For Pragmatic RCTs, Patient-Reported ‘Hard’ Events Fall Short

Patient-reported events in pragmatic RCTs contribute less to clinical endpoint analysis than what’s yielded from electronic health record (EHR) and claims data, an analysis of the ADAPTABLE trial shows.

“Many of us are interested in understanding data sources in clinical trials and which ones are most effective, because obviously every type of data source that you utilize can be costly,” Jennifer A. Rymer, MD (Duke Clinical Research Institute, Durham, NC), told TCTMD. “So you want to make sure you're getting the most number of events and getting adequate clinical event ascertainment from the sources that you have.”

Participant-reported data in the ADAPTABLE trial was collected through an internet portal platform that consented patients and collected information from them about risk factors, medications, and events. This information was used alongside EHR and claims data by the trialists. ADAPTABLE compared high- and low-dose aspirin in patients with atherosclerotic cardiovascular disease (ASCVD), finding that 81 mg aspirin was just as effective as 325 mg at preventing major cardiovascular events in the full cohort and in a recently-published sex-specific analyses.

In their paper published this week in JAMA Cardiology, Rymer and the ADAPTABLE investigators conclude that for all endpoints (all-cause death, hospitalization for MI, hospitalization for stroke, major bleeding, or a composite), the contribution of participant-reported data was less than 10% of clinical events not otherwise found in EHR or claims data.

To TCTMD, Rymer said the findings suggest that “the most complete ascertainments are going to come from claims and EHR, but particularly from claims. Many clinical trials, at least in the cardiovascular space, will be in a Medicare population, so claims data are very helpful in that sense.”

Pragmatic but Often Problematic

The analysis included 15,006 ADAPTABLE participants, the majority of whom had participant-reported data plus one other source of data (EHR or claims), with fewer than one-third of patients having all three sources available.

For those with EHR and claims data, with or without participant-reported data, 92% to 100% of events could be identified in the EHR or the claims data. For the composite endpoint, EHR data contributed 71% of events and claims contributed 28%.

Rymer said for endpoints like those in ADAPTABLE, patient recall may be part of the problem, but another important issue pertains to the need to obtain authorized releases for hospitalizations that patients report.

“That's challenging and tough, and I think we miss a lot based on having to go back and maybe not being able to get in touch with them [or] they may not be willing to consent for additional things at that point,” she said. One possible takeaway from that, she added, is to try to get all of that information and consent at the initial intake.

As for what the findings mean for ongoing pragmatic trials, Rymer and colleagues say if access to EHR and claims data is available, thought should be given to “the respondent burden relative to the low yield of participant portal (or other self-reported) data.”

But Rymer said the findings don’t diminish the importance of participant-reported information in certain situations and for softer outcomes like health-related quality of life.

“We don't want to diminish the patient's voice, but we all know that when patients are sick or they're being hospitalized [for] myocardial infarction or stroke, they may not remember all of the data or endpoints, or they may not even fully understand what happened if there's not clear communication or if they're very ill during that hospitalization,” she added. “What we’re saying is the best bang for your buck in terms of ascertainment is claims data and/or EHR.”