MUAC as a criterion for discharge from feeding programmes    
MUAC as a criterion for discharge from feeding programmes Claire Schofield 22.08.98
MUAC as a criterion for discharge from feeding programmes Rita Bhatia 27.08.98
Re: MUAC as a criterion for discharge from feeding programmes Alain Mourey 21.08.98
Re: MUAC as a criterion for discharge from feeding programmes Michael Golden 23.08.98
no title Ron Waldman 23.08.98
Re: MUAC as a criterion for discharge from feeding programmes Michael Golden 23.08.98
MUAC, Rambo and survival Andre Briend 28.08.98
Re: MUAC, Rambo and survival and bed-scales Steve Collins 31.08.98
Re: MUAC as a criterion for discharge from feeding programmes Bart Burkhalter 28.08.98
Z scores vs % median Andre Briend 28.08.98
Re: Z scores vs % median Judith McGuire 28.08.98
Re: Z scores vs % median Pierre Traissac 28.08.98
Re: MUAC as a criterion for discharge from feeding programmes Rita Bhatia 28.08.98
Historical note on Z Steve Hansch 29.08.98
Re: MUAC, Z score and % median Tim Cole 30.08.98
Re: MUAC, Z score and % median Andre Briend 31.08.98

Date: Sat, 22 Aug 1998 16:31:06 +0100

From: Claire Schofield <>

Subject: Ngonut: MUAC as a criterion for discharge from feeding programmes


Dear ngonuts,

Much has been written about the uses of MUAC as a screening tool for entry to feeding programmes and also as a predictor of mortality. I write to ask if anyone knows of publications or has knowledge or experience of the use of MUAC in assessing readiness for discharge from therapeutic feeding programmes.

Many thanks

Claire Schofield

E.C.Schofield, Lecturer

Public Health Nutrition Unit,

London School of Hygiene and Tropical Medicine 49-51 Bedford Square, London WC1B 3DP.

Tel 0171 299 4600 Fax 0171 299 4666

Date: Thu, 27 Aug 1998 10:00:07 +0200

From: Rita Bhatia <>

Subject: Ngonut: MUAC as a criterion for discharge from feeding programmes -Reply



You have raised a very important issue. MUAC in Refugee situation is not being used as a entry criteria in Feeding Programmes , may be some NGOs are still using it in very crtical situations. MUAC for screening with a cutoff point and those eligible are then measured for entry in programmes using wt/ht Indicators.

I havenot come across any programme where this is the case.

Some one can enlighten us all.


From: "Alain Mourey" <>

Date: Fri, 28 Aug 1998 09:10:01 +0100

Subject: Re: Ngonut: MUAC as a criterion for discharge from feeding programmes


Dear Claire,

The ICRC does not use MUAC but MUAC / Ht both as a criteria for admission

in and discharge from therapeutic feeding centres. The usual criteria for

admission is AC/Ht < 75 % whilst for discharge it is > 85%. However for

discharge, we also use Wt/Ht > 90 %. I did not have time yet to dig into

all the data we have concerning the correlation between Wt/Ht amd AC/Ht.

However I know that AC/Ht is a far better criteria for admission than

WT/Ht, though the follow up is done with weight since we have catch up

criteria expressed in g / kg / day and not in MUAC mm / day. As for the

discharge criteria, provided that the child is cleared of diseases and

oedema, Wt/Ht is probably as good as AC/Ht and perhaps even better. Jenny

McMahon is working on data we have from Angola and is also going to look at

those we are gathering now in Sudan. Provided we have time to analyse them,

we will have a better picture on how Wt and AC correlate from the time of

admission through catch up and until discharge. I would suspect that the

correlation is going to be different from one situation to another, owing

to the different ecological conditions in which PEM develops, i.e.

