The RECOVERY trial sets a new standard in clinical trials, the challenge now is to make it business as usual. Becky McCall reports.
Nearly 1 year ago to the week, the first of 39 000 patients with COVID-19 was recruited into a national clinical trial that changed the course of the pandemic. Faced with a soaring death toll from a previously unknown disease, researchers combined the magic of randomisation (a means of reducing bias in treatment trials) with the scale of routine data collection to find effective therapies, all at an astonishing speed. In so doing, the UK’s flagship RECOVERY (Randomised Evaluation of COVID-19 Therapy) trial rewrote the textbook on how clinical trials can be done.
During the first wave of the pandemic, health-care staff could offer little more than symptomatic relief, care, and comfort. “Under desperate circumstances, it was clear that we knew nothing about COVID. There were treatments being used for which we had no idea if they worked or not”, remarked Martin Landray, one of the leads on the RECOVERY trial and Professor of Medicine and Epidemiology at the University of Oxford.
A trial on this scale would be a challenge at the best of times, “but we had to do so in a health system that was overwhelmed,” Landray said, presenting at the recent Association of Physicians of the UK and Ireland annual meeting. “How does one do a trial in those circumstances?”
But in its first 100 days last Spring, RECOVERY recruited 12 000 patients. Every acute care hospital in the UK participated. Key to this success, Landray noted, was to avoid complexity and hold fast to a streamlined and efficient game plan. “Keep it really simple and focus on what you need to know and not on optional extras”, he said, adding that, “the first priority had to be finding treatments to save lives and do so quickly”.
Showing a trial design dating from the 1980s, Landray highlighted the one-page consent form, the one-page case report form, and the simple randomisation to four different drugs: hydroxychloroquine, azithromycin, dexamethasone, and lopinavir–ritonavir.
Data used in RECOVERY were drawn from routine sources, including information from National Health Service (NHS) hospitals (diagnoses, procedures, and discharges), civil care (cause-specific mortality), primary care (demographics and previous medical history), critical care, dialysis audits and registries, and new COVID-19-specific data from the UK’s Lighthouse Labs. The Health Data Research Hub for Clinical Trials, known as NHS DigiTrials, was central to these data-sourcing activities.
“The challenge was to access and bring together about 25 different datasets, each with a different format and some of which had never been used for research before”, said Landray. Ventilator support provides a case in point. Information is constructed from over 300 000 data items from seven different sources, while information on deaths was collated from nine different sources, he explained.
The latest dataset used to create analyses for review by the independent data monitoring committee included about 100 000 rows of data from the manually completed case report form and more than 300 000 rows from the linked routine data sources.
These datasets come in a variety of formats and from a variety of sources. Each required careful evaluation, data mapping, and processing in order to create datasets that meet the Clinical Data Interchange Standards Consortium standards and are suitable for trial analysis, explained Landray. “Deterministic algorithms were developed to create summary phenotypic variables, for example, date of death, or duration of hospital stay of relevance to the protocol.”
Convinced of the impact of this data-driven approach, Landray stressed that it has huge advantages over traditional methods. “We have been able to reduce manual data collection to a single, one-page form, reducing the burden on NHS staff.”
Equally important, routine data collection ensures complete and comprehensive follow-up of trial participants even if they move hospital, improving feasibility and efficiency, and simultaneously driving up the quality of study results, Landray added.
In contrast, the RECOVERY trial found clear evidence that several other treatments are ineffective for patients hospitalised with COVID-19, namely, hydroxychloroquine, lopinavir–ritonavir, azithromycin, convalescent plasma, and colchicine. Results on aspirin are due shortly.
“When I started RECOVERY, nobody believed it was possible to draw on routine data in a way that could inform such a large trial”, said Landray, who says COVID-19 has served as a catalyst to demonstrate how it can be done. “RECOVERY sets a new standard in clinical trials. The challenge going forward is making that business as usual.”
Deepak L Bhatt, Professor of Medicine at Harvard Medical School, has run numerous clinical trials. “RECOVERY is a brilliant example of an efficient and effective platform to conduct important trials quickly”, he noted. “I think the COVID pandemic, in general, will catalyse several durable improvements in the conduct of clinical trials, in particular enhancing their efficiency.”
But he cautioned that not all ongoing COVID-19 trials have been as carefully constructed as RECOVERY. “Many are small, underpowered studies that provide quick but inconclusive answers.”
Bhatt also noted that trials of completely new drugs rather than known ones usually require more detailed data collection. Endpoints are often non-fatal events and subjective in nature, and involve key safety data—rare events may be missed unless conventional protocols are followed, said Bhatt. “In fact, relatively few things we do in medicine reduce mortality, but reductions in pain and improvements in quality of life are also important but typically require more focused data collection.”
Landray pointed out that to apply the trial design and methods used in RECOVERY more widely will take a concerted effort from the entire medical research community and beyond. “We need to think about the world differently”, he stressed. “This will involve new collaborations and partnerships, and experts are needed who understand the mathematics behind machine learning and can combine that with understanding of medical imaging. How do we combine these different forms of data and expertise to generate new knowledge?”
“We need to generate a new cadre of scientists who might be trained in one discipline [computational and statistical science] but can converse and understand the clinical and biomedical context and vice versa”, said Landray.
Also, importantly, regulatory and government approaches need to reflect these changes to clinical trials. “And, of course, the trust of patients, the public, and clinicians in this endeavour is pivotal”, Landray pointed out.
He advises against the use of observational studies to investigate treatment effects. Often people turn to such analyses because randomised trials are seen as too difficult, too time-consuming, or too costly, said Landray, adding that RECOVERY suggests otherwise. “RECOVERY has shown us that the combination of a simple coin toss plus the richness and scale of routine health-care data, can provide rapid, compelling, and widely applicable results, distinguishing clearly between those treatments that genuinely benefit patients and those that do not.”
However, with RECOVERY, as with many COVID-19 trials, all resources—redeployment of staff, funding, and networks—were used to find solutions in a pandemic scenario. But researchers working in fields of health care that are currently considered less urgent often hit a wall of intractable issues around governance (access to, and protection of, patient data), which is limiting the potential benefits derived from these data.
Cancer registry data takes 18 months to collate, but that is too long in a situation like a pandemic, said Morris. “We’ve used novel, rapid data feeds instead which are of lower quality, but they still provide valuable info.”
However, Morris highlights how governance issues are limiting work that could spot cancers and save lives. The Hub research team used to report the interval cancers not found by the bowel cancer screening programme. “To do this you need to link cancer registry and screening data. But we can’t get this linkage done now so no-one is looking at the interval cancers”, she noted, asking, “is the risk of researchers using data greater than the risk of cancers being missed and non-one knowing about? Apparently so”.
If necessity is the mother of invention, then the needs of treating COVID-19 have propelled clinical trial design towards innovation and greater efficiency. “The way routine data has been used in RECOVERY needs to go beyond pandemics and tackle the enormous burden of common disease on patients, their families, and health systems worldwide.”, said Landray.
Published: April 08, 2021
© 2021 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license. Published by Elsevier Ltd.
Access this article on ScienceDirect