Dear Meeta, The ENN recently facilitated a consultation on behalf of Save the Children UK, ACF and UNHCR that looked at the use of WFH and MUAC in SAM programming. The report can be found here: [url]http://www.ennonline.net/resources/920[/url]. Best wishes, Tamsin

Dear Meeta, I had raised a similar question a little over a year ago and Mr Tamsin Walters replied "I do not think there is a clear answer to the ratio between MUAC and W/H cases detected, as they measure different things and are not directly comparable." Just as much as I understand his answer I also understand why you ask- In Nutritional surveys we use both MUAC and W/H index for identification of severe and moderate wasting. With MUAC we detect fewer cases but with W/H index we capture more. Emergency projects are often planned on the basis of MUAC but once projects roll out on ground more cases are detected as per Z score for W/H and this has resource implications. I also share the link which he shared with me http://www.en-net.org.uk/question/328.aspx There is a suggestion in this for younger children. In our project in Sundarbans we had a 0.6% SAM as per MUAC less than 11.5 cms and 6.8% below -3SD WHZ score. In a recent survey conducted in our project areas in north Bangladesh following floods we were surprised to find that MUAC detected more cases of GAM (SAM + MAM) than WFH - score with a prevalence of 30% by MUAC and 20% by WFH. This matches the generally held view point that MUAC measures thinness and is more sensitive to child mortality in emergencies. In most Indian contexts where it is a protracted crisis perhaps it is best to include both categories for treatment but prioritise cases as per MUAC . I hope this is of some help. I also follow the discussion with interest and hope experts on the subject would reply. Many thanks Charulatha

Thanks Dr. Charulata. Actually I have same results my SAM is coming high as per WFH as compared to MUAC. Warm regards Meeta

Dear Mark, You mention a database of SMART Surveys- We are based in eastern India and will soon do a SMART survey in the deltaic region of West Bengal. Does your database include surveys from anywhere in India and if yes what does it say about the MUAC and WFH question? Many thanks for your very many useful responses. Warm Regards Charulatha

I have nothing from India but I do have data from Afghanistan, Myanmar (Burma), Pakistan, and Sri Lanka. Here is what I get for these countries for SAM by WHZ (WHZ < -3) using the WGS reference and SAM by MUAC (MUAC < 115 mm) both ignoring oedema: MEAN Country Survey WHZ < -3 (%) MUAC < 115 (%) WHZ -? MUAC DIFFERENCE ----------- ------ ------------ -------------- ---------- ---------- Afghanistan 1 2.15 2.04 0.11 -1.04 2 1.49 1.03 0.46 3 1.49 1.03 0.46 4 0.21 0.97 -0.76 5 1.76 3.08 -1.32 6 1.62 2.92 -1.30 7 6.51 10.53 -4.02 8 2.29 2.65 -0.36 9 2.39 5.42 -3.03 10 2.48 3.24 -0.76 11 1.74 2.07 -0.33 12 0.65 0.86 -0.21 13 0.76 0.65 0.11 14 2.21 2.53 -0.32 15 0.75 1.38 -0.63 16 2.23 0.85 1.38 17 2.70 2.80 -0.10 18 3.45 5.12 -1.67 19 7.05 8.28 -1.23 20 2.43 2.43 0.00 21 6.86 3.68 3.18 22 2.11 1.89 0.22 23 1.15 1.46 -0.31 24 3.45 9.61 -6.16 25 5.00 9.27 -4.27 26 1.37 1.05 0.32 27 1.71 1.32 0.39 28 1.90 4.10 -2.20 29 1.29 1.72 -0.43 30 0.90 2.55 -1.65 31 0.78 5.93 -5.15 32 4.53 5.80 -1.27 33 1.36 2.52 -1.16 34 1.43 2.32 -0.89 35 4.90 8.37 -3.47 ----------- ------ ------------ -------------- ---------- ---------- Burma 1 1.51 3.23 -1.72 -0.55 2 1.49 1.81 -0.32 3 0.64 2.03 -1.39 4 5.29 6.75 -1.46 5 6.40 5.72 0.68 6 4.54 3.43 1.11 7 3.66 4.44 -0.78 ----------- ------ ------------ -------------- ---------- ---------- Pakistan 1 1.98 1.10 0.88 0.10 2 3.46 3.34 0.12 3 1.77 3.53 -1.76 4 1.71 2.06 -0.35 5 2.92 3.52 -0.60 6 4.56 2.44 2.12 7 2.61 1.31 1.30 8 0.87 1.96 -1.09 9 1.08 0.76 0.32 ----------- ------ ------------ -------------- ---------- ---------- Sri Lanka 1 3.12 1.22 1.90 2.21 2 1.79 0.14 1.65 3 4.78 1.71 3.07 ----------- ------ ------------ -------------- ---------- ---------- As with the data from Ethiopia (see my previous message) we see within-country differences. BTW : I forgot to mention the infection angle ... we's expect to see MUAC rapidly depressed by infection but not central mass (so WHZ would lag). Thank you for your kind comment.

