Can height-adjusted cut-offs improve MUAC’s utility as an assessment tool?
By Michel Van Herp, An Verwulgen,
Bérengère Leurquin, and Pascale
Delchevalerie
Michael Ven Herp, Bérengère Leurquin, An Verwulgen & Pascale Delchevalerie
Michael Ven Herp is a medical doctor
and epidemiologist and has ten years
field experience with MSF in South
America and Africa. He is currently an
epidemiologist in the medical department
of MSF-B, involved in infectious
diseases and surveys encompassing
nutrition, mortality, access to health
care, efficacy and compliance.
Bérengère Leurquin is a paediatric nurse
and has 4 years field experience with
MSF in Angola, South-Sudan,
Afghanistan, Ivory-Coast. For the past
two years she has she has worked as a
nutrition advisor in the medical department
of MSF-B.
An Verwulgen is a paediatric nurse, and
has six and a half years field experience
with MSF in Kenya, South Sudan,
Angola, and Rwanda. She analysed the
databases under the guidance of Michel
Van Herp.
Pascale Delchevalerie is a paediatric
nurse and epidemiologist. She has 10
years field experience with MSF in medical
and nutritional programmes in Asia
and Africa. For the past 2 and a half
years, she has worked as a nutrition
advisor in the medical department of
MSF-B.
Thanks to the field teams who carried out the surveys and
to Nancy Harris for her support in editing the text.
This field article looks at the relationship between
MUAC and WH and finding that MUAC correlated
most closely with W/H indices in shorter children,
suggests the possibility of height specific
MUAC cut-offs, distinguishing between assessment
and screening.
While quick and convenient to use,
Mid Upper Arm Circumference
(MUAC) measurement has limitations
in the assessment of malnutrition.
In certain contexts, notably the Horn
of Africa, MUAC sometimes underestimates the
problem on an individual and global basis compared
to other anthropometric indices.
Médecins Sans Frontières (MSF) first noticed
this in nutritional surveys in Ethiopia (Somali
region),1 where MUAC underestimated prevalence
of malnutrition compared to weight for
height (W/H) indicators. In Chad, a MUAC <
125 mm was used by home visitors to refer children
for W/H measurements and consideration
for admission to a feeding centre. However,
after an anthropometric survey,2 calculation of
the sensitivity of the 125 mm cut-off showed
that 65.8 %, of all malnourished children and
42.9 % of severely malnourished were missed,
as compared to W/H % of the median (WHM).
With a cut-off of 135 mm, 32.9 % were still
missed. These findings resulted in an adjustment
of the screening strategy: the cut-off was
increased to 145 mm, and W/H was calculated
on the spot to exclude children who would not
meet admission criteria in the feeding centre.
Another example comes from Niger (2005),
where rapid MUAC assessment was undertaken
without other anthropometric indices being
measured. The prevalence of malnutrition identified
by this method was at odds with information
from other sources, e.g. nutritional programmes
reporting a significant increase in
admissions, suggesting greater prevalence of
malnutrition than that identified by MUAC
assessment.
These findings raised questions about the
validity of MUAC in rapid assessments (to estimate
the nutritional status of populations) and
as a screening tool (to select individuals requiring
further assessment, e.g. W/H measurements,
to determine eligibility for admission to
nutritional programmes).
How were MUAC cut-offs in current use
selected?
An Epicentre study of 64 nutritional surveys3 (n=34,933) showed that the MUAC mean
increases with age: 132mm at 6 months (~ -1Z of
the National Centre for Health Statistics
(NCHS) reference of 143 mm), 148mm at 59
months (~ -1.5Z of the NCHS ref of 174 mm). It
is therefore difficult to select a single cut-off for
all children under 5 years of age.
An MSF-Holland analysis of five nutritional
surveys4 (n=2,656) demonstrated a clear correlation
between MUAC and W/H as expressed by
Z-score (WHZ), with better correlation in populations
with higher prevalence of acute malnutrition
(r = 0.65 if ?10%, r = 0.57 if < 5%).
