A re-evaluation of data from the Third International Stroke Trial has demonstrated that while the study’s primary endpoint remained robust against potential bias, multiple secondary endpoints were sensitive to missing-data assumptions and outcome assessment methods. The study, conducted by researchers Ravi Garg, Gabriel Torrealba-Acosta, and Pitchaiah Mandava and published in BMC Medical Research Methodology, analyzed the publicly available dataset from the trial. The Third International Stroke Trial represents the largest randomized controlled trial evaluating intravenous alteplase for acute ischemic stroke and contributes substantially to international meta-analyses and treatment guidelines.
In the investigated dataset, ninety-six participants, representing 3.2 percent of the trial cohort, had missing six-month Oxford Handicap Scale scores. The investigators observed that these individuals with missing outcome data presented with modestly worse prognostic profiles at baseline and at seven days compared to participants with complete records.
To rigorously evaluate the potential influence of missing data, the researchers performed multiple imputation separately across treatment arms under missing-at-random assumptions. They followed this with a tipping-point sensitivity analysis under missing-not-at-random assumptions. The tipping point was defined as the number of one-level upward shifts in Oxford Handicap Scale scores within the alteplase arm required to yield a null effect for ordinal endpoints or to cause a transition from an event to a non-event for binary endpoints.
The sensitivity analyses revealed notable variations in stability across endpoints. Under multiple imputation, the estimate for the endpoint defined by an Oxford Handicap Scale score below 3 crossed the null. Under an alteplase-arm-specific missing-data assumption, the binary endpoint of an Oxford Handicap Scale score below 2 tipped after the reclassification of only six participants. Conversely, both ordinal endpoints evaluated in the study remained robust under these missing-data assumptions.
The authors also evaluated the validity of the six-month Oxford Handicap Scale score by calculating Spearman’s rank correlation with baseline National Institutes of Health Stroke Scale scores stratified by outcome assessor, including patients, friends, and physicians, anchored to prespecified thresholds. Treatment-effect heterogeneity by assessor was evaluated using adjusted regression models and the ICEMAN instrument. The validity analysis showed lower-than-prespecified correlations for patient-assessed scores across all endpoints, whereas friend and physician assessor correlations were indeterminate due to wider confidence intervals. The analysis also identified moderate credibility of treatment-effect heterogeneity across assessor types, with patient assessors showing the largest treatment effect across all endpoints.
Applying the Cochrane Risk of Bias 2 instrument demonstrated that risk of bias ratings varied considerably depending on the specific endpoint assessed. The primary endpoint proved to be the most robust, earning a low risk of bias rating across both evaluated domains. In contrast, the secondary endpoint of an Oxford Handicap Scale score below 2 exhibited a high risk of bias in both domains, while the score distributions of 0 to 3 versus 4 to 6 and 0 to 6 received high risk of bias ratings specifically for the measurement of the outcome. The authors concluded that these findings warrant cautious, endpoint-specific interpretation of trial estimates and underscore the necessity of sensitivity analyses and standardized outcome assessment in open-label stroke trials.
The findings underscore the broader clinical relevance of scrutinizing large-scale stroke trial datasets, particularly given that the Third International Stroke Trial, cataloged under registration ISRCTN25765518, provides evidence that informs international guidelines and meta-analyses for intravenous alteplase in acute ischemic stroke.
Subject of Research: Bias assessment and sensitivity analysis in the Third International Stroke Trial
Article Title: Assessment of bias in the third international stroke trial (IST-3)
Article References: Garg, R., Torrealba-Acosta, G., & Mandava, P. (2026). Assessment of bias in the third international stroke trial (IST-3). BMC Medical Research Methodology. https://doi.org/10.1186/s12874-026-03017-6
Image Credits: AI Generated
DOI: 10.1186/s12874-026-03017-6
Keywords: acute ischemic stroke, alteplase, IST-3, Oxford Handicap Scale, risk of bias, multiple imputation, sensitivity analysis, clinical methodology, Assessment, bias, third, international
News Source: Ophelia Keating. (November 10, 2026). IST-3 Analysis Finds Primary Outcome Robust While Secondary Endpoints Show Bias. Scienmag.