Just to follow-up ... SAM is rare and the reliability (precision) of estimates from SMART type surveys will be poor and the figures above will be subject to a lot of sampling variation.

Hi This is indeed a major issue, particularly if you consider that the discrepancy between MUAC and WHZ diagnosis is much higher than the difference in the percentages. In fact, in most surveys, only a minority of SAM children (as defined by one criteria or the other) are diagnosed by both WFH and MUAC. This diagnosis discrepancy is subject to considerable variations, even within countries, as outlined by Mark Myatt. Current hypothesis to explain where the discrepancy comes from, not restricted to the "body shape" explanation, are outlined in the document mentioned by Tamsin Walters above. All these hypothesis need further investigations to better understand how far the nutritional needs of these children actually differ. To have an idea of what has already been observed in India in terms of overlap between the indicators and difference in percentages, as well as an interesting alternative explanatory hypothesis, you might find useful this article here: http://www.indianpediatrics.net/jan2013/jan-154-155.htm Best regards Benjamin

Beware! WHZ is not a gold-standard. It is just a crude anthropometric measure (as is MUAC) used to identify cases of a disease we call SAM. The terms "nutritional" status and "anthropometric status" are often used interchangeably. Nutritional status refers to the internal state of an individual as it relates to the availability and utilisation of nutrients at the cellular level. This state cannot be observed directly so observable indicators are used instead. There are a range of observable indicators (biochemical, clinical, and anthropometric) of nutritional status, none of which taken alone or in combination are capable of providing a full picture of an individual's nutritional status. There is, therefore, no single “gold-standard” indicator of nutritional status. Nutritional status can be usefully defined at the individual, as opposed to the cellular, level as the ratio of nutrient reserves (muscle and fat) to the nutrient requirements of organs (brain, liver, heart, kidneys, lungs, &c.). It is generally recognised that muscle plays a special role as a nutrient reserve during infection and that infection is a major aetiological factor in acute undernutrition. W/H expresses the relationship between weight and height. In children, about 4% of weight is nutrient reserves in muscle. About 96% of weight is, therefore, unrelated to nutrient reserves. Height is almost completely unrelated to the nutrient requirements of organs. MUAC, however, is directly related to muscle mass and is, therefore, a direct measure of nutrient reserves. The limited evidence that is currently available suggests that an index known as the lean-mass ratio (LMR), the ratio of the estimated mass of the limbs to the estimated mass of the trunk, is the best anthropometric indicator of nutritional status. The available evidence suggests that MUAC uncorrected for age or height is a better indicator of nutritional status than all other practical indicators and that weight-for-height is not associated with LMR and is the worst practical indicator of nutritional status. An alternative to examining the association between an anthropometric indicator and nutritional status is to examine the prognostic or predictive value (i.e. of predicting death) of various indicators. When this has been done, W/H has been consistently shown to be least effective predictor of mortality and that, at high specificities, MUAC is superior to both height-for-age and weight-for-age. In terms of indicators that are practical to collect in developmental and emergency settings, MUAC has the best claim to be a practicable “gold-standard” of nutritional status.

