Commentary: What can we learn from the continuing regulatory focus on the thiazolidinediones?BMJ 2010; 341 doi: https://doi.org/10.1136/bmj.c4812 (Published 06 September 2010) Cite this as: BMJ 2010;341:c4812
- Nick Freemantle, professor of clinical epidemiology and biostatistics
- 1School of Health and Population Sciences, University of Birmingham, Birmingham
Should rosiglitazone be withdrawn from use? An answer is hampered by the inadequacies of the data collected to date. In particular, regulators have tolerated loss to follow-up in trials designed to examine the effects of rosiglitazone on surrogate outcomes (haemoglobin A1c) despite relying on these same trials to provide evidence on safety (mortality and morbidity). The more robust regulatory approach required for new diabetes drugs introduced in late 2008 should prevent this situation recurring in diabetes, but if we are to avoid similar problems in other clinical areas, we need a far more widespread overhaul in the standards of regulatory trials.
Few would disagree that large scale randomised trials provide the most reliable evidence on the safety of interventions.1 Through randomisation, both known and unknown biases are distributed between the experimental groups by the play of chance.2 Because of random allocation of participants, any observed difference between groups on an outcome of interest may be attributed either to the play of chance or to the random treatment allocation. If chance is unlikely (for example, if the confidence intervals are relatively narrow and the corresponding P value small) the only plausible alternative is that the randomised treatment is responsible for the difference.