Real-world evidence can better inform diabetes, cardiometabolic care
Click Here to Manage Email Alerts
CHICAGO — The use of real-world evidence together with randomized controlled trial data can provide further insight into the effectiveness and safety of drugs for diabetes and cardiovascular disease in real-life clinical settings, according to a speaker at the Cardiometabolic Health Congress.
Randomized controlled trials remain the gold standard for the initial assessment of efficacy and safety for any therapy, but real-world evidence can answer some questions that are difficult or impossible to answer with a randomized controlled study design, Lawrence Blonde, MD, FACP, MACE, director of the Ochsner Diabetes Clinical Research Unit at the Frank Riddick Diabetes Institute at Ochsner Medical Center in New Orleans, said during a presentation. Real-world data — information on patient characteristics, care and outcomes derived from electronic health records, claims databases and patient registries — can inform patient care and outcomes research, safety surveillance, research on health care systems and well-controlled effectiveness studies, Blonde said.
“Real-world data does not always make real-world evidence, but real-world evidence usually starts with real-world data,” Blonde said. “Real-world evidence can confirm or add to and/or complement randomized controlled trial evidence.”
Overlooked patients
Many real-world patients with type 2 diabetes would not be eligible for the landmark randomized controlled trials that assessed glucose-lowering therapies, Blonde said. In a literature review published in Diabetic Medicine that assessed the proportion of people with type 2 diabetes living in Scotland who met eligibility criteria for inclusion in five major diabetes trials using the Scottish Care Information database (n = 180,590), only a fraction of those individuals were eligible for inclusion.
“It ranges from as low as 3.5% for PROactive to as high as 35.7% for ADVANCE,” Blonde said. “You can see why there is sometimes a question about the generalizability of [randomized controlled trial] results.”
In an analysis of the real-world use of GLP-1 receptor agonists and DPP-IV inhibitors compared with randomized controlled trial data published in Diabetes Care, researchers found that, in the real-world studies, patients had a 0.51% lower reduction in HbA1c with GLP-1 receptor agonists and a 0.81% lower reduction in HbA1c for DPP-IV inhibitors compared with randomized controlled trial data, Blonde said.
“The patients in the real-world studies did not do as well as in the randomized controlled trials, and this has been seen in many studies,” Blonde said. “Here, using a database, they had ways of assessing adherence, and, in doing that, they found that poor medication adherence accounted for about three-quarters of the gap between the real-world studies and the expected randomized controlled trial results.”
A pivotal analysis that increased the interest in real-world studies was CVD-REAL, Blonde said. Researchers analyzed effects of SGLT2 inhibitors compared with other glucose-lowering drugs on HF hospitalization and all-cause mortality in the first large-scale, real-world comparative effectiveness study of patients with type 2 diabetes with and without CVD. In that study, researchers found that the CV event rate was lower among patients with established CVD treated with an SGLT2 inhibitor compared with other glucose-lowering drugs.
“This suggested a possible CV benefit for a broader population of type 2 patients, and not just those seen with CVD,” Blonde said.
The CVD-REAL study, Blonde noted, also has limitations of residual, unmeasured confounding; however, the results were similar across countries and “remarkably stable” in multiple sensitivity analyses.
Real-world limitations
There are limitations to real-world evidence, Blonde said, including risk for misclassification or misdiagnosis, measurement errors, nonstandardized data elements, selection biases, detection biases and residual confounding. The complex U.S. health care system can be a reason for missing data in real-world studies, he said. Pragmatic, randomized real-world studies, he said, which use initial randomization to control for selection bias followed by a study that adheres as close to real-world practice as possible, have helped to reduce bias in such studies.
When randomization is not possible, propensity-score matching can also minimize confounding factors by distributing measured confounders equally between two groups to improve precision of estimates of treatment effects, Blonde said. Additionally, predictive, modeling machine learning, which uses deep-learning algorithms to recognize patterns, can also improve data integration, Blonde said.
Blonde said a good-quality, real-world study should have a clearly stated research question, a well-conceived study design, appropriate study populations, interventions and endpoints, as well as generalizability to a broad patient population, while also clearing outlining any study limitations.
The FDA’s Real-World Evidence Program, Blonde said, notes that such studies can generate hypotheses for future randomized controlled trials, identify drug development tools such as biomarkers and assess trial feasibility. Additionally, the American Diabetes Association held the first symposium focused on the use of real-world data to improve the prevention and care of diabetes-related outcomes in 2018.
Blonde said the evolution of new approaches, such as machine learning, along with the availability of more comprehensive datasets, will improve real-world evidence.
“Together with randomized controlled trials, real-world evidence provides further insight into the effectiveness and safety of drugs in real-life clinical settings,” Blonde said. – by Regina Schaffer
References:
Blonde L. Real-world evidence in cardiometabolic health. Presented at: Cardiometabolic Health Congress; Oct. 10-13, 2019; Chicago.
Carls GS, et al. Diabetes Care. 2017; doi:10.2337/dc162725.
Kosiborod M, et al. Circulation. 2017; doi:10.1161/CIRCULATIONAHA.117.029190.
Saunders C, et al. Diabet Med. 2013; doi:10.1111/dme.12047.
Disclosure: Blonde reports he has received grant or research support from Janssen, Lexicon, Merck, Novo Nordisk and Sanofi, and honoraria from AstraZeneca, Gilead, Janssen, Merck, Novo Nordisk and Sanofi.