline variables, and detect end points
large-scale registry will truly reflect the overall patient population, particularly if the registry is integrated into the healthcare system. However, observational studies have limited value for assessing differences in therapeutic options owing to selection bias and confounding factors, both measured (for which adjustment is possible) and unmeasured (for which adjustment is not possible). Observational studies are, therefore, not sufficiently reliable to estimate modest treatment effects, and should not be used for this purpose. For many treatments, including ones that are commonly used, randomized trials are highly unlikely to be performed owing to complex study procedures, small patient populations, lack of funding opportunities for academia, or lack of financial incentives for industry. For these patient subsets, results from observational studies might be the only data available to evaluate safety and efficacy.14 For the same reasons, many prospective randomized trials are not adequately powered for important clinical end points, and surrogate end points that have not been sufficiently validated are often used. An observational study might also provide important data for reliable power calculations that are needed to achieve clinically meaningful results, particularly in low-risk populations and in trials to evaluate low-frequency events.
Registry-based randomized trials
A prospective trial requires several important elements to be successful, including the identification of eligible patients, attainment of patient consent, random assignment of treatment, collection of baseline variables, and the detection and adjudication of clinical end points (Box 2). A clinical registry can be used for some, and ideally most, of these aspects. A registry can be used for collection of baseline variables and to identify eligible patients for the trial, and by using interactive methodology, can also actively propose enrolment in a trial. By incorporating randomization in a clinical quality registry, some of the critical attributes of a prospective randomized trial can be combined with the practical features of a large-scale clinical registry including the main strengths of minimally-selected consecutive enrolment and automated and efficient patient identification and followup (Table 1).15,16 We describe such a prospective randomized trial built on a registry platform as a ‘prospective registry-based randomized clinical trial’ (RRCT). An RRCT can efficiently and effectively be used to assess hard clinical end points in large patient cohorts, and is well suited for open-label evaluation of commonly used therapeutic alternatives in settings with existing registries. A simple trial design with open-label randomization, limited monitoring of trial processes, data focused on key outcomes, and no centralized event adjudication is inexpensive, but has limitations and is not suitable for all types of trials. For drugs or medical devices that require comprehensive safety reporting or intense pharmacokinetic or pharmacodynamic modelling, or trials that require strictly defined end points, a traditional trial methodology including blinding of treatment alternatives, dedicated follow-up, systematic monitoring, and formal adjudication is needed to ensure patient safety and sufficient quality of end point reporting. However, the linking of numerous functionalities to a clinical registry might still be possible. To date, RRCTs are being used to evaluate treatments, strategies, and devices or acute-phase pharmacological agents, but no strict limit exists as to what therapy can be evaluated with the registry link, as long as patient safety is assured and existing regulations are followed. RRCTs with efficient and streamlined trial conduct might also be used for the evaluation of pharmaceutical agents for new indications. Depending on the trial hypotheses, outcome variables, and specific
2 | ADVANCE ONLINE PUBLICATION
aims, the registry can be used for most of the typical trial functionalities or for collection of baseline variables only. Benefits of the linkage to a registry might include the capacity to identify patients, to enrol larger proportions of patients, and to conduct long-term follow-up at low cost. Another important strength of an RRCT is the capacity to describe and follow up the complete reference population, that is, all patients who are eligible but nonrandomized, as well as noneligible individuals.
Low-cost, streamlined design
Conducting prospective randomized trials can be tedious and expensive. Conse quently, owing to both cost and the regulatory burden, many new and existing interventions are not evaluated, and the trials that are conducted are smaller and less informative than they might otherwise be. The RRCT design solves some of these cost issues. A large reduction in cost can be achieved by using existing hospital and registry networks. The TASTE trial17,18 involved 7,400 patients and was performed and completed with >90% cost saving compared with similar randomized trials with a conventional design. Internet-based platforms, current systems for patient identification, baseline and study data collection, and event detection can all be used from existing databases rather than arranging and rebuilding electronic records and other infrastructures for every new trial. Other cost-saving aspects of an RRCT include the limited number of trial-specific patient visits and fewer problems with patient retention. Furthermore, cost can be reduced with the use of existing collaboration networks for site selection, web-based start-up meetings, and focused risk-based monitoring of critical variables, excluding only those variables that do not affect the main outcome of the trial. However, a RRCT should still be based on appropriate power calculations to enrol enough patients and acquire sufficient outcome events to demonstrate effectiveness. The link to a registry Box 2 | Collection of outcome variables ■■ Detection of outcomes can be performed by a registry or a separate method such as telephone calls or outpatient visits ■■ Short-term and long-term all-cause mortality is a reliable hard end point that does not require definition and has low susceptibility to ascertainment bias ■■ Outcome variables can be adjudicated if subject to uncertainty
www.nature.com/nrcardio © 2015 Macmillan Publishers Limited. All rights reserved
PERSPECTIVES can facilitate an accurate estimation of event rates, and the event rate of the enrolled and nonenrolled patients can be followed up to place the trial population in the context of an unselected population.