especialy owing to the presence or absence of infectious and parasitic

diseases. As an exemple, in Tigray in 1985, on the highlands and at the

beginning of the rain season, most children with AC/Ht < 75 % would still

have Wt/Ht quite above 80 % because of infectious and parasitic diseases

(swollen bellies and effect of infections on water retention), masking the

severity of their malnutrition . Using Wt/Ht < 80% as a criteria for

admission meant that many of the kids were so sick with parasitic and

infectious diseases and so malnourished that many died in the centre. Using

then AC/Ht < 75 %, we admitted children earlier than with Wt/Ht < 80 %,

giving them a better chance to survive. On the contrary, in the lowlands of

Erythrea, at the same period, where it was dry and warm, with a nomadic

population mostly affected by hunger but not overburdened by infectious and

parasitic diseases, many of the children with AC/Ht < 75 % also had their

Wt/Ht < 80 % and most of these were still in good health and simply needed

food. There were also some children with Wt/Ht < 60 % and most of them

survived after being admitted in the TFC, whilst in Tigray, none of them

survived. This eventually to say that there is certainly not one truth

concerning discharge and admission criteria and that common sense and

clinical examination is probably as important as a given cut off point for

a given indicator.


Best regards

Alain Mourey

Date: Sun, 23 Aug 1998 16:05:05 +0100

From: Michael Golden <>

Subject: Re: Ngonut: MUAC as a criterion for discharge from feeding programmes -Reply


Dear Rita,

Action Contre la Faim does use MUAC as a criterion both on admission to TFC and on discharge.

Also for admission to SFC and discharge.


TFC admission

Presence of nutritional oedema


Weight-for-height < 70% NCHS standard


MUAC < 110 mm (for children => 75 cm)


TFC discharge to special follow-up clinic for 6 months with supplemental feeding

For initially oedematous patients, 15 after start of loss of oedema plus absence of nutritional oedema for at least 7 days


7 days after reaching weight-for-height => 85% NCHS standard


MUAC => 120 mm


SFC admission

Weight-for-height >=70% and Weight-for-height <80%


MUAC >= 110 and MUAC <120 (for children over 6 months even if <75 cm)


SFC discharge (dry ration)

Weight-for-height >= 85% for 15 days


MUAC => 120 mm for 15 days


(note there are different criteria for <6 month old, adolescents and adults in both the TFC and SFC, which we are evaluating at the moment - and we are trying to develop criteria for malnourished pregnant women, based upon MUAC, but progress on this is slow because we do not have a sufficient data base as yet)

From:, Ron Waldman


Mike, Rita, et al.

Wasn't there any literature, even gray, regarding performance of MUAC

vs. WFH and WFA at some point in the mid-late '80s? I'll be happy to

try to dig some up. The ACF criteria look interesting (how has the

relationship between WFH 70%-79% and MUAC 110mm-119mm established? do

they correlate well?). Is there any data to suggest whether one

parameter improves more rapidly than another? Can one expect longer or

shorter stays in TF/SF centers depending on which parameter one uses for

admission/discharge? The ACF criteria require either to be met for

admission, but both for discharge, obviously the most conservative

approach. Does this warrant further discussion/consideration? I look

forward to hearing more.


Ron Waldman

Date: Sun, 23 Aug 1998 18:32:41 +0100

From: Michael Golden <>

Subject: Re: Ngonut: MUAC as a criterion for discharge from feeding programmes -Reply


Dear Ron,

Yes of course there needs to be lots more discussion - or should I say data collection and analysis rather than discussion!

We have deliberately adopted a conservative approach - although the extra patients that are admitted because of addition of the MUAC criterion is not excessive - it varies from site to site quite considerably (depending strongly on the mean age and the prevalence of stunting in the patients presenting for admission).

The discharge criteria for oedematous patients are based upon biochemical measures of the length of time it takes for glutathione, NADPH/NADP ratios and intracellular potassium to become normal (about 14 days from start of loss of oedema) - and interestingly knemometry shows that it takes about this length of time also for the children to start to gain length as well - so everything points to the oedematous child being biochemically vulnerable for at least 14 days even if his/her oedema is lost in a couple of days.

For the anthropometric criteria, most children continue to gain in the 7 days after they have reached 85% so that the actual discharge weight is higher than this - we are following the children up for 6 months after discharge and the relapse rate is very low with these discharge criteria - clearly, it does not benefit the patients much if we rapidly get them up to 85% and send them out only to have to readmit them or have them die - so the real test of whether we have got the correct discharge criteria or not lies in what happens to them subsequently! The data are a bit scrappy, because of losses to follow-up, this is due to the difficulty in getting the resources and staff to go out and find what has happened to the losses-to-follow-up. We intend to try to get donor support for just such a project.

