Abstract
Background
Using Mid-Upper Arm Circumference (MUAC) in place of Body Mass Index (BMI) may be preferable for identifying malnutrition in various situations, especially in resource-poor settings. The primary objective of this study was to determine MUAC cut-offs corresponding to BMI < 20 kg/m2, < 18.5 kg/m2, < 17 kg/m2 and < 16 kg/m2 in a cohort of patients with tuberculosis from West Africa. The secondary objective was to examine the prognostic value of MUAC cut-offs in predicting mortality at two months of tuberculosis treatment. The aim was to propose unisex MUAC cut-offs that could be used to identify malnutrition and to identify patients at increased risk of dying.
Methods
This prospective, observational cohort study was conducted from 2003 to 2022. Diagnostic accuracy of MUAC to identify BMI cut-offs was assessed for every 0.5 cm in the range < 20.0 cm to < 27.0 cm. Area under the receiver operating characteristic curves (AUROCCs), sensitivity (SENS), specificity (SPEC), false negative and false positive were determined. Cox proportional hazard model was used to examine the association between MUAC cut-offs and mortality.
Results
Data from 2,098 patients were included. MUAC was found to be excellent in its ability to identify BMI cut-offs with AUROCCs close to 0.9. The MUAC cut-offs that best corresponded to BMI < 20 kg/m2, < 18.5 kg/m2, < 17 kg/m2 and < 16 kg/m2 were < 25 cm (SENS 75.1%, SPEC 83.0%), < 24 cm (SENS 75.1%, SPEC 80.0%), < 23 cm (SENS 77.9%, SPEC 83.2%) and < 22.5 cm (SENS 80.3%, SPEC 81.9%). Mortality risk significantly increased with MUAC values < 22 cm.
Conclusions
MUAC cut-off < 25 cm was proposed to be used in place of BMI < 20 kg/m2 to identify malnourishment and MUAC cut-off < 22 cm was proposed to identify patients at increased risk of dying and thus in need of further attention.
Supplementary Information
The online version contains supplementary material available at 10.1186/s12889-025-25880-6.
Keywords: MUAC, BMI, Anthropometry, Nutritional status, Tuberculosis, Mortality
Introduction
Body Mass Index (BMI) cut-offs are used in various settings. This includes screening populations for malnutrition and contributing to malnutrition diagnosis assessments like The Global Leadership Initiative on Malnutrition (GLIM) [1]. It is also used to determine nutrition therapy eligibility in famine relief or among patients with tuberculosis (TB) and human immunodeficiency virus (HIV) in resource-poor countries.
Measurement and calculation of BMI require good-quality, calibrated equipment and acceptable health literacy and numeracy skills of healthcare workers. It is not possible to assess BMI in the most ill patients, who cannot stand up for body height and weight measurements, or patients with oedema. In resource-poor settings, BMI is therefore not always the optimal assessment tool.
Mid-upper arm circumference (MUAC), on the other hand, is a faster, and simpler measure of nutritional status usable by a broader population. However, globally recognized, standard MUAC cut-offs for adults do not exist. In a meta-analysis from 2019, Tang et al. [2, 3] proposed a range of global MUAC cut-offs (≤ 23.5 cm to ≤ 25.0 cm) corresponding to BMI < 18.5 kg/m2. MUAC cut-offs corresponding to other commonly used BMI cut-offs are less examined, and information on the ability of various MUAC cut-offs to predict morbidity or mortality is scarce. The latter is important when choosing clinically relevant MUAC cut-offs.
The primary objective of this study was to determine MUAC cut-offs corresponding to BMI cut-offs of < 20 kg/m2, < 18.5 kg/m2, < 17 kg/m2, and < 16 kg/m2, in a cohort of patients with TB from West Africa. The secondary objective was to examine the prognostic value of MUAC cut-offs in predicting mortality during the first two months of TB treatment.
The aim was to propose unisex MUAC cut-offs that could be used to identify malnutrition and to identify patients at increased risk of dying and thus in need of further attention. Unisex cut-offs were chosen because of the simplicity, which is often preferred in a clinical or urgent setting. Analyses were stratified by sex and HIV co-infection to explore any differences and the potential consequences hereof. Alternatives to multiple BMI cut-offs were explored corresponding to the traditional use of BMI cut-offs in the clinical setting: BMI cut-offs of < 20 kg/m2 and < 18.5 kg/m2 are used in the GLIM criteria for the diagnosis of malnutrition [1], and BMI cut-offs of < 17 kg/m2 and < 16 kg/m2 are used to identify individuals at increased risk of either morbidity or mortality. In this paper, the term malnutrition will be used synonymously to underweight, recognizing that underweight is only one of the manifestations of malnutrition.
Methods
Study design and setting
The current prospective, observational study was conducted at the Bandim Health Project (BHP) [4], a health and demographic surveillance site located in the capital of Guinea-Bissau, West Africa. The BHP covers a well-defined study area of six suburbs in Bissau with a population of approximately 100,000. Patients with TB are registered at the three health centres located in the study area and at the national TB reference hospital located adjacent to the study area. The long-running Bandim TB cohort was established in 1996. Overall TB incidence rate in the area was 294 per 100,000 person-years in 2004 and 273 in 2020 [5].
