Overweight and obesity in childhood and adolescence is one of the greatest public health problems facing most countries in the world, which has increased dramatically in recent decades(1,Reference Abarca-Gómez, Abdeen and Hamid2) . The global prevalence of obesity in children and adolescents was 5·6 % in girls and 7·8 % in boys, respectively, in 2016(Reference Abarca-Gómez, Abdeen and Hamid2). Childhood obesity can persist into adulthood, leading to an increased risk of chronic non-communicable diseases and premature mortality(Reference Avgerinos, Spyrou and Mantzoros3–Reference Han and Lean6). Timely recognition and early diagnosis of overweight and obesity in young people are essential to mitigate the short- and long-term health risks associated with excess body weight, especially fat mass(Reference Simmonds, Llewellyn and Owen7,Reference Llewellyn, Simmonds and Owen8) . Health care professionals need to use an accurate and efficient screening tool to diagnose overweight and obesity for early intervention(Reference Javed, Jumean and Murad9).
According to the WHO, obesity is defined as an abnormal or excessive fat accumulation that may impair health(10). Total body fat can be accurately measured using several methods such as hydrostatic weighing, air displacement plethysmography, deuterium oxide dilution or dual-energy X-ray absorptiometry, which estimates total body fat by measuring total body water and based on the absorption patterns of X-rays(Reference Wells and Fewtrell11,Reference Cornier, Després and Davis12) . However, these methods are laboratory-based, expensive, time-consuming and not feasible for routine use(Reference Simmonds, Llewellyn and Owen13).
Alternatively, there are a range of anthropometric measurements that are simple and inexpensive options for overweight and obesity screening in children and adolescents(14). However, the results of anthropometric measurements need to be interpreted using reference standards to define overweight and obesity in young people. While the BMI z-score is a widely utilised method to identify those with overweight and obesity in epidemiological studies but has limited applicability in routine clinical practice(Reference Abarca-Gómez, Abdeen and Hamid2,Reference Pischon, Boeing and Hoffmann15) . Simple and inexpensive alternatives to BMI z-scoring would be helpful to promote screening and early identification, especially in low- and middle-income countries with limited health care resources(Reference Prentice16).
The mid-upper arm circumference (MUAC) has been proposed as one such alternative to screen for overweight and obesity in children and adolescents(Reference Chaput, Katzmarzyk and Barnes17–Reference Sisay, Haile and Hassen20). It is a simple measure commonly used to screen for undernutrition in infants and children aged 6–59 months(Reference Myatt, Khara and Collins21) as well as thinness and severe thinness in adolescents(Reference Sethi, Gupta and Pedgaonkar22,Reference Sisay, Haile and Hassen23) .The existing evidence of the usefulness of the MUAC against the BMI Z-score among children and adolescents is limited and unclear(Reference Chaput, Katzmarzyk and Barnes17–Reference Sisay, Haile and Hassen20,Reference Shinsugi, Gunasekara and Takimoto24–Reference Lu, Wang and Lou26) . Therefore, this systematic review and meta-analysis aim to summarise the currently available evidence on the performance of MUAC to identify children and adolescents with overweight and obesity.
Methods
This systematic review and meta-analysis were conducted per the Preferred Reporting Items for Systematic Reviews and Meta-Analyses for Diagnostic Test (PRISMA-DTA) statement (see online supplementary material, Supplemental Table S1)(Reference McInnes, Moher and Thombs27). The protocol of this systematic review and meta-analysis was registered on PROSPERO (Registration number CRD42020183148) and published(Reference Sisay, Hassen and Gebreyesus28); minor deviations from the original protocol have also been explained (see online supplementary material, Supplemental Table S2).
Review question
This systematic review aims to generate an evidence summary to answer the following review question: what is the diagnostic performance of the MUAC assessment for diagnosing overweight and obesity in children and adolescents.
Search strategy and study selection
A systematic search was performed in PubMed, EMBASE, SCOPUS, Cochrane Library, Web of Science and CINAHL database of references for peer-reviewed articles. To retrieve grey literature, systematic search was performed using Google Scholar. The databases were systematically searched from their inception to December 10, 2021. A detailed search strategy is provided in the supplementary material (see online supplementary material, Supplemental Table S3).
All articles identified through the systematic search databases were imported into EndNote as a single library. Duplicate articles from the searches were verified and removed. The remaining articles were imported into rayyan.QCRI.org(Reference Ouzzani, Hammady and Fedorowicz29), a web-based tool that facilitates screening and collaboration among researchers.