The rate of increase in MUAC during recovery lags behind the increase in weight: why this should be we have not investigated. However, it is not unusual to have a 6-18 month old patient who fulfills WFH discharge criterion before the MUAC criterion - These tend to be the stunted and younger children whom I think are the most vulnerable after discharge if they have not repleated their muscle mass (do not have the follow-up data to support this supposition) - however, I think that we should air on the side of caution for this particular group of patients.

Yes, there is quite a literature on the prognostic ability of MUAC - within this age group, it is consistently better than WFH in determining the risk of death - Andre Briend summarized these findings on an ngonut message from about a year ago- perhaps he will up date this for us. This seems to be because of the effect of biasing the intake towards making it easier for younger children to meet the criteria for admission - this performs well because there is a strong age effect on mortality risk and so such a selection bias is to be welcomed. It is really only in this younger age group that we admit patients on a MUAC criterion where they do not fulfill the WFH criterion whereas with the older patients they are more likely to fulfill the WFH than the MUAC criterion (for those that do not fulfill both) - we have not yet run the logistic regressions selecting only those patients that fulfill one of the criteria but not the other and to look at these groups separately. So sorry cannot yet give the information that is really required - but "soon come".

I am unaware of any dataset where the patients who have a MUAC below and a WFH above cut-offs, or visa versa, have been looked at as a separate group in community studies - although appropriate datasets exist we have not had access to them yet. There are buckets of data of WFH and MUAC in the same children in the community - but we do not have subsequent mortality data.

We are going to try to get these data from SFC situation - but this will take some time.

Incidentally, one of the ACF students (Carlos Navarro) in now looking at MUAC and BMI as prognostic indicators for adults admitted to TFCs - the first very preliminary analyses seem to show that the MUAC ROC curves are better than other anthropometric parameters (BMI) for adults as well as children - we have also confirmed Steve Collins finding that the most important adult/adolescent prognostic criterion is the inability to stand to have weight taken (we are looking at these patients MUAC with especial interest to see if its specificity and sensitivity for this subgroup is reasonable): also for female adolescents MUAC seems to be good but not males adolescents - I do not yet understand this difference - but we do not have details of pubertal staging for these patients which may explain the differences.

Hope these observations are of some interest - it is very much on-going analysis and the required data are not yet "in the bag" to fully justify all the decisions that we have taken.

Prof. Michael H.N.Golden | Tel +44 (1224) 681 818 ext 52793/53014 Dept of Medicine and Therapeutics | Tel(direct) +44 (1224) 663 123 527 93 Univ of Aberdeen, Foresterhill | Fax +44 (1224) 699 884 AB9 2ZD. Scotland, (UK) |

INTERNET - PERSONAL............. =




Date: Fri, 28 Aug 1998 15:38:12 +0200

From: (Andre' BRIEND)

Subject: Ngonut: MUAC, Rambo and survival


Dear Ngonuts,

Sorry for coming back on MUAC.

Mike says :

" Yes, there is quite a literature on the prognostic ability of MUAC - within this age group, it is consistently better than WFH in determining the risk of death..."


"This seems to be because of the effect of biasing the intake towards making it easier for younger children to meet the criteria for admission - this performs well because there is a strong age effect on mortality risk and so such a selection bias is to be welcomed. "

This is the 'official explanation' that you can even find in the WHO manual on anthropometry. May be true. I have argued for years for an alternative explanation that I am glad to repeat for Ngo'nuts who missed my mail last year (??).

I suggest that survival is not related to the ratio of different anthropometric indices with the NCHS norm (which is what all classical nutritional indices measure) but to the body composition. In a nutshell, I argue that survival is related to the balance between i) organs consuming energy and with a fast protein turnover (in practice brain for energy + internal organs) ii) and muscle mass. Muscle is known to act as a source of fuel during fasting and / or infection. My argument is that MUAC might better reflect this balance than classical indices because it is a measure directly related to muscle, the most variable component of this ratio.

