Predictive modeling key to improved clinical trials, patient care

April 28, 2016 – The DCRI’s Michael Pencina, PhD, and Eric Peterson, MD, MPH, discuss the promise of improved predictive models in a recent editorial.

The growing importance of precision medicine will require researchers to think about clinical trials in a new way, argue DCRI Executive Director Eric Peterson, MD, MPH, and Direction of Biostatistics Michael Pencina, PhD, in a new editorial.

The editorial, which accompanies a new study by Robert Yeh, MD, MSc, of Harvard University, appears in the latest issue of the Journal of the American Medical Association.

michael-pencina-newsMany clinical trials are conducted in the service of evidence-based care, Peterson and Pencina write, yet they often use a “one-size-fits-all” approach. Despite the prevalence of guidelines and prediction models, clinicians are frequently unsure how to apply them to all of their patients.

Yeh’s study is noteworthy because it constructs more complex predictive models to better interpret the results of a clinical trial. The goal of Yeh and his colleagues was to identify which patients who had received drug-eluting stents and were assigned to receive extended thienopyridine treatment had the most or least favorable absolute benefit-risk ratio. More studies, Peterson and Pencina argue, should take this approach.

“Precision medicine calls for the customization of health care, with medical decisions tailored to the individual patient,” they write. “Sometimes precision medicine can identify a single variable such as a gene or biomarker that can successfully differentiate individuals who benefit or are harmed by a given treatment.

“However, in many situations, the outcomes of intervention are associated with multipleeric-peterson-news variables. In these instances, statistical risk prediction models can estimate the likely implications of a therapeutic intervention and thereby assist medical decision making. Specifically, these models can simultaneously aggregate multiple patient characteristics into a simplified ‘risk prediction score’ that can provide individualized probabilities of outcome with or without treatment.”

Better prediction models are key to the future of precision medicine and clinical trials, they conclude, but such models must be put into practice to be truly useful. Both clinicians and patients should be able to understand all of the possibilities offered by these new, more nuanced models.