Participants
Adults aged 18–69 years living in the study area, who were newly diagnosed with drug-susceptible TB between November 2003 and August 2022 were considered for the study. Only patients with anthropometric measurements taken during the first two weeks of treatment were eligible. This selection was made to evaluate the predictive ability of baseline MUAC measurements for assessing the risk of mortality at 2 months of treatment. Exclusion criteria were: (i) pregnancy, (ii) missing MUAC values and (iii) extreme values of MUAC or BMI measurements (identified by scatterplot). For the primary objective, patients with oedema were excluded. Oedema affects body weight and BMI, thus questioning the validity of those measurements. However, oedema is often a symptom of more advanced disease, and patients with oedema were therefore included in the mortality analyses for the secondary objective. Oedema may, in extreme cases, influence MUAC measurements, however, this was considered a highly unlikely occurrence among study participants.
Variables
Body height was measured to the nearest cm using a roll-up tape measure with the patient barefoot. Body weight was measured to the nearest 0.1 kg using a digital bathroom scale with the patient barefoot and wearing minimal clothing. MUAC was measured at the midpoint between the acromion and olecranon on the left arm, with the arm hanging loosely. A non-stretchable tape measure (TALC, Harpenden, UK) measuring to the nearest 2 mm was used. BMI was calculated as body weight in kg divided by body height in m squared. Classification of BMI groups was according to World Health Organization guidelines [6]. TB treatment outcome (either cured, treatment completed, treatment failure, death, loss to follow-up or not evaluated) and date of outcome was assessed at the end of treatment [7]. Patients were screened for HIV co-infection using Determine HIV-1/2 assay (Abbott Laboratories, Tokyo, Japan) and positive tests were confirmed with Bioline HIV 1/2 assay (Abbott Laboratories, Tokyo, Japan) or First response HIV 1–2.0 card test (Premier Medical Corporation, Sarigam, India).
Sample size
Sample size was calculated using the diagsampsi command in STATA. Estimating a specificity of 80% with a 0.05 width of the 95% Confidence Interval and a malnutrition prevalence of 51% [8], would require a sample size of minimum 502 patients.
Statistical methods
Data were entered in Dbase 5.0 (dataBased Inc, Vestal, NY, USA), and statistical analyses were performed in Stata SE 11.2 (Stata Corporation, College Station, TX, USA). Difference in prevalence of categorical variables between men and women (sex assigned at birth) were assessed using the χ2 test. Continuous variables with a normal distribution were compared using the Student’s t-test, and continuous variables with a non-normal distribution were compared using the Mann-Whitney Wilcoxon rank test.
Scatterplot was used to examine the correlation between MUAC and BMI. Locally weighted scatterplot smoothing (lowess) regression was used to fit a smooth curve through the scatterplot and create a visual presentation of the potential curvilinear relationship between MUAC and BMI. Pearson’s correlation coefficient between MUAC and BMI was calculated.
The diagnostic accuracy of MUAC to identify BMI < 20 kg/m2, < 18.5 kg/m2, < 17 kg/m2 and < 16 kg/m2 was assessed. Sensitivity (SENS), Specificity (SPEC), False Negative (FN) and False Positive (FP) were calculated for every 0.5 cm MUAC cut-offs in the range < 20.0 cm to < 27.0 cm. Receiver operating characteristic (ROC) curves were plotted and area under the receiver operating characteristic curves (AUROCC) were determined. Analyses were stratified by sex and HIV-coinfection. A specificity and sensitivity of minimum 70% were considered acceptable, however, the cut-off with the highest sensitivity while maintaining a minimum specificity of 80% was considered the most optimal cut-off value.
Cox proportional hazard model was used to examine the association between a series of MUAC cut-offs (< 20 cm; [20 < 21) cm; [21 < 22) cm; [22 < 23) cm; [23 < 24) cm; [24 < 25) cm) and mortality during the first two months of TB treatment. MUAC between ≥ 25 cm to < 28.5 cm was considered the reference group. Overweight or obese patients with MUAC > 28.5 cm were not included in mortality analyses. Exit date was defined as the date of death (event of interest), transfer to another treatment centre, transfer to a drug-resistant treatment regimen, or date last seen alive, whichever came first during the first 60 days of TB treatment. Mortality was a priori assumed to be affected by age, sex, oedema and HIV-coinfection, and analyses were adjusted accordingly.
Results
Between November 2003 and August 2022, 3,264 patients with drug-susceptible TB aged 18–69 years were registered in the study area. A flowchart of inclusions and exclusions is presented in Fig. 1. In total, 2,098 were included in the study, of which 84 had oedema.
Fig. 1.
Flowchart
Table 1 presents baseline characteristics and anthropometric values as well as treatment outcomes of the study population both in total and stratified by sex. Malnutrition prevalence was high, and almost one out of four patients were known to be HIV-coinfected. Of the 137 patients who died, 61 (45%) did so during the first two months of treatment (33 men, 28 women).
Table 1.