This has several implications.

a) This would suggest that all papers comparing MUAC with WFH presented as a 'golden standard' of measure of nutritional status based on anthropometry are misleading. Hundreds of papers have shown that MUAC correlates poorly with WFH and concluded that for this reason MUAC has no value. Yes, MUAC correlate poorly with WFH, but is more closely related to survival, which is what is needed for targeting programmes.

b) This hypothesis put the relation between age and survival in a new perspective: one may argue that young children have an increased risk of dying as a result of their body composition (big brain, small muscles). The adult body composition, especially 'Rambo type' (big muscle, and ... small brain) is the best adapted to survival in terms of metabolism. More seriously, this suggests that the argument that MUAC should not be used to assess malnutrition because it is related to age is not valid : young age and malnutrition may well have an effect on survival by the same mechanism, viz an insufficient muscle mass in relation to other organs, and there is no strong reason to attempt to guess which is the predominant effect causing low muac. Note that management of malnutrition has a lot in common with feeding younger children (need of frequent feeds to prevent hypothermia, hypoglycemia especially).

c) This interpretation suggests that it is useless to correct MUAC for age for targeting programmes (if one puts apart the very young who has small muac... but small brain to sustain. I am quite happy with the change introduced by ACF for below 6 months). Indices using MUAC in combination with height (sorry for ICRC, long standing disagreement...) or age are usually less effective than MUAC alone to assess the risk of dying; The best evidence is given by a table (Table 18) in the WHO anthropometry manual arguing... for correction for age and showing that MUAC for age is less sensitive than MUAC by itself.

The debate on the use of MUAC has been obscured by irrelevant statistical considerations. It looks smarter to use computer generated Z score rather than simple MUAC. Sorry, but muac is probably clinically more relevant.

My apologies to those who heard my theories too often... if they read this mail up to this point;


Dr. Andre' Briend

INSERM U 290, Hopital Saint Lazare

107 rue du Faubourg Saint Denis, 75 475 Paris Cedex 10, France

tel 33-1-45 23 24 07 tel (direct) 33-1-48 00 56 04

fax 33-1-47 70 28 35

Date: Mon, 31 Aug 1998 09:23:57 -0400

From: (Steve Collins)

Subject: Re: Ngonut: MUAC, Rambo and survival and bed-scales



I an very glad that you repeated your ideas on MUAC as I must have

missed the first post last year. I find them very interesting and in

accordance with some observations that I have collected.

My data comes from adults in Somalia, South Sudan and Angola where we

collected BMI and MUAC data. Unfortunately follow up of the patients

was not possible, so all I have is a cross sectional analysis of the

relationship between MUAC BMI plus other confounders such as oedema,

sex, race etc. Not as useful as longitudinal data but there are a

couple of things which might be of interest and in support of what you

say about MUAC in children.

The relationship between MUAC and BMI does not appear to be the same

in well nourished and under-nourished adults. In malnourished adults

the slope of the regression line of MUAC (y) on BMI (x) is usually

steeper the more malnourished the adults are. This is true except for

at the extremes of malnutrition (BMI <11-12 in Nilotic people

therefore probably 12-13 in more normal body shapes) This "S" shaped

relationship would accord with your ideas about MUAC being more

relevant than Weight for Height or BMI because it is focusing on

protein reserves. These differences might be related to the fact that

during acute energy deficiency tissue with a low metabolic rate, such

as subcutaneous fat and skeletal muscle, is preferentially

catabolised. This would tend to decrease the MUAC relatively faster

than the loss of weight. This effect would be further enhanced by the

increased catabolism of skeletal muscle caused by the higher rates of

infection in acute severe malnutrition (due to the immuno-suppression)

and the reduction in physical work using the arms that occurs during

famine (when people often have little to do but wait for food aid).

