Ferrous sulfate raised haemoglobin by 0.42 g/dL (95% CI 0.10-0.74; 8 RCTs) and ferritin by 12.61 ng/mL (8.29-16.92; 7 RCTs) against control in active women and girls, most in groups whose mean starting ferritin was under 20 ng/mL (ferritin +14.16); across 42 diet and supplement trials none of the 15 interventions that measured VO2max, lactate or inflammation changed them significantly, by the authors' summary. Read in full
Main caveat: The authors': short trials (34 of 42 under 12 weeks) and small samples (36 of 42 under 50 people); reproductive history and contraception seldom recorded.
McCarthy and colleagues (University of Illinois Chicago) reviewed every interventional trial they could find in which a diet or dietary supplement was tested on iron status in physically active females of reproductive age, and pooled the trials that allowed it.
From 6759 records they included 42 trials, run on five continents with 5 to 171 participants each: 18 of oral iron alone, 9 combining oral iron with another dietary component (vitamin C, milk, a probiotic, lactoferrin), 5 of food-based iron, 5 of dietary or lifestyle change and 5 of other supplements. Only ferrous sulfate had enough parallel-arm RCTs to pool.
Against control, it raised haemoglobin by 0.42 g/dL (95% CI 0.10-0.74; 8 RCTs; I² 56.6%) and ferritin by 12.61 ng/mL (8.29-16.92; 7 RCTs; I² 57.38%), and leave-one-out analyses kept both significant.
By baseline status, the gains were significant in trials whose mean ferritin was under 20 ng/mL (ferritin +14.16, haemoglobin +0.49) and not in the single trial with mean ferritin 20-35 ng/mL; the two trials with normal mean status were mixed. Longer trials went with smaller ferritin gains and, in the joint model, larger haemoglobin gains; dose was not a significant predictor.
Outside the meta-analysis, oral iron with or without vitamin C and some iron-rich foods improved iron markers, other supplements (colostrum, pea or soy protein, a carbohydrate-protein drink) mostly did not, and a ketogenic diet lowered haemoglobin.
The authors conclude that ferrous sulfate improves haemoglobin and ferritin in active females, most clearly from low stores, and that care should be personalised.
Participants
Eligibility: females aged 12-52 who are physically active, defined as McKay "tier 2" or above, training at least three times a week and identifying with a sport; pregnant or lactating women and studies of males only or not split by sex excluded.
Included: 42 trials; participant ages 12-45; sample sizes 5 to 171; 17 trials in North America, 11 in Europe, 7 in Asia, 4 in Australia, 3 in Africa.
Pooled ferrous sulfate RCTs (haemoglobin, from Table 3): rowers (n = 31), mixed competitive sports (31), runners (18), military trainees (171), volleyball players (22), intercollegiate athletes (26, haemoglobin only), high-school cross-country runners (14) and mixed sports (15); 328 participants for haemoglobin and 302 for ferritin (our sums).
Reproductive health: contraceptive use and reproductive history were recorded in only a few studies.
Protocol
Search: five databases to March 2026, with a university librarian; 4318 duplicates removed, 2441 titles and abstracts screened, 231 full texts, 42 included.
Inclusion: single-arm, parallel or crossover, randomised or non-randomised interventional trials of a diet or dietary supplement with at least one iron marker (haemoglobin, haematocrit, ferritin or serum iron); observational studies, animal or cell work and reviews excluded. Two independent reviewers at each stage (Covidence).
Risk of bias: Cochrane RoB 2 (randomisation, deviations, missing data, outcome measurement, reporting), applied to all 42 trials, randomised or not.
Meta-analysis: parallel-arm RCTs with arm-level pre and post means, SDs and sizes, ferrous sulfate as the only iron compound; mean differences, random effects (rma in metafor), I², τ² and Cochran's Q; leave-one-out; subgroups by mean baseline ferritin (normal > 35 ng/mL; stage 1 iron deficiency < 35; stage 2, iron deficiency without anaemia, < 20); meta-regression on dose, duration and baseline status. Medians converted to means by Wan's method; g/L converted to g/dL.
Funding and interests: no funding received; the authors declare no conflicts of interest.
Key Results
Ferrous sulfate against control (random effects):
| Outcome | Mean difference (95% CI) | Trials | Heterogeneity |
|---|---|---|---|
| Haemoglobin | +0.42 g/dL (0.10-0.74), p = 0.0109 | 8 RCTs | I² 56.6%, τ² 0.11, Q = 15.42 (p = 0.031) |
| Ferritin | +12.61 ng/mL (8.29-16.92), p < 0.0001 | 7 RCTs | I² 57.38%, τ² 17.66, Q = 13.99 (p = 0.0297) |
| Ferritin, baseline mean < 20 ng/mL (iron deficiency without anaemia) | +14.16 ng/mL (9.22-19.10), p < 0.0001 | subgroup | — |
| Haemoglobin, baseline mean < 20 ng/mL | +0.49 g/dL (0.1651-0.8244), p = 0.0033 | subgroup | — |
| Ferritin, baseline mean 20-35 ng/mL | positive, not significant | 1 trial | — |
| Ferritin, normal baseline (> 35 ng/mL) | mixed: one significant gain, one with wide uncertainty | 2 trials | — |
Sensitivity: leave-one-out ferritin estimates 11.28 to 13.6 ng/mL, all significant; for haemoglobin one trial (Mielgo-Ayuso) was moderately influential, and the result held.
