Iron Supplementation Affects Hematologic Biomarker Concentrations and Pregnancy Outcomes among Iron-Deficient Tanzanian Women^,,

Abstract

Background: Iron deficiency is a highly prevalent micronutrient abnormality and the most common cause of anemia globally, worsening the burden of adverse pregnancy and child outcomes.

Objective: We sought to evaluate the response of hematologic biomarkers to iron supplementation and to examine the predictors of the response to iron supplementation among iron-deficient pregnant women.

Methods: We identified 600 iron-deficient (serum ferritin ≤12 μg/L) pregnant women, aged 18–45 y, presenting to 2 antenatal clinics in Dar es Salaam, Tanzania using rapid ferritin screening tests, and prospectively followed them through delivery and postpartum. All women received 60 mg Fe and 0.25 mg folate daily from enrollment until delivery. Proportions meeting the thresholds representing deficient hematologic status including hemoglobin <110 g/L, ferritin ≤12 μg/L, serum soluble transferrin receptor (sTfR) >4.4 mg/L, zinc protoporphyrin (ZPP) >70 mmol/L, or hepcidin ≤13.3 μg/L at baseline and delivery were assessed. The prospective change in biomarker concentration and the influence of baseline hematologic status on the change in biomarker concentrations were assessed. Regression models were estimated to assess the relation of change in biomarker concentrations and pregnancy outcomes.

Results: There was significant improvement in maternal biomarker concentrations between baseline and delivery, with increases in the concentrations of hemoglobin (mean difference: 15.2 g/L; 95% CI: 13.2, 17.2 g/L), serum ferritin (51.6 μg/L; 95% CI: 49.5, 58.8 μg/L), and serum hepcidin (14.0 μg/L; 95% CI: 12.4, 15.6 μg/L) and decreases in sTfR (−1.7 mg/L; 95% CI: −2.0, −1.3 mg/L) and ZPP (−17.8 mmol/L; 95% CI: −32.1, 3.5 mmol/L). The proportions of participants with low hemoglobin, ferritin, and hepcidin were 73%, 93%, and 99%, respectively, at baseline and 34%, 12%, and 46%, respectively, at delivery. The improvements in biomarker concentrations were significantly greater among participants with poor hematologic status at baseline – up to 12.1 g/L and 14.5 μg/L for hemoglobin and ferritin concentrations, respectively. For every 10-g/L increase in hemoglobin concentration, there was a 24% reduced risk of perinatal mortality (RR = 0.76; 95% CI: 0.59, 0.99) and a 23% reduced risk of early infant mortality (RR = 0.77; 95% CI: 0.60, 0.99). The risk of anemia at delivery despite supplementation was predicted by baseline anemia (RR = 2.11; 95% CI: 1.39, 3.18) and improvements in ferritin concentration were more likely to be observed in participants who took iron supplements for up to 90 d (RR = 1.41; 95% CI: 1.13, 1.76).

Conclusion: Iron supplementation decreases the risk of maternal anemia and increases the likelihood of infant survival among iron-deficient Tanzanian pregnant women. Interventions to promote increased duration and adherence to iron supplements may also provide greater health benefits.

Introduction

Iron deficiency, the most common nutritional abnormality globally and a leading risk factor for disease, is commonly found among pregnant women and children due to increased demand for maternal and fetal metabolism and growth during childhood (1, 2). In resource-limited settings, infectious diseases, especially malaria, worsen anemia and iron deficiency by inducing hemolysis and anemia of inflammation and impairing iron absorption and distribution (1). Iron deficiency anemia substantially increases the risk of low birth weight, preterm births, and early infant mortality (2). Although prenatal iron supplementation has been shown to reduce the risk of many of these conditions (3), the burden of iron deficiency anemia remains very high, and coverage for iron-folic supplementation remains very low in many countries– ∼10% in Sub-Saharan Africa including Tanzania (4, 5).