Single MUAC cut-offs of 125mm (global)
and 110mm (severe) were proposed for all children
under 5 years (regardless of ethnicity), and
it was further suggested that these criteria
could be used for quick assessment to assess the
need for a survey. However, due to a significant
number of false positives and negatives with
MUAC, it was recommended that MUAC not
be used as an entry criterion for nutritional
rehabilitation programmes, but it could be used
as the first stage of a two step screening process
with a relatively higher cut-off (135mm). In
1994, a meeting of all MSF medical directors ratified
these MUAC cut-offs for use in rapid
assessments.
Finally, an Epicentre analysis of 114 nutritional
surveys (December 2002; n=66,446) compared
MUAC (< 125mm) to WHZ < -2 or < 80%
WHM. Overall, MUAC overestimated malnutrition
as compared to WHZ in 45% of the surveys,
and underestimated malnutrition as compared
to WHM in 8 %. In the analysis by age
group, the under-estimation was greater in children
? 29 months and also in boys (13%) than
girls (4%). Greater underestimation also
occurred in higher prevalence contexts.
All of the above studies were based on onetime
surveys. The effect of situational factors
such as hunger gaps, epidemics, chronic malnutrition,
etc., and the evolution of correlations
between MUAC and other indices over time
were not examined. Nor did these studies
assess the possible effect of selection criteria age
(6-59m) versus height (65-110cm) on outcomes.
Our study provides a preliminary exploration
of these factors.
Objectives and method
The objective of this study was to further
analyse existing surveys, conducted in areas
where MUAC underestimates malnutrition, with the aim of assessing the relationship
between MUAC and WH indices over time. A
secondary research question arose from the
finding that MUAC correlated most closely
with W/H indices in shorter children, thus suggesting
the possibility of height specific MUAC
cut-offs.
Twelve databases of nutritional surveys (n =
10,226) were used for this analysis. The surveys
were conducted in Denan/Ethiopia (6) between
May 2000 and September 2001, Korma and Serif
Umra/Darfur (4) in 2004/2005 and Tine and
Iriba/Chad (2) in 2003/2004. The standard
methodology of UNHCR/WFP/MSF was followed.
Only children with a length/height
between 65cm and 110cm were included. Data
analysis was undertaken using the computer
program EPI INFO 6, version 6.04d.
The cut-off point of W/H - 2 Z-score, or 80%
of median, or oedema, was used to classify
global acute malnutrition. Height for Age
(H/A) - 2 Z-score or 90% of the median5 was
used to determine global chronic malnutrition.
MUAC was compared with global acute malnutrition
expressed in W/H in % of median or
oedema and Z scores.6
In the Chad and Korma surveys, no children
with oedema were found. In the other surveys,
the % of oedema was between 1.3 and 0.1, with
a total of 37 oedema cases out of 10,226. These
low numbers should not influence the results.
Findings
Overall, for the majority of the surveys (9/12),
the under-estimation of MUAC prevalence
compared to WHM prevalence was statistically
significant. Malnutrition prevalence as per
WHM varied from 23% to 41% for Denan, 8% to
13% for Sudan and from 12% to 18% for Chad.
The Denan curve was the most informative
because it entailed the longest time series.
Except for the first survey, the difference
between prevalence in MUAC and W/H was
statistically significant. The crude mortality rate
(CMR) and under fives mortality rate (U5MR)
was extremely high in the first survey, but
returned to normal in the following ones (see
Table 1 and Figure 1).
Influence of age or height
When the surveys were further analysed on the
basis of age, it was noted that the discrepancy
between MUAC and W/H occurs mainly in
children above 2 years (see Figure 2). In Denan,
the under five mortality was very high leading
to under-representation of the 6-24 months age
group in the sample (8-15 % instead than 33%),
and, as a consequence, influencing underestimation
of MUAC due to a greater preponderance
of older children in the sample. The same
phenomenon occurred to a lesser extent in the
Darfur and Chad series (± 20% of 6-24 months).