Meta-regression, ferritin: dose β 0.0109 ng/mL per mg of iron (−0.0004 to 0.0221, p = 0.0578, not significant; dose explained 49.91% of heterogeneity); duration β −1.44 ng/mL per week (−2.44 to −0.45, p = 0.0046), longer trials going with smaller gains. In the joint model (dose, duration, baseline status) no single predictor was significant (duration β −1.14, p = 0.063; dose β 0.005, p = 0.35).
Meta-regression, haemoglobin: dose β 0.0003 g/dL per mg (−0.0009 to 0.0016, p = 0.6); duration β 0.11 g/dL per week (−0.0074 to 0.2329, p = 0.0658); in the joint model duration was the only independent predictor (β 0.16 g/dL per week, 0.02 to 0.31, p = 0.031).
Risk of bias (RoB 2, authors): low overall across domains; 16 of 42 trials had some concerns or high risk in randomisation and 9 of 42 in deviations from the intended intervention. All 8 pooled ferrous sulfate RCTs are rated low in every domain (Table 2).
Narrative findings (not pooled): in some trials oral iron held ferritin steady while it fell in the placebo group (military trainees and volleyball players among them), which the authors read as preventing a decline rather than refilling stores; vitamin C, probiotic or prebiotic additions to oral iron gave favourable iron results; iron-fortified food bars, an iron-rich diet and an iron-rich porridge improved some markers; colostrum, pea powder, soy protein and a carbohydrate-protein drink did not; a 4-week ketogenic diet lowered haemoglobin and Ramadan fasting left it unchanged.
Performance and inflammation: the authors state that none of the 15 interventions measuring VO2max, lactate, hepcidin, CRP or IL-6 changed them significantly.Our reading of Table 3 is less clean: it marks a significant VO2max increase against control in one non-randomised running trial (LaManca 1993), within-group lactate falls in two trials (Fogelholm 1992, in the placebo arm; Schoene 1983, in the iron arm), and significant differences against control in IL-6 with hydrogen-rich water (Sha 2018) and in CRP with fortified food bars (Karl 2010).No performance outcome was pooled.
Practical Application
When a clinician treats a measured iron deficiency in an active woman, ferrous sulfate raises ferritin by about 12-14 ng/mL and haemoglobin by about 0.4-0.5 g/dL against control in trials of 2-11 weeks (the pooled trials, from Table 3), most from the lowest stores (mean ferritin under 20 ng/mL); it is not shown to improve VO2max. The dose, the schedule and the decision to treat stay with the athlete's physician; this is not treatment advice.
Test before iron still holds. The gains were significant where mean ferritin started under 20 ng/mL and not shown in the one trial starting at 20-35 or consistently in the two with normal mean stores; this matches SYNTH-97's "treat only a measured deficiency".
Where this differs from the guide's evidence: SYNTH-97, citing REF-1409 (13 RCTs), says haemoglobin did not change significantly with oral iron (in REF-1409 a standardised pooling of 12 trials, SMD 1.31, 95% CI −0.29 to 2.93, I² 97%); this ferrous-sulfate-only pooling in active females found a small significant rise (0.42 g/dL, CI 0.10-0.74) with moderate heterogeneity. Both agree that ferritin rises and that a VO2max gain is not shown.
Do not judge a course too early or by ferritin alone. Longer trials went with larger haemoglobin gains and smaller ferritin gains, which the authors read as stores filling; SYNTH-97's re-test after 6-8 weeks is consistent with this, though the review did not test re-test timing.
Food and other supplements: iron-rich foods and fortified foods improved some markers in small, mostly non-randomised trials; protein powders, colostrum and a carbohydrate-protein drink did not, and a ketogenic diet lowered haemoglobin. The authors' suggestion to take iron in the morning or within 30 min after morning exercise is their inference from hepcidin's daily rhythm; no included trial tested timing.
Limitations
The authors': short trials (34 of 42 under 12 weeks) and small samples (36 of 42 under 50 people); reproductive history and contraception seldom recorded; dietary iron and vitamin C intake rarely controlled; baseline iron status often not accounted for; activity level and type varied, some trials introducing new training; supplement timing seldom reported; stratified analyses underpowered.
Certainty: no GRADE or other certainty rating. RoB 2 was applied to non-randomised and single-arm trials, for which it was not designed, so the "low overall" rating covers designs it does not fit (our reading).
Dose: Table 3 gives some doses as ferrous sulfate salt (for example 100 or 975 mg) and others as elemental iron (for example 50 or 105 mg), and the text states that dosages ranged from 100 to 975 mg daily; the meta-regression is expressed "per mg of iron" without saying which, so the dose result cannot be turned into a milligram figure (our reading).
Text against tables: the Discussion says a greater dose was associated with a greater ferritin response, but the dose term was not significant (p = 0.0578) and lost weight in the joint model (p = 0.35); the claim that no performance outcome changed sits against Table 3's VO2max mark for one non-randomised trial.
Population: the pooled trials mix military trainees in basic training, high-school runners and collegiate and club athletes, from 1979 to 2015, with baseline status from normal to iron deficient; one pooled trial used no-treatment rather than placebo control.
Not read: the supplementary search strings; the forest-plot values exist in the text only as the pooled summaries.
How this entry was checked and changed
Why this level
the pooled part is a meta-analysis of RCTs whose risk of bias we read (all 8 pooled trials rated low in every domain by the authors), held to A- by mostly small trials (14-171 per trial) and moderate heterogeneity with no certainty grading; no marker step, because haemoglobin and ferritin are the iron-status outcomes this entry is about (rule 3), the same reading as the base's other oral-iron meta-analysis