The flattening out at the extremes of starvation would again be

expected as peripheral stores are exhausted and the body turns to more

central reserves. If so this would indicate a level of starvation of

great functional significance as below such a threshold a shift to

more centrally located catabolism would increase the rate at which

organ function decreased and would therefore greatly worsen the

prognosis and increase the need for specialized nutritional


Such a non linear relationship between MUAC and BMI during acute

malnutrition also has practical implications that need to be taken

into consideration when MUAC cut-offs suitable for use in severe acute

malnutrition are stipulated. Recently the renewed academic interest

in severe adult malnutrition has seen, in the absence of much data

from acutely starving adults in famines, a tendency to derive MUAC

cut-offs proposed as suitable for use in famine relief), by

extrapolating data from normal populations, using a z score approach.

However if this changing relationship between BMI and MUAC during

acute under-nutriton is true, the distribution of MUAC values in a

population containing both normally nourished and some acutely

malnourished adults, would not be normal but would be skewed.

Therefore deriving cut-offs from z scores would be inappropriate and

certainly less functionally significant than using data form acutely

malnourished adults.

MUAC cut-offs equivalent to a BMI of 13, derived from better

nourished populations (e.g. James et al / Ferro-luzzi et al) were 2 -

2.5 cm higher than cut-offs derived from our Concern data sets which

included several hundred acutely malnourished adults. If we had used

these "theoretically derived" cut-offs as entry criteria, in place of

ones based upon severely malnourished patients the numbers of patients

admitted into Concerns' adult TFCs would have been increased by up to

50% - a major step to take in the absence of a sound theoretical or

empirical basis. At the moment the paucity of data means that the

validity of such cut-offs has yet to be tested, so rock on ACF and

their research into adult indicators in famine.

Again I fully agree with you thesis that the acceptance of indicators

based upon weight and height as gold standards is inappropriate. BMI

exhibits large intRA-population variation due to differences in leg

length : back length (cormic index) which cannot be corrected for other

than by evaluating the cormic index in each individual case. Anybody

who has tried to get even a height and weight measurement on each

patient in a long que of severely malnourished adults awaiting admission

into a feeding centers, will realise the impossibility of such

measurements at the height of a famine relief response (precisely the

time when severe adult malnutrition and mortality tends to occur).

Also, as Mike mentioned, from the Concern data we have found that the

inability to stand is itself often the best predictor of death - making

BMI rather an irrelevance in the most severe cases. On a lighter not a

response that I recently had from one reviewer that I should have used

bed-scales during a relief programme in Ayod, South Sudan illustrates

the gulf that sometimes exists between the practicalities in the field

and what is considered important in the "towers".

Hope this was of interest


steve collins

Date: Fri, 28 Aug 1998 08:08:38 -0400

From: "Bart Burkhalter" <>

Subject: Re: Ngonut: MUAC as a criterion for discharge from feedingprogrammes -Reply


See Pelletier DL, The relationship between child anthropometrry and mortality in developing countries:Implications for policy, programs and future research, J of Nutrition 124 (10S), 1994 Oct: pp 2073S-2076S and Table 18. It is a nice little meta-analysis of MUAC, W/A and H/A.

Date: Fri, 28 Aug 1998 15:39:45 +0200

From: (Andre' BRIEND)

Subject: Ngonut: Z scores vs % median


Dear NGO'nuts,

A short comment to Rita query re: Z scores.

Z scores where introduced more than 10 years ago purely for theoretical reasons. For obscure reasons, statisticians like the concept and sold the idea to nutritionist whithout checking whether it does help in practice. To my knowledge, this was done only recently by two studies (abstracts below) which showed that % of median was at least as good, if not better, to assess the risk of dying in malnourished children. Conclusion: using Z score will not help to optimise programmes aiming at reducing mortality.

There has been a lot of discussions in NGO circles after it was 'discovered' that 'Z scroes are more sensitive to detect malnutrition'. This point is irrelevant, because indices can be made sensitive by chosing different cut offs.

In any case, I never heard of a programme becoming more effective after switching to Z score. In my opinion, a lot of energy was wasted to promote Z score at field level.



Sachdev, H.P., Satyanarayana, L., Kumar, S., and Puri, R.K. Classification of nutritional status as 'Z score' or per cent of reference median--does it alter mortality prediction in malnourished children? Int.J.Epidemiol.

21(5):916-921, 1992.

