Malnutrition is common in hospitalized patients, but current measures of nutritional status are limited. Because levels of somatomedins are regulated by nutrition, the utility of somatomedin-C measurement in nutritional assessment was studied. Thirty-seven malnourished patients had measurement of somatomedin-C and conventional nutritional indexes. In 28 patients seen before therapy, the somatomedin-C level was reduced (38 percent of normal) and was lower than the albumin level (66 percent, p less than 0.01), transferrin level (59 percent, p less than 0.05), and lymphocyte count (43 percent, p = NS). Somatomedin-C level was lowest with combined "kwashiorkor-marasmus" (25 percent of normal) and also reduced with "kwashiorkor" (51 percent) or "marasmus" (57 percent) alone. Somatomedin-C was correlated with albumin, transferrin, and lymphocyte count (p less than 0.02 for each). In six patients given nutritional therapy, somatomedin-C levels rose by more than 70 percent in each (mean increase 181 percent), whereas lymphocyte counts rose in four (increase 78 percent for all patients), transferrin levels rose in four (increase 33 percent), and albumin levels rose in one (-6 percent). In 20 patients with detailed dietary analysis, only somatomedin-C was correlated with intake of protein and calories (p less than 0.005 for each). Somatomedin-C may be a sensitive marker of malnutrition and the response to nutritional therapy.