To determine whether clinical marasmus occurs in small groups of children from easily recognizable high-risk households, the authors conducted a case-control study to identify risk indicators that may be used in targeted interventions. Cases were children whose mid-upper arm circumference measured less than 110 mm, and controls were children matched for age and sex with arm circumferences greater than 120 mm. Between June 1988 and June 1989, 164 such pairs of children aged 1-4 years were studied in Matlab, Bangladesh. Conditional logistic regression analysis showed an increased risk of marasmus among children from families with other children under 5 years of age (odds ratio = 2.80, 95% confidence interval 1.56-5.02) and children who consumed formula foods (odds ratio = 18.81, 95% confidence interval 4.15-85). Higher maternal education was associated with reduced risk of marasmus. Further examination of these risk indicators suggests that the resources saved through targeting fewer households will be negated by missing many children with marasmus. The authors conclude that the application of targeted interventions against marasmus, using the risk approach, is unlikely to be efficient.