Well I think I am aware of that. Precise assessment of nutritional status and needs, as well as assessment of body composition, are currently not available for young children, especially for those in pathophysiological conditions. So there is indeed very limited evidence regarding the link between anthropometry and nutritional status/needs.And everybody should be aware of that. As we should all be aware of the limits of the evidence behind the link between anthropometry and mortality (few cohort studies from the 80's, comparing different indicators than the ones in use today, in a quite different mortality context). In fact, the only precise assessment of nutritional needs of which I am aware has been performed some decades ago by Mike Golden and colleagues, for SAM children which were, at this time, defined by a low WFH (old NCHS reference) and/or nutritional oedema. I think that this assessment is the origin of the formulation of F75, F100 and later of the RUTF. Now, the extent to which the children admitted today for SAM treatment (mainly low MUAC and stunted children, as you probably know that low WFH only children are very often excluded from "MUAC only" programming or practices) 1) have the same nutritional needs for their short-term nutritional rehabilitation, as well as for their healthy growth, than the ones which have been cured by Golden et al. some decades ago 2) should be considered as in greater need for being admitted to the SAM treatment than the low WFH children, who will be excluded from MUAC-only programming and practices ... are important issues, which necessitate further investigation. Particlularly regarding the bad overlap between the diagnosis by MUAC and WFH, which was the initial subject of this thread.

First, Benjamin, I was not replying solely for your benefit. I am also not aware that I had strayed from the topic of this thread. My response was to the statement that WHZ is a "gold-standard" in a short paper that you explicitly linked as "useful" in your post above. I feel free to criticise a sloppy paper (at least it was short!) for making an unwarranted claim just as you feel free to dismiss a series (several is a pretty large number of cohort studies ... more than a "few") good papers by leading scientists from the 1980s just because they were from the 1980's (in facts, some were from the 1990s and gave the same results). If you really feel that this evidence is worthless then you should press for funding to do these cohort studies again. The problem with this is that you can't because such studies would be unethical today. It seem to me to be a bit silly to dismiss the only evidence that we have or are likely to have and to fall back on "hand-waving" arguments. The argument "I have no evidence but I really don't like your evidence (which is from credible sources) so I'll just ignore it" is, IMO, pretty weak. That said, we do have the opportunity of the MUAC-only programs. We can ethically "deny treatment" to low WHZ children but high MUAC children as the best evidence we have is that low WHZ with high MUAC is not dangerous (of course you have to believe evidence not unwarranted assertion). We have some evidence. When we follow-up these low WHZ but high MUAC kids how have been "denied treatment" they do not appear to be in danger. Why not do that rather than spuriously criticise data that contradicts a cherished dogma. I agree this is an important issue so I ask ... "Where is your evidence?". We cannot settle this without evidence. We have evidence. You don't like it so you have to provide better evidence. That is the way these things are suppose to work. BTW, when did Golden and partners do their work? ... 1976 (Picou-Golden Mix - F75 precursor .. F75 is "tweaked" PG-Mix) to 1987 (Golden-Morris - F100 test in Jamaica). We could extend this to 1988 (Golden publishes his type I / type II hypothesis). It's all 1980s and earlier. Why quote this work from a period when you clearly believe that data from the 1980's is bound to mislead us? It is after all a mere handful of clinical trials from a long time ago. I do not take this view. Golden's evidence is (IMO) pretty compelling. The cohort data is also pretty compelling.

First, please do not mis-interprete my words: I do not "dismiss" the available evidence, nor say that anybody is "silly". Talking about "cherrish dogma", I have not been myself for the last 10 years a passionated advocate of the use of one indicator against the other, in what you have done a wonderful job. No, I am just refering to the balanced document mentioned by Tamsin here above. Secondly, the Indian Pediatrics paper which I posted is the only one reporting about the discrepancy between SAM indicators in an indian context, which was the initial question of this thread. These are important data, whatever you think about the authors' worries about the low overlap with WFH (they never call it a gold standard), and whatever you think about their interpretation of the strong association between low MUAC and stunting. By the way, they are not the first authors to hypothesize this association, whose influence in the diagnosis of SAM by low MUAC should clearly be investigated. I am not just pointing out some (among others - again refer to the doc mentioned by Tamsin) important limits and gaps in the evidence you present, I am clearly calling and acting for more research in this field, in this post as well as in the last one. It is good to have your encouragements. Benjamin