Two independent reviewers (BGS and BRJ) conducted the title and abstract screening and included articles for the full-text review. Disagreement between the two reviewers was resolved by inviting the third reviewer (HYH) to make the final decision. The following inclusion criteria were used:
-
1. Population: children or adolescents aged 2–19 years.
-
2. Index test: studies that assessed the diagnostic performance of MUAC as an index test to identify children and adolescents with overweight/obesity.
-
3. Comparator: compared to reference standards such as BMI z-score, weight to height, waist circumference, skinfold thickness, dual-energy X-ray absorptiometry, air-displacement plethysmography, bioelectrical impedance and hydro densitometry.
-
4. Outcome: overweight and obesity.
-
5. Study design: observational studies including cross-sectional, cohort and case-control were included.
-
6. Language: studies published in any language were included.
-
7. Year of publication: no restriction was made based on the year of publication.
Studies that fulfill any of the following criteria were excluded:
-
1. Articles available only in abstract form, letters, reviews, commentaries, editorials, case series.
-
2. Duplicate publication of the same study
Data extraction
Two independent reviewers (BGS and BRJ) have extracted the following information from included studies, using pilot-tested data collection form: first author’s name, year of publication, country or region, funding source, study design, total sample size, number of males and females, response rate, age of study participants, MUAC cut-off values, reference standard, diagnostic criteria of overweight and obesity (reference standard), sensitivity, specificity, AUC, positive and negative likelihood ratio, prevalence of overweight and obesity, positive and negative predictive values. The extracted data by independent reviewers were compared and any discrepancy was resolved by consensus. When relevant information was missing from the article, we contacted the primary authors twice via e-mail.
Risk of bias and certainty of evidence assessment
Two independent reviewers (BGS and BRJ) assessed the risk of bias and applicability of the included studies using Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2)(Reference Whiting, Rutjes and Westwood30). We have also assessed the certainty of evidence for relevant outcomes using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach for diagnostic tests(Reference Schünemann, Oxman and Brozek31). Discrepancies among reviewers on individual items were resolved by discussion and consensus.
Statistical analysis
The extracted data were exported to STATA/SE Version 16 for further processing and analysis. We have summarised the diagnostic test accuracy by creating a 2 × 2 table for each study. We have performed a graphical descriptive analysis of the included studies. We have reported coupled forest plots (sensitivity and specificity separately, along with the 95 % CI), and we provided a graphical representation of studies in the summary receiver operating characteristic (SROC) curve (sensitivity against 1 – specificity).
Among the included studies, those that report the number of true positive, true negative, false positive, false negative or values that are required to calculate them are included in the meta-analysis.
We have used the hierarchical summary receiver-operating characteristic curve model to produce SROC curves(Reference Rutter and Gatsonis32). The AUC – SROC curve values were used to describe test accuracy. We evaluated the discriminatory power by the AUC-SROC using values proposed by Swets(Reference Swets33), with ≤ 0·5 considered to have no discriminatory power, > 0·5 and ≤ 0·7 to have low discriminatory power, > 0·7 and ≤ 0·9 to have good discriminatory power and 1 to be a perfect test. In addition, we have estimated the pooled sensitivity, specificity, positive likelihood ratio and negative likelihood ratio to complement the findings of SROC. Subgroup analysis was performed for boys, girls, children (2–9 years), adolescents (10–19 years), with overweight and obesity.
We assessed the heterogeneity of diagnostic test parameters by visual inspection of the paired forest plots and SROC plots. One of the major sources of heterogenicity in diagnostic accuracy systematic review and meta-analysis is the use of different thresholds. We have assessed the presence of a threshold effect using spearman’s correlation coefficient between the logit of sensitivity and the logit of 1 – specificity(Reference Moses, Shapiro and Littenberg34). We have explored potential sources of bias including sex, cut-off point, age group and weight status (overweight, obesity) variables. To further investigate heterogeneity, we performed subgroup analysis for boys, girls, children (2–9 years), adolescents (10–19 years), children and adolescents with overweight, children and adolescents with obesity. To assess possible publication bias, we used Deeks’ funnel plot, with Deeks’ asymmetry test, where P < 0·05 was considered as significant asymmetry(Reference Deeks, Macaskill and Irwig35).
