The intervention agents included a placebo (Melojel® corn starch) and three forms of iron: FeSO₄·7H₂O (FS), IHAT, and ASP. Reagent grade FS was used. In Phase I, one arm also included a multiple micronutrient powder [MNP; MixMe™ Vitamin and Mineral Powder (DSM Nutritional Products, Geneva, Switzerland) (29)] the contents of which are based on recommendations by UNICEF/WHO/WFP for one RDA of 15 vitamins and minerals.

IHAT, developed by the Medical Research Council Elsie Widdowson Laboratory, Cambridge, UK, is composed of three GRAS substances: iron hydroxide, tartaric acid and adipic acid. It is a nano-particulate iron supplement, recently approved by the European Commission as a novel food (2022/1373; Aug. 5, 2022). IHAT is a tartrate-modified, nano-dispersed Fe(III) oxo-hydroxide with similar functional properties and small primary particle size (~2 nm) to the iron form found in the ferritin core (i.e., ferrihydrite) (30), and has been designed to have a benign side-effect profile by withholding any unabsorbed iron from redox activity and the gut flora (22, 30–33). Insoluble in the gut lumen (31, 32), IHAT appears poorly utilized by enteric bacteria (22); thus, it may be less irritating to the gut and less pro-inflammatory than FS. IHAT-iron appears to enter the metabolic iron pool more slowly than Fe-iron (33), suggesting a lesser post-absorptive surge in pro-oxidative NTBI. In female adults who were iron-deficient, the bioavailability of IHAT 75% that of FS (33). The European Food Safety Authority found the no observed adverse effect level of IHAT to be 231 mg (77 mg Fe) per kg body weight per day (34).

ASP (Aspiron™ Natural Koji Iron), a dried, iron-enriched Koji biomass containing 8–10% iron, was developed by Cura Global Health Inc. (Ames, IA). Koji is a GRAS constituent of food considered safe by the FAO/WHO Committee on Food Additives and (23, 24). Its bioavailability for the rat is of 60% of FS (25). Studies with female adults who were healthy and non-anemic with marginal iron stores (serum ferritin <40 μg/L) showed the absorption of ASP-Fe to be comparable to that of FS-Fe, but slower to appear in the plasma where it persisted longer, produced less NTBI, and produced fewer side effects (27, 35, 36).

Each of these agents was approved for human use. Each was encapsulated in opaque, double-zero gelatin capsules identical in size, color, and weight. This was done by the Richardson Centre for Functional Foods and Nutraceuticals, University of Manitoba (Winnipeg, Manitoba, Canada), which is licensed as a Health Canada Natural Health Product Site with the ability to produce good manufacturing practices and Hazard Analysis and Critical Control Points certified products.

The two phases of this study were conducted sequentially, i.e., Phase I (Figure 1) was completed in its entirety prior to the initiation of Phase II (Figure 2). For each phase of the study, a randomization scheme for entering participants was generated using SIMD-oriented Fast Mersenne Twister algorithm (v. 1.5.1)-based randomization functions in the StatsBase.jl Julia package,¹ stored, and known only to the data manager who held the unblinded key under lock and key data collection for the entire study. For Phase I, 160 participants were randomly assigned to one of the six treatment arms at the time they were enrolled in the study (Figure 1): placebo; FS daily (60 mg iron/d); FS weekly (420 mg iron/wk); FS + MNP (60 mg iron/d); IHAT (60 mg iron/d); and ASP (60 mg iron/d). The dose of 60 mg iron /d was based on the WHO recommendation for daily supplementation for non-anemic, pregnant females (30–60 mg iron/d) (37). A separate treatment consisting of weekly supplementation of 420 mg iron in four capsules taken once weekly; this dose regimen has been recommended as feasible, cost-effective in population-base programs (38–40), and may produce fewer gastrointestinal complaints than daily FS doses. In addition, a treatment involving the co-administration of FS and MNP was included to examine the safety of this common practice, as components of MNP, particularly ascorbic acid, have been shown to promote the enteric absorption of ferrous iron (41). The corn starch placebo provided a control for prospective changes over time, effects of daily ingesting capsules, and any time-related and/or and seasonal effects. For Phase II, 86 participants were randomly assigned to one of three treatment arms (Figure 2): IHAT (120 mg iron/d); ASP (120 mg iron/d); and FS (120 mg iron/d). All groups received four physically similar capsules to ensure blinding of the treatments. The dose of 120 mg Fe/day was based on the report of the Food and Nutrition Board (42), which suggest low adverse effects levels (constipation, diarrhea, black stools, gut irritation) of FS at that dose. Because this phase comprised direct comparisons between the three sources of iron, placebo control was not included.

