1119 IH=incidence rate. Table: Risk estimates according to type of disorder and study drug were substantially lower than those previously reported for flucloxacillin (about 1 per 13 000)2 and erythromycin (about 1 per 28 000)’ and similar to the risk for oxytetracycline (about 1 per 49 000)-’ in studies based on the GPRD that used identical methods. There were only 5 subjects who developed acute idiopathic blood disorders (2 exposed to co-trimoxazole); all recovered. This low risk is in sharp contrast to the risk of blood disorders (27 per 10 332) associated with sulphasalazine in a study based on the GPRD that used identical methods.4 7 subjects developed erythema multiforme or Stevens Johnson syndrome (4 exposed to co- trimoxazole) ; all recovered. There was 1 case of toxic epidermal necrolysis (in a recipient of cephalexin). 5 subjects developed acute renal disorders (1 exposed to co- trimoxazole) ; none were attributed clinically to a study drug. The table summarises the results. We conclude that the risks for the serious illnesses studied are small for the three study drugs and similar to the risks for many other antibiotics. *Hershel Jick, Laura E Derby Boston Collaborative Drug Surveillance Program, Boston University Medical Center, Lexington, MA 02173, USA 1 Jick H, Derby LE. A large population-based follow-up study of trimethoprim/sulfamethoxazole, tnmethopnm, and cephalexin users for certain uncommon senous drug toxicity. Pharmacotherapy (in press). 2 Derby LE, Jick H, Henry DA, Dean AD. Cholestatic hepatitis associated with flucloxacillin. Med J Aust 1993; 158: 596-600. 3 Derby LE, Jick H, Henry DA, Dean AD. Erythromycin-associated cholestatic hepatitis. Med J Aust 1993; 158: 600-02. 4 Jick H, Myers MW, Dean AD. Sulfasalazine and mesalazine associated blood disorders. Pharmacotherapy 1995; 15: 176-81. 5 Jick H, Derby LE, García Rodriguez LA, Jick SS, Dean AD. Nonsteroidal anti-inflammatory drugs and certain rare serious adverse effects: a cohort study. Pharmacotherapy 1993; 13: 212-17. Use of mid-upper-arm circumference for nutritional screening of refugees SIR-In refugee camps, nutritional anthropometry is used for two quite distinct purposes. First, it is used to provide an epidemiological description of the refugee’s nutritional status. Here, a population sample is taken-not all children are measured. Second, it can identify individual children at increased risk of death for inclusion in special programmes- usually referred to as screening. In this case, all vulnerable children need to be assessed. Bern and Nathanail (March 11, p 631) seem to confuse these two purposes. They do not use screening for selection of beneficiaries, but use it only for epidemiological assessment. Although weight-for-height is generally accepted for descriptive epidemiology, it is not necessarily the most efficient for identification of individuals at risk of death, and hence for screening. Bern and Nathanail conclude that it is not useful to measure mid-upper-arm-circumference (MUAC) as a filter before weight and height measurement of refugee children, because many of the children who have a low weight-for- height would be eliminated from consideration during the initial screen. As they clearly show, different groups of children are selected as malnourished on the basis of MUAC and weight-for-height criteria; only 39% of younger and 20% of older children with a low weight-for-height have a low MUAC. Thus, MUAC does not function as a valid substitute for weight-for-height; by the same token, weight- for-height does not function as a substitute for MUAC. Which should be the gold standard when screening individuals? If the most meaningful outcome measure is mortality then the predictive value of MUAC has been repeatedly shown to be better than weight-for-height.1-3 Thus, if selection for targeted intervention is the goal of the assessment, and only one measure can be made, then one would predict that more, but different, deaths would be averted if MUAC was the sole criterion rather than weight- for-height. We agree that this is partly because weight-for- height tends to underestimate malnutrition in younger and overestimate malnutrition in older children, relative to MUAC.4 However, a selection bias towards the younger children could be an advantage in practice, for they are the most vulnerable. That a substantial number of children selected by MUAC might be classified as only moderately malnourished by weight-for-height might not be important, since moderate and mild malnutrition are associated with the highest absolute number of deaths.5 Clearly, different populations are being selected by the two procedures and both are related to mortality; it seems logical, therefore, to use both criteria to select children for special treatment. However, unlike a hospital or research setting, the practical difficulties of making the measurements, recording the results, and calculating the nutritional status for every child in a refugee camp are crucial. MUAC is not only quick and easy to measure but also it is easily interpreted on the spot. Weight-for-height requires at least two measurers to make two different measurements and then comparison of the values with a standard chart or computer calculation to reach a decision about an individual child. MUAC is not only more closely related to mortality, but also is a much more practical measurement than weight-for-height. For these reasons we advocate that MUAC alone is used to screen children in refugee camps. With severely malnourished children a major difficulty with MUAC, as a selection criterion for admission, is that weight gain (change in weight-for-height) is used to assess progress and weight-for-height is the discharge criterion: it is not appropriate to have different criteria for admission, assessment of progress, and discharge. This difficulty does not arise in a refugee camp setting where very large numbers of children are being selected for supplemental feeding. In some supplemental feeding programmes all children under 5 years are given extra food. This is almost certainly a much more effective strategy for averting death and deterioration to severe malnutrition than to select on the