Many thanks to the contributors to this discussion and to Mark and Benjamin for attempting to present the best available evidence to respond to Meeta. Although this topic tends to raise the passions, we should all be careful of making strong positive or dismissive assertions against a body of published work that needs to be understood within its limitations and nuanced appropriately. Much of the evidence-base was discussed at length during the consultation mentioned in my previous post, where many of the above issues were raised. These are represented in the paper I included the link to [url]http://www.ennonline.net/resources/920[/url]. The question itself has no clear, final answer at this time, because, as both Benjamin and Mark have pointed out, the relationship between WFH and MUAC varies between contexts for a variety of reasons, many of which are not well understood. As they both mention, there is a need to examine data from other surveys in your specific context to be able to hypothesise about the discrepancy you are most likely to see between the two measures. It is now accepted that there is no "gold standard" for measuring acute malnutrition using anthropometric indicators, so it is problematic to assess the predictive power of MUAC against WFH, as was done in the Indian paper, however, the data may still be useful for you in your context, Meeta, and the association between MUAC and stunting might be interesting for further research.

Sorry, I have to take issue with you when you say "WFH (they never call it a gold standard)" as the article has:

MUAC as a screening tool should not be identifying less children than WHZ (the ‘gold standard’) (emphasis added).

Take a look. If you start from the premise that WHZ is a gold-standard then any indicator will inevitably suffer in comparison. As Tamsin writes:

It is now accepted that there is no "gold standard" for measuring acute malnutrition using anthropometric indicators, so it is problematic to assess the predictive power of MUAC against WFH, as was done in the Indian paper ...

My post regarding this paper was only to point out that there is no "gold standard" and, of all the practicable anthropometric indices, MUAC comes closer to that then does WHZ. The preface to the post was "Beware!" because I wanted to point out a flawed premise of the paper. There was no intention to provoke. Tamsin is correct when she says that the data presented in the third paragraph of the Indian Pediatrics letter do speak directly to the question of differences between prevalence in one Indian population. If you look [url=http://www.brixtonhealth.com/MyattBodyShape.pdf]here (Figure 1)[/url] you will see that, in Ethiopia at least, the difference varies within country. For those interested:

Eveleth PB, Tanner JM. 1990. Worldwide Variation in Human Growth (2nd Ed.). Cambridge: Cambridge University Press.

provides an overview and copious data on variability in anthropometry. MUAC has been criticised for selecting stunted children. This is another reason for favouring MUAC as low H/A children are also at elevated risk of mortality (most studies put this higher than for W/H). The propensity for MUAC to select younger children (also a morality risk) means that this stunting can be reversed. This is one reasons (the other being oedema) for the long-standing observation of reducing W/H in TFP patients during recovery which is sometimes called "catch-up" or "compensatory" growth (note that you will only see this if you monitor both weight and height and not use height at admission for all WHZ calculations as is frequently done).

OK, the term "gold standard" is mentioned into brackets in this article, and furthermore between parenthesis, to reflect the worries of the authors regarding the fact that low MUAC badly capture low WFH, the historical anthropometric way of identifying acute malnutrition. So again, I don't think that this is a good reason to dismiss the article. Thanks for highlighting the good points of the article, finally, in your last post. As I said earlier the extent to which a treatment which has been designed to cover the nutritional needs of low WFH children is covering the needs of low MUAC and stunted children (for short-term weight gain, and clinical rehabilitation, as well as for long term healthy growth) is a key area of investigation which deserves much more attention than what has been done so far. This will be my last post on this thread.

