Acute Osteomyelitis in Children

n engl j med 370;14 nejm.org april 3, 2014 1364 association between BMI and mortality was similar in persons with a high level of physical activity and those with a low level (P = 0.19 for interaction in women and P = 0.37 for interaction in men). In a previous study, we found that BMI and physical activity were independently associated with the risk of death.1 In the Aerobics Center Longitudinal Study, a high BMI was the most important modifiable risk factor for low cardiorespiratory fitness, and obesity was found to offset the benefits of physical activity on fitness.2 Furthermore, weight loss through lifestyle interventions was highly effective in improving physical fitness.3 In response to Logue et al.: although our Nurses’ Health Study and Health Professionals Follow-up Study cohorts were enrolled in 1976 and 1986, respectively, our analyses included incident diabetes cases through January 1, 2010. Thus, our cohorts reflected contemporary populations of patients with diabetes. Because body weight is substantially influenced by disease severity and methods of treatment, it is important to use BMI before or at the time of a diabetes diagnosis. As shown in our data and those of others, weight loss is common in patients with diabetes, even shortly before the diagnosis. Therefore, use of the postdiagnosis weight increases the potential for reverse-causation biases. Dixon and Kral note that overweight and obesity might be protective against premature death in older populations by providing metabolic reserves, although this hypothesis has yet to be tested. However, among older participants (≥65 years of age) at diabetes diagnosis, we found no survival advantage associated with overweight or obesity. BMI is a less valid measure of excess body fat in elderly populations than in younger populations, owing to differential loss of muscle mass related to sarcopenia and increased frailty. Future studies involving elderly populations should pay particular attention to these methodologic issues, especially weight loss due to chronic diseases, and should also include measures of bodyfat distribution such as waist circumference.