BMI is a time-intensive measurement to assess nutritional status. Mid-upper arm circumference (MUAC) has been studied as a proxy for BMI in adults, but there is no consensus on its optimal use.

We calculated sensitivity, specificity and area under receiver operating characteristic curve (AUROCC) of MUAC for BMI < 18·5, <17 and <16 kg/m2. We designed a system using two MUAC cut-offs, with a healthy (non-thin) ‘green’ group, a ‘yellow’ group requiring BMI measurement and a ‘red’ group who could proceed directly to treatment for thinness.

We retrospectively analysed monitoring data collected by the International Committee of the Red Cross in places of detention.

11 917 male detainees in eight African countries.

MUAC had excellent discriminatory ability with AUROCC: 0·87, 0·90 and 0·92 for BMI < 18·5, BMI < 17 and BMI < 16 kg/m2, respectively. An upper cut-off of MUAC 25·5 cm to exclude healthy detainees would result in 64 % fewer detainees requiring BMI screening and had sensitivity 77 % (95 % CI 69·4, 84·7) and specificity 79·6 % (95 % CI 72·6, 86·5) for BMI < 18·5 kg/m2. A lower cut-off of MUAC < 21·0 cm had sensitivity 25·4 % (95 % CI 11·7, 39·1) and specificity 99·0 % (95 % CI 97·9, 100·0) for BMI < 16 kg/m2. An additional 50 kg weight requirement improved specificity to 99·6 % (95 % CI 99·0, 100·0) with similar sensitivity.

A MUAC cut-off of 25·5 cm, above which detainees are classified as healthy and below receive further screening, would result in significant time savings. A cut-off of <21·0 cm and weight <50 kg can identify some detainees with BMI < 16 kg/m2 who require immediate treatment.