A systematic review and meta-analysis of primary studies was conducted in accordance with the Preferred Reporting Items for Systematic and Meta-Analysis-2020 (PRISMA-2020) guideline [30], as shown in Supplementary Material S1. The study also followed a pre-designed systematic review protocol that was registered in the International Prospective Register of Systematic Reviews (PROSPERO)—registration number CRD42023476402. Following the preliminary search, the scope of the review became narrowed to determining the effects of amaranth-containing food products on hemoglobin concentration.

As we aimed to assess the best available evidence related to the effectiveness of amaranth, interventional studies were targeted to be included in the review. Accordingly, a Randomized Controlled Trial (RCT), Quasi-Experimental Study (QE) and pre-post observational studies aimed to assess the effectiveness of amaranth-based food intervention on individuals aged 6 months or older regardless of their health condition were included in this review. To minimize bias in selecting studies [31], both published and unpublished studies available online from January 2000 to April 2024 were targeted in a comprehensive database search.

Additional eligibility criteria included population, intervention types, comparator and outcome. Amaranth-containing food can be consumed by different age groups, including young children and adults [22, 32]. Children older than 6 months should start complementary food [33]. Therefore, the population in this study consists of individuals over 6 months of age, regardless of their health status or any demographic characteristics. In terms of intervention type, studies involving food interventions that included amaranth as a key ingredient, regardless of specific food type, were targeted. The intervention could have been implemented in different settings like households, schools, and communities. Comparisons were made either against other functional foods or in the absence of any interventions for the control group.

The main outcome of this study was to determine the effect of amaranth on anemia, which is assessed by hemoglobin levels. Anemia’s diagnosis and severity are determined by the cut-off value of hemoglobin concentration, which is influenced by factors like age, sex, health status, and altitude [34]. Hence, studies that assessed the levels of hemoglobin in both the intervention and control groups were sought by measuring the mean and standard deviation of hemoglobin before and after the intervention. Then mean difference was generated to see the effect of the intervention on hemoglobin concentration, as recommended in the Cochrane Handbook [31].

In converse, studies assessing amaranth’s nutritive value, qualitative investigation, book chapters, compressive reviews, systematic reviews, studies that missed reporting hemoglobin before and after the intervention in both the intervention and control groups and studies published in another language other than English were excluded from the review.

Online databases like PubMed, Scopus, Cochrane, Embase, Google Scholar, AJOL, and Web of Science were searched for the records. Furthermore, ProQuest’s dissertation and gray literature from search engines like Google were used to search unpublished studies. The initial search was done compressively using the following keywords: “effectiveness,” “efficacy,” “amaranth,” “amaranthus,” “amaranthus cruentus OR amaranthus caudatus OR amaranthus hypochondriacus,” “food OR snack,” and “intervention.” Then the search was updated by adding keywords like hemoglobin OR Anemia OR nutrition and by indexing dated between 1 January 2000 and 13 April 2024. The reference lists of eligible studies were also evaluated for potential articles based on selection criteria. The database search strategies used are provided as a supplementary document (Supplementary Material S2).

The bibliographies of articles identified from all databases were recorded and exported to EndNote [35] and stored in a respective database labelled group. MY and AE used the Covidence Systematic Review online software ¹ to independently review and screen articles by title, abstract and full text based on eligibility criteria. Any discrepancy was resolved by the third reviewer, AZ, and through discussion.

Methodological quality was assessed using the JBI systematic review critical appraisal tool for systematic review of effectiveness, which has a separate format to examine the quality of RCT and quasi-experimental studies [36]. To ensure the consistency of the data, the assessment was done independently by MY and AE using a table. In a situation where the design of the study was not clear, we decided on the protocol of the study, particularly based on randomization and the presence of a control group. Any discrepancy was resolved by discussing and reviewing the article together.

Data were extracted from eligible studies in Excel using a pre-structure sheet by two independent reviewers (MY and AE) focusing on study design, year of publication, type of study, sample size, type of food intervention (drink, meal, snack, juice or else), part of amaranth used (leaves or grain), amaranth species, population, comparator food, follow up time, frequency of meal, statistical analysis, findings and conclusion. Any discrepancy was resolved by discussing and extracting the data together.

The characteristics of included studies, like the type of intervention, target population, the health status of the study participants, species, used amaranth part (leaves or grain), frequency of consumption and follow-up period, were synthesized and described in the table. Unreported standard deviation was obtained using the square root of n times the width of the confidence interval divided by 3.92 [29].

