Differences in psychometric properties, cut-off scores, and outcomes between the Barthel Index and Modified Rankin Scale in pharmacotherapy-based stroke trials: systematic literature review

Objective: Review published clinical trial studies on pharmacological treatment of stroke using both the Barthel Index (BI) and Modified Rankin Scale (MRS) as outcome measures, and to highlight the differences in psychometric properties and cut-off scores through a systematic review. Methods: A systematic literature search on stroke studies involving a pharmacological treatment was conducted between 1955-2008. Key words included Barthel index, Rankin, modified Rankin, pharmacotherapy, validity, reliability, responsiveness, sensitivity, specificity, outcomes, psychometrics, prediction, randomized clinical trials, analysis, and stroke. All search terms were limited to Medical Subjects Headings (MESH) terms, English-language abstracts, and human subjects. Results: Overall, 44 studies were identified, six studies comparing the psychometric properties of the BI and the MRS, 24 studies on use of both the BI and the MRS in clinical stroke trials involving a pharmacological treatment, and 14 studies reviewed the cut-off scores and statistical issues related to scale selection. Most studies measured outcomes at 90 days after initiating therapy although differences were observed in this lag time. There was inconsistency in cut-off points used for both scales in the studies. There was no apparent relation between time to initiation of stroke therapy and outcomes measured by the BI and the MRS. The time window ranged from 3 hours to 72 hours although most of the studies reported outcomes after therapy initiation within 3-6 hours of stroke onset. BI may not be an appropriate scale to measure treatment effects due to the inherent ceiling and floor effects. Use of total distribution scores on the scales rather that dichotomizing or trichotomizing the scales has been favored recently. In mild to moderate stroke patients, the MRS seems to detect small and significant treatment effect changes as compared to the BI. Since most stroke studies try to exhibit the effects of treatment within 3 hours after symptom onset, the MRS might be more relevant to clinicians and patients receiving early intervention. Key limitations of this review are absence of studies that might have been identified through databases other than PubMed and MEDLINE and exclusion of non-pharmacological stroke trials that used the BI and the MRS for outcome measurement. Conclusions: Despite the lack of uniformity in the cut-off points used in the trials, the follow-up time after administration of therapy, and the amount of time within which treatment is initiated after onset of stroke symptoms, the MRS seems to be more sensitive and responsive as compared to the BI in measuring stroke disability. However, more studies are required to differentiate the BI and the MRS that would help in selecting a scale that would appropriately capture outcomes among stroke patients.