The IP article takes the view that WHZ is a "gold standard" and the ncriticise MUAC for not being WHZ. The same data could be used to make the exact opposite point. The entire approach of reifying either W/H or MUAC is deeply flawed. A more useful approach is to compare indicators with regard as to how well they predict functional outcome. In child survival programs "functional outcome" is mortality. When we do this we find that none of the practicable anthropometric indicators (i.e. MUAC, MUAC/A, MUAC/H, W/H, H/A, W/A) is very good. We have to "pick the best of a bad lot". These turn out to be MUAC and W/A. A consistent finding is that W/H is the worst of a bad lot (See Pelletier's 1994 review). As for history as a gold standard ... James Joyce has "History is a nightmare from which I am trying to awake". Historically (and for very much longer than W/H has been used), slavery was a key component of agricultural production and social organisation. I do not think this a good argument for the retention of slavery as a "gold standard" for agricultural production or social organisation or for the introduction of labour camps. I suppose "historical" varies by what you take to be the start of history. Taking the mid-20th century as a pivot point. Before this we have a collection of single measure indicators such a height and chest circumference (usually by age on in a single birth cohort) and Quetelet's Index (now called BMI) in adults as well as some biochemical methods and subjective clinical assessment. After this we have W/A dominant for many years before being supplanted by MUAC and MUAC/H in the late 1960s (in response to famines in Africa) with W/H coming later. The 1980s and 1990s cohort studies were undertaken as the utility of W/H above W/A and MUAC was (and still is) questionable (See Pelletier's 1994 review). Historically, then, things are subject to change. I do, however, rather like the idea of W/H being condemned to the "dustbin of history" (Trotsky's phrase). The CTC research program purposely included low MUAC children as a target patient group. In some cohorts, low MUAC children formed the majority of cases (in some CMAM programs they are the vast majority of cases). The CTC protocol was, therefore, designed and tested to meet the needs of low MUAC children as well as low W/H children. This is the protocol we now use for most SAM cases. It is, admittedly, an adaptation of the older TFC protocols but the adaptations were made with both low MUAC and low W/H children in the test populations. The Pelletier review is:

Pelletier DL. The relationship between child anthropometry and mortality in developing countries: implications for policy, programs and future research. J Nutr 1994; 124(10 suppl):2047S–81S

Here the approach is test anthropometry against mortality.

I have used the WFH measures and MUAC. I am often asked what is the difference between WFH and BMI. Can anyone help?

They are both problematic and antiquated indexes. WFH is a "standardised" index. A large dataset (the reference dataset) of children is taken and a subset of children for each sex and height (e.g. males of height = 95 cm) is taken and the mean (or median) and standard deviation is calculated. The process is a little more involved as interpolation is applied so we can have usefully precise means and standard deviations for all heights with a merely large rather than a truly massive dataset. Also, we usually pick healthy (and wealthy) individuals so as to create an aspirational reference. For each possible height we have a mean (M) and a standard deviation (S). We can calculate a weight-for-height z-score using: z = X - M ----- S were X is the measured weight, M is the mean weight for the measured height in the reference, and S is the standard deviation associated with M. Our case-definitions with WHZ are statistical in nature. A WHZ of -2 corresponds (e.g.) the (approximately) the 2.3rd percentile (i.e. only 2.3% of the children of the same sex and height in the reference population have a lower weight. Other approaches are to use percentage of median or to use percentiles (this is, underneath, the same as using z-scores) The problem with this reference approach is the selection of an appropriate reference. Since weight for height is influenced by body shape which is influences by diet, genetics, altitude, temperature, and other factors we may end up with a reference that leads to high levels of misdiagnosis in some populations. This means that the functional consequences of a child having a given WHZ varies considerably between populations. BMI takes a different approach. It is calculated as weight divided by height squared. BMI is a good proxy for fat percentage in many populations but has problems with the use of universal thresholds because it is also influenced by body shape. Local thresholds may have to be used for BMI to be useful for diagnosis. Singapore (e.g.) uses BMI >= 27.5 in adults as an "overweight - at risk" category whilst Hong Kong uses BMI >= 23.0. Recent work from the USA suggest that "at risk" may be better set at BMI >= 30.0 rather than the current BMI >= 25.0. So, again, we have the problem of BMI meaning different things in different populations. If BMI is used for diagnosis then local standards may be used or BMI corrected for body shape. If it is used to measure change (as in a surveillance system) then no correction is needed as bias can be taken to be consistent between survey rounds. Other objection to using BMI is that it is of little use in pregnancy, thresholds have different functional significance at different ages and in sick individuals, and difficulty of measurement in the elderly and the disabled. There is also a "misspecified model" problem in that it ignores basic physical laws in assuming that mass increase with the square of linear dimensions rather than with the cube of linear dimensions. This means that larger persons have higher BMIs than smaller persons even if they have the same body shape and body composition. BMI is falling out of favour and being replaced by circumferential measures that are more strongly associated with functional outcome such waist:hip ratio for diagnosis and waist circumference in obesity treatment programs. I hope this of some use.