An understanding of the individual-level determinants of effective supplementation may improve detection of iron deficiency and delivery of iron supplementation interventions. Although nutritional and genetic factors have been suggested to influence the response to oral supplements (6–8), there is limited understanding of the magnitude of response to supplementation in malaria-endemic settings and the factors driving it. The choice of hematologic biomarkers to assess the effectiveness of supplementation in these settings is also unclear. Hemoglobin is affected by several factors other than iron status, thus limiting its sensitivity and specificity, and its use in combination with other biomarkers, notably ferritin, has been recommended (9, 10). Concentrations of ferritin, the best surrogate measure of body iron reserves, and zinc protoporphyrin (ZPP)¹¹, which represents iron-deficient erythropoiesis, are altered by inflammation, making the distinction of malarial inflammation from anemia of chronic disease challenging in endemic settings (10). Serum soluble transferrin reception (sTfR) is less influenced by inflammation but it requires standardization to be suitable for use on a large scale in clinical or program settings (10). Hepcidin, measured in blood or urine, is more responsive to erythropoietic demands for iron than to inflammation, especially in a setting of iron deficiency, although its suitability for large-scale use is poorly understood (11, 12).

The aim of the study was to evaluate the response of hematologic biomarkers as well as pregnancy and neonatal outcomes to iron supplementation and to examine the predictors of the response to iron supplementation among iron-deficient pregnant women.

Methods

Study design and population.

This was a prospective cohort study of pregnant women presenting to 2 antenatal clinics in Dar es-Salaam, Tanzania. Participants were recruited between January 2012 and March 2013 and followed up monthly until delivery and 6 wk postpartum. They were eligible if they were 18–45 y old, HIV negative, in their first or second pregnancy, presenting for antenatal care before 28 wk gestation, planning to stay in Dar es Salaam until delivery, and iron-deficient at baseline–serum ferritin concentrations ≤12 μg/L following a confirmatory test (see later) (13). The participants received a capsule containing 60 mg Fe (200 mg of ferrous sulfate) and 0.25 mg folate (Tishcon Corporation) to be taken 1 time/d from enrollment until delivery and intermittent preventive treatment for malaria (IPTp-SP; 1500 mg sulfadoxine, 75 mg pyremethamine) ≥2 times during the course of pregnancy as per standard care in Tanzania (14). They were followed up monthly at clinical study sites until delivery and the end of the 6-wk postnatal period.

At enrollment, monthly pregnancy visits, delivery, and 6-wk postpartum visits, participants underwent detailed assessments in the manner described in a clinical trial of iron supplementation in the same setting (15). Participants were provided a month’s supply of the iron/folate capsules at each monthly visit. Study staff collected used regimen bottles at each visit and counted remaining pills. Maternal blood samples were collected at enrollment and at delivery, or within the first 48 h postpartum. Vouchers for bed nets were issued through a government program at prenatal clinics. Incident cases of clinical malaria with positive peripheral blood smear were treated, according to Tanzanian Ministry of Health guidelines. Screening and confirmatory tests for HIV status and iron deficiency were conducted before participants were enrolled. Participants with rapid ferritin between 10 and 20 μg/L had a serum ferritin test (Cobas Integra; Roche Diagnistics) conducted at the laboratory and were enrolled if ferritin ≤12 μg/L. Enrollment and delivery blood samples were tested at the research laboratory for a complete blood count (AcT5 Diff AL; Beckman Coulter), serum ferritin (Cobas Integra), C-reactive protein (CRP; Roche Diagnostics), soluble transferrin receptor (Roche Diagnostics), hepcidin (EIA-5258, version 4.1; DRG International Inc), ZPP, and α₁-acid glycoprotein (AGP). In 7% of enrolled participants, the rapid test incorrectly classified participants as having ferritin <10 μg/L when compared with later testing from same-day blood samples at the research laboratory. These participants were followed nonetheless, and all subsequent analyses were based on the confirmatory testing in the research laboratory. Participants with extremely low undetectable concentrations of biomarkers were assigned the lowest detectable concentrations.

