In Reply We appreciate the interest of Doehner and Audebert in our article1 on body mass index (BMI) and short-term stroke survival. However, they fail to relate to our article’s essential and central point: BMI in stroke patients is significantly associated with stroke severity (ie, higher BMI is associated with lesser stroke severity). This means that strokes associated with obesity are less severe than strokes of other causes (eg, atrial fibrillation). Stroke severity is the all-important determinant of short-term stroke survival.2 Thus, stroke severity is essential to examine differences in mortality after stroke between BMI groups. Without that information, survival of obese stroke patients inevitably appears to be better than that of normal-weight stroke patients. However, when studying differences in stroke survival between BMI groups, information on stroke severity cannot stand alone. This information is useful only if the deaths studied are due to the index stroke for which information on stroke severity is available. If death is due to diseases other than the index stroke (eg, cardiac disease, recurrent stroke, cancer, or infection) for which information on disease severity is not available, the information on index stroke severity is of no value because the death does not necessarily relate to the index stroke. As stated in our article, Doehner and Audebert controlled for index stroke severity but did not ensure that the death was also due to the index stroke. A large proportion of their patients died of diseases other than the index stroke (eg, cardiac diseases and cancer) for which they had no information on severity of the disease leading to death. We have also published studies on the obesity paradox based on all-cause mortality3,4 but we have now realized that it is a methodological fault to conclude on survival between groups without knowing the severity of the illness that led to death in the groups examined.
Dehlendorff C, Andersen KK, Olsen TS. The Impact of Body Weight on Mortality After Stroke—Reply. JAMA Neurol. 2015;72(1):127-128. doi:10.1001/jamaneurol.2014.3634