Baseline characteristics and treatment outcome of the study population in total and stratified by sex
| Characteristic | Total (n = 2,098) |
Men (n = 1,409) |
Women (n = 689) |
p-valuea |
|---|---|---|---|---|
| Age, years, median (IQR) | 33 (26;42) | 33 (26;42) | 31 (25;43) | 0.485 |
| MUAC, cm, mean (sd) | 24.1 (± 3.2) | 24.3 (± 2.9) | 23.7 (± 3.7) | < 0.001 |
| Body weight, kg, mean (sd) | 52.8 (± 9.1) | 54.8 (± 8.3) | 48.6 (± 9.2) | < 0.001 |
| BMI, kg/m2, mean (sd) | 18.6 (± 2.9) | 18.5 (± 2.6) | 18.7 (± 3.4) | 0.078 |
| BMI categories, kg/m2, n (%) | ||||
|
<16 kg/m2 [16 < 17) kg/m2 [17 < 18.5) kg/m2 [18.5 < 20) kg/m2 [20 < 25) kg/m2 [25 < 30) kg/m2 ≥30 kg/m2 Missing (n) |
325 (16) 255 (13) 486 (24) 433 (22) 445 (22) 54 (3) 8 (0) 92 |
199 (15) 179 (13) 336 (25) 326 (24) 293 (22) 20 (1) 3 (0) 53 |
126 (19) 76 (12) 150 (23) 107 (17) 152 (23) 34 (5) 5 (1) 39 |
< 0.001 |
| HIV-status, n (%) | ||||
|
Uninfected Infected Unknown |
1,422 (68) 493 (23) 183 (9) |
1,034 (74) 257 (18) 118 (8) |
388 (56) 236 (34) 65 (10) |
< 0.001 |
| Type of TB, n (%) | ||||
|
Bacteriologically confirmed Clinically diagnosed Extrapulmonary TB |
1,613 (77) 436 (21) 49 (2) |
1,116 (79) 261 (19) 32 (2) |
497 (72) 175 (25) 17 (3) |
0.002 |
| TB treatment outcome, n (%) | ||||
|
Cured Treatment completed Treatment failure Loss-to-follow-up Death Not evaluated (n) |
980 (48) 648 (32) 14 (1) 241 (12) 137 (7) 78 |
659 (48) 432 (32) 10 (1) 188 (14) 69 (5) 51 |
321 (48) 216 (33) 4 (1) 53 (8) 68 (10) 27 |
< 0.001 |
IQR Interquartile Range, sd Standard Deviation, MUAC Mid-Upper Arm Circumference, BMI Body Mass Index, HIV Human Immunodeficiency Virus; TB Tuberculosis
ap-value for difference between men and women
Correlation between MUAC and BMI
MUAC was highly correlated with BMI, with a Pearson’s correlation coefficient of 0.79. The correlation between MUAC and BMI stratified by sex and HIV-coinfection is illustrated in Additional Files 1 and 5. A curvilinear relationship between MUAC and BMI was observed, with a slope increase for MUAC values > 29 cm. Women tended to have slightly higher MUAC values relative to their BMI compared with men.
MUAC cut-offs identifying BMI < 20 kg/m2, < 18.5 kg/m2 < 17 kg/m2, and < 16 kg/m2
ROC curves and AUROCCs illustrated excellent diagnostic ability of MUAC identifying low BMI, with all AUROCCs close to 0.90 (Fig. 2). ROC curves and AUROCCs stratified by sex and HIV-coinfection are available in Additional Files 2 and 6. SENS, SPEC, FN and FP for every 0.5 cm MUAC cut-off in the range < 20.0 cm to < 27.0 cm is presented in Table 2 (Data stratified by sex and HIV-coinfection is presented in Additional Files 3, 4, 7 and 8).
Fig. 2.
ROC curves and AUROCCs for MUAC cut-offs identifying BMI cut-offsROC Receiver Operating Characteristics; AUROCC Area Under the Receiver Operating Characteristics Curve; MUAC Mid-Upper Arm Circumference; BMI Body Mass Index
Table 2.