At enrollment, informed consent was obtained and instructions for stool sample collection were reviewed. The participant was then randomly assigned to a treatment group and was scheduled for a baseline, pre-intervention visit (wk 0) at which a stool sample and a fasting blood sample (7 ml) were collected, and the participant was then given a standard continental breakfast (English muffin or bagel, spreads, coffee, juice, fruit excluding orange products, and teas). Immediately following the breakfast, the participant was given the assigned blinded intervention agent for the baseline (wk 0) dose. Instructions for the intervention were reviewed, and the participant was sent home with a pre-loaded calendar pack containing 90 capsules (three to be taken each day for 30 days). For the weekly FS group, treatment concealment was deemed impractical and therefore was not used; each participant received 16 capsules (four to be taken once weekly for 4 weeks). Each instructed to take all their scheduled capsules within a few minutes, was also given a compliance calendar to record capsule intake and was instructed to return the calendar with any unused capsules at the final study visit. Adherence to supplement protocol was assessed by counting the number of capsules remaining in the returned planner. Two hours after the breakfast, a second, post-prandial (PP) blood sample (7 ml) was drawn for use in bacterial proliferation assays.

The study was powered to detect treatment effects on the following endpoints:

Also addressed were the following secondary endpoints for which the study was not powered:

Data were managed using a REDCap (61) database and handled in compliance with HIPAA and 21CFR11. Participant identifiers were recorded in separate electronic case reports and were maintained separately from research data. Each participant’s identity and records of telephone pre-screening, final screening, and study data were kept in locked files or eCRF, with only the PI (GFC), Study Physician (JBM), and Study Coordinators having access to participant identifying information if needed. Otherwise, only de-identified data were made available to study co-investigators. Data entered into the database underwent external validation checks; meta-data files and data dictionaries were used to provide information necessary for proper use and understanding of the data files. Laboratory specimens were identified by unique study codes; the master list linking the identities of participants and specimens was kept in a locked file and on a separate, password-protected database on a secure server. REDCap was used for edit resolution; trouble shooting data entry problems; an audit trail of database editing; performing range checks for cleaning, daily backup, cleaning of transitional databases; and, ultimately, transferring master databases for statistical analysis. At key points in data acquisition, clean-up and analysis, an additional off-site copy of the data was stored using the LabArchives electronic laboratory notebook software (LabArchives LLC, San Marcos, CA, USA). Access to the study database was restricted until the study was completed and unblinded. Protected participant health information was automatically de-identified for all viewing and data exports.

The sample sizes used in both phases were based on power calculations for each primary outcome as previously described (26). For malarial infectivity, a standard deviation (SD) of 0.202 was estimated from published studies with P. falciparum infectivity (62). Based on an independent two-sample t-test with 0.05 type I error, it was determined that 25 participants per group would yield 90% power to detect a 20% reduction in parasite infectivity. For bacterial proliferation, between-participant estimates of SD for doubling time during the exponential phase of growth were estimated for several species from published data (59). This indicated that, with 25 participants per treatment group and 0.01 type I error (adjusted for planned analyses with five species), power was >0.99 for a minimal detectable difference in doubling time of 1 h. Mean fecal calprotectin was reported previously to be 1.9 μg/g in participants with minimal inflammation (irritable bowel syndrome) and 27.6 μg/g in participants with mild inflammation (collagenous colitis) with a SD of 1.3 in log concentration (63). A sample size of 25 participants per group was estimated to allow the detection of a 3.5-fold change in fecal calprotectin values with 90% power at a type I error level of 0.05.