The objective of the study was to evaluate, using a prospective cohort study, whether classification of nutritional status by 'Z score' or per cent of reference median alters the prediction of death in malnourished children.

The subjects were children with diarrhoea requiring hospitalization due to moderate or severe dehydration and/or associated complications. There were 382 participants under 5 years of age, of whom 37 died (cases), 320 were discharged in a satisfactory condition (controls) and 25 left before diarrhoea was completely cured (lost to follow-up--excluded) . Rehydrated weight and recumbent length (under 2 years) or standing height were recorded and the three indices (weight for age, height for age and weight for height) derived as both 'Z scores' and per cents of reference National Centre for Health Statistics (NCHS) medians. Logistic regression, sensitivity specificity curves and Zda test for normalized distances were used to compare the relative utility of these two classification methods in identifying children with a high risk of dying. The per cent of reference median and 'Z scores' were highly correlated (r = 0.9540, 0.9787 and 0.9667, respectively). Both methods yielded virtually identical results in predicting death of malnourished children for all the three indices. It was concluded that 'Z score' and per cent classification of nutrition are equally efficient in predicting death of malnourished children

Prudhon, C., Briend, A., Laurier, D., Golden, M.H., and Mary, J.Y.

Comparison of weight- and height-based indices for assessing the risk of death in severely malnourished children. Am.J.Epidemiol. 144:116-123, 1996.

To compare the effectiveness of treating malnourished children in different centers, the authors believe there is a need to have a simple method of adjusting mortality rates so that differences in the nutritional status of the children are taken into account.The authors compared different anthropometric indices based on weight and height to predict the risk of death among severely malnourished children.Anthropometric data from 1,047 children who survived were compared with those of 147 children who died during treatment in therapeutic feeding centers set up in African countries in 1993.The optimal ratio of weight to height determined by logistic regression was weight (kg)/height (m)1.74 (95% confidence interval of beta estimate 1.65-1.84).The receiver operating curves (sensitivity vs.specificity) showed that the body mass index (weight (kg)/height (m)2), optimal ratio of weight to height, and weight/height index expressed as the percentage of the median of the National Center for Health Statistics' standard were equivalent and superior to the weight/height index expressed as the z score of the National Center for Health Statistics' standard to predict death.As the optimal ratio of weight to height is easier to calculate than the weight/height index expressed as the percentage of the median or z score and does not depend upon either standards or tables, the optimal ratio of weight to height could be conveniently used to adjust mortality rates for nutritional status in therapeutic feeding centers.

Date: Fri, 28 Aug 1998 07:39:22 +0100


Subject: Re: Ngonut: Z scores vs % median


Dr. Briend throws the baby out with the bathwater when he denigrates Z-scores because they don't predict mortality. There are primarily three ways in which anthropometry is used (only one of which potentially involves mortality): for screening (including screening in life-threatening situations), for surveys and socioeconomic analysis, and for growth promotion. While Z-scores of weight and height may not (in Dr. Briend's opinion) be optimal for the first purpose, particularly in life-threatening situations, they are invaluable for the second purpose and quite useful for the third. Part of our job as advocates, program managers, and policy makers is to evaluate whether we're making progress, to prioritize among regions, districts, etc., and to identify factors which are important determinants of malnutrition. For these purposes, Z scores are extremely useful ... far more valid than %s or centiles. Being able to cross disciplinary lines analytically has been crucial to making nutrition credible and z-scores have made this possible. It's not saving lives but it can lead to increased attention to nutrition and better programs.

Please don't call into question a useful tool and by doing so undermine the credibility of our whole field of endeavor. Can't you just promote the right tool for the right job and let those of us who use other tools do our own jobs well?!

Judith McGuire

LCSHD; I 7-207

World Bank

1818 H St. NW

Washington, DC 20433

Phone: 202-473-3452

Fax: 202-522-1201


Date: Fri, 28 Aug 1998 17:11:05 +0200

From: Pierre TRAISSAC <>

Subject: Re: Ngonut: Z scores vs % median


A short comment on Andre Briend's post about Z-scores :

>Z scores where introduced more than 10 years ago purely for theoretical

>reasons. For obscure reasons, statisticians like the concept and sold the

> idea to nutritionist whithout checking whether it does help in practice.

