Here is the fully adjusted model, stratified by survival status and containing all suspected confounding variables. Baseline age was inversely associated with body weight as each year increase in age was associated with a body weight lower by about 0.75 lbs in survivors and over 1 pound in non-survivors. Hence, older people started off the follow-up period weighing less than their younger counterparts, and the magnitude of this inverse association was greater in non-survivors compared with survivors. The estimates for follow-up time and its second-order term represent the markers for aging in this model. The negative regression coefficients for the linear and quadratic terms among survivors and non-survivors indicate a curvilinear slope for weight loss during follow-up. Among the survivors, however, the magnitude of weight loss was significantly dependent upon the presence of a chronic condition, whereas the magnitude of weight loss among non-survivors was not. The statistically significant quadratic term for "time" in the modeling suggests that the amount of weight loss accelerated in both survivors and non-survivors over the follow-up. As indicated by the interaction term between total activity score and follow-up time (TAS*Time), each 1-unit increase in total activity score significantly attenuated aging-related weight loss by nearly 1 pound among the entire cohort. Again, however, the independent relation between physical activity and attenuated weight loss is evident primarily among those who did not survive. The 3-way interaction term among physical activity, follow-up time, and chronic conditions was statistically significant among the survivors, but not among non-survivors. Thus, among survivors, physical activity attenuates weight loss ONLY in the presence of chronic conditions, whereas among non-survivors, physical activity significantly attenuates weight loss whether or not chronic disease is present.