Results
Selection of studies
The search strategy results in a total of 3039 references. Of these, 2041 were duplicates, resulting in 998 articles. After screening titles and abstracts, 963 studies were excluded. Thirty-five papers were retained for review after full-text evaluation. Of these, fourteen articles were excluded with reasons; did not report measures of diagnostic performance; different target condition; was conducted on age group outside of the scope of this review. Therefore, twenty-one studies met the inclusion criteria and were included in the systematic review, however, ten studies with sufficient information are included in the meta-analysis. These eleven articles reported neither the required parameters for meta-analyses (true positive, true negative, false negative or false positive) nor other parameters to calculate them (prevalence, positive predictive value or negative predictive value). Figure 1 shows the detailed description of the article screening process.
Study selection flow chart
Characteristics of studies
Table 1 shows characteristics of studies included in the systematic review. All of the studies were cross-sectional and were published from 2013 to 2021. They were conducted in twenty-one countries including Brazil(Reference Dumith, Muraro and Monteiro36), China(Reference Lu, Wang and Lou37), Netherlands(Reference Talma, Van Dommelen and Schweizer38), Ethiopia(Reference Sisay, Haile and Hassen20), India(Reference Jaiswal, Bansal and Agarwal39–Reference Nitika41), Indonesia(Reference Ayu, Aditiawati and Anzar42), Nigeria(Reference Orimadegun43,Reference Oriaifo, Abiodun and Atimati44) , Pakistan(Reference Asif, Aslam and Altaf25,Reference Asif, Aslam and Altaf45) , Seri Lanka(Reference Shinsugi, Gunasekara and Takimoto24), South Africa(Reference Craig, Bland and Ndirangu46,Reference Otitoola, Oldewage-Theron and Egal47) , Thailand(Reference Rerksuppaphol and Rerksuppaphol48), Turkey(Reference Mazicioǧlu, Hatipoǧlu and Öztürk49,Reference Ozturk, Cicek and Mazicioglu50) , Trinidad and Tobago(Reference Ramcharitar-Bourne, Nichols and Badrie51) and twelve countries(Reference Chaput, Katzmarzyk and Barnes17) (Australia, Brazil, Canada, China, Colombia, Finland, India, Kenya, Portugal, South Africa, the UK and the USA) (n 1). The number of participants varied substantially between studies (range from 211 to 31 471), with a pooled population of 54 381 children and adolescents.
Characteristics of included studies
BIA, bioelectrical impedance analysis; WC, waist circumference.
Studies used different reference methods: three studies used bioelectrical impendency(Reference Craig, Bland and Ndirangu18,Reference Oriaifo, Abiodun and Atimati44) ; two used waist circumferences(Reference Mazicioǧlu, Hatipoǧlu and Öztürk49,Reference Ozturk, Cicek and Mazicioglu52) ; the rest used BMI. Most studies used an 85th percentile (Z score > 1 + sd) cut-off of BMI growth curves for overweight and a 95th percentile cut-off (Z score > 2 + sd) of BMI curve for obesity. Both studies that use bioelectrical impendency classify participants with 85th percentile as overweight(Reference Craig, Bland and Ndirangu18,Reference Oriaifo, Abiodun and Atimati44) . The cut-off points of MUAC values for defining overweight and obesity varied substantially between studies (see online supplementary material, Supplemental Table S4).
Meta-analysis
A meta-analysis was conducted to assess the performance of MUAC for identifying overweight and obesity in children and adolescents. The meta-analysis showed that a pooled AUC of 0·92 (95 % CI 0·89, 0·94), sensitivity of 85·2 (95 % CI 77·3, 90·6) and a pooled specificity of 85·6 (95 % CI 81·0, 98·2) for boys and girls, respectively. In boys, MUAC showed a pooled AUC of 0·92 (0·89, 0·94), sensitivity of 84·4 (95 % CI 84·6, 90·8) and a pooled specificity of 86·0 (95 % CI 79·2, 90·8) (Fig. 2 and Table 2). As for girls, MUAC showed a pooled AUC of 0·93 (95 % CI 0·90, 0·95), sensitivity of 86·4 (95 % CI 79·8, 91·0) and pooled specificity of 86·6 (95 % CI 82·2, 90·1) (Fig. 3 and Table 2).
