Sie sind auf Seite 1von 6

Clinical Nutrition (2003) 22(3): 307312

r 2003 Elsevier Science Ltd. All rights reserved.


doi:10.1016/S0261-5614(03)00009-8

ORIGINAL ARTICLE

A comparison of mid upper arm circumference, body mass


index and weight loss as indices of undernutrition in acutely
hospitalized patients
JEREMY POWELL-TUCK,* ENID M. HENNESSYy
Department of Human Nutrition, UK, wDepartment of Wolfson Institute of Preventive Medicine, St Bartholomews and the Royal
London Hospital School of Medicine and Dentistry,Whitechapel, London E11BB, UK (Correspondence to: JP-T, Adult and Paediatric
Gastroenterology,The Royal London Hospital, London EI IBR, UK)
n

Abstract A nutritional supplementation trial (Vlaming et al., Clin Nutr 2001; 20: 517) enabled us to assess the nutrition
of 1561patients on emergency admission to hospital. Patients acutely admitted to the 15 relevant medical, surgical and
orthopaedic wards were identied. Mid upper arm circumference (MUAC) measurements were obtained in 95% (848 m,
635f) patients. For clinical reasons, Body mass index (BMI) was assessable in only 44% patients (408 m, 285f). Data on
three month weight loss were obtainable in 509 patients. These measurements combined to demonstrate that 18.3% of
patients were undernourished (At least one of : BMIo20 kg/m2 or MUACo25 cm or loss of weight Z10%).There was a
close relationship between BMI and MUAC. Regression equations (excluding age)were for men : BMI=1.01 MUACF
4.7, (R2=0.76), and for women BMI=1.10  MUAC6.7, (R2=0.76). After adjustment for age, weight loss Z10% was the
most signicant of the three as a predictor of mortality. Among patients in whom weight loss was not recorded MUAC
was a signicant predictor of mortality either alone (P=0.002) or after adjustment for BMI (P=0.007), but BMI was not
signicant. All three measures, even when adjusted for age and sex, were poor predictors of hospital stay although
MUAC was signicant in the larger group with a MUAC measure (R2=0.7% Po0.001). MUAC correlates closely with
BMI, is easier to measure and predicts poor outcome better.
r 2003 Elsevier Science Ltd. All rights reserved.

interventions in the trial were small or non-existent, and


the condence intervals for difference in mean length of
stay for supplements vs placebo were narrow, (2 to 1.2
days) we used all the available data to assess the
measurements validity as predictors for hospital length
of stay and mortality.

Key words: body mass index; mid upper arm circumference; weight loss; anthropometrics; nutritional assessment; nutritional screening

Introduction
Body mass index (BMI) and percentage weight loss are
recommended and widely used to assess the nutrition of
patients entering hospital. However it is not always
possible to measure weight or height in the acutely ill
because patients may not be able to stand or leave their
bed. A nutritional supplementation trial (1) gave us the
opportunity to assess the nutrition of 1561 patients on
emergency admission to medical (1097), surgical (335)
and orthopaedic (129) services of the Royal London
Hospital, a teaching hospital in Londons socially
deprived East End. We sought to compare measurements of the mid upper arm circumference with the BMI
and the percentage weight loss.
The purpose of such screening measures is to identify
patients likely to benet from enhanced nutritional care
by improved outcome. They will thus be of most use in
the acutely hospitalized patient if they can be shown to
relate to outcome. Because the benet of the nutritional

Methods
Two research dietitians employed for the purposes of a
nutritional supplementation trial (1), tried to include
all consenting patients admitted as an emergency to
General Medicine, Surgery or Orthopaedics on Sundays
to Thursdays, from June 1997January 1999. Patients
admitted to the 15 relevant wards were identied by lists
generated by the admissions department. Patients
electively admitted to medical or orthopaedic wards
were excluded. Patients were, whenever clinically possible, weighed and their height and mid non-dominant
upper arm circumference (MUAC) measured. An
estimate of weight 3 months prior to admission was
made from the notes and by questioning the patient and
their relatives, and percentage weight loss calculated.
Height was measured using Seca model 220 ward
307

