The thresholds approach is problematic since prevalence has different meanings in different contexts. A simple example is that 10% before harvest at the end of the "lean period" may not be too worrying but 10% a month after harvest is very worrying. Prevalence should always be interpreted in context. Context can be provided by (e.g.) the use of disease, agricultural, and food availability calendars; food-security / household economy data; and market monitoring. There are additional problems when using W/H since this indicator (regardless of whether NCHS or WGS) is known to lack consistency between populations. Studies that have looked at body-shape and W/H indicate that W/H tends to underestimate the prevalence of acute undernutrition in Central American populations and overestimates prevalence in North African pastoralist populations. Overestimation is also probable in Indian and Myanmar populations. These findings, together with the well-described poor prognostic performance of W/H compared to all other anthropometric indices, suggest that MUAC is a better indicator. There has been some work comparing prevalence by W/H using NCHS and WGS references. Much of this is available on the WWW. The principal finding of this work is that WGS prevalence tends to be higher than NCHS global prevalence and considerably higher SAM prevalence. Should thresholds be revised? I think that this will not be useful unless guidance on interpretation WRT context is also given. Is the WGS reference useful? I am not so sure WRT W/H.

Dear Caroline, First, I think that these thresholds should be abandoned. Consider a camp of 1000 children with 10% GAM - here are slightly less than 100MAM and a handful of SAM kids. For a city like Nairobi, Dhaka, Lagos etc -there are 2,000,000 kids - if there is a 5% GAM level then there are 100.000 MAM and sufficient SAM to overwhelm all paediatric services (OTP included). An emergency is called for the camp and the 100,000 are ignored as "acceptable" - it is not equitable or even ethical. I define an emergency in terms of the capacity of the normal services to cope with the caseload. This has many important implications - resources are directed to the greatest number of malnourished, NGOs should strengthen local capacity to take over if they want to leave. The absolute numbers should be reported in survey results etc. %prevalence should also be reported as a measure of the general nutritional status of that population - but not to plan intervention or call an emergency. Second, there is no a priori reason why deviation from normal by a multiple of the distribution of the normal (Zscore) should qualify one for admission for treatment. What we really want is to admit those who need admission and exclude those that do not. For this we need risk of death (if not admitted) data. The Prudon Index is the only risk of death data that we have -and the recent analysis by MSF (Niger data) shows that this seems to hold up in population data. One of the great things about MUAC is that it is biased (appropriately) towards the younger more vulnerable child whereas Z-scores are deliberately age/height neutral. Third, when we used UNISEX w/h admission criteria about equal numbers of boys and girls were admitted - and they had the same mortality experience (so they were equally at risk when admitted with unisex criteria). With the new stanards, a girl of a certain height has to be much lighter than a boy of the same height to get admitted. The standards discriminate against girls. Some programs are finding that they are admitting many more boys than girls with the new standards. The new standards are much better than the old ones, mainly becaue they discriminate less againt the younger child. So we strongly advocate for their introduction. BUT we have also been advocating for the use of the BOYS chart for both sexes - in other words use the boys chart as a unisex chart. This admits the boys appropriately and does not discriminate against the girls. Another good thing about MUAC is that it is indeed UNISEX! But please only use ABSOLUTE MUAC and do not use muac for age or hight which would remove the good bias towards the younger child. Cheers Mike

Mike's message contains a lot of sense. I think that we need to be careful when using example comparisons such small camp vs. large free-living population. In the camp setting we almost always have a clear duty of care toward the population and a prevalence of (e.g.) 10% outside of reception camps or the initial establishment of camps is indicative of poor general ration distribution procedures, poor infection control, &c. These are things that we should be getting right in this context. This is a somewaht niggling point as Mike is correct in stating that these tresholds need to be abandoned and we should evaluate emergencies with regard to the magnitude of need and the ability of existing services to meet need. Looking at things this way is likley to improve targetting but may also lead to more sustainable programming in which supoort is given to existing services rather than a knee-jerk leap to parrallel and vertcial interventions. Mike's comments on MUAC are, I think correct. Low MUAC is strongly predictive of near-term mortality and is treatable. I second Mike's obseravation that MUAC should not be corrected for height or age. I am interested in what Mike says about the Prudhon Index. I am not very familiar with the use of this index. My understanding is that the Prudhon Index is an "audit" tool intended to create a standard against which a program can be compared. In this application, the investigator uses the index to create an expectation of mortality from characteristics of the program population and then compares the observed number of deaths to the expected number of deaths. If observed > expected then further investigation and remedial action is indicated. The index was developed as a way of predicting a number of deaths in a TFC patient cohort with an upper threshold at admisison of, I believe, 70% W/H median. This is about deaths IF ADMITTED not "death if (not) admitted". This is not to say that the index should not be used for prevalence estimation by survey. Such an index /might/ fix some of the weaknesses in the W/H case-definition and, given that calculating WHZ in a survey dataset is computerised the calculations involved are no argument against its use and an alternative to WHZ. I do think, however, that examination and possible recalibration of the index on a wider population would be required before its use in this role. Suitable data should be available from the many previous cohort studies examining the relationship between anthropometry and mortality. One concern that I have with the Prudon Index is that it is basically a sum of a modified BMI (i.e. the log of weight divided by height raised to some power) and oedema with some constants thrown in passed through a logistic linking function (to yield a value between 0 and 1 as a probability of near-term death). This suggests, to me, that it will suffer a body-shape bias and overestimate probabilities for log-legged / short-trunked individuals.

