That is a very interesting observation.
Low BMI is associated with with increased morbidity and mortality. In women of childbearing age, low BMI is associated with increased risk of having low birth weight babies. BMI is highly correlated with both fat and fat-free mass although the strength of association varies with age, sex, and ethnicity.
Most research has concentrated on the use of BMI for estimating the prevalence of chronic undernutrition in stable populations. This is a very different role to screening for acute undernutrition to decide admission to supplementary and therapeutic feeding programs. The assumption in NGO manuals and many academic articles that BMI is an appropriate indicator for screening for acute undernutrition has not been tested. The BMI threshold for severe (i.e. “Grade III”) chronic undernutrition (i.e. BMI < 16) may not reflect the severity of acute undernutrition requiring specialised treatment. Extremely low BMI values were (e.g.) observed in Somalia during the 1992 emergency. This prompted a downward revision of the BMI threshold to BMI < 13 to denote severe wasting. This revision did not account for the Somali long-legs / short-trunk phenotype which is an important explanatory factor behind the very low BMI values observed (see below). This threshold is, therefore. probably inappropriately low and lacking in case-finding sensitivity.
Both acute and chronic undernutrition present as low BMI but the process leading to a low BMI may be acute or chronic and the examination of a single BMI value does not allow these two very different conditions to be differentiated from each other. Apart from this major problem, there are several other problems associated with the use of BMI as an anthropometric index. These problems are discussed below.
We have to be very careful using BMI because factors other than nutritional status determine the functional significance of BMI values. The most important of these is body shape. Body shape is frequently determined by the sitting height to standing height ratio (SSR):
SSR = Sitting Height / Standing Height
This index varies considerably both between and within populations. International comparisons have found average SSR to vary between 0.48 (in Australian aborigines) and 0.55 (in Japan). This range translates into differences in BMI due to body shape alone of over 6 BMI units. In one Australian aborigine population, for example, the SSR was found to vary between 0.41 and 0.53. This is larger than the worldwide variation in average SSR and translates into differences in BMI in excess of 10 BMI units due to body shape alone. When BMI is used to assess an individual (as in screening for admission to SFP or TFP), the calculated BMI should be adjusted using a correction factor based on their SSR. This requires an additional measurement (i.e. of sitting height) and several additional calculations (i.e. to calculate SSR and to apply a correction to the calculated BMI). Without this adjustment, sensitivity and specificity of diagnoses based on BMI thresholds may be low.
Other factors influencing BMI are diurnal variability in height, diurnal variability in weight, increased error in measuring height in acutely malnourished individuals, loss of height during starvation, problems with oedema and ascites, problems with measuring height in older persons, problems with weight in pregnant and lactating women.
MUAC with clinical signs appears to be the better option. The use of MUAC for this purpose was recommended by United Nations Forum for Nutrition in July 2000 (see this guideline).