The hepcidin assay was checked in nonanemic and healthy individuals to confirm hepcidin-25 activity. Six levels of standards (0–5) and low and high controls were each included in duplicate in every assay run. The optical units of the standards were used to plot the standard curve and to calculate the concentration of hepcidin-25 for each test sample. The 5–95% range of the hepcidin assay in normal apparently healthy adults is 13.3–54.4 μg/L (16). ZPP within red blood cells was measured by hematoflourimeter by quantifying the fluorescence emitted by ZPP during excitation of the sample using a light source filtered to produce a set wavelength of 415 nm. For participants with hemolysis, washed red blood cells were used to exclude plasma bilirubin interference. Controls were included in every assay run to ensure accuracy and reliability of results. AGP tests were conducted using Cobas Integra 400 Plus analyzer, based on the principle of immunoturbidimetric assay. Precinorm protein and precipath protein controls were included in every assay run. The analyzer was recalibrated daily using calibrator for automated systems (Cfas) proteins before the sample run. The biomarker tests possessed between-run precisions (CV) of 3.4–7.8%, 2.0–2.2%, 2.7–2.9%, 1.7–2.3%, and 11.5–14.6% for ferritin, sTfR, CRP, AGP, and hepcidin, respectively.

Increases in ferritin, CRP, and AGP due to inflammation follow different patterns, and correction of ferritin concentration for inflammation is therefore necessary (17). All participants were categorized into 4 groups based on concentrations of CRP and AGP: reference/no inflammation (CRP <5 μg/mL, AGP <1.0 g/L); incubating (elevated CRP only, ≥5 μg/mL); early convalescence (elevated CRP and elevated AGP, ≥1 g/L); and late convalescence (elevated AGP only). We corrected ferritin for inflammation by multiplying values by the ratio of the median ferritin concentration for the reference group over the median ferritin concentration for each inflammatory group (17).

The mean ± SD values of the hemoglobin, ferritin, ZPP, hepcidin, and sTfR at baseline and delivery and the mean difference ± SEM were calculated. To estimate the proportions meeting prespecified threshold concentrations and estimate differences in biomarker concentrations within baseline groups, we selected thresholds representing deficient hematologic status based on recommendations from previous studies: hemoglobin <110 g/L, ferritin ≤12 μg/L, sTfR >4.4 mg/L, ZPP >70 mmol/L, or hepcidin ≤13.3 μg/L (10, 16, 18–20). Almost all participants in our study had hepcidin concentrations <13.3 μg/L at baseline. To assess whether the changes in biomarker concentrations were significant, we obtained P values from Wilcoxon's rank-sum test, which does not require normally distributed data. We obtained P values for between-group change in biomarker concentrations among those with normal and deficient baseline concentrations using Oldham’s test. The null hypothesis of no baseline effect was tested, and the SDs of the measurement errors were assumed to be equal at baseline and delivery (21–23).

We evaluated the relation of change in biomarker concentrations and pregnancy outcomes, using binomial regression and obtained risk ratio estimates (24).Outcomes considered include low birth weight (<2500 g), small for gestational age [<10th percentile for gestational age, based on the INTERGROWTH (International Fetal and Newborn Growth Consortium for the 21^st Century) standard] (25), preterm birth (<37 wk gestational age), perinatal mortality (stillbirths and neonatal deaths in first week of life), early infant deaths (in the first 6 wk of life), delivery anemia (hemoglobin <110 g/L), and moderate to severe anemia (hemoglobin <100 g/L). Outcomes for twin pregnancies (n = 6 pairs) were analyzed as a single outcome in the main analysis (15) and participants with twin deliveries excluded in sensitivity analysis assessing the risk of adverse clinical outcomes. Ferritin concentrations were scaled by 10 μg/L for better interpretability. In a few instances, the model did not converge and log-Poisson models, which provide consistent but not fully efficient estimates of the RR and its CIs, were used (26). Linear regression models were also estimated for continuous delivery outcomes–gestational age at delivery in weeks, hemoglobin concentrations at delivery, and birth weight in grams, and robust regression whenever the assumption of normality did not hold (27).

Binomial regression models were estimated to evaluate the univariate associations of delivery anemia, iron deficiency, and up to 10-g/L increase in hemoglobin and 10-μg/L increase in ferritin concentration with baseline and follow-up covariates. Variables that were significant at P < 0.20 were selected for entry into the final models. These variables, along with others known to be associated with iron status and pregnancy outcomes (28–34), were adjusted for in multivariate models.Variables considered for inclusion were age (18–25, 26–35, or >36 y), parity (0 or 1), gestational age at enrollment (weeks), years of formal education (0–