A total of 3097 articles were retrieved from the databases and registers; two articles were identified using citation search. After duplicate removal, 2171 articles were screened, and 24 were considered for full-text review, from which 10 articles were excluded (Online resource Table 2), from the 14 articles left, 13 articles were included for the systematic review, and one was an ongoing trial (Fig. 1).

PRISMA flow diagram of trial selection

The 13 trials included [42, 43, 55,56,57,58,59,60,61,62,63,64,65] accounted for a total of 910 participants. Six studies were conducted in adult population, including adult women of reproductive age (n = 273) [42, 57, 62,63,64,65] and men (n = 49) [56, 63], three studies in children (n = 154) [43, 55, 60], two in female adolescents (n = 200) [58, 59] and two in pregnant women (n = 232) [61, 62]. The trials were conducted in Norway [56, 61, 62, 64, 65], Sweden [57, 63], Mexico [55, 58, 59], United States [42], Guatemala [60] and Peru [43]. Most of the evidence was indirect as only one study administered HI as a single supplement (n = 1) [42], while the others combined HI with NHI formulations through supplements (n = 7) [43, 56, 61,62,63,64,65], or use dfortified foods with HI  (n = 5) [55, 57,58,59,60]. Baseline iron stores were different among studies, where participants initiate either with variable iron stores (low to moderate or normal levels) (n = 9) [42, 55, 57,58,59,60,61,62,63] or with iron deficiency (n = 4) [43, 56, 64, 65]. The follow-up intervention period ranged from 7 to 36 weeks. One ongoing trial was found, using HI polypeptide compared with ferrous salts in Gambian children aged 6 to 12 months with IDA [66]. Details about the characteristics of the included trials are provided in Table 1 and the summary of the outcomes are available in Tables 2 and 3.

The overall risk of bias assessment showed that most study outcomes presented some concerns due to a lack of available study protocols to assess selection of reported results [43, 55, 56, 58,59,60,61,62,63,64, 64]. Other studies presented a high risk of bias because outcomes reported to be measured in methodology could not be found in the manuscript or with data requests to the authors [42, 62]. Some studies presented some concerns in the randomization process since the precise randomization methods were not clearly described [42, 43, 57, 61, 62, 64]. A high risk of bias due to missing outcome data was also present; two studies excluded or lost participants due to infections or intolerance, and presented unequal proportions of participants in their groups at endpoint [55, 57]. Some studies showed some concerns in deviations from intended interventions arising from manufacturing problems or using an open-label design [43, 57, 61, 65] (Online resource Table 3).

All included studies reported this outcome [42, 43, 55,56,57,58,59,60,61,62,63,64,65]. However, the study by Swain et al. [42] did not provide quantitative data, instead stating that no difference was observed between the groups [42]. The meta-analysis showed no statistically significant difference comparing HI with NHI (MD 0.22 g/dL; 95% CI: − 0.05; 0.48; I2 = 80%; Fig. 2). The subgroup analysis of the overall population showed no difference between treatments for children (MD 0.54 g/dL; 95% CI: − 0.39; 1.46; I2 = 90%), female adolescents (MD 0.36 g/dL; 95% CI: − 0.32; 1.05; I2 = 89%), pregnant women (MD 0.07 g/dL; 95% CI: − 0.41; 0.54; I2 = 70%) and women of reproductive age (MD 0.04 g/dL; 95% CI: − 0.19; 0.27; I2 = 0%). However, point estimates were heterogeneous for children, female adolescents, and pregnant women population subgroups. The test for subgroup differences was not statistically significant for this analysis (p = 0.64; Fig. 2).

Forest plot comparing hemoglobin mean difference between HI vs NHI across population subgroups

Conversely, when comparing subgroups based on the iron status of children and women of reproductive age, the analysis revealed a statistically significant difference among children (p < 0.01