As age assessment is often imprecise, children
are usually selected on the basis of height,
which serves as a proxy for age. Typically 65 -
110 cm represents children of 6 months to 5
years in age and 85cm represents age 24 months
(this is also the cut-off point above which children
are measured standing). However, in
countries affected by chronic malnutrition,
there will be a discrepancy between age and
height. In the Denan series, with a cut-off of 85
cm, 27% of the children were declared to be
above 24 months and 41% equal to 24 months
(age rounding effect). With a cut-off at 80 cm,
12.5% were declared to be above 24 months and
42.5% equal to 24 months. In this context, with
the second cut-off, fewer children > 24 months
are included, while most children ? 24 months
are captured (see Figure 3 for exact numbers).
For the groups < 85cm and < 80 cm, MUAC
underestimated malnutrition compared to
W/H in the majority of the surveys (11/12) but
the discrepancy was not statistically significant.
In the < 80cm group, the MUAC prevalence
curve lies closer to the WHM prevalence curve
compared with the height group < 85cm. For
the group of ? 85 cm and ? 80cm, the MUAC
generally underestimated malnutrition compared
to WH and in both groups, 10/12 surveys
this difference was statistically significant.
These findings led us to consider whether
assessment tools should be adapted to height,
to achieve better sensitivity. Accordingly, malnutrition
prevalence in different height groups
was compared to different MUAC cut-offs to
find the cut-off point that lies closest to the
WHM, with results as follows:
- In areas of very high prevalence, e.g. Denan,
(23.3 - 40.7 % WHM), the best cut-off was
125 or 126mm for children ? 80 or 85cm,
respectively, and 133 or 136mm for children
> 80 or 85cm, respectively.
- In areas of high prevalence, e.g. Tine (18.1%),
the closest cut-off was 127mm for height ?
80cm and 137mm for height > 80cm.
- In areas of moderate prevalence, e.g. Iriba
and Darfur (11-13%), the best cut off was
125-127mm for height ? 80cm, 125-128 mm
for ? 85cm and 133-137mm for height > 80
or 133-138mm for height > 85cm.
- In areas of lower prevalence, e.g. Korma
(8.3%), the best cut-off was 127mm for
height ? 80cm or 128mm for height ? 85cm
and 137mm for height > 80cm or 138mm for
height > 85cm.
Thus there is an inverse relationship between
MUAC cut-offs and prevalence of global acute
malnutrition, with lower MUAC values in high
prevalence areas.
However, the variation inside the same
height group is small (125 to 128 mm for shorter
children and 133 to 138 mm for taller ones).
Therefore, in practice, we could use a single cutoff
per height group.
Effect of chronic malnutrition
During the analysis of the 80 and 85 cm height
groups, we explored the role of chronic malnutrition.
The prevalence of chronic malnutrition
(as defined by HAZ) was found to be high in all
the surveys (for example, between 30 and 40%
in the Denan series), with the highest overall
(45.4%) in the second survey in Serif Umra.
In the Denan surveys, we examined H/A by
height group and noticed that the prevalence of
chronic malnutrition was higher in the group of
children < 80cm (from 64.8 % to 80.2%) than in
the group ? 80cm (16.6% to 28%). This is in contrast
to more common findings of increased
prevalence with increasing age (highest in 24-36
months group) and may be linked to the
measles outbreak that occurred in 2000.
Conclusions and recommendations
It is clear that the use of a single MUAC cut-off,
for all children 65 to 110 cm in height, may produce
prevalence data in some contexts that correlates
poorly with the prevalence of global
acute malnutrition found using other anthropometric
indices.
Diagram 1
Using the MUAC with different cut-offs (see
Diagram 1) represents one solution. However,
deeper analysis of the context for each situation
is also required. Since one cut-off point is not
enough for the entire age range, we propose a
'simplified quack-stick',7 with different cut-offs
for children ? 85 cm and those > 85 cm. This cutoff
should be shifted to 80 cm in areas where
chronic malnutrition is high. Since we are often
confronted by situations of moderate/high
prevalence of malnutrition, the cut-offs of 125
mm and 135 mm were chosen.
The choice of cut-off point for MUAC will be
determined by the objective of the enquiry:
assessment versus screening.