308

BMI, MUAC AND WEIGHT LOSS

stadiometers, weight using regularly checked ward


scales; using a non-stretch tape measure MUAC was
measured at the mid point between acromion and
olecranon.

minority of the patients, results are shown for the binary


variables BMI o20 (or not), MUAC o25 (or not); they
were also calculated for oMUAC equivalent to a BMI
of 20, adjusted for age. All analyses used Stata 7.

Statistical methods
Results

MUAC, BMI and weight loss were compared independently for their ability to predict either death during the
period of hospital admission using multiple logistic
regression, or length of stay in hospital using multiple
regression. The strong effects of age over 60, and being
male and over 60 were adjusted for length of stayage
over 60 was adjusted for mortality.
Analyses were performed among patients with measures for both BMI and MUAC for comparison
purposes, and also among patients with MUAC alone
to show whether results for MUAC held in a larger data
set. Analysis was repeated among patients in whom
weight loss of more than 10% was not recorded.
Because length of stay was not normally distributed
and had signicant outliers, the regressions were
performed using values truncated at 21 days (so that
patients staying longer than this were reclassied as 21
days), and also using loge (length of stay).
BMIs of 20 and 18.5 are standard measures of
thinness. BMI of less than 20 is widely accepted as
underweight (2), particularly in well-developed countries
and 18.5 is recommended as a practical lower limit for
most populations representing the third centile of a
population with median BMI of 23 (3).
In order to establish MUAC values equivalent to
BMIs of 18.5 and 20 two regressions were performed,
BMI on MUAC and MUAC on BMI, separately by sex
and adjusted for age. The averages of the two predicted
MUAC values associated with BMIs of 18.5 and 20 were
calculated for 55 year olds were taken as the equivalent
MUAC value.
Because we were interested in those who had lost
weight or were underweight or thin, and these were a

The sample analysed


For the statistical analysis of 1561 study members 1559
had an outcome measure. Of these 1559, 78 had no
MUAC, 867 had no BMI, and 11 had neither
measurement. Those missing both were not considered
in the analysis. In practice because of the clinical state
of the patient weight and/or height could often not be
measured. Thus MUAC measurements (Table 1)
were obtained in 95% (848 m, 635f ) patients and BMI
(Table 2) could be assessed in 44% (408 m, 285f). Weight
loss data were successfully obtained in 509 patients.
Weight loss information was not obtainable/available
for 1052 patients. Consistent with weight and height
being difcult to measure in the sickest patients, those
missing a BMI were more likely to die (Po0.01) as were
those missing weight loss information (Po0.1). By
contrast the smaller proportion of those missing a
MUAC measure were less likely to die (P 0:27).
In this dataset BMI differed slightly by sex (not quite
signicantly) but not by age. In contrast MUAC was not
associated with sex but was associated with age.

Prevalence of weight loss and low BMI and MUAC


The distribution of the BMI and MUAC measurements
are compared with representative UK general populations in Tables 1 and 2. These measures are compared in
their ability, alone or in combination, to demonstrate
undernutrition in this acutely ill population in Table 3.
Combined they demonstrate that 18.3% of patients were

Table 1 Distribution of mid upper arm circumference measurements compared with a general population (7)
Total n 1483
o25
2527.5
2730
3032.5
432.5

217
268
315
321
362

Male n 848

(14.6%)
(18.1)
(21.2%)
(21.6%)
(24.5%)

106
152
195
209
186

Gen pop male n 1158 (%)

(12.5%)
(17.9%)
(23%)
(24.7%)
(21.9%)

Female n 635

3
14
28
31
24

111
116
120
112
176

Gen pop female n 1161 (%)

(17.5%)
(18.3%)
(18.9%)
(17.6%)
(27.7%)

