Exclusion criteria

The search excluded paper on adults and adolescents and was limited to children under five. Articles prior to 1980 were not considered. Languages included English and French. Articles had to assess whether the changes or trend observed at population level were statistically significant. All study designs were included.

Search strategy

The peer-reviewed literature search was conducted using Embase, Global Health and Medline. The search was performed in November 2015 by a single reviewer using the following terms to capture two concepts: 1) Undernutrition: ((arm or midarm or mid-arm or mid-upper arm) and circumference) or MUAC or weight-for-height or weight-for-length or WHZ or WHM or weight- for- age or WAZ or height- for-age or length-for-age or HAZ or kwashiorkor or oedema or WAM or HAM or weight or height or anthropometry or anthropometric indices or anthropometric indicators or stunting or wasting or acute malnutrition or marasmus or underweight AND 2) Nutrition surveillance: Nutrition$ assessment or nutrition$ survey or nutrition$ surveillance or nutrition$ situation or malnutrition prevalence or nutrition$ monitoring or nutrition$ screening or nutrition$ evaluation or nutrition$ early warning system or nutrition$ change or nutrition$ variation or nutrition$ impact or season$ change or season$ variation.

Grey literature undertaken by searching the following websites: Emergency Nutrition Network (ENN), The United Nations System Standing Committee on Nutrition and the Community-based Management of Acute Malnutrition (CMAM) forum [18–20].

Data extraction, analysis and reporting

Returned citations were downloaded to Endnote software and a five-stage screening process applied (see Fig. 1). Articles that met the inclusion criteria were selected and data abstracted in an excel sheet. The following data were extracted from each paper: i) study authors, ii) year; iii) study country and collection period, iv) setting, (v) type of study, vi) sample size, vii) age group, viii) independent variables, ix) dependent variables, x) reference and unit, xi) outcome of the study. The outcome of the study included prevalence, means and Odd Ratios (OR) with associated p-values. Descriptive analysis was used and the systematic review methodology adheres to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [21]. We did not register the review protocol and this review does not include a bias analysis. Papers included were too different to be able to do a synthesis and very few of them had as objective to assess performance of different anthropometric measurements or indices in the detection of change in the nutritional status of the study population.

Performance of nutritional measurement/indices to detect changes in nutritional situations

A total of 4563 articles were retrieved from the electronic database search (1837 articles from Embase, 1102 from articles Global Health and 1624 articles from Medline) out of which 1033 duplicates were excluded leaving 3530 articles to review. A large majority of articles (3137, 89%) were found out of topic and 39 articles were left for full review. A total of 17 articles met the inclusion criteria. All potential articles found in the grey literature had been published and therefore included in the above search. An additional 4 papers were added after snowballing. A total of 21 articles were included in this review. Figure 1 flow diagram summarises the search.

Out of the 21 studies included, 4 (19%) compared the change in mean and/or prevalence of several measurements/indices. Benefice et al. [25] presented variations of mean MUAC, WFH, TSF and MC in a longitudinal study in rural Senegal. Mean WFH/L was the only index that was not changing significantly and TSF showed the largest differences. Briend et al. [15] examined the effect of seasonal change as well as the change between the first two years of the study and the last two on weight, WFA, H/LFA, WFH/L, and MUAC. This study revealed that Weight, MUAC and WFH were the nutritional indices that changed the most between seasons. Mean H/LFA, WFA and MUAC were significantly higher during last 2 years. This was more pronounced for HFA. In a study assessing the fluctuations of the mean weight, height increment, WFA, HFA, WFH, MUAC-for-Age and TSF-for-Age, Brown et al. [26] found that TSF-for-Age had the greatest seasonal change. WFA and MUAC-for-Age followed the same patterns and magnitude while WFH had greater range but similar coefficient of variation as MUAC and smaller than WFA. Finally, Garenne et al. [27] study looked at seasonal changes of mean WFA, WFH, MUAC, TSF, weight, height, BMI, MC, SSF and HC. The highest contrast value was observed for mean MUAC which made it the best measurement for the detection of short term changes. Mean weight, height and head circumference had the highest responsiveness. Responsiveness was defined as a measure of the change over a semester compared with the variation of the indicator in the population (change divided by the standard deviation of the same indicator). These indices were the most appropriate to monitor growth velocity of children in a stable situation (Table 4).

Unsurprisingly, H/LFA was mainly out of phase compared to other measures of undernutrition and was a good measure of long term change (Briend et al. [15], Brown et al. [26], Huong et al. [23], Marin et al. [28], Martin-Prevel et al. [29], Miller et al. [30], Panter-Brick et al. [31]) (Table 4).

The capacity to detect change in the nutritional status of the population did not seem to differ whether the anthropometric measurements/indices were treated as continuous or binary. However, the sample size requirement differs whether assessing the mean of a continuous variable or looking at the prevalence of a binary variable. Using means allows for smaller sample size which has important implications in terms of logistics, costs and timeliness. This was confirmed in Briend et al. [15] paper.

Simplicity

Any index that includes an age component requires that age be ascertained accurately and it is widely acknowledged that determining age correctly is problematic in many developing countries [33–36]. The use of multi-component indices (i.e. WFA, HFA, WFH, MUAC-for-Age, TSF-for-Age) is usually more complex [33, 37]. Moreover, transporting and carrying weight scales as well as height/length board is more logistically challenging than for MUAC tapes.

Acceptability

The measurement of weight, height and MUAC is widely accepted and commonly performed in nutrition surveillance and interventions. A study reported that younger children tended to become upset and agitated during both weight and height measurement but not during MUAC measurement [37]. TSF index is not currently used for surveillance or programming and would probably not be as acceptable as the measures above as it requires the measurement of the width of a fold of skin taken over the triceps muscle using a skinfold caliper.

Cost

The measurement of height and weight requires fairly costly equipment [37–39] while the MUAC tool – a tape measure – is cheap and easy to transport. A caliper is also relatively costly and may be harder to procure.

Independence of age

One way to ascertain age-independence is to adjust indices for age (i.e. WFA, HFA, MUAC-for-Age) but the issue of the accuracy of age remains. MUAC is relatively age and sex independent among 1–5 year olds [32, 37, 40–43] as well as WFH [39, 44]. It was also shown that MUAC alone, without correction for age, was a better predictor of death than indices based on height, weight and age [41–43, 45].

Reliability and accuracy

Although weight and height alone were shown to be more precisely measured [46–49], it was reported that MUAC has a better reliability than WFH and shows better performance in screening programmes [50]. It was also shown that in field conditions, minimally trained workers make fewer and smaller errors in screening children with MUAC than with WFH [37]. Indices usually require finding values in tally sheets or calculations that can lead to further errors. A recent paper shows that MUAC is more reliable than WFH [51] and another that MUAC outperforms weight-based measures of nutritional status in children with diarrhoea [52]. It was also shown to be less affected by dehydration than WFH [53]. As mentioned above, any index requiring the age (i.e. WFA or HFA) of the child is likely to be less accurate.

Sensitivity and specificity (to mortality)

MUAC is increasingly recognised as a very useful index of nutritional status [50]. There is a consensus that MUAC is a better predictor of mortality than WFH [40, 45, 54–60] and it was recently reported that using MUAC alone is preferable for identifying high-risk malnourished children [61].

These findings are consistent with the increasing interest in MUAC-only nutrition programming and use for admission and discharge to feeding programmes [62–66]. This concordance makes the findings of this review applicable and of interest to international policy makers and programme managers.