In some contexts, before carrying out MUAC
assessment, an estimate of prevalence by W/H
measurement must be obtained. If this is not
available from past surveys, a preliminary survey
would be worthwhile to determine population
adapted cut-offs. Lot Quality Assurance Sampling (LQAS)8 surveys with these MUAC cut-offs could ensue for
further assessments or follow-up. In a case where only rapid MUAC
assessment information is available, consideration should be given to
basing operational decisions on the prevalence in shorter children (<
80-85cm) instead of up to 110 cm, as our studies suggest better correlation
with other indices in this age group.
Figure 1 - Comparison of malnutrition prevalence expressed in MUAC or W/H, MSF-B nutritional surveys, Denan, Ethiopia, 2000-2001
For screening purposes, a nutritional survey should be performed
first, and adapted MUAC cut-offs then chosen, for use in subsequent
screening. This choice should be based on the need to have sufficient
sensitivity to capture the majority of malnourished children, while
minimizing false positives (children who fail W/H entry criteria for
nutrition programmes). Unnecessary referrals exact a high social cost
for families (travel, family disruption etc) and discredit the screening
team. Where we believe that social cost may be important, implementation
of W/H measurement on the spot following MUAC screening,
for selected children, will differentiate between those who do and
those who do not meet admission criteria, and hence reduce this social
cost.
For selection criteria to a Supplementary Feeding Programme
(SFP), consideration should be given to utilising both MUAC and
W/H (as they detect different physiopathological situations in a child)
and to adapt entry criteria to MUAC < 125mm or WHM <80%.
Considerations for future studies
In nutritional survey analysis:
Figure 2 - Comparison of MUAC < 125 mm, W/H < -2 Z score and W/H < 80% prevalence by age group, MSF-B nutritional surveys, Denan,
Ethiopia, 2000-2001.
- Consider stratifying analysis on the basis of height rather than
reported age.
- Consider presenting graph of height distribution as well as age
distribution. Age rounding may bias age distribution.
- MUAC measurements may be influenced by quality of technique.
MUAC is quick and easy to measure but needs training and
supervision to be precise. Presenting a graph of the MUAC
ditribution may help assess the quality of measurements, influencing
the validity of findings.
- Calculate and present prevalence of chronic malnutrition (H/A).
- Interpretation of results should consider the role of context:
mortality rates, past and present epidemics, population characteristics
(nomads, refugees, internally displaced persons (IDPs)),
seasonal variations or practices (hunger gap, Ramadan) etc.
Figure 3 - Discordance between reported age and height in MSF-B
nutritional surveys, Denan, Ethiopia, 2000-2001.
Further studies to explore these factors and their influence on malnutrition
indices would be worthwhile. These investigations were done
with the NCHS growth references. It would be interesting to repeat a
similar analysis of the dataset with the new WHO growth standards to
see if similar results are obtained. However, as the growth standards
are only presented as Z-scores, they will need to be expressed as % of
the median in order to allow comparison.