14
27
27
17
15

Table 2 Body mass index measurements: study population vs the general population (10)
BMI kg/m2

Total
Study group
n=692

r=20
42025
42530
430

116
254
208
114

(16.8%)
(36.7%)
(30%)
(16.5%)

Male
Study group
n=408
67
150
134
57

(16%)
(37%)
(33%)
(14%)

Female
Gen pop
n=3114 (%)
6
40
42
12

Study group
n=285
49
105
74
57

(17%)
(37%)
(26%)
(20%)

Gen pop
n=3430 (%)
8
47
29
16

CLINICAL NUTRITION

309

Table 3 Numbers and percentages classied as undernourished using three parameters alone or in combination: weight loss Z10%,
BMIo20 kg/m2, MUAC o25 cm
Number with data in
category n=1561

% of population
classied
undernourished

Number (%) classied


by BMI but not the
given measure

BMI alone
MUAC alone
wt loss alone

693
1483
508

115/1561 (7.4%)
217/1561 (13.9%)
50/1561 (3.2%)

41 (2.6%)
99 (6.3%)

143 (9.2%)

199 (12.7%)

33 (2.1%)
31 (2.0%)

BMI or MUAC
BMI or wt loss
MUAC or wt loss
any of three

1551
870
1541
1553

258/1561
149/1561
249/1561
286/1561

37 (2.4%)

137 (8.8%)

28 (1.8%)

(16.5%)
(9.5%)
(16.0%)
(18.3%)

40

40

35

35

MUAC

MUAC

Measure

30

25

20

20

20

25

30

35

15

20

25

30

35

BMI

BMI
Fig. 1. MUAC vs BMIall patients.

Number (%) classied


by weight loss but not
the given measure

30

25

15

Number (%) classied


by MUAC but not the
given measure

Fig. 2. MUAC vs BMIfemales.

undernourished (At least one of : BMIo20 kg/m2 or


MUACo25 cm or loss of weight Z10%).
40

Comparison of MUAC and BMI

The regression equations for men were:


BMI=1.02  MUAC+0.03  age6.7, Po0.0000
(R2=0.77).
and for women:
BMI=1.10  MUAC+0.023  age8.0, Po0.0000
(R2=0.76).
Removing the small contribution of age from
the equations demonstrates the relationship between
BMI and MUAC in all patients, males and females
(Figs. 13):
BMI=1.06 MUAC5.8 (all) Po0.0000 (R2=0.75).
BMI=1.01 MUAC4.7 (males) Po0.0000 (R2=0.76).
BMI=1.10
MUAC6.7
(females) Po0.0000
(R2=0.76).

MUAC

35

While weight loss can be seen in the overweight, BMI


and MUAC are comparable in that they only identify
people who are underweight or thin. Their relationship
was therefore of interest. The relationship between BMI
and MUAC was different in men and women (P 0:006
for sex) and was affected by age.

30

25

20

15

20

25

30

35

BMI
Fig. 3. MUAC vs BMImales.

A BMI of 20 was found to be equivalent to a MUAC


of 24.8 in a man of 55 years and 24.6 in a woman of
55 years (+0.265 for a 65-year old and pro rata for all
ages). The equivalent gures for a BMI of 18.5 were
MUAC of 23.2 for men and 23.0 for women. Thus, for
clinical purposes, especially when patients cannot easily
be weighed or their height measured MUAC-5 (cm)

310

BMI, MUAC AND WEIGHT LOSS

can be used to provide an approximate estimate of BMI


(kg/m2). The centile for these MUAC values are close to
the centiles for BMIs of 18.5 and 20.
If we regard BMI Z20 and MUAC Z25 as not thin,
BMI Z18.5 and o20 and MUAC Z23.5 and o25 cm
thin, and BMI o18.5 or MUAC o23.5 very thin, we
can compare how BMI and MUAC classied the
patients (Table 4). Though agreement in general was
reasonable seven patients classied not thin by BMI
were classied very thin by MUAC, and two patients
classied very thin by BMI were classied not thin by
MUAC.
Comparison of the effectiveness of BMI, percentage
weight loss and MUAC as indices of hospital
undernutrition
Tables 3 and 4 compare how the measures identify
patients as undernourished and Table 3 provides an