Summary receiver operating characteristic (SROC) curve for boys
Diagnostic performance of mid-upper arm circumference
Summary receiver operating characteristic (SROC) curve for girls
We have further explored the diagnostic performance of the MUAC by weight status. The highest discriminatory ability was observed among children and adolescents with obesity, resulting in a pooled AUC of 0·95 (95 % CI 0·92, 0·96), sensitivity of 89·4 (95 % CI 80·2, 94·6) and a pooled specificity of 88·9 (95 % CI 83·4, 92·8).
Moreover, we have analysed the diagnostic performance of MUAC by the age of participants 2–9 years and 10–19 years. The MUAC showed higher performance in adolescents between the age of 10–19 compared to those aged 2–9 years with an AUC of 0·96 (0·93, 0·97) v. 0·88 (95 % CI 0·85, 0·90).
Risk of bias and publication bias
The full results of the risk of bias and applicability were assessed using QUADAS-2 is shown in supplementary material (see online supplementary material, Supplemental Table S5). The study design and procedure were homogeneous and almost all met all QUADAS-2 domains. One study was classified as ‘high risk’ in the question of applicability domain (Index test) because the reference population they use to identify children with overweight and obesity was inappropriate. All included studies had a ‘high risk’ of bias domain (reference standard) because they used BMI as their reference standard, which is not a golden standard to measure excess adiposity. Even though BMI is highly correlated with excess adiposity, it misclassifies a significant number of children and adolescents(Reference Javed, Jumean and Murad9). Except for one(Reference Sisay, Haile and Hassen20), none of the included studies reported the time interval between performing the index test and reference standard. However, it is unlikely that any time delay between conducting the index test and the reference standard would introduce bias. The description of the index tests and reference standards was adequately reported. Evidence of publication bias was not observed by Deek’s test (P = 0·71) (see online supplementary material, Supplemental Fig. S1).
Investigation of heterogenicity
Visual inspection of the paired forest plot and SROC curve revealed substantial heterogeneity. (Fig. 4) A Spearman’s rank correlation test showed the presence of a threshold effect r = 0·63. We have examined the influence of covariate sex, age group and weight status. We have observed that MUAC cut-off (P-value = 0·00), age group (P-value = 0·00) and weight status (P-value = 0·05) significantly contribute to the heterogenicity of the study. However, sex of participants did not contribute to the heterogenicity (P-value = 0·10). To further explore sources of heterogeneity, we performed subgroup analyses by sex, age group (children, adolescents) and weight status (overweight, obese). The results suggest that weight status might contribute to the heterogeneity among studies. Since pooling sensitivity and specificity are more reliable in the absence of a threshold effect, the findings should be primarily judged based on the SROC curve.
Paired forest plot of sensitivity and specificity with 95 % CI
Certainty of evidence assessment
We rated the certainty of evidence of the pooled studies and considered it as moderate for all pooled measures of diagnostic accuracy. The reasons for downgrading the certainty of evidence included the marked heterogenicity observed(Reference Schünemann, Oxman and Brozek31). The result of certainty of evidence assessment is available on supplementary material (see online supplementary material, Supplemental Table S6).
Discussion
This meta-analysis showed that when assessing the diagnostic performance of MUAC compared with BMI defined overweight and obesity. MUAC has high sensitivity and specificity, correctly identifying about 86·4 % of children and adolescents with BMI defined overweight and obese. In addition, MUAC has a greater discriminatory ability in identifying children and adolescents with BMI-defined obesity than those with BMI-defined overweight. However, no sufficient evidence on the performance of MUAC compared with gold standard measures of adiposity.
The MUAC was initially developed to screen and diagnose under-five children with moderate and severe acute malnutrition(Reference Myatt, Khara and Collins21). Moreover, MUAC has also been used to identify adolescents and women of reprobative age with thinness and severe thinness(Reference Sethi, Gupta and Pedgaonkar22,Reference Sisay, Haile and Hassen23) . In recent years, MUAC has been explored as an alternative screening tool for children and adolescents with overweight and obesity(Reference Chaput, Katzmarzyk and Barnes17–Reference Sisay, Haile and Hassen20,Reference Shinsugi, Gunasekara and Takimoto24–Reference Lu, Wang and Lou26) . MUAC has attractive characteristics that make it desirable as a simple screening tool.
To the best of our knowledge, this is the first systematic review and meta-analysis to examine the discriminatory performance of MUAC to identify children and adolescents with overweight and obesity considering age group, sex and weight status. The previous systematic review had included MUAC as one of the uncommonly used measures of overweight and obesity. However, this systematic review has not assessed the discriminatory ability of MUAC in identifying children and adolescents with overweight and obesity(Reference Ranasinghe, Jayawardena and Gamage53).
