MUAC cutoff to screen SAM

This question comes up from time to time on EN-NET.

Age-independence is a useful attribute of a SAM case-definition. When you look at MUAC in children you can see that it increases with increasing age quick quickly before 12 months or age and more slowly afer that. MUAC is, therefore, not independent of age. This is what you observe and comment upon.

The issue is, however, a bit more complicated than this. Therapeutic feeding programs, are, I think best understood as child survival programs. We treat SAM to avoid death not to treat thinness. In this case we would be most interested in the relation of an indicator to mortality risk. For MUAC we see that fixed MUAC thresholds have similar mortality risk over a wide range of ages. This is not universally the case with (e.g.) W/H.

MUAC does select more younger children than older children. This is not a flaw. It is what we want because the younger children selected are at higher near term mortality risk that can be reduced by therapeutic feeding.

A simple single threshold allows programming with community-based volunteers doing case-finding and mothers doing MUAC at home. This allows high spatial and temporal coverage of case-finding activities that are not possible with other indicators.

See this article for a discussion and review of SAM case-definitions.

I hope this helps.

I am not sure I quite understand the question here.

Fixed MUAC thresholds have similar mortality risk over a wide range of ages. This means that the we should see a similar false positive rate over a wide range of ages for a fixed MUAC threshold. I do not think that "there will be more false positives amongst the younger ones". Am I missing something imprtant here?

I'd like to pick up on "false positives" ... I like to think of this as a "provider error". By this I mean that the provider uses resources when they may not be needed. Such an error is in favour of the child. I do not mind making this error as much as I mind making the opposite error (i.e. to refuse resources when they are needed). From this perspective "false positive" errors are benign errors.

The use of MUAC without correction for age is based on a pragmatic approach. It relies on the observation that there are no solid data to claim that any anthropometric measure really reflects nutritional status for which there is no clear objective definition. The idea is to select malnourished children selected by whatever criterion and which are most at risk of death. Studies carried out at the community level on untreated children have shown consistently that MUAC is best at that and that correction for age does not improve risk assessment. See articles below on the lack of effect of age correction:

Briend A, Zimicki S. Validation of arm circumference as an indicator of risk of death in one to four year old children. Nutr Res 1986;6:249–61.

Rasmussen J, Andersen A, Fisker AB, Ravn H, Sodemann M, Rodrigues A, Benn CS, Aaby P. Mid-upperarm-circumference and mid-upper-arm circumference z-score: the best predictor of mortality? Eur J Clin Nutr 2012;66:998–1003.

I am confused too by concern about “false positive” raised by Mija. In the context of risk of death assessment, false positive are children selected by MUAC who would survive in absence of intervention. When the objective is to reduce mortality, it is necessary to use high cut off which will be associated with many false positive based on this definition. This is not a real concern: there is no more false positive under this definition in young children as mentioned by Mark.

This risk approach leads to the selection of children who are quite different from the traditionally used normative approach which selects children deviating from the WHO standard. The discrepancy between the two methods has been a topic of discussion for the last 20 years at least. The main questions to raise when choosing between the two methods is: what is your priority objective ? do you want to prevent death ? Or to have children be within accepted deviation of growth standards ?

One of the potential problems of the risk approach is that it may select children with a high risk of death for other reasons than acute malnutrition and may not respond to treatment. More specifically, it has been objected that using MUAC without correction for age can lead to the selection of young and stunted children who will not respond to treatment and may be even exposed to overfeeding and have excess fat deposition if treated with RUTF. This lead some experts to recommend to exclude from treatment low MUAC children who are below some minimal height. This recommendation, however, was not supported by any solid data. Current evidence suggests that short children respond to treatment just as well as taller children. See:

Dale NM, Myatt M, Prudhon C, Briend A. Using mid-upper arm circumference to end treatment of severe acute malnutrition leads to higher weight gains in the most malnourished children. PLoS One. 2013;8(2):e55404.

Fabiansen C, Phelan KP, Cichon B, Ritz C, Briend A, Michaelsen KF, Friis H, Shepherd S.
Short children with a low midupper arm circumference respond to food supplementation: an observational study from Burkina Faso. Am J Clin Nutr. 2016 Feb;103(2):415-21.