Table 4 Classication of patients as not thin, thin or very thin by


MUAC and BMI in patients with both measurements

MUAC Z25 cm
MUAC Z23.5 and
o25 cm (thin)
MUAC o23.5 cm
(very thin)

BMIZ20
kg/m2

BMIZ18.5 and
o20 (thin)

BMIo18.5
(very thin)

474
17

29
11

2
9

13

29

estimate of the proportion which would be classied


undernourished (or missed) on the basis of combinations of the other indices. MUAC would classify more
patients as undernourished than BMI and fewer would
be missed.
Predictive power on hospital mortality and length of stay
of categorization by weight loss, BMI or MUAC
Hospital mortality: In Table 5a the power of weight loss,
equal to or in excess of 10%, to predict outcome during
the index hospital admission can be seen. Though
weight loss could only be estimated in a minority of
patients it was a relatively strong predictor of mortality
(P 0:001).
The ability of low BMI or MUAC to predict outcome
among those patients in whom weight loss was either
not present or not measurable is shown in Table 5b. In
590 subjects with both BMI and MUAC measurements
MUAC o25 predicted mortality signicantly when
corrections were made for age. BMI showed a similar
trend but its prediction was weaker and statistically not
signicant. There was a statistically signicant effect of
MUAC after adjustment for BMIOdds ratio=7.7,
95% condence interval (1.8, 34), P 0:007: Using cut
off points for MUAC adjusted for age and sex and
equivalent to a BMI of 20 gave no greater predictive
power than using MUAC of 25. As before it was a better
predictor than BMI.

Table 5
(a) Predictive power of being categorised undernourished by weight loss Z10%
Subjects with valid weight loss measures n-509

All subjects n=1559

Length of stay (days) truncated at 21 days and adjusted being older than 60 years and being male and over 60
Coefcient
P-value
Coefcient
P-value
added R2
(95% CI)
(95%CI)

added R2

Estimatedn
values

Weight loss Z10%

0.066

0.2%

1053

P-value

added
pseudo-R2

Estimatedn
values

0.015

1.1%

1053

Length of stay (days) truncated at 21 days and adjusted being older than 60 years and being male and over 60
Coeff
P-value
coeff
P-value
added R2
(95% CI)
(95% CI)

added R2

Estimatedn
values

BMI o20
MUAC o25

1.85 (0.13, 3.58)

0.035

Death in hospital, adjusted for being older than 60 years


Odds ratio
P-value
(95% CI)
Weight loss Z10%

7.09 (2.15, 23.4)

0.001

0.8%
added pseudo-R2
7.0%

1.61 (0.10, 3.32)


Odds ratio
(95% CI)
3.41 (1.26, 9.21)

(b) Predictive power of being categorised with low MUAC or BMI in those without a measured weight loss of 410%
Subjects with both BMI and MUAC measures n=590

0.50 (0.72, 1.73)


1.01 (0.26, 2.28)

0.42
0.12

Death in hospital, adjusted for being older than 60 years


Odds ratio
P-value
(95% CI)
BMI o20
MUAC o25
n

2.80 (0.82, 9.60)


6.43 (2.00, 20.1)

0.10
0.002

0.1%
0.4%
added pseudo-R2
2.0%
7.5%

All subjects n=1509

0.55 (1.79, 0.69)


1.55 (0.65, 2.46)

0.38
o0.001

0%
0.7%

868
78

Odds ratio
(95% CI)

P-value

added
pseudo-R2

Estimatedn
values

0.63
o0.001

0.5%
4.5%

1.30 (0.45, 3.72)


4.18 (2.33, 7.47)

Number of cases with missing informationreclassied as weight losso10%, BMIZ20 and MUAC Z25 respectively.