The Prudhon index is primarily an audit tool to assess the quality of care that is given to SAM children in a program. The question I addressed is whether simple WFH Z-scores are an appropriate measure for admission and suggested that if we use a weight and height index then it should be related to risk of death (I agree if not admitted). There is a need to develop such an index (at least for those outside the height range where MUAC is useful). Intuitively, it would be better to have "metabolic weight" for height, or surface area for height (which are both related to such things as drug dosage, toxic effects etc) - at any rate, some function, such as that derived by Prudhon that biases the index towards the more vulnerable younger child and does not strive to be age and height neutral. Yes, body shape will have an effect - perhaps on all anthropometry, including MUAC? Fat patterning between the arms and trunk is different in different ethnic groups - and is also affected by the ambient temperature during growth - these are unexplored areas as far as I can see. Mike

Sorry. I misunderstood your intented meaning. There are validated formulae for calculating BSA from weight alone which have been developed for drug-dosage calculation. I have recently done some work on this for the WHO for ARv drug-dose calculation and have a draft paper prepared if you'd like to see it. As for body shape, MUAC, and W/H ... I did a study for ENCU / UNICEF which was published in last months Annals of Human Biology. This found that W/H is strongly associated with body shape (SSR) and MUAC only weakly associated with body shape (and not at all if case-status such as MUAC < 125 mm is used).

To sum up this discussion, the main point reiterated by all is that prevalence data needs to be interpreted in context to ensure appropriate response, however it remains an important measure of the general nutritional status of a population. It is unclear whether any group is currently working on revision of the thresholds in line with the new WHO standards and discussants have raised the issues of usefulness and appropriate interpretation of thresholds, while there is also increasing research and data on use of MUAC.

Fiona Watson offers a further perspective to the discussion on use of thresholds and the use of appropriate tools for decision-making: While the use of thresholds to determine interventions has been criticised, there remain strong arguments for maintaining some kind of framework to help guide response. In an emergency, information is frequently lacking, biased or simply wrong. Donors, governments and humanitarian agencies are under pressure to respond quickly and sometimes on a large scale to prevent a crisis from deepening. In a world where numerous factors determine response (of which only one may be absolute need), it is essential to try to establish some kind of common scale, which helps to identify need and which can also be used to call donors to account. A number of frameworks have been developed to classify emergencies. The most well known is probably the Integrated Food Security and Humanitarian Phase Classification (IPC). See www.ipcinfo.org for more information. These frameworks attempt to classify the severity of an emergency situation by using the quantifiable (numeric) outcomes of mortality and malnutrition as measures of severity, and linking these with qualitative (descriptive) indicators of food security. They also set thresholds for response. Clearly, where there is reliable information about the size of a population, the prevalence of malnutrition, and where resources for support are unrestricted, it would be ideal to take decisions about interventions on a case by case. This is rarely reality, however. In the current climate of highly politicised aid, some kind of classification system for provoking response and ensuring accountablity is needed. Using malnutrition rates as one among a number of other indicators can help to focus donors and ensure that response is timely and appropriate.

hello, It is my first visit on this site, so I arrive a little bit late in the discussion. We intend to work , with Epicentre, on the revision of the tresholds according to new WHO standards, as our opreations asks us to get guidance, although we strongly agree that these tresholds should be interpreted in the context and are only one indicator among others to describe a situation; they should never be used solely to take decision. I agree with Mike that these % should be also looked together with absolute numbers of expected malnourished children. To refine our analysis of the situation, we also started to look systematically at prevalence (and mortality) among under 2 years to complement the picture and help decision making.