Thank you So is WFH standards used by the WHO based essentially on children say from the US?

The WHO Growth Standards (WGS) uses data from the The WHO Multicentre Growth Reference Study (MGRS). The MGRS collected primary growth data and related information from about 8500 children from the Brazil, Ghana, India, Norway, Oman, and the USA. The WGS is a "criteria reference" in that for a child's data to be in the reference dataset they needed to meet a strict set of criteria. There were criteria for study sites and criteria for individual children. You can see the criteria in [url=http://www.who.int/childgrowth/mgrs/en/fnb_planning_p15_26.pdf]this article[/url]. When you look at the criteria you can ask if this reference is what you want. Are there consequences to the use of these criteria? Some of the criteria are exclusive. A site and child can meet all (but one) of the criteria for inclusion but will be excluded because they are / live more than 1500 metres above sea level. People who live at higher elevations tend to have broad "barrel" chests and short limbs. These body shapes are associated with high WHZ. In these populations we can expect to see a lot of false negatives for wasting and a lot of false positives for overweight. The exclusion of mobile populations excludes pastoralists who in warm climate countries often have short narrow trunks and long limbs. These body shapes are associated with low WHZ. In these populations we can expect to see a lot of false positives for wasting and a lot of false negatives for overweight. I seriously wonder if it is legitimate to elevate a set of data from the wealthiest and most educated (which in some MGRS sites means the mercantile and political elites) as a world standard. I think, however, that there is a bigger issue with the WGS. An indicator designed to represent ideal growth may not be ideal for the purpose of identifying children, or populations, requiring emergency nutrition interventions. The primary aim of most anthropometric surveys is to identify populations in need of emergency nutrition interventions. The primary aim of interventions treating acute malnutrition is to prevent mortality. In this context, the most useful case-definition will be one that can identify individuals who are at high risk of dying if they remain untreated but would be likely to survive if treated in an appropriate nutrition support program. The WGS approach equates failure to look like the affluent ideal as being indicative of seriously elevated mortality risk. That is absence of the ideal must be a very bad thing rather than a merely ordinary thing. This is a sort of category error. We usually like to use case definitions that define illness rather than case-definitions that operate in the back-to-front manner of "your growth is not ideal therefore you are really sick"). This realisation has led a number of workers (me among them!) to argue that the utility of case-definitions for malnutrition are defined more by their ability to reflect mortality risk than than their ability to reflect ideal growth. Studies examining the prognostic or predictive value (i.e. of predicting death) of various anthropometric indicators have consistently reported weight-for-height to be the least effective predictor of mortality and that, at high specificities, MUAC is superior to both height-for-age and weight-for-age. In terms of indicators that are practical to collect in developmental and emergency settings, MUAC uncorrected for age, height, or sex) has the best claim to being a "standard” of nutrition-associated mortality risk. I hope this is of some use.

On the issue of WHO growth standards and the common (mis)interpretation that automatically 'small=unhealthy', would be interested in people's thoughts on the below paper. I commonly hear that because WHO standards are based on breastfed infants, there will be less problems assessing breastfed infants using the new charts. Not at all what we found in this small RCT: (the full text is open access and free to download from journal website) [url]http://www.ncbi.nlm.nih.gov/pubmed/24134409[/url] Interpretation of World Health Organization growth charts for assessing infant malnutrition: A randomised controlled trial What is already known on this topic? 1) WHO growth standards are based on a breastfed population and are technically superior to NCHS growth references. 2) However, more infants aged <6 months fall below anthropometric thresholds for malnutrition. 3) Implications for exclusive breastfeeding have been debated but lack an evidence base. What this paper adds? 1) Health-care workers take insufficient account of linear growth trend, clinical and feeding status when interpreting low weight-for-age growth chart plots. 2) They are hence more concerned about small infants when assessed using World Health Organization-based rather than the National Center for Health Statistics-based growth charts: this risks interrupting exclusive breastfeeding. 3) To prevent inappropriate management, guidelines and training should emphasise the importance of growth trend, breastfeeding adequacy and clinical status.