Meta-analysis was conducted using Stata version 17 to generate the pooled effect of amaranth-containing food on hemoglobin concentration. The standardized mean difference (Hedges’ g) was used to estimate the overall effect size considering variation in the population, intervention, and measurement of outcome. The pooled estimate was summarized and presented in a forest plot. A random effect model was used considering the characteristics of studies and observed heterogeneity. Publication bias was assessed using a funnel plot and Egger’s test. In response to a significant small study effect, trim-and-fill estimation was done and described using a trim-and-fill funnel plot [37]. A subgroup analysis was done to examine the source of heterogeneity.

Certainty of evidence and summary of findings on the effectiveness of amaranth-containing food on hemoglobin concentration was assessed using a Grading of Recommendations, Assessment, Development and Evaluation (GRADE) approach and the summary of findings table is presented as a supplementary document (Supplementary Material S3).

A systematic search of studies was conducted to access published and unpublished studies from databases like Embase, Scopus, PubMed, Cochrane and AJOL. Furthermore, Google Scholar and ProQuest were used for articles and dissertations. A total of 1,032 articles were found from database and citation search, and only 10 articles were found to fulfill the inclusion criteria from 70 overall records screened in full text, as depicted in Figure 1.

Methodological quality was assessed using the JBI systematic review critical appraisal tool for a systematic review of effectiveness review of RCT and QE studies [36]. The baseline characteristics of comparison groups in some of the QE studies were different in some of the included quasi-experiment study designs [38, 39]. The study design was not clearly specified as RCT or quasi-experimental [28, 39]. As a result, we considered the study design to be quasi-experimental design based on the study protocol. One QE study did not discuss the treatment other than the intervention given for the comparison group. Furthermore, the statistical analysis of the included studies was not appropriate to compare differences between groups (Table 1) [28].

The assessment of the methodological quality of RCT showed that allocation concealment was only described by half of the included studies. Most studies did not report information regarding the blinding of the study of participants, treatment providers and outcome assessors [41–43]. One study discussed that all measurements were blind. However, the authors did not describe blinding participants, those delivering treatment and outcome assessors [44]. The baseline characteristics of study participants were also different in some of the studies [32, 42]. Regarding the statistical analysis, some studies did not report the approach they followed as an intention-to-treat or per protocol (Table 2) [41, 43, 45, 46]

A total of ten studies were included, of which 60% them were RCT and 40% were QE studies. Most studies were published between 2012 and 2019 from five countries which are Indonesia [3], Kenya [3], South Africa [1], Ethiopia [1] and India (one published and one unpublished). About 30% of included studies were community based and 20% were school based. Most of the studies (70%) included a younger population, and one study included post-partum women without age specification. The total sample size was 1,225 (614 in the intervention and 611 in the control group), ranging from 20 to 334 and follow-up period from 2 to 72 weeks. Furthermore, about 70% of included studies used leaves of amaranths to prepare the intervention like juice, herbal drinks, porridge and bread (Table 3).

The minimum standardized mean difference for the effect of amaranth-containing food on hemoglobin was zero (95%CI: −0.29, 0.29) [42] and the largest effect size was 1.46 (0.51, 2.41) [28]. The forest plot indicates the presence of heterogeneity where I² is 57.1%. To investigate the possible reason behind observed variation, subgroup analysis was conducted, and it showed that there is a higher heterogeneity among studies that recruited anemic subjects (86.02%) than healthy subjects (no heterogeneity). Similarly, heterogeneity was higher among quasi-experimental studies (75.3%) than RCT (36.3%). Studies that used amaranth leaves showed a higher heterogeneity (71.7%) than those that used amaranth grain (67.3%).

The standardize mean hemoglobin concentration difference between the intervention and control groups found to be positive, with the overall pooled effect of amaranth-containing food on hemoglobin concentration being 0.08 (95% CI: −0.11, 0.26). The statistical test of the overall effect size is not different from zero (p-value 0.4330). The studies by Singh et al. [28] and Muliani et al. [40] may affected the findings with a low sample size and large effect size (Figure 2). To identify the most influential studies, a leave-one-out graph was constructed, and it was found that omitting Muliani et al. [40] and Singh et al. [28] decreased the overall effect size by 0.07.

The asymmetric distribution of the study in the funnel plot indicated the presence of publication bias (Supplementary Material S4). Furthermore, egger’s test of small study effect (p-value < 0.001) suggests the presence of a small study effect. The trim-and-fill method which used to correct publication bias by imputing two studies where only one falls within alpha < 0.05. The trim-and-fill estimator changes the direction and magnitude of the overall effect size −0.01 (95%CI: −0.32, 0.3) (Supplementary Material S4).

The effect of the intervention was compared across different groups such as study design, population and amaranth part used. Accordingly, the result showed a significant positive effect of the intervention for postpartum women, 0.85 (95%CI; 0.12, 1.58) (Figure 3).