A paper by Fabiansen et al, about to be published will even show with body composition data (using D2O dilution technique) that treated short children mainly put on lean tissues and that the concern of making them fat by treating with RUTF them is totally unfounded. The practice of using a minimum height to eliminate young and stunted children when using MUAC should be abandoned.

A more genuine concern re. the use of MUAC uncorrected for age is the upper age limit to which it can be applied. Clearly the current cut off of 115 mm cannot be used in adolescents or adults. What to do after the age of 3 to 5 years is a current topic of research.

I hope this helps,

Dear Mija,

Thanks for your kind comments.

The curve showing the relationship between MUAC and mortality does increase sharply below 115 mm, but it starts to increase from around 120-130 mm. See the curve p 17 at:

http://www.who.int/nutrition/publications/severemalnutrition/FNB_0379-5721_A_review.pdf?ua=1

This curve is based on mortality, but you would get the same pattern with relative risk. See this study showing an increase in relative risk with decreasing MUAC below 130 mm:

Briend A, Wojtyniak B, Rowland MG. Arm circumference and other factors in children at high risk of death in rural Bangladesh. Lancet. 1987 Sep 26;2(8561):725-8.

Based on this, some argue that children with a MUAC below 120 or 125 should also receive some treatment. This has been tested in Sierra Leone with some success. See:

Maust A, Koroma AS, Abla C, Molokwu N, Ryan KN, Singh L, Manary MJ. Severe and Moderate Acute Malnutrition Can Be Successfully Managed with an Integrated Protocol in Sierra Leone. J Nutr. 2015 Nov;145(11):2604-9

This promising approach is currently further investigated.

To my knowledge, no one looked at the curves relating MUAC to risk of death for different age groups. This may not be so easy to do because to look at this you need to break down the population in small subsamples and the association will be subject to a lot of random variation. In absence of effect of a correction for age on risk assessment, I assume the risk should be similar in different age groups.

In my post yesterday, I forgot to mention an other paper which also shows that risk assessment does not improve when using MUAC-for-age compared to unadjusted MUAC. Here is the reference:

Bairagi R. On validity of some anthropometric indicators as predictors of mortality. Am J Clin Nutr. 1981 Nov;34(11):2592-4.

By the way, this paper was the first to have shown the superiority of MUAC to detect children with a high risk of death.

I did some work on this with the WHO and others almost ten years ago. I speak for myself here.

My recall is that the shift from 110 mm to 115 mm was based on more than one consideration.

The primary consideration was mortality risk. The decision was made on the grounds of both absolute risk and on the rapid change of risk below 115 mm. The "elbow" in the plot of mortality risk by MUAC occurs between 120 mm and 115 mm. This can be seen here:

Since CMAM is a child survival program we wanted high sensitivity as the consequence of missing cases is failure to prevent preventable deaths. This led us to a decision to revise the 110 mm threshold upwards.

We also needed to balance sensitivity and capacity. MUAC is usually normally distributed and small changes in thresholds will lead to large changes in numbers identified. For example, in a population of 10,000 children with a mean MUAC of 140 mm with SD = 14 mm we identify:

    MUAC < 110 mm    161 children
    MUAC < 115 mm    371 children
    MUAC < 120 mm    766 children

Each 5 mm increase in the threshold more than doubles the number of children identified. Selecting 120 mm would have resulted in a c. 500% increase in case numbers. This was too big an increase to argue for at the time for MoH delivered programming. This, and the pattern we saw in the plot (above) caused us to select 115 mm. Some NGOs (e.g. MSF) adopted the 120 mm threshold. Newer intervention models (e.g. COMPAS) use MUAC < 125 mm and vary treatment intensity by MUAC with good results and controlling costs, workloads, and crowding at delivery points.

We could have chosen 116 mm or 117.5 mm but the convention is to use 5 mm steps and this was reflected in the mortality reports available to us. We were wary of squeezing out the middle (i.e. MAM) and had no remit to advise on revising case-definitions.

We could, given sufficient data (now unethical to collect) have selected age-specific thresholds. This would have eliminated the simplicity which is a key feature of MUAC. Also, available evidence showed (and still shows) that adding age or height to MUAC added complication but did not improve case-finding sensitivity or specificity.