Table 1 | Advantages and disadvantages of different clinical trial designs Clinical trial design
Advantages
Disadvantages
Registry studies and observational studies
Ideal for description of standards Unselected patient populations (generalizable cohorts) Large number of events allows for the identification of rare events Inexpensive
Data quality is variable and questionable Cannot be used for comparative outcomes research Confounding factors cannot be adjusted for, despite advanced statistical models
Randomized clinical trials
Well-designed studies with adequate power (gold-standard clinical design) Removes confounding factors
Highly selected populations owing to specific inclusion and exclusion criteria Often performed at specialized study centres Often include surrogate end points Requires long time to plan and complete Expensive Often sponsored by industry (only studies with economic interest will be performed)
Registry-based randomized clinical trials
Randomization removes potential confounding Less-selected patient populations Large number of events allows for the identification of rare events Simple design Inexpensive
Data quality might be variable and questionable Variables might not be well-defined Limited possibility for collection of detailed safety reporting, biospecimens, and pharmacokinetics or pharmacodynamic indices
The TASTE experience
The RRCT concept was first implemented within the SWEDEHEART registry, 9 in which manual thrombus aspiration was prospectively evaluated as an adjunctive treatment to primary percutaneous coronary intervention (PCI) for acute myocardial infarction, with mortality as the primary end point.17 The TASTE trial18 was a multicentre, prospective, randomized, controlled, open-label clinical trial involving patients with ST-segment elevation myocardial infarction (STEMI) undergoing PCI. The national comprehensive SWEDEHEART registry 9,19 was used for all aspects of the TASTE trial. The registry contains data for all consecutive patients from all coronary intervention centres in Sweden and Iceland. SWEDEHEART is funded solely by national health authorities and provides immediate and continuous feedback on processes and quality-of-care measures. Baseline and procedural data are entered into the registry directly online. Validation of source data against electronic health records has an overall agreement of 95%.9,19 The online registry was used to identify patients suitable for inclusion in TASTE by confirming that the patients were >18 years of age, that PCI was planned because of STEMI, and that they had not previously been included in the trial. The treating clinician had to confirm only a few inclusion criteria, such as correspondence between an electrocardiogram finding and stenosed artery, and obtain verbal consent from the patient. The ethics committee requested a written confirmation of the consent from the patient as soon as possible after the procedure. Patients were randomly assigned treatment directly in the registry clinical report form, and no additional studyrelated activities for staff or patients were included in the study. All end points were automatically collected from national registries. No study-specific monitoring of data or event adjudication was performed during the trial. Consequently, all-cause mortality collected from the national population registry was chosen as the primary end point of the TASTE trial. Secondary end points such as stent thrombosis, repeat myocardial infarction, stroke, and readmission to hospital for heart failure were collected
from the SCAAR/SWEDEHEART registry and from the mandatory national hospital admission registry. In total, 7,244 patients were randomly assigned in the study.20 During 2 years and 9 months of enrolment, 61% of all patients who presented with STEMI and were referred for PCI, and 82% of all patients who were potentially eligible for enrolment in Sweden and Iceland were randomly assigned to receive either thrombus aspiration as an adjunct to PCI, or PCI alone (Figure 1). Patients who were not randomly assigned included those unable to give informed verbal consent and, therefore, did not fulfil the inclusion criteria. No patients were lost to follow-up for the primary end point owing to automated, personalized identification number tracking. The trial showed no significant difference in the primary end point of all-cause mortality at 30 days or at 1 year.18 While the cost of such a trial is subsidized by the existing registry and willingness of investigators to participate for minimal monetary compensation, the additional cost involved in establishing and administering the SCAAR/SWEDEHEART registries was ~US$400,000, compared with tens of millions of dollars for a study of equivalent size using a traditional industry‑funded trial model.18
Other registry-based trials