Diagnostic accuracy of every 0.5 cm MUAC cut-off to identify BMI cut-offs
| BMI < 20 kg/m2 | BMI < 18.5 kg/m2 | BMI < 17 kg/m2 | BMI < 16 kg/m2 | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| MUAC | SENS (%) | SPEC (%) | FN (%) | FP (%) | SENS (%) | SPEC (%) | FN (%) | FP (%) | SENS (%) | SPEC (%) | FN (%) | FP (%) | SENS (%) | SPEC (%) | FN (%) | FP (%) |
| < 20.0 | 11.1 | 100 | 72 | 0 | 15.5 | 99.9 | 49 | 1 | 26.9 | 99.3 | 23 | 6 | 41.5 | 98.2 | 10 | 19 |
| < 20.5 | 15.1 | 99.6 | 72 | 1 | 20.9 | 99.5 | 47 | 2 | 35.9 | 98.6 | 21 | 9 | 52.3 | 96.6 | 9 | 25 |
| < 21.0 | 19.6 | 99.0 | 71 | 2 | 26.7 | 98.6 | 46 | 4 | 44.3 | 97.1 | 19 | 14 | 61.9 | 94.2 | 7 | 33 |
| < 21.5 | 26.4 | 98.8 | 69 | 1 | 35.8 | 98.0 | 43 | 5 | 55.5 | 94.5 | 16 | 20 | 71.1 | 89.9 | 6 | 42 |
| < 22.0 | 28.7 | 98.4 | 68 | 2 | 38.6 | 97.1 | 42 | 6 | 58.6 | 93.1 | 15 | 22 | 73.9 | 88.2 | 5 | 45 |
| < 22.5 | 37.0 | 97.6 | 66 | 2 | 48.5 | 94.8 | 38 | 9 | 69.1 | 88.4 | 12 | 29 | 80.3 | 81.9 | 4 | 54 |
| < 23.0 | 44.8 | 96.3 | 63 | 3 | 57.6 | 91.8 | 34 | 11 | 77.9 | 83.2 | 10 | 35 | 85.5 | 75.4 | 4 | 60 |
| < 23.5 | 55.2 | 93.1 | 59 | 4 | 68.6 | 86.0 | 29 | 15 | 86.7 | 74.8 | 7 | 42 | 91.4 | 66.3 | 2 | 66 |
| < 24.0 | 62.4 | 89.4 | 55 | 5 | 75.1 | 80.0 | 26 | 19 | 90.7 | 67.5 | 5 | 47 | 92.9 | 59.1 | 2 | 69 |
| < 24.5 | 71.8 | 84.0 | 50 | 7 | 83.5 | 71.6 | 21 | 23 | 94.1 | 57.2 | 4 | 53 | 95.4 | 49.6 | 2 | 73 |
| < 25.0 | 75.1 | 83.0 | 47 | 7 | 86.5 | 69.3 | 18 | 24 | 95.0 | 53.7 | 4 | 55 | 96.3 | 46.6 | 2 | 74 |
| < 25.5 | 81.7 | 76.3 | 41 | 9 | 91.3 | 60.4 | 14 | 28 | 97.1 | 45.2 | 3 | 58 | 97.5 | 38.9 | 1 | 76 |
| < 26.0 | 86.1 | 70.4 | 37 | 10 | 93.7 | 53.0 | 12 | 31 | 97.6 | 38.6 | 2 | 61 | 98.2 | 33.3 | 1 | 78 |
| < 26.5 | 91.4 | 59.2 | 30 | 13 | 96.4 | 41.6 | 9 | 35 | 98.1 | 29.3 | 3 | 64 | 98.2 | 25.2 | 1 | 80 |
| < 27.0 | 94.1 | 52.7 | 25 | 15 | 97.5 | 34.5 | 8 | 37 | 98.6 | 24.0 | 3 | 65 | 98.5 | 20.8 | 1 | 81 |
MUAC Mid-Upper Arm Circumference, BMI Body Mass Index, SENS Sensitivity, SPEC Specificity, FN False Negative, FP False Positive
The MUAC cut-off with the highest SENS for a SPEC ≥ 80% corresponding to BMI < 20 kg/m2 was < 25 cm (acceptable range with a minimum SENS and SPEC of 70% were < 24.5 cm to < 26 cm). The MUAC cut-offs that best corresponded to BMI < 18.5 kg/m2, < 17 kg/m2 and < 16 kg/m2 were < 24 cm (acceptable range < 24 cm to < 24.5 cm), < 23 cm (acceptable range < 23 cm to < 23.5 cm) and < 22.5 cm (acceptable range < 21.5 cm to < 23 cm) respectively. Cut-offs for men were equivalent to the cut-offs for men and women combined, however, the acceptable ranges were shifted 0 to 0.5 cm toward higher MUAC cut-offs. Cut-offs for women were consequently 0.5 cm lower than the best estimated cut-offs for men and women combined. Correspondingly, the acceptable ranges were shifted 0 to 1.5 cm toward lower MUAC cut-offs. MUAC cut-offs for patients with TB living with HIV differed slightly; the MUAC cut-off corresponding to BMI < 18.5 kg/m2 was 0.5 cm lower and the MUAC cut-off corresponding to BMI < 16 kg/m2 was 1 cm lower than the unisex cut-offs. The acceptable ranges were also shifted towards slightly lower MUAC values.
The best estimated MUAC cut-off (< 25 cm) for BMI < 20 kg/m2 showed a high proportion of FN (47%) but a low proportion of FP (7%). Contrary to this, the best estimated MUAC cut-off (< 22.5 cm) for BMI < 16 kg/m2 showed a low proportion of FN (4%) and a high proportion of FP (54%). Thus, using MUAC < 25 cm in place of BMI < 20 kg/m2 would result in several patients with BMI below 20 kg/m2 not being considered malnourished. Using MUAC in place of BMI < 16 kg/m2 would result in several patients with BMI above 16 kg/m2 being considered malnourished.