Just to expand in the previous comment by Mark regarding WFH and BMI. Both are just methods/strategies that seek to remove the contribution of length/height to the differences in weight observed among individuals/children, to be able to rank according to heaviness. Each method uses a different mathematical approach to remove this contribution. Mark has kindly explained a simplified version of the method behind WFH, but glossed over that for BMI. Adolphe Quetelet back in the 19th century worked out that to remove the contribution of height to weight variance, you needed to divide weight by height elevated to the power of the regression coefficient of their association, when both values were log-transformed. That is, if you log-transform weight and height and perform a simple OLS, the regression coefficient was at most times (for adults) around a factor of two. Given that Quetelet was more interested in ranking large sample of individuals than sensitively and specifically select a small sub-sample, this method was a good approximation. Fast-forward to now, and we know many more details, such that the regression coefficient between log-transformed height and weight varies through development, at times closer to two, at times closer to four; that when separating lean from fat mass, the regression coefficient between log-transformed lean and height is more robustly closer to the value of two, even throughout development, whilst that of log-transformed fat and height shows a great variance. So contrary to Mark assertion, BMI indexes better lean mass variance than fat mass variance, especially among those on borderline categories such as that between overweight and obesity. BMI becomes better at indexing fatness once you move towards very high values (>30). The best way to visualise the relation between BMI and lean and fat mass is to look at what are called “Hattori charts”. Now all anthropometric indicators, as well as the indices created based in their relationships are likely to change with phenotypical changes among individuals driven by environment, gender and age. Mark has already mentioned the impact of two aspects such as body shape and proportions on WFH, but these factors also affect body circumferences. For instance, waist circumference is largely modified by height, whereby taller individuals will present greater waist circumferences, hence now the suggestion of developing a waist circumference for height indicator. This is also true for waist to hip ratio, an indicator that seems to be better at ranking fertility than obesity among women. MUAC is not excluded, for example similar MUAC values taken from equally tall and equally old children are very likely to index different body composition values if children are from Guatemala than if they are from Ethiopia or from India, as there are known phenotypical body composition differences between the populations living in the above mentioned countries. How these differences are likely to be translated into functional outcomes or survival, to my knowledge, is still unknown and arguably disputed.

Thanks to Carlos. I am not sure that I said much about BMI's utility regarding fat mass and lean mass. BMI (as the Quetelet Index) was never very popular and was replaced by the "ponderal index" (AKA "Rohrer's Index") which is weight divided by the cube of height. This was preferred to BMI because the cubing of height yields an index with the proper dimensionality and because it is more robust to changes in stature than BMI. The Quetelet index was not popular until Keys et al (1972) showed that, amongst several ratios of weight and height, it was the best proxy for fat as a proportion of body weight and renamed the index "BMI". Keys et al (1972) ere explicit that in stating that BMI is not useful for individual diagnosis. It did, however, become used for individual diagnosis. This was done at the time of an "obesity scare" (they come and go) in the advanced capitalist economies. The motivation for using BMI was fat. BMI was better at this task than the ponderal index. It seems to me that if your interest is to adjust weight for height then the ponderal index (or similar - there are bio-physical models that suggest an exponent around 2.5 might be better than 3) might be preferred but if your interest is in fat then BMI is to be preferred. MUAC is not free from problems of body shape. In Ethiopia (e.g.) is was found that uncorrected MUAC (as we use in CMAM porgrams) was associated with body shape but that the strength of association was weak and there was no effect on case-status. When we look at the cohort studies from the 1980's and early 1990's we see a consistent pattern. Mortality increases exponentially with declining MUAC, with small increases in mortality at intermediate MUAC values (i.e. between 110 mm and 130 mm) and large increases in mortality at MUAC values below 110 mm. There is little between-study variation in the observed relationships and mortality, despite the fact that the studies were undertaken by different teams in different locations at different times with varying lengths of follow-up and inconsistent censoring of accidental deaths. This congruence of results is not as strong with W/H. This suggests that MUAC is relatively independent (in terms of functional outcome) of factors such as body-shape. Now we have effective treatment for acute malnutrition it is very difficult to collect more data on outcomes since we know that withholding treatment will, for a large number of children, result in death. We have to make do with data that are "sufficient for action but insufficient to satisfy the intellect".