A number of stent trials performed by investigators in the Danish research network (the SORT OUT trials) have involved the
NATURE REVIEWS | CARDIOLOGY
random allocation of patients using a postal or interactive voice randomization system. In the SORT OUT studies, data was collected in a specific clinical report form, but national registries were used to detect clinical events such as death, myocardial infarction, and revascularization.21,22 Detected events were thereafter adjudicated with standard methodology. This trial design has two important advantages: the ability to describe baseline demographics and clinical outcomes in all patients treated during the study period, and not only those included in the randomized controlled trial; and systematic end-point detection through registries without additional patient visits. A registry-based trial methodology was also used in the SAFE‑PCI for Women trial in the USA, which incorporated a randomized trial into the existing cardiovascular research infrastructure of the NIH National Cardiovascular Data Registry™ CathPCI Registry.23 SAFE‑PCI for Women24,25 was a prospective, randomized trial to compare radial and femoral artery access in women undergoing PCI. The primary efficacy end point was a composite of bleeding or vascular complication requiring intervention. The trial design had two major advantages over traditional methodologies: using a clinical registry for identification of trial sites and investigators, and providing trial data. The registry enabled the identification of operators and sites that could include patients whose risk of femoral bleeding or vascular complications was balanced by a risk ADVANCE ONLINE PUBLICATION | 3
© 2015 Macmillan Publishers Limited. All rights reserved
PERSPECTIVES 14,000 12,000
All primary PCIs Randomly allocated in trial
Patients
10,000 8,000 6,000 4,000 n = 7,244 61% of all patients 82% of all eligible patients
2,000
Ju n J e Se Au uly pt gu e s O mb t No cto er Devember ce be Ja mb r Fe nuaer br r y u M ar y ar Apch r M il a Ju y ne Se Au July pt gu em st NoOct ber o Devember ce be Ja mb r Fe nuaer br r y u M ar y ar Apch r M il Ju ay ne Se A Jul u pt gu y e s O mb t No cto er Devember ce be Ja mb r Fe nuaer br r y u M ar y ar ch
0
2010
2011
Date
2012
2013
Nature Reviews | Cardiology Figure 1 | Inclusion rate in the TASTE trial.18 All primary PCI procedures in Sweden between June 2010 and March 2013 are displayed in blue, and patients enrolled in the TASTE trial in red, during the enrolment period of 2.5 years. Abbreviations: PCI, percutaneous coronary intervention.
of procedural failure with radial access, or those at risk of vascular complications from both radial and femoral approaches. Patient demographics, medical history, concomitant medications, procedural details, and index hospitalization clinical outcomes are routinely coded into the CathPCI registry data collection form using standardized data elements and electronically captured without the involvement of the study site coordinator. For the SAFE‑PCI for Women trial itself, extra information specific to vascular access and bleeding definitions that were not in the CathPCI Registry was obtained using additional electronic case report form pages.24 Overall, this approach reduced the site coordinator workload by ~65% per patient, compared with the traditional study clinical report forms.25 The resulting SAFE‑PCI database was designed to be compliant with regulatory requirements for investigational new drug or device exemption studies. The trial was successful, but the cost savings was far less than that in the TASTE trial, owing to the lack of full integration of this registry into clinical care and necessary redundancies. The DETO2X‑AMI trial26 is the first RRCT to assess the efficacy of a pharmacological strategy using the SWEDEHEART registry as a basis for randomization, baseline variable collection, and primary outcome detection. Investigators in this trial are evaluating whether supplemental oxygen, as compared with normal room air, can reduce all-cause mortality in 6,600 patients with suspected or confirmed myocardial infarction. Patient
randomization is performed in ambulances, emergency departments, coronary care units, and catheterization laboratories, by means of an easily accessible online randomization module in which a 10‑digit patient personal identification number is entered. Inclusion criteria, such as symptoms indicating myocardial infarction, oxygen saturation >90%, age >30 years, and clinically significant electrocardiogram changes or elevated troponin levels, and exclusion criteria such as cardiac arrest before inclusion, need for ambulatory continuous oxygen therapy, and inability to consent, need to be confirmed before a patient is allocated to one of the treatment alternatives. The randomization module automatically connects with the SWEDEHEART registry for collection of name and sex, and all data are automatically