Validation for mortality
Overall, for every 1.5 kg/m2 BMI point decrease (from 20 kg/m2 to 18.5 kg/m2 to 17 kg/m2) there was a 1 cm decrease in the estimated optimal MUAC cut-off (from 25 cm to 24 cm to 23 cm). For the purpose of mortality analyses, this trend was assumed to continue with decreasing BMI and MUAC values. Hazard ratio for mortality risk was assessed after 2 months of TB treatment for each 1 cm MUAC cut-off point from < 25 cm to < 20 cm, see Table 3. Patients with MUAC < 22 cm had a significantly increased risk of dying compared with patients with MUAC between ≥ 25 cm to < 28.5 cm. The risk of death increased with decreasing MUAC values.
Table 3.
MUAC cut-offs as a prognostic factor for death at 2 months of tuberculosis treatment
| MUAC | Patients, n (%) | Deaths, n (%) | Crude HRa (95% CI) |
Adjusted HRa, b (95% CI) |
|---|---|---|---|---|
| < 20 cm | 190 (10) | 19 (31) | 9.82 (4.13–23.35) | 6.83 (2.80–16.66) |
| [20 < 21) cm | 143 (7) | 9 (15) | 5.91 (2.20–15.87) | 4.89 (1.81–13.23) |
| [21 < 22) cm | 151 (8) | 7 (12) | 4.29 (1.50–12.22) | 3.70 (1.29–10.62) |
| [22 < 23) cm | 265 (14) | 5 (8) | 1.72 (0.55–5.42) | 1.59 (0.50–5.00) |
| [23 < 24) cm | 310 (16) | 7 (12) | 2.08 (0.73–5.92) | 2.00 (0.70–5.74) |
| [24 < 25) cm | 226 (12) | 6 (10) | 2.43 (0.82–7.24) | 2.43 (0.82–7.23) |
| [25 < 28.5) cm | 637 (33) | 7 (12) | Reference | Reference |
MUAC Mid-Upper Arm Circumference, HR Hazard Ratio, CI Confidence Interval
aA total of 1,922 patients and 60 deaths: time at risk analysed was 3,723 person-months
bAdjusted for sex, age, oedema and Human Immunodeficiency Virus-coinfection
Discussion
The current study found that MUAC was excellent in its ability to identify BMI cut-offs and that mortality risk increased with decreasing MUAC values. The unisex MUAC cut-offs that best corresponded to BMI < 20 kg/m2, < 18.5 kg/m2, < 17 kg/m2 and < 16 kg/m2 were < 25 cm, < 24 cm, < 23 cm and < 22.5 cm respectively. Best estimates for women were consequently 0.5 cm lower than best estimates for men. Best estimates for patients with TB living with HIV were 0 cm to 1 cm lower than unisex best estimates. These small sex- and disease-related differences may lead to underestimation of malnutrition among women and patients living with HIV and overestimation of malnutrition among men when using MUAC cut-offs. However, the differences were minimal and did not justify proposing specific sex- and disease-stratified cut-offs.
Choosing a cut-off is a trade-off between a high sensitivity and a high specificity. A high sensitivity is usually preferred for screening purposes to identify as many at-risk individuals as possible. In contrast, a high specificity is usually preferred for diagnostic purposes, so that only the ones in need of treatment receives treatment. A comparison of cut-offs proposed by different studies using different sensitivity and specificity trade-offs can therefore be challenging.
Tang et al. compiled 20 datasets from Africa (Guinea-Bissau, Namibia, South Africa, Malawi, Zambia), USA, Argentina and Asia (India, Bangladesh, Vietnam), totalling 13,835 subjects [2, 3]. They conducted an individual participant data meta-analysis and proposed a MUAC cut-off ≤ 24 cm (SENS 84.1%, SPEC 83.2%) corresponding to BMI < 18.5 kg/m2. The proposed cut-off is in agreement with the current study as well as results from studies from South Africa [9] (< 23.7 cm, SENS 89.3%, SPEC 82.9%), India [10] (≤ 23.2 cm, SENS 89.0%, SPEC 82.0%) and Spain [11] (≤ 24 cm, SENS 85.8%, SPEC 84.4%). Others found higher MUAC values corresponding to BMI < 18.5 kg/m2: <25.5 cm (SENS 77.0%, SPEC 79.6%) in African male detainees [12] and ≤ 25.5 cm (SENS 96.0%, SPEC 54.0%) in South Sudan [13]. A few studies have also examined MUAC cut-offs corresponding to BMI < 16 kg/m2. A study on African male detainees [12] found < 24 cm (SENS 80.9, SPEC 86.3), a study from South Africa [9] proposed < 22.6 cm (SENS 91.6%, SPEC 91.7%), and a study from South Sudan [14] proposed < 22.5 cm (SENS 100%, SPEC 33%). Two of the studies valued a different sensitivity and specificity trade-off and thus proposed different best estimated cut-off values for use than the ones reported here [11, 12].
Only one other study has examined MUAC cut-offs corresponding to BMI cut-offs in a population of patients with TB. White et al.[15] found that MUAC cut-offs of 20.5 cm (SENS 89%, SPEC 84%) for men and 18.5 cm (SENS 91%, SPEC 89%) for women corresponded to BMI < 17 kg/m2 and MUAC cut-offs of 19.5 cm (SENS 89%, SPEC 82%) for men and 18 cm (SENS 100%, SPEC 86%) for women corresponded to BMI < 16 kg/m2. Their cut-offs were lower than the cut-offs identified in the current study. A possible explanation for this could be ethnicity.