An intention-to-treat analysis was performed with all participants according to their randomization assignment. Analysis excluding participants with low supplement adherence was also performed, however did not substantially affect results; thus, the results from the intention-to-treat approach are presented. Participant characteristics and dietary intake were summarized for each intervention group using mean, median, or frequency, along with SD, range, or interquartile range, as appropriate. Statistical testing was structured to test the study hypotheses and objectives in a series of planned pair-wise comparisons. The formal hypotheses of the study (26) were tested by evaluating the following outcomes: metabolic responses as assessed by biomarkers of iron status; the susceptibility of participant RBCs to infection by P. falciparum as assessed ex vivo; the ability of participant plasma to facilitate proliferation of selected species of pathogenic bacteria as assessed ex vivo; and inflammatory responses as assessed by fecal biomarkers of inflammation. Treatment effects were evaluated by the following sets of pairwise comparisons:

Phase I (six treatments each providing 60 mg Fe/day):

An alpha level was designated for each set of comparisons to achieve a familywise error rate of 0.05 for each outcome. Models were fit using the lmer4 package in R v. 4.2 and p values from pairwise comparisons were adjusted based on Tukey’s HSD method applied in the emmeans package v. 1.7.4-1 (64). Four-week changes were assessed using linear models adjusted for age, sex, and BMI. A participant-specific random effect was added to the models for outcomes that were assessed both at fasting and 2-h PP. Additionally, noninferiority tests were performed for the primary outcomes in Phase I. These tests were not prespecified but were added to the analysis to characterize comparisons of the supplemental iron forms to FS daily, given the interpretation of these comparisons as one-sided equivalence hypotheses. A conservative noninferiority margin was set at 50% of the 4 wk. change between placebo and FS daily. Data were evaluated graphically to assess modeling assumptions including outlier detection, normality, and linearity with covariates. Log transformations were used when appropriate to satisfy normality assumptions.

Susceptibility to malarial infection. The magnetic isolation of mature parasites yielded nearly homogeneous populations of schizonts without complications of uninfected erythrocytes. Schizont morphology, as examined microscopically, showed no effects of participant iron supplementation; merozoites released from purified schizonts were able to invade erythrocytes directly. That merozoites were also able to re-invade participant erythrocytes (data not shown), demonstrated normal development of parasites under these conditions. That no form or level of iron affected malarial invasion of participant erythrocytes (Tables 2, 8) suggests that iron supplementation of these iron-replete adults did not affect their apparent risk to clinical malaria infection. This finding is consistent with minimal effects on plasma NTBI, which has been proposed to increase the intensity of malarial infections (19, 20). Still, the possibility that the ex vivo assay may not fully reflect susceptibility to infection in clinical settings cannot be excluded. This study did not address the role in the malarial life cycle of hepatocytes in which schizonts are produced from sporozoites delivered by Anopheles injection into the bloodstream. Therefore, these results cannot be taken as dispositive of the hypothesis that iron supplementation of young iron-replete individuals can enhance clinical malarial infection as proposed by Clark et al. (62).

Iron status. The participants in both phases of this study were iron-replete, as indicated by their normal hemoglobin levels, their estimated dietary iron intakes, which exceeded the recommended daily allowance (8 mg) (42), and their baseline values of biomarkers of iron status (Tables 5, 11). In no case did any form of supplemental iron, at either the 60 mg/day or 120 mg/day dose level, increase iron status (e.g., transferrin saturation remained ca. 30%). That no increase in serum iron or transferrin saturation was detected 2 h after iron dosing (data not shown), is as expected for iron-replete individuals within the timeframe of these studies, suggesting that absorption was low for all iron supplements and implying that iron treatment did not increase NTBI in the plasma (19) and that lower gut exposure to non-absorbed iron was significant.

While the Food and Nutrition Board (42) suggested 120 mg Fe/day as the LOAEL for FS, a more recent meta-analysis of more than 40 clinical trials (65) found that the prevalence of gastrointestinal complaints associated with FS administration appears to be idiosyncratic and not related to dose. Accordingly, the observation that, in Phase II of this study, neither IHAT nor ASP providing 120 mg iron/day produced a pattern of reported gastrointestinal symptoms different from FS is not surprising, as the insolubility of each in the post-duodenal gut would suggest that neither form has substantial interactions with those epithelial surfaces.

This was a rigorously designed study that addressed the effects of multiple modalities of iron supplementation of iron-replete individuals on key aspects of malarial infectivity, bacterial infections, and gut inflammation. Both phases of the study were powered on these primary outcomes which, thus, had type I error control. Limitations include that we did not perform power calculations for other outcomes, i.e., iron status parameters, markers of gut inflammation besides fecal calprotectin, and GI side effects, or noninferiority tests. That iron supplementation did not affect malarial infection of erythrocytes does not imply that such treatment may not affect susceptibility to malarial infection in clinical settings, as we did not access effects on hepatocytes, which are also important in the cycle of malarial infection.