transferred to the registry. All relevant trial data are obtained directly from the registry with no other documentation needed. Mortality data obtained outside the hospital is collected by merging with the Swedish Population Registry,27 which includes information of the vital status of all Swedish citizens. Merging with additional public registries, including the National Patient Registry, which contains information on hospital discharge, allows for the collection of other clinical end point data. However, a trial design that uses national registries as the only basis for follow-up is limited by the lack of formal central adjudication of clinical events. The DETO2X‑AMI trial26 also has a potential for ascertainment bias, owing to its open-label design. Consequently,
4 | ADVANCE ONLINE PUBLICATION
1‑year all-cause mortality from the Swedish Population Registry was chosen as the primary end point. The estimated completion of enrolment in this study is early 2016. Finally, the VALIDATE-SWEDEHEART trial28 is an open-label, multicentre, prospective, RRCT involving patients with myocardial infarction and undergoing PCI. The trial aim is to determine whether bivalirudin is superior to heparin alone in reducing death, myocardial infarction, and major bleeding events in patients with STEMI or NSTEMI, pretreated with ticagrelor or prasugrel, and undergoing PCI. Investigators will enrol 3,000 patients with STEMI and 3,000 patients with NSTEMI undergoing PCI. VALIDATESWEDEHEART will use a hybrid RRCT design, whereby randomization is performed and baseline data are collected from the SWEDEHEART registry.28 Clinical end point data are collected through national registries and phone calls after 7 days and 180 days, and adjudicated by a blinded central end-point committee. Long-term outcomes will also be accumulated from later events recorded in the registry or in public databases. The estimated completion of enrolment is early 2016. RRCTs are currently promoted by the National Institutes of Health6 and the Patient-Centered Outcomes Research Institute in the USA as priorities to address public-health needs.6,29
Conclusions
Study randomization controls for confounding baseline differences and is an important feature for studies designed to compare treatment effect. By including a randomization module in a large, all-inclusive clinical registry with unselected consecutive enrolment, some of the most important features of a prospective randomized trial can be combined with the inclusiveness and efficiencies of a large-scale all-comers clinical registry. Such trials can be described as a prospective RRCT and are a powerful and highly cost-effective tool to establish clinical evidence in many areas of medicine that might not otherwise be appropriately evaluated. The integration of clinical trials and real-world practice is critical to determine which treatments will improve public health efficiently. Uppsala Clinical Research Center, Uppsala University Hospital, Dag Hammarskjölds väg 14B, 75237 Uppsala, Sweden (S.J.). The Duke Clinical Research Institute, Duke University Medical Center, 2301 Erwin Road, Durham, NC 27710, USA (S.V.R., C.B.G.). Correspondence to: S.J. [email protected]
www.nature.com/nrcardio © 2015 Macmillan Publishers Limited. All rights reserved
PERSPECTIVES 1.
Steg, P. G., James, S. K. & Gersh, B. J. 2012 ESC STEMI guidelines and reperfusion therapy: evidence-based recommendations, ensuring optimal patient management. Heart 99, 1156–1157 (2013). 2. O’Gara, P. T. et al. 2013 ACCF/AHA guideline for the management of ST-elevation myocardial infarction: a report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines. Circulation 127, e362–e425 (2013). 3. Tricoci, P., Allen, J. M., Kramer, J. M., Califf, R. M. & Smith, S. C. Jr. Scientific evidence underlying the ACC/AHA clinical practice guidelines. JAMA 301, 831–841 (2009). 4. Wiviott, S. D. et al. Prasugrel versus clopidogrel in patients with acute coronary syndromes. N. Eng. J. Med. 357, 2001–2015 (2007). 5. Carlsson, J. et al. Outcome of drug-eluting versus bare-metal stenting used according to on- and off-label criteria. J. Am. Coll. Cardiol. 53, 1389–1398 (2009). 6. Reith, C. et al. Randomized clinical trials— removing unnecessary obstacles. N. Eng. J. Med. 369, 1061–1065 (2013). 7. Duley, L. et al. Specific barriers to the conduct of randomized trials. Clin. Trials 5, 40–48 (2008). 8. Eapen, Z. J., Lauer, M. S. & Temple, R. J. The imperative of overcoming barriers to the conduct of large, simple trials. JAMA 311, 1397–1398 (2014). 9. Jernberg, T. et al. The Swedish web-system for enhancement and development of evidencebased care in heart disease evaluated according to recommended therapies (SWEDEHEART). Heart 96, 1617–1621 (2010). 10. James, S. K. et al. Long-term safety and efficacy of drug-eluting versus bare-metal stents in Sweden. N. Eng. J. Med. 360, 1933–1945 (2009).
