A longitudinal analysis of weight change in Alzheimer's disease.

Abstract

To the Editor: Recently, the Journal published the results of our study in which we explored the association between Alzheimer's Disease (AD) and weight change. We used data collected by the Consortium to Establish a Registry for Alzheimer's Disease (CERAD),1 a nationwide, longitudinal study of subjects with AD and cognitively normal controls developed to standardize the assessment of AD.2 We found evidence to support the association of AD with weight loss, especially when considering a weight loss of 5% or more and episodic weight loss. However, the average percent weight change per year approached but did not reach statistical significance (P = .07). At that time we did not have enough data to take full advantage of this longitudinal data set. However, with the availability of additional CERAD data (2 more years of follow-up), we were able to estimate the average weight change per year (the trajectory of weight change) using hierarchical linear models, a repeated measures model allowing for a varying number of observations per person.3 There were 1094 probable AD subjects; 33 .5% of these had three or more weight measurements. Even those who had only one weight measurement were included in this analysis because they influenced the estimated entry weight. Of the 425 control subjects, 61.2% had three or more weight measurements. The demographic characteristics for the two groups were similar to those outlined in our previous study. Again, the AD cases were older and tended to weigh less than controls upon entry into the study. The estimated weight change per year for both AD cases and controls was small, less than 1 lb. per year. However, AD cases lost weight at a significantly faster rate than controls (P < .02), controlling for age, sex, race, marital status, and education (see Table 1). These findings lend additional support to the association between weight loss and AD. By using longitudinal statistical techniques, each weight measurement contributed to the overall trend of weight change for each subject, allowing a more accurate model of weight change over time. These statistical techniques allowed us to take full advantage of the data that had been collected without any manipulation of data or exclusion of data. In this analysis the general tendency in AD subjects was toward weight loss. Even so, the estimated average weight change per year was very small (-0.7 lbs/year), and the difference between cases and controls was small (-0.6 lbs/year, see Table 1). Several factors may account for this. In our previous study we found a significant amount of variability in weight over time, with many subjects gaining 5% or more of their initial weight, and much of the weight loss that occurred was episodic. Such factors would tend to minimize the average weight change per year.