To summarize the findings from the abovementioned studies, MUAC values between ≤ 23.2 cm and ≤ 25.5 cm were found to correspond to BMI < 18.5 kg/m2 and MUAC values between < 22.5 cm and < 24 cm were found to correspond to BMI < 16 kg/m2 (except for in the study by White et al.[15]). Those cut-offs are comparable to the cut-offs identified in the current study. The overlap of MUAC values best corresponding to the two BMI cut-offs of < 18.5 kg/m2 and < 16 kg/m2 supports the observation from the current study, that there are only small differences between MUAC cut-offs for adjacent BMI cut-offs (1 cm MUAC difference for every 1.5 kg/m2 BMI point difference). Thus, it does not appear feasible or clinically relevant to use MUAC cut-offs to distinguish between mild, moderate, and severe malnutrition. Rather, it should be considered to choose only a few MUAC cut-offs that align with key BMI cut-offs to identify malnutrition and at-risk patients.
In a paper by Lee et al.[16], examining the same patients with TB as White et al., they found a significant increased odds ratio for death for men with MUAC < 20.5 cm compared with ≥ 20.5 cm at 28 days of hospitalization. The odds ratio for death for women was not statistically significant. Gustafson et al.[17] examined mortality at 8 months among patients with TB living in Guinea-Bissau in 1996–2001. Mortality rate ratio significantly increased with MUAC values < 20 cm, but a strong tendency of increased mortality risk was also observed for MUAC < 21 cm, compared with values ≥ 24.1 cm. Two studies examined MUACs ability to predict mortality among people living with HIV in Tanzania and Guinea-Bissau respectively. Liu et al.[18] found significantly increased relative risks for MUAC cut-offs < 25 cm compared with ≥ 27 cm at 3 months, and Oliveira et al.[19] found a significantly increased hazard ratio for MUAC < 25 cm compared with ≥ 25 cm at 6 months. Overall, the studies on MUAC and TB mortality found varying degrees of increased mortality risk with MUAC values below 25 cm. The same was observed in the current study. This demonstrates that all MUAC cut-offs from 25 cm and below are clinically relevant for patients with TB.
A major strength of this study is the large sample size of patients with TB, who also had a high prevalence of malnutrition. The large sample size allowed for an investigation into potential sex- and disease-specific differences. The Bandim TB cohort followed patients throughout their treatment period and made it possible to evaluate the ability of MUAC cut-offs in predicting mortality. This strengthens the interpretation of the clinical relevance of MUAC cut-offs.
This study also had some limitations. Selection bias is present as some patients died before being included in the cohort, and others were too ill to be included in the cohort. It is reasonable to assume those patients would have been moderately to severely malnourished with an increased risk of dying and including them would have strengthened the mortality analyses. Mortality risk may have changed over time during the 19 years of data collection. However, antituberculosis treatment was available during the entire study period, and adjustment for year of treatment in mortality analyses did not change results (data not shown). Another limitation to the mortality analyses is the fact that patients received antituberculosis treatment during the observation time. It is possible that the antituberculosis treatment reduced the number of fatalities during the observation time. An important limitation to this study is that MUAC is compared with BMI, which is an imperfect measure of malnutrition [20, 21]. Comparing MUAC with more advanced measures was not feasible in this resource-poor setting.
The results from the current study were generally aligned with the findings of other studies, including the large globally-oriented meta-analysis by Tang et al. and other studies from African nations [2, 9, 12, 13, 17–19]. Thus, the findings are assumed to be generalizable to the sub-Saharan region and perhaps beyond.
Based on the results of the current study it is proposed to use MUAC < 25 cm as an alternative to BMI < 20 kg/m2 to identify malnutrition, e.g. as part of the GLIM criteria for the diagnosis of malnutrition. Furthermore, it is proposed to use MUAC < 22 cm, roughly corresponding to BMI < 16 kg/m2, to identify patients at an increased risk of morbidity and mortality.
The MUAC cut-off < 25 cm demonstrated a low proportion of false positives and thus accurately excluded adequately nourished patients, while maintaining acceptable sensitivity for detecting those who were genuinely malnourished. The MUAC cut-off < 22 cm showed a very low proportion of false negatives, effectively identifying those with more severe malnutrition. Importantly, mortality risk increased with decreasing MUAC values, supporting the use of MUAC < 22 cm to identify patients at higher risk of a poor outcome, who might benefit from interventions like nutrition therapy and hospitalization. Choosing these two cut-offs will balance operational feasibility with diagnostic relevance in a public health and clinical setting. However, it is possible that an even lower MUAC cut-off is needed in famine relief, where many people are affected but resources are extremely limited [14]. The proposed unisex MUAC cut-offs will favour women, which is also the case for the current BMI cut-offs [20].