868
78

CLINICAL NUTRITION

MUAC had the advantage of being measurable in


many more patients than BMI or weight loss.When this
is taken into account its predictive power was still more
statistically signicant (Odds ratio 4.18, Po0.0001
Table 5b).
Length of stay: Among those in whom it could be
measured weight loss was a weak but signicant
predictor of length of stay (Table 5a). If analysis is
restricted to patients with data for both BMI and
MUAC (Table 5b), neither measure predicted length of
stay signicantly; but if the larger number of patients
(all subjects column) in whom MUAC could be
measured is analysed MUAC predicted truncated length
of stay weakly but signicantly (added R2=0.7%,
Po0.001). The adjusted MUACs equivalent to a BMI
of 20 gave no better predictions than MUAC o25.
In summary: Thus weight loss and MUAC were
signicant predictors of mortality and weaker predictors
of length of stay. This was not demonstrated for BMI.
The use of other truncations (14 and 100 days) gave
similar patterns with P-values similar for 14 days but
smaller for 100 days because the variability in length of
stay was considerably larger. Similar effects were found
when loge (length of stay) was used as the outcome.

Discussion
Undernutrition in hospital patients is often unrecognized and there is a need for a simple means of screening
to facilitate targeted nutritional intervention. BMI has
been emphasized as an objective anthropometric criterion
(4). The British Association for Parenteral and Enteral
Nutrition (BAPEN) recommends that the measurements
used for screening are based upon the patients weight for
height (BMI) and their percentage weight loss (5). This
study conrms the high prevalence of undernutrition
among patients admitted to acute services though the
proportions with weight loss, low BMI, or low MUAC
are less than those observed elsewhere (6).
An important nding in this study was the difculty
our dietitians had in obtaining BMI soon after admission in many of these acutely ill patients. This was
despite dedicated time, and strong motivation by virtue
of the clinical study being undertaken. The reason was
nearly always that patients could not be taken out of
their beds for weighing or could not stand for height
measurement. Thus BMI has major drawbacks in the
early assessment of the acutely ill.
Our study was conned to the rst days of acute
hospital admission. During early acute illness missing
data can be expected to be associated with poor outcome
because we cannot estimate BMI and weight loss in those
too sick to be weighed or have their height measured.
Missing values can be established from Table 3.
Reecting the difculties of measuring the sickest
patients, missing BMI in this cohort was clearly

311

associated with longer stays and a greater risk of death


in hospital (OR=2.4, Po0.005 for all outcomes),
whereas missing weight loss data was marginally
associated with death OR=1.7 P=0.098. Missing
MUAC data by contrast, a much less common problem,
although not signicant, was associated with better
outcomesometimes patients had gone home before it
could be measured.
Even though weight loss could often not be assessed,
those with weight loss Z10% were at signicantly
higher risk of death than those with weight loss o10%
or no information (OR=3.4 P 0:015) and the effect
among those with data was even stronger (OR 7.1
P 0:001). The MUAC measures were signicant
predictors of death whether (P 0:001) or not
(P 0:002) those with missing data were included in
the analysis. In this study BMI o20 was not a
signicant predictor of hospital mortality.
BMI o20 was not a signicant predictor of length of
stay either including or excluding patients who could
not be classied by this measure. Weight loss Z10%
was a weak (R2=0.8%, signicant (P 0:035) predictor
if only patients with this measure were analysed; it
became still weaker and statistically insignicant
when all patients were included. MUAC was a weak
(R2=0.7%) signicant (Po0.001) predictor of length of
stay if all patients with this measure were considered.
In this study therefore an estimated weight loss of 10%
or more is conrmed as an important nutritional measure
with signicant implications for hospital outcome.
However its use is limited by a need to be able to weigh
the patient and to know previous weights. BMI has
similar disadvantages, depending on the need to weigh
the patient and measure height which are often not
possible or appropraite in the early stages of acute
admission to hospital. MUAC, for which British (7) and
American (8, 9) norms exist, is measured with inexpensive
equipment readily supplied to every ward, and does not
require the patient to get out of bed or stand. While it is
practical to train staff to measure MUAC, it is not
practical or cost effective to place delicate Holtain skin
calipers in non-specialist clinical areas subject to rapid
staff turnover, nor is it likely to be easy to train a large
nursing staff to use such calipers accurately. We have
shown no advantage in using calculated MUAC values
for age and sex appropriate to the BMI cut off points;
hence the use of a single MUAC measure can be
recommended which requires no calculations by busy
nurses and reduces the risk of error compared for
example with derived measurements like arm muscle
circumference or BMI. MUAC emerges as a useful
measure of nutritional status which is applicable in nearly
all acutely ill patients. It correlates closely with BMI but
misses fewer patients and, unlike BMI, predicts outcome
signicantly. It is a simple measure which requires neither
mathematical derivation nor expensive equipment and
we believe it should be used much more frequently as a
routine assessment of nutritonal status in acutely ill