For further information, contact:
Pascale Delchevalerie, MSF, medical department, Rue Dupré 94, 1090
Brussels, Belgium, email: pascale.delchevalerie@brussels.msf.org
| |
Mortality |
All children |
| CMR * |
U5MR* |
N° |
W/H<-2
Z score
or
oedema |
W/H <
80% or
oedema |
Muac
<125 |
MUAC<125 /WH<80% |
r_ |
| sens
% |
spec
% |
PPV |
MUAC/
WHM |
| Denan |
|
|
5286 |
42.4 |
32.6 |
19.6 |
46 |
93 |
0.76 |
0.47 |
| 05/00 |
27.5 |
8.9 |
765 |
52.9 |
40.7 |
37.0 |
73 |
87 |
0.79 |
0.58 |
| 08/00 |
1,2** |
0,4** |
897 |
43.7 |
33.8 |
23.1 |
51 |
91 |
0.75 |
0.49 |
| 10/00 |
0,27** |
0,13** |
914 |
40.8 |
31.3 |
15.5 |
36 |
94 |
0.72 |
0.43 |
| 02/01 |
0,25** |
0,15** |
906 |
37.2 |
27.2 |
13.0 |
31 |
94 |
0.64 |
0.29 |
| 04/01 |
0,25** |
0,12** |
902 |
51.2 |
40.5 |
20.7 |
43 |
94 |
0.84 |
0.41 |
| 09/01 |
0,27** |
0,1** |
902 |
30.0 |
23.3 |
10.8 |
34 |
96 |
0.74 |
0.55 |
| Korma |
|
|
1835 |
16.6 |
10.7 |
4.4 |
32 |
99 |
0.80 |
0.46 |
| 04/05 |
2.2 |
1.3 |
923 |
19.6 |
13.1 |
6.4 |
40 |
99 |
0.81 |
0.48 |
| 12/05 |
2.2 |
0.8 |
912 |
13.5 |
8.3 |
2.4 |
22 |
89 |
0.32 |
0.19 |
| Serif Umra |
1.8 |
0.8 |
1741 |
17.5 |
13.0 |
7.3 |
33 |
97 |
0.61 |
0.31 |
| 10/04 |
1.2 |
0.8 |
865 |
14.7 |
11.2 |
8.0 |
41 |
96 |
0.58 |
0.28 |
| 06/05 |
0.2 |
3 |
876 |
13.5 |
10.7 |
6.6 |
34 |
97 |
0.55 |
0.34 |
| Tine |
2.2 |
1.3 |
536 |
29.5 |
18.1 |
6.5 |
28 |
98 |
0.77 |
0.37 |
| Iriba |
|
|
828 |
19.6 |
12.2 |
7.7 |
40 |
97 |
0.63 |
0.38 |
| * .. / 10.000 / day |
| ** mortality data collected prospectively by Community Heath Workers (CHWs)
during a specific period |
1 Nutritional surveys and retrospective mortality assessment,
Denan, Ogaden, Ethiopia, MSF-Belgium. May 2000 to
September 2001.
2 Jeroen Beijnsberger, Michel Van Herp, Pascale
Delchevalerie. Enquête Nutritionnelle, de Mortalité
Rétrospective et de Sécurité Alimentaire, camps de réfugiés
Soudanais de Touloum et Iridimi, région d'Iriba, Est-Tchad,
Octobre 2004.
3 Manoncourt S, Coulombier D, Pécoul B, Desvé G, Moren A
(Epicentre, MSF- France), Comparaison des informations
fournies par différents indicateurs de malnutrition aiguë calculés
au cours de 64 enquêtes nutritionnelles réalisées
auprès de populations en situation précaire, MSF Medical
News, vol 3, N°2, May 1993.
4 Koert Ritmeijer, Arine Valstar, Austen Davis (MSF -
Holland), 'Middle Upper Arm Circumference (MUAC): what are
the uses and constraints at the field level; a meta analysis of
MSF survey data', MSF Medical News, vol 3, N°2, May 1993.
5 Issues in the assessment of nutritional status using anthropometry.
Bulletin of the WHO 1994; 72:273-83
6 A comparison with z scores was included in the beginning,
however on further discussion and analysis, comparison was
only done with % of the median that is typically used in the
field.
7 A quack stick is an instrument currently available for measuring
MUAC for height.
8 Lot quality assurance sampling (LQAS) is a methodology that originated in the manufacturing
industry and has been applied to health contexts, such as immunisation coverage.
Sub-populations are divided into 'lots' and the sample size is the number of units that are
selected from each lot. Before sampling, a decision must be made on the number of
'defective' items, e.g. children not immunised, that will deem a 'lot' unacceptable, which in
turn will influence sample size. Since the response for each sample is binary, i.e. acceptable
or non-acceptable, smaller samples are required compared to other survey methods.
By combining information from different 'lots', the LQAS method offers a less conventional
means of stratified sampling. (WHO/V&B/0126(2001))
Taken from Field Exchange Issue 30, April 2007
http://fex.ennonline.net/30/canheightadjustedcutoffs.aspx