Current issues of ACP Journal Club are published in Annals of Internal Medicine


Stroke units and mortality: a meta-analysis

ACP J Club. 1994 Mar-April;120:32. doi:10.7326/ACPJC-1994-120-2-032

Source Citation

Langhorne P, Williams BO, Gilchrist W, Howie K. Do stroke units save lives? Lancet. 1993 Aug 14;342:395-8.



To compare mortality in patients who have had an acute stroke and were treated either in specialized stroke units or in general medical or neurology wards.

Data Sources

Studies were identified using Index Medicus and MEDLINE databases up to January 1993, lists of conference abstracts, author contact, and bibliographies of relevant studies, textbooks, and review articles.

Study Selection

Studies were selected if they were randomized controlled trials comparing patient management in specialized stroke units with usual care in medical or neurology wards. Acute stroke was defined as a focal neurologic deficit due to cerebrovascular disease, excluding subarachnoid hemorrhage and subdural hematoma. A stroke unit was defined as a multidisciplinary team of specialists in stroke care. Studies that compared physiotherapy techniques or late rehabilitation were excluded.

Data Extraction

Data were extracted on mortality within 17 weeks of stroke, mortality at final review, details of the study population, intervention, time to entry to the stroke unit, and time in the unit. Functional outcome could not be analyzed because of the various outcome measures assessed. Cause of death was not usually available.

Main Results

10 published trials with 766 patients treated in a stroke unit and 820 patients given usual care were included. 6 studies showed that a higher proportion of patients treated in specialized stroke units received remedial therapy (physiotherapy, occupational therapy, and speech therapy) than patients receiving care in medical and neurology wards; 3 studies showed that patients in stroke units received remedial therapy sooner than patients receiving usual care. Within 4 months of stroke, 148 patients (19%) in the stroke unit and 212 patients (26%) receiving usual care died; at 12 months, the corresponding rates of death were 29% (203 of 702 patients) and 35% (263 of 758 patients). Using intention-to-treat analysis, the patients treated in a stroke unit, compared with patients treated in medical and neurology wards, had a reduced risk for death during the first 4 months after stroke (odds ratio [OR], 0.72; 95% CI, 0.56 to 0.92) and during the first 12 months (OR, 0.79; CI, 0.63 to 0.99). Subgroup analysis showed similar reductions in mortality for trials studying various types of stroke units (stroke wards, stroke teams, intensive rehabilitation, and comprehensive rehabilitation). Tests for heterogeneity were nonsignificant.


Patients with acute stroke treated by specialized stroke units had a lower mortality at 4 and 12 months than patients given care in medical and neurology wards.

Source of funding: Not stated

For article reprint: Dr. P. Langhorne, Department of Clinical Neurosciences, Western General Hospital, Edinburgh EH4 2XU, United Kingdom. FAX 44-31-343-6148.


In their 1992 review, Sandercock and Willems (1) concluded that no effective treatment existed for acute ischemic stroke. The meta-analysis by Langhorne and colleagues contradicts this conclusion and raises at least 3 questions, as suggested in the accompanying editorial by Donnan (2): Which patients were selected, what was the intervention, and was mortality the most important outcome in these trials? As in any trial, select groups were enrolled; however, the selection criteria varied too much across the trials to be easily summarized. What aspect of care applied by a multidisciplinary team of specialists led to reduced mortality? Was the effective intervention as simple as giving more attention to prophylaxis for deep venous thrombosis or as complex as giving more medical surveillance, which in some of the trials extended for months? Treatments should "add life to your years, not years to your life" (3). Functional outcomes varied so much across the trials that they could not be summarized in the meta-analysis. Nonetheless, the trend was toward improved function, not just improved survival.

This meta-analysis fails to resolve the following questions: Which patients might benefit from treatment in a stroke unit; what would constitute the essential components of these units; and should these units proliferate? Advice for practitioners based on this study is difficult. Advice for clinical investigators is easy. Answers require additional studies of what the meta-analysis suggests is currently the most effective treatment for ischemic stroke. Future studies need to better define the interventions, including costs, and the benefits, especially with respect to functional outcome.

Will Longstreth, MD, MPH
University of Washington Seattle, Washington, USA