Whether or not one deals with anthropometric measures such as weight or height, calculating Z-zcores is a very common way to assess how far a value is from the center of the distribution independently of the unit of measurement and taking into account the variance of the distribution, which enables to make comparisons across distributions with different variances.

As for using Z-scores as opposed to % of the median or centiles to assess the nutritionnal status of individuals there are a quite a number of *technical* advantages most notably :

- The same interval between centiles correspond to different variation according to the place in the distribution (the scale is not linear).

It is thus incorrect to compute any summary statistics such as mean or standard deviation for centiles.

- as for %median, the main flaw is that there is no exact correspondence with a fixed point of the distribution so you don't have a uniform criteria for different indices of for the same indice across different age classes because the variances are different. For example as weight and height have different distributions (variances) , in the reference population 80% of the median for Weight for height is about -2 Z scores as you have to have 90% of the median to reach -2 Z-scores for height-for age.

As for whether the whole Z-score and -2 cut-point were really promoted by statisticians, I don't know all the whereabouts but I guess if real statisticians had been in charge they would have choosen a nice -1.96 cutpoint, which would have resulted in a neat 5%/2=2.5% reference percentage instead of the weird 2.27% :-) To add fuel to the debate I may add that (in the spirit of Mora's Standardized Prevalence) mathematicians have many ways to compare distributions more refined than comparing the % under a certain cutpoint. Thus we may hint the cut-point idea was on the contrary something that was initiated by people with a strong emphasis on practicality and "fieldability".

As for using Z-scores in the field, I guess nowadays it does not make a big difference to compute Z-scores instead of % median or centiles once you have the anthropometric measurements.

I do agree with Andre about the irrelevant so-called better sensitivity of Z-scores. Sensitivity has not only to do with the indices but mainly with the choosen cutpoints.

Ref :

WHO (1995). Physical status : the use and interpretation of anthropometry. WHO Technical Report Series n° 854, Geneva.

Pierre Traissac


ORSTOM - Nutrition

fax : (33) 4 67 54 78 00 BP 5045 tel : (33) 4 67 41 61 70


www : FRANCE

Date: Fri, 28 Aug 1998 09:58:38 +0200

From: Rita Bhatia <>

Subject: Re: Ngonut: MUAC as a criterion for discharge from feeding programmes -Reply



I have three publications on MUAC

1. MUAC what are the uses and constraints at the field level: A meta-analysis of MSF survey data, Koert Ritmeijer, Arine Valstar, Austen Davis MSF Holland, this was published in MSF news I think when? MSF -H can enlighten us.

Conclusions of this study is There is a clear correlation between MUAC and W/H. a CUTOFF POINT OF 120-125mm was found to be most appropriate. However MUAC and W/H indicators select different children.

MUAC prevalence estimates correspond better to W/H when expressed as Z score than % of median.

<125 =<-2sd and ,-

<-120= ,<-3sd

However the inclination of prevalence of malnutrition defined by MUAC between 120 and 125 indicate the difficulties in defining one appropriate cut-off point and illustrates how prone MUAC is to measurement error.

Therefore, MUAC can not be used as a reliable measure of prevalence of acute malnutrition, but it can be used for quick screening to provide rough estimate and further need of carrying out a survey.

This based on total sample of 2656 Children from 5 countries refugees, displaced and locals

Bangladesh, Liberia, Mozambique, Rwanda and Rwanda Burndeses Refugees.

2. Other publication The development of a MUAC for Height reference, including a comparison to other Nutritional Status Screening Indicators published in WHO Bulletin 1997 pages 333 -341 by Onis, Yip, Grummer-Strawn, and Mei.

followed by another one on MUAC for Age recommended by WHO in 1997 WHO bulletin. Onis, Yip and Mei.

None of these publication mention using MUAC as criteria for Admission and Discharge.

On the issue of Criteria for Admission and Discharge we are struggling with whether to use % of median W/H or Z scores.


From: steve hansch

Date: Sat, 29 Aug 1998 18:41:26 +0100 (BST)


Historical note on Z

Conversion to Z scores in nutritional monitoring, in the US, with CDC fostering the transition around 1985-87.