This meta-analysis showed that MUAC has a high discriminatory ability to identify both boys and girls with BMI defined overweight and obesity with a comparable value of sensitivity and specificity. However, concerning weight status, MUAC has superior discriminatory ability among children and adolescents with obesity than those who are BMI defined overweight. Similar findings have been reported by a meta-analysis conducted on the performance of neck circumference(Reference Kroll, Mastroeni and Czarnobay54). This might be due to the fact that sensitivity and specificity of screening tools depend on the spectrum of a condition, in this case, the degree of adiposity; those who are obese are more likely to be easily identified by screening tools than those who are overweight.
An ideal anthropometric measurement to identify adolescents with overweight and obesity should be easy to use, accurate and reliable(Reference Roper55). MUAC fulfil almost all characteristics; MUAC is simple to use since its measurement requires only a non-stretchable MUAC tape; MUAC measurement is easy to interpret since it does not require the use of an additional reference chart. Moreover, MUAC can easily be used by an illiterate person if it is colour-coded. Traffic light colours of red (obese), amber (overweight) and green (normal weight) may also be considered by non-numerate field workers in developing countries to facilitate screening(Reference Chaput, Katzmarzyk and Barnes17). The other important characteristic of ideal measurement is its accuracy in identifying overweight and obesity among children and adolescents. As we have observed in this meta-analysis, MUAC has high accuracy in identifying overweight and obesity. However, in this meta-analysis, all included articles compare the discriminatory ability of MUAC against BMI defined overweight and obesity. Even though BMI is highly correlated with percent body fat, it does not differentiate between lean mass and fat mass(Reference Freedman, Horlick and Berenson56). Only two studies have compared the discriminatory ability of MUAC with total body fat which was measured with bioelectrical impendency. A high level of discriminatory ability of MUAC has been reported by both studies(Reference Craig, Bland and Ndirangu18,Reference Oriaifo, Abiodun and Atimati44) . None of the studies have assessed the discriminatory performance of MUAC with the four-compartment model, the known golden standard to measure total body fat. There is a need for future researches to explore the performance of MUAC to identify overweight and obesity, ideally as defined by the ‘golden standard’ measure of total body fat.
The major limitation of this meta-analysis was the presence of marked heterogenicity. First, a diagnostic threshold bias was identified as a cause of heterogeneity in the pooled results. In this meta-analysis, there was no consistent cut-off value. To overcome this limitation, we have used a hierarchal SROC curve that accounts for the threshold effect. Furthermore, we have conducted a sub-group analysis to reduce heterogeneity, we calculated measures of diagnostic performance according to age group, sex or weight status. Second, we have attempted to reach the corresponding authors of articles with insufficient data (prevalence, sample size by category) to be included in the meta-analysis, but without success. The main strength of this study was the rigorous statistical methods used to pool data across diagnostic accuracy studies. The other strength of this study is the comprehensive search strategy in several electronic databases.
Conclusions
In conclusion, MUAC has high sensitivity and specificity compared with BMI defined overweight and obesity. However, there is no sufficient evidence on the performance of MUAC compared to gold standard measures of adiposity. There is a need for future studies to evaluate the diagnostic performance compared with the ‘golden standard’ measure of total body fat and evaluate the predictive value of the MUAC measurement for developing weight-related complications.
Acknowledgements
Acknowledgements: We would like to express our deepest gratitude to authors that respond to our request for additional information and explanations. Financial support: This research received no specific grant from any funding agency, commercial or not-for-profit sectors. Conflict of interest: There are no conflicts of interest. Authorship: B.G.S. was the lead author of the manuscript including study conception, design, conduct, statistical analysis and writing of the first draft. S.H.G., H.Y.H., B.R.J. and E.A. were involved in formulating a research question and critical revision of the manuscript. B.G.S. and E.A. performed systematic searches of bibliographic databases. H.Y.H. and B.R.J. were also involved in the screening of potentially relevant articles for review, risk of bias assessment, data extraction and data analysis. H.Y.H., B.R.J. and E.A. have all critical reviewed and edited the manuscript. All authors read and approved the final manuscript. None of the authors reported a conflict of interest related to the study. Ethics of human subject participation: Not applicable since this study is systematic review and meta-analysis.
Supplementary material
For supplementary material/s referred to in this article, please visit https://doi.org/10.1017/S1368980022000143
Open access
