Does Additional MRI Help CT-Diagnosed Stroke Patients?

Patients with acute ischemic stroke (AIS) did not have worse outcomes when a CT-only imaging strategy determined the course of their treatment, according to a small observational study.

Those skipping MRI met the -7.50% threshold for noninferiority, even faring numerically better in rates of death and impaired functional status at hospital discharge compared with propensity-matched peers who had undergone subsequent MRI (modified Rankin Scale score of 3-6: 42.3% vs 48.0%), reported William Powers, MD, of Duke University Medical Center in Durham, North Carolina.

Similarly, noninferiority was supported by the longer-term outcome of stroke or death at 1 year among those discharged alive (19.5% with MRI vs 12.5% without; relative risk 1.14, 95% CI 0.86-1.50), meeting the 0.725 relative risk criterion, the authors noted in JAMA Network Open.

These results suggest that costly routine MRI may not be justified in this setting, despite its wide prevalence.

Powers and colleagues based the study on 123 matched pairs of initially CT-imaged stroke patients with and without additional MRI at the UNC Hospitals Comprehensive Stroke Center.

“Our data strictly apply to the use of MRI in addition to an initial CT in patients hospitalized with AIS and not to other situations for which MRI may be used, such as choice of initial imaging, transient ischemic attacks, uncertain diagnoses, and in persons awakening with stroke symptoms,” they cautioned.

What’s more, only 6.5% of matched patients underwent endovascular treatment.

“Further research is needed to determine which patients hospitalized with AIS benefit from MRI,” Powers and team noted. “More than 90% of patients with AIS receive MRI in addition to CT with few data to determine whether there is an associated benefit with patient outcomes.”

In an invited commentary, Michael Teitcher, MD, and Jose Biller, MD, both of Loyola University Chicago Stritch School of Medicine in Maywood, Illinois, called the implications of this study “substantial.”

“As stewards of health care resources, clinicians should be asking whether the additional information provided by diagnostic tests meaningfully affects patient outcomes,” they wrote.

MRI is thought to have added value for the selection of treatment, and is currently recommended by American stroke guidelines as a reasonable choice after initial imaging of the head in cases where the initial scan did not show infarction or did not provide enough information.

“Of course, there are circumstances in which additional MRI is still justified. But at a minimum, these results should give the health care practitioners reason to pause and reconsider routine use of CT plus MRI,” Teitcher and Biller urged.

On the other hand, Bruce Campbell, PhD, of Royal Melbourne Hospital in Australia, argued that “asking whether the imaging alters mortality may not be the right question.”

“In the case of MRI after stroke, it may be to confirm the diagnosis, clarify stroke localization (e.g., to confirm likelihood of a stenosis being symptomatic), provide hints of stroke causes, determine timing of anticoagulation, assess the severity of underlying small vessel disease, or assess prognosis,” he wrote in a separate editorial.

“Asking any of these valid clinical questions may help improve clinical care by targeting subsequent therapies most appropriately,” he noted.

Campbell added that very few proven stroke interventions significantly reduce death, going only so far as to reduce disability or recurrent stroke. “It should therefore not be surprising that a diagnostic test performed during the acute admission would have no association with mortality,” he pointed out.

This retrospective observational study relied on electronic medical records of adults hospitalized with AIS who had their admission diagnosis based on CT. Admissions were limited to those from January 2015 to December 2017.

From 508 eligible patients, 246 were selected for propensity-matched analysis. Median age was 68 years, and 53% were men. The two study groups were well matched, except for differences in history of coronary artery disease and chronic kidney disease.

At the UNC stroke center, admission diagnosis was made by the center’s neurology resident on call. Residents could order MRI on their own without prior attending consultation. For 111 of the 123 MRIs, there was no specified indication other than stroke or neurological symptoms.

“Data are from a single tertiary referral academic medical center with continuous coverage by in-house neurology residents and subspecialty-trained vascular neurologists. Our findings may not be generalizable to other settings,” Powers and colleagues acknowledged.

Furthermore, the nonrandomized study design left room for unmeasured confounding and bias.

Teitcher and Biller called for future prospective studies on the benefit of MRI in stroke, while Campbell warned that “the practicality of obtaining such evidence is less certain, given the potentially very large sample sizes required to address small minimally important clinical differences and confounding variables.”