We agree with Drs. Shemie and Fontela1 that not all potential confounding variables could be considered in our study.2 We did adjust for the strongest baseline clinical predictors of death in this population,3,4 but as we indicated in our interpretation, variables that could not be evaluated may have also contributed to the observed variation (e.g., regional differences in referral patterns, case-mix, religious beliefs or personal preferences). We agree that the evolution of brain injury over time often affects recommendations by physicians to withdraw life-sustaining therapies. However, we also observed large inter-hospital variability in rates of death due to withdrawal of life-sustaining therapies during the first three days of care — a time frame that is arguably too early to form accurate predictions about neuroprognosis. The observed inter-hospital variation cannot likely be explained by systematic differences in the temporal evolution of the brain injury across centres as suggested, and we remain concerned that differences in practice patterns are also likely to be responsible.
Although our study conclusions may generate discomfort, we believe that variation in physicians’ perceptions of neurologic prognosis for patients with severe traumatic brain injury likely contribute to the observed variability in rates of death following the withdrawal of life-sustaining therapies. We believe our study helps highlight the need for high-quality research to better inform neuroprognostication, so that we can help families decide when to continue — and when to stop — life-sustaining treatments for these patients.