Some of the work we did is covered in this report. Note that the recommendation to use proportional weight gain to decide discharge made in this report has already been overturned and MUAC is now used for this purpose.

Following up on Andre's "To my knowledge, no one looked at the curves relating MUAC to risk of death for different age groups. This may not be so easy to do because to look at this you need to break down the population in small subsamples and the association will be subject to a lot of random variation. In absence of effect of a correction for age on risk assessment, I assume the risk should be similar in different age groups". I think:

    Briend A, Wojtyniak B, Rowland MGM. Arm circumference and other factors
    in children at high risk of death in rural Bangladesh. Lancet 1987; 2(8561):725–8

    Vella V, Tomkins A, Ndiku J, Marshal T, Cortinovis I. Anthropometry as a
    predictor for mortality among Ugandan children, allowing for socio-economic
    variables. Eur J Clin Nutr 1994; 48:189–97.

    Briend A, Garenne M, Maire B, Fontaine O, Dieng K. Nutritional status, age and
    survival: the muscle mass hypothesis. Eur J Clin Nutr 1989; 43:715–26

    Berkley J, Mwangi I, Griffiths K, Ahmed I, Mithwani S, English M, Newton C,
    Maitland K. Assessment of severe malnutrition among hospitalized children in
    rural Kenya: comparison of weight for height and mid upper arm circumference.
    JAMA 2005; 294:591–7

    Berkley J, Newton C, Maitland K. Severe malnutrition assessment in children
    in rural Kenya. JAMA 2005; 294:2577

Show that the predictive power of MUAC (i.e., the power of predicting mortality) is independent of age even in children below 1 year of age.

I hope this is of some use.

Not so much THE history. Just a history from one person involved in the process for several years. I think the process was a rational one.

It may be more than you expect as the expected (i.e. Normal) distribution has exponential properties with small changes in 'x' leading to quite large changes in 'y'.

From my post above we'd expect a 371 / 161 = 2.304348 fold increase in case-numbers.

I think theer are two approaches that you can use to find this in your specific context. A model based approach assuming a normal distribution might be useful. We'd need to find mean and SD for MUAC. You can do this from survey (e.g. SMART, MICS, DHS) data or from the WGS reference tables. To make use of this data we would use a PROBIT approach.

Assuming Mean MUAC = 142 mm (SD = 14.5 mm) and define SAM as MUAC < 115 mm then the proportion of the population with SAM would (using EXCEL) be:

        =NORM.DIST(115, 142, 14.5, TRUE)

which gives:

        0.031296685 (3.13%)

If we change the SAM definition to MUAC < 120 mm we use:

        =NORM.DIST(1120, 142, 14.5, TRUE)

which gives:

        0.064602876 (6.45%)

If coverage remains te same we'd expect cae number to increase by:

        0.064602876 / 0.031296685 = 2.064208

Case numbers double (really a little bit more than double). That is quite a larg change for a 5 mm change in the MUAC case-definition.

You can also do the calculation from survey data using classical estimator. I did this using data from a SMART survey from Kabul, Afghanistan. Using MUAC < 115 mm I found prevalence to be 2.04%. Using MUAC < 120 I found prevalence to be 3.39%. In this context we expect a 1.66 fold increase in case numbers. YOu don't need to work with prevalence. You cn use case numbers from tables. I get:

        > table(x$muac < 115)

        FALSE  TRUE 
            866    18 

        > table(x$muac < 120)

        FALSE  TRUE 
            854    30 

and 30/18 = 1.67. Close enough for jazz!

I think I'd prefer to use the survey data approach as it makes no grand assumption about the distribution of MUAC and it is easy to incude oedema in the SAM case definition (which I think you should do). With the Kabul data I get

        > table(x$muac < 115 | x$oedema == 1)

        FALSE  TRUE 
            862    22 

        > table(x$muac < 120 | x$oedema == 1)

        FALSE  TRUE 
            851    33 

That is a 33 / 22 = 1.5 fold increase in cases.

I hope this is of some use to you.

Oops ... I have 1120 above ... that should have been 120!