INTRODUCTION

The modified Rankin Scale (mRS) is the most frequently used primary outcome measure in acute stroke research despite significant interobserver variability in assigning grades. We assessed the use of central blinded adjudication of the mRS based on a video recording of an interview in the PRECIOUS trial.

PATIENTS AND METHODS

PRECIOUS was an international, randomized, open-label, clinical trial with blinded outcome assessment of preventive treatment with metoclopramide, paracetamol, and ceftriaxone in elderly patients with acute stroke. Trained local investigators interviewed patients or their representatives and graded functional outcome at 90 days after stroke with the mRS. In each participating country, a video recording of the interview was assessed by three blinded, independent adjudicators. The primary outcome of the present substudy was interobserver agreement between the local mRS score and the median score of the three central adjudicators for patients alive at 90 days, assessed with Cohen's kappa and quadratic weighted kappa statistics. The difference between treatment effect estimates based on local and central adjudication was a secondary outcome.

RESULTS

Of 1493 patients enrolled in PRECIOUS, 1471 were included in this analysis. At 90 days, 1117 patients (75.9%) were alive and had both a central and local assessment; 28 participants did not have a central mRS score. Interobserver agreement was seen in 829 (74.2%) patients and was substantial (kappa of 0.68; 95% CI 0.65-0.71). Disagreement occurred more often in patients with a central mRS score of 0-2 (OR 2.24; 95% CI 1.14-4.24). Treatment effects were neutral for all three study drugs and did not differ between central and local adjudication.

DISCUSSION AND CONCLUSION

Central adjudication of the mRS based on a video recording is feasible in a large international, randomized stroke trial. This ensures blinding of the outcome assessment. In this neutral trial, the impact of central adjudication on the precision of effect size estimates could not be assessed.