© 1981 American Society for Clinical Nutrition Nutritional anthropometry and mortality risk Dear Sir:

The recent paper by Chen et al. ( I) pro-vides a very useful definition of the relative performance of different nutritional indica-tons in defining mortality risk among pre-school age children in Bangladesh. An im-portant observation in their study was that arm circumference for age (followed closely by weight for age) offered the best discnimi-nation of mortality risk, performing better than arm circumference for height. This ob-servation raised for us the question of whether simple arm circumference without age correction might do as well, and might be more practical by avoiding the need to identify age. In order to explore this point we reviewed available data from the same area in Bang-ladesh collected in 1970 to 1972 which had been used to examine the mortality risk as-sociated with arm circumference for height (QUAC stick) categories (2). Simple arm cm-cumference categories were chosen for each TABLE I age group to identify a prevalence of indica-ton positive children similar to the prevalence of QUAC stick I (moderate-severe malnoun-ished) children. The analysis (Table 1) mdi-cated that simple arm circumference gener-ally provided better discrimination of relative mortality risk as well as higher sensitivity and specificity when compared with QUAC stick categories at similar prevalence, although dif-ferences in sensitivity at similar specificity were not statistically significant (3). Assess-ment of simple arm circumference in relation to mortality also showed the same threshold phenomena for increased mortality risk de-scribed by Chen et al. ( 1) (Fig. I). In a previous study (4) one of us found that simple arm circumference had better sensitiv-ity/specificity performance in identifying weight for age deficit than did arm circum-fenence for height. In that study correcting the arm circumference for age provided no significant improvement in sensitivity and specificity. These observations and the data presented above suggest that simple arm cir Comparison ofarm circumference and QUAC stick for identifying mortality ris