Further studies are needed to explore potential ethnic differences in MUAC. In order to identify MUAC cut-offs that are clinically relevant in other settings and situations, other studies examining the association between MUAC and morbidity and mortality are needed. Using the proposed MUAC cut-offs in intervention studies as an inclusion criterion or an effect measure would further strengthen the use of MUAC cut-offs in place of BMI.
In conclusion, the current study demonstrated that MUAC was excellent in its ability to identify BMI cut-offs and thus could be used as an alternative to BMI. A unisex MUAC cut-off < 25 cm was proposed to be used in place of BMI < 20 kg/m2 to identify malnourishment. A unisex MUAC cut-off < 22 cm was proposed to identify malnourished patients at increased risk of dying, who would benefit from e.g. nutrition intervention.
Supplementary Information
Additional File 1: Scatterplot and regression lines of the correlation between MUAC and BMI stratified by sex. Additional File 2: ROC curves and AUROCCs for MUAC cut-offs identifying BMI cut-offs stratified by sex. Additional File 3: Diagnostic accuracy of every 0.5 cm MUAC cut-off to identify BMI cut-offs for men. Additional File 4: Diagnostic accuracy of every 0.5 cm MUAC cut-off to identify BMI cut-offs for women. Additional File 5: Scatterplot and regression lines of the correlation between MUAC and BMI stratified by HIV-coinfection. Additional File 6: ROC curves and AUROCCs for MUAC cut-offs identifying BMI cut-offs stratified by HIV-coinfection. Additional File 7: Diagnostic accuracy of every 0.5 cm MUAC cut-off to identify BMI cut-offs for patients with TB who tested negative for HIV. Additional File 8: Diagnostic accuracy of every 0.5 cm MUAC cut-off to identify BMI cut-offs for patients with TB living with HIV
Acknowledgements
The authors extend their gratitude to patients participating in the cohort and to the dedicated field and research assistants, who have worked with the Bandim TB Cohort over the years.
Abbreviations
- BMI
Body Mass Index
- GLIM
The Global Leadership Initiative on Malnutrition
- TB
Tuberculosis
- HIV
Human Immunodeficiency Virus
- MUAC
Mid-Upper Arm Circumference
- BHP
The Bandim Health Project
- SENS
Sensitivity
- SPEC
Specificity
- FN
False Negative
- FP
False Positive
- ROC
Receiver Operating Characteristic
- AUROCC
Area Under the Receiver Operating Characteristic Curve
Authors’ contributions
CBP and CW conceptualized the study. CBP, AS and FR supervised data collection. CBP performed the analyses and drafted the manuscript. All authors provided critical feedback and contributed to the final version of the manuscript.
Funding
The Bandim TB Cohort have been supported by The European and Developing Countries Clinical Trials Partnership (EDCTP) (grant number JP.2009.10800.06), Novo Nordisk Foundation (grant number NNF15OC0018034) and Aarhus University Research Foundation. The funders had no role in study design, data collection and analyses, decision to publish, or preparation of the manuscript.
Data availability
The dataset used and analysed during the current study is available from the corresponding author on reasonable request.
Declarations
Ethics approval and consent to participate
The study abided by the Declaration of Helsinki. Written informed consent by signature or fingerprint if illiterate, was provided by all subjects before inclusion. The Bandim TB cohort studies were permitted by the National Committee of Ethics in Health in Bissau, Guinea-Bissau (MINSAP 220405).
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Jensen GL, Cederholm T, Correia M, Gonzalez MC, Fukushima R, Higashiguchi T, et al. GLIM criteria for the diagnosis of malnutrition: a consensus report from the global clinical nutrition community. J Parenter Enteral Nutr. 2019;43(1):32–40. [DOI] [PubMed] [Google Scholar]
- 2.Tang AM, Chung M, Dong KR, Bahwere P, Bose K, Chakraborty R, et al. Determining a global mid-upper arm circumference cut-off to assess underweight in adults (men and non-pregnant women). Public Health Nutr. 2020;23(17):3104–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Tang AM, Chung M, Dong KR, Wanke C, Bahwere P, Bose K et al. Determining a Global Mid-Upper Arm Circumference Cutoff to Assess Underweight in Adults (Men and Nonpregnant Women). Available at https://www.fantaproject.org/sites/default/files/resources/Global-MUAC-Cutoffs-nonPregnant-Adults-Jun2017.pdf; 2017. Accessed 05 Mar 2024.
- 4.Bandim Health Project. Available at https://www.bandim.org/. Accessed 05 Mar 2024.
- 5.Bohlbro AS, Mendes AM, Sifna A, Gomes V, Rudolf F, Wejse C. Incidence of pulmonary tuberculosis in suburban Bissau, Guinea-Bissau between 2004 and 2020: a prospective cohort study. Infection. 2023;51(4):955–66. [DOI] [PubMed] [Google Scholar]
- 6.World Health Organization. Management of severe malnutrition: A manual for physicians and other senior health workers. Available at https://iris.who.int/bitstream/handle/10665/41999/a57361-eng.pdf?sequence=1; 1999. Accessed 05 Mar 2024.
