Sources of measurement variation in child anthropometry in the Malawi maternal and child nutrition study

An examination of intra- and inter-observer error in child anthropometry and variation between two skinfold calipers was conducted as part of a large prospective study of maternal and child nutrition in Malawi, Central Africa. Measurements include length, weight, head and arm circumferences, triceps and subscapular skinfolds, and estimated cross-sectional arm fat and arm muscle areas. Total observer error, expressed as a percentage of total variance in each trait, is ⩽3.2% for length, weight, and head circumference; 7% for arm circumference; 10% for arm fat area; and 16-20% for skinfolds and arm muscle area. When total observer error is divided into the two additive components studied, intraobserver error represents approximately 50% or more of total observer error for most traits. An exception is the subscapular skinfold for which intra-observer represents 38% of total observer error. Errors of this magnitude will cause an attenuation of approximately 10% in the correlation coefficients between anthropometric and other variables, and a loss of power in attempting to detect associations between anthropometric variables and either their determinants or outcomes. However, the effects of errors of this magnitude are minor compared to those encountered in studies correlating nutrient intake with other variables. This study also finds that the SlimGuide caliper systematically overestimates skinfolds at low readings and underestimates skinfolds at high readingss relative to the Lange caliper.