11. Lagerqvist, B. et al. Long-term outcomes with drug-eluting stents versus bare-metal stents in Sweden. N. Eng. J. Med. 356, 1009–1019 (2007). 12. Bondesson, P. et al. Comparison of two drugeluting balloons: a report from the SCAAR registry. EuroIntervention 8, 444–449 (2012). 13. Graham, D. J. et al. Cardiovascular, bleeding, and mortality risks in elderly Medicare patients treated with dabigatran or warfarin for nonvalvular atrial fibrillation. Circulation 131, 157–164 (2015). 14. Benson, K. & Hartz, A. J. A comparison of observational studies and randomized, controlled trials. N. Eng. J. Med. 342, 1878–1886 (2000). 15. James, S., Frobert, O. & Lagerqvist, B. Cardiovascular registries: a novel platform for randomised clinical trials. Heart 98, 1329–1331 (2012). 16. Lauer, M. S. & D’Agostino, R. B. Sr. The randomized registry trial—the next disruptive technology in clinical research? N. Eng. J. Med. 369, 1579–1581 (2013). 17. Fröbert, O. et al. Thrombus Aspiration in STElevation myocardial infarction in Scandinavia (TASTE trial): a multicenter, prospective, randomized, controlled clinical registry trial based on the Swedish angiography and angioplasty registry (SCAAR) platform. Study design and rationale. Am. Heart J. 160, 1042–1048 (2010). 18. Lagerqvist, B. et al. Outcomes 1 year after thrombus aspiration for myocardial infarction. N. Eng. J. Med. 371, 1111–1120 (2014). 19. Harnek, J. et al. The 2011 outcome from the Swedish Health Care Registry on Heart Disease (SWEDEHEART). Scand. Cardiovas. J. 47 (Suppl.), 1–10 (2013). 20. Fröbert, O. et al. Thrombus aspiration during ST‑segment elevation myocardial infarction. N. Eng. J. Med. 369, 1587–1597 (2013).
NATURE REVIEWS | CARDIOLOGY
21. Thuesen, L. et al. Event detection using population-based health care databases in randomized clinical trials: a novel research tool in interventional cardiology. Clin. Epidemiol. 19, 357–361 (2013). 22. Devereaux, P. J. & Yusuf, S. When it comes to trials, do we get what we pay for? N. Eng. J. Med. 369, 1962–1963 (2013). 23. Thuesen, L. et al. Event detection using population-based health care databases in randomized clinical trials: a novel research tool in interventional cardiology. Clin. Epidemiol. 5, 357–361 (2013). 24. Moussa, I. et al. The NCDR CathPCI Registry: a US national perspective on care and outcomes for percutaneous coronary intervention. Heart 99, 297–303 (2013). 25. Hess, C. N. et al. Embedding a randomized clinical trial into an ongoing registry infrastructure: unique opportunities for efficiency in design of the Study of Access site For Enhancement of Percutaneous Coronary Intervention for Women (SAFE-PCI for Women). Am.Heart J. 166, 421–428 (2013). 26. Hofmann, R. et al. DETermination of the role of Oxygen in suspected Acute Myocardial Infarction trial. Am. Heart J. 167, 322–328 (2014). 27. Swedish Population Registry. Swedish Tax Agency [online], http://www.skatteverket.se (2015). 28. US National Library of Medicine. ClinicalTrials.gov [online], https://clinicaltrials.gov/ct2/show/ NCT02311231 (2015). 29. Patient-Centered Outcomes Research Institute. Research We Support [online], http://www. pcori.org/content/research-we-support (2014). Author contributions All the authors researched data for the article, substantially contributed to discussion of content, wrote, and reviewed and edited the manuscript before submission.
ADVANCE ONLINE PUBLICATION | 5 © 2015 Macmillan Publishers Limited. All rights reserved