- 7.World Health Organization. Definitions and reporting framework for tuberculosis – 2013 revision: updated December 2014 and January 2020. Available at https://iris.who.int/bitstream/handle/10665/79199/9789241505345_eng.pdf?sequence=1; 2013. Accessed 05 Mar 2024.
- 8.Patsche CB, Rudolf F, Mogensen SW, Sifna A, Gomes VF, Byberg S, et al. Low prevalence of malnourishment among household contacts of patients with tuberculosis in Guinea-Bissau. Int J Tuberc Lung Dis. 2017;21(6):664–9. [DOI] [PubMed] [Google Scholar]
- 9.Van Tonder E, Mace L, Steenkamp L, Tydeman-Edwards R, Gerber K, Friskin D. Mid-upper arm circumference (MUAC) as a feasible tool in detecting adult malnutrition. S Afr J Clin Nutr. 2019;32(4):93–8. [Google Scholar]
- 10.Das A, Saimala G, Reddy N, Mishra P, Giri R, Kumar A, et al. Mid-upper arm circumference as a substitute of the body mass index for assessment of nutritional status among adult and adolescent females: learning from an impoverished Indian state. Public Health. 2020;179:68–75. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Benítez Brito N, Suárez Llanos JP, Fuentes Ferrer M, Oliva García JG, Delgado Brito I, Pereyra-García Castro F, et al. Relationship between mid-upper arm circumference and body mass index in inpatients. PLoS One. 2016;11(8):e0160480. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Philpott DC, Belchior-Bellino V, Ververs M. Use of mid-upper arm circumference to screen for thinness among sub-Saharan African male detainees. Public Health Nutr. 2021;24(15):4777–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Musa IR, Omar SM, Adam I. Mid-upper arm circumference as a substitute for body mass index in the assessment of nutritional status among adults in Eastern Sudan. BMC Public Health. 2022;22(1):2056. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Collins S. Using middle upper arm circumference to assess severe adult malnutrition during famine. JAMA. 1996;276(5):391–5. [DOI] [PubMed] [Google Scholar]
- 15.White LV, Lee N, Marin FP, Saludar NR, Edwards T, Cox SE. Performance of alternative measures to body mass index in the assessment of moderate and severe under-nutrition among acutely unwell patients hospitalized in a TB ward in the philippines: A cross-sectional study. PLoS One. 2019;14(5):e0215968. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Lee N, White LV, Marin FP, Saludar NR, Solante MB, Tactacan-Abrenica RJC, et al. Mid-upper arm circumference predicts death in adult patients admitted to a TB ward in the philippines: A prospective cohort study. PLoS One. 2019;14(6):e0218193. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Gustafson P, Gomes VF, Vieira CS, Samb B, Nauclér A, Aaby P, et al. Clinical predictors for death in HIV-positive and HIV-negative tuberculosis patients in Guinea-Bissau. Infection. 2007;35(2):69–80. [DOI] [PubMed] [Google Scholar]
- 18.Liu E, Spiegelman D, Semu H, Hawkins C, Chalamilla G, Aveika A, et al. Nutritional status and mortality among HIV-infected patients receiving antiretroviral therapy in Tanzania. J Infect Dis. 2011;204(2):282–90. [DOI] [PubMed] [Google Scholar]
- 19.Oliveira I, Andersen A, Furtado A, Medina C, da Silva D, da Silva ZJ et al. Assessment of simple risk markers for early mortality among HIV-infected patients in Guinea-Bissau: a cohort study. BMJ Open. 2012;2:e001587 [DOI] [PMC free article] [PubMed]
- 20.James WP, Ferro-Luzzi A, Waterlow JC. Definition of chronic energy deficiency in adults. Report of a working party of the international dietary energy consultative group. Eur J Clin Nutr. 1988;42(12):969–81. [PubMed] [Google Scholar]
- 21.Maalouf-Manasseh Z, Remancus S, Milner E, Fenlason L, Quick T, Patsche CB, et al. Global mid-upper arm circumference cut-offs for adults: a call to action. Public Health Nutr. 2020;23(17):3114–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Additional File 1: Scatterplot and regression lines of the correlation between MUAC and BMI stratified by sex. Additional File 2: ROC curves and AUROCCs for MUAC cut-offs identifying BMI cut-offs stratified by sex. Additional File 3: Diagnostic accuracy of every 0.5 cm MUAC cut-off to identify BMI cut-offs for men. Additional File 4: Diagnostic accuracy of every 0.5 cm MUAC cut-off to identify BMI cut-offs for women. Additional File 5: Scatterplot and regression lines of the correlation between MUAC and BMI stratified by HIV-coinfection. Additional File 6: ROC curves and AUROCCs for MUAC cut-offs identifying BMI cut-offs stratified by HIV-coinfection. Additional File 7: Diagnostic accuracy of every 0.5 cm MUAC cut-off to identify BMI cut-offs for patients with TB who tested negative for HIV. Additional File 8: Diagnostic accuracy of every 0.5 cm MUAC cut-off to identify BMI cut-offs for patients with TB living with HIV
Data Availability Statement
The dataset used and analysed during the current study is available from the corresponding author on reasonable request.