312

BMI, MUAC AND WEIGHT LOSS

patients. It is a simple index which correlates closely to


BMI and which helps draw attention to undernutrition in
acutely ill patients in whom measurement of weight and
height may be inappropriate or impossible.

3.

Acknowledgements

4.

Ms Shona Vlaming, Ms Anzonette Biehler, Dr Crawford PJamieson,


Dr Santanu Chattophadhyay, Ms Claire Archer, and Ms Katie
Durman were all active investigators in the original study from which
the data for this analysis are derived. Drs O.A. Obeid and Sandra
Warrington and Ms A. Farrell also contributed to that study.
The study was supported by grants from the Responsive Funding
Group of the North Thames Regional Health Authority NHS R&D
and Abbott Laboratories Ltd. We thank the medical, surgical and
orthopaedic consultant staff of Barts and the London NHS Trust for
allowing us to include their patients and the ward nurses of the fteen
wards taking part.

2.

5.
6.
7.
8.
9.
10.

References
1. Vlaming S, Biehler A, Hennessy E M et al. Should the food intake
of patients admitted to acute hospital services be routinely
Submission date: 25 September 2002 Accepted: 20 January 2003

supplemented? A randomised placebo controlled trial. Clin Nutr


2001; 20: 517526
Garrow J S, Webster J. Quetelets index (W/H2 ) as a measure of
fatness. Int J Obesity 1985; 9: 147153
James W, Francois P J. The choice of cut-off point for
distinguishing normal body weights from underweight or chronic
energy deciency in adults. Eur J Clin Nutr 1994; 48(suppl 3):
S179S184
James W P T, Ferro-Luzzi A, Waterlow JC. Denition of chronic
energy deciency in adults. Report of working party of IDECG.
Eur J Clin Nutr 1988; 42: 969981
Sizer T (ed.). Standards and Guidelines for Nutritional Support of
Patients in Hospitals. Maidenhead: BAPEN, 1996.
McWhirter J, Pennington C. Incidence and recognition of
malnutrition in hospital. Br Med J 1994; 308: 945948
Gregory J, Foster K, Tyler H, Wiseman M. The Dietary and
Nutritional Survey of British Adults. London: HMSO, 1990
Frisancho A R. New norms of upper limb fat and muscle areas
for assessment of nutritional status. Am J Clin Nutr 1981; 34:
25402545
Bishop C W, Bowen P E, Ritchey S J. Norms for nutritional
assessment of American adults by upper arm anthropometry. Am
J Clin Nutr 1981; 34: 25302539
Breeze E, Maidment A, Bennett N, Flatley J, Carey S. Health
Survey for England, 1992. London: HMSO, 1994