Use of different outcome measures in randomised studies of malignant glioma can significantly alter the interpretation of time to progression: reanalysis of the MRC BR2 study

Abstract

The Medical Research Council (MRC) BR2 study [1] is a randomised trial of two doses of cranial radiation for patients with malignant glioma. We reanalysed data to examine the effect of using change in ranked scales of neurological status (MRC Neurological Status Scale) and performance (World Health Organisation Scale: WHO) to determine progression rather than clinician's impression. Four hundred and seventy four patients were studied. Where clinicians recorded no progression, ranked scales frequently documented progression (MRC 13%; WHO 13%). Where clinicians recorded progression, ranked scales frequently did not alter (MRC 33%; WHO 30%) or occasionally improved (MRC 5%; WHO 3%). When analysing time to progression based on a variety of measures, the estimated difference between treatments was most extreme (hazard ratio 0.81, logrank p = 0.04) when change in WHO status was used, and least extreme when change in MRC neurological status was used (hazard ratio 0.99, p = 0.94). This study highlights how different outcome measures can significantly alter the interpretation of randomised studies.