By 1987 we were fully using Z scores at the Massachussets Commonwealth Department of Public Health as a clearer way of tracking individual children at different ages. The goal in these monitoring programs was not the same in an emergency -- we were not attempting to survey large at-risk populations to triage and target very-high-risk children. We were, rather, attempting to maintain ongoing surveillance for a large stable population.

The Z score concept helps, in an NCHS-like population like the US, in comprehending how children track against various growth curves, as they grow older. % of median was always a bit of a bizarre concept, as you got such wildly different numbers of children depending on whether it was Wt/HT or Wt/Age.

take care and good luck,

steve hansch

Congressional Hunger Center

Date: Sun, 30 Aug 1998 11:17:51 +0100

From: Tim Cole <>

Subject: Ngonut: Re: MUAC, Z score and % median


It is worth being clear about the pros and cons of MUAC, MUAC as % of median and MUAC Z score. They all have their place.

Mean MUAC changes with age, so an individual value can't be assessed unless the child's age is known. % of median achieves this in a simple way, but it doesn't adjust for the SD of MUAC as % of median. It can be shown that this SD is the same as the coefficient of variation (CV) of MUAC, which is about 8% say (this is a wild guess). So a MUAC 92% of the median is about 1 SD below the mean, which corresponds to a MUAC Z score of -1.

So if the CV is constant, then % of median and Z score are broadly equivalent.

A very crude formula to convert from one to the other is

Z score = (% of median - 100)/CV

If the CV changes with age, or differs between two measurements (e.g. MUAC vs weight) then the two cannot be compared directly.

Steve Hansch's comment that

>% of median was always a bit of a bizarre concept, as you got such wildly

>>different numbers of children depending on whether it was Wt/Ht or Wt/Age

reflects the different CVs of Wt/Ht (8%) and Wt/Age (12%).

[The formula above assumes that MUAC is Normally distributed. In fact it is skew, like weight, which complicates things. My LMS method deals with this, but it is too complicated and irrelevant for the discussion here.]

The advantage of MUAC Z score over MUAC is that, converted to a centile, it expresses an individual child's value in the context of the distribution for the child's age.

André Briend points out that this is not always what we want. For mortality, MUAC is more informative than MUAC Z score. This means that adjusting for age actually throws away information - how can this be?

To explain, imagine a mortality cut-off for MUAC of say 12.5 cm. In younger children, where mean MUAC is relatively low, 12.5 cm corresponds to a relatively high centile. In older children, the same cut-off is a much lower centile.

Now imagine that the mortality risk associated with a given centile is greater in younger children, as they are more vulnerable. So a relatively high centile in young children, and a much lower centile in older children, both have the same mortality risk.

But these two centiles correspond to the same MUAC value. So in this case, MUAC is more useful than MUAC Z score for predicting mortality.

Best wishes to all,

Tim Cole tel: +44 1223 426356 fax: +44 1223 426617 MRC Dunn Nutrition Centre, Downhams Lane, Milton Road, Cambridge CB4 1XJ, UK

Date: Mon, 31 Aug 1998 10:24:56 +0200 (METDST)

From: (Andre' BRIEND)

Subject: Re: MUAC, Z score and % median


Dear NGO'nuts,


I fully agree with Tim's comments. My arguments with muscle mass aims

to explain what we see in physiological terms, but a statistical explanation

is also needed. Tim is indeed in a better position than me to make this point.


Please note that I have nothing against Z score in general. I know it is a

useful concept in many situations. I know it is widely used in the US (and

in France) for following up children individually. I argue though that the

practical benefits of the introduction of Z scores at the field level in the

emergency context are dubious at best and not worth the time spent

explaining the pros and cons of each index at each nutritional meeting for

the last 10 years + papers, books, manuals, field guides, etc...on the




Dr. Andre' Briend

INSERM U 290, Hopital Saint Lazare

107 rue du Faubourg Saint Denis, 75 475 Paris Cedex 10, France

tel 33-1-45 23 24 07 tel (direct) 33-1-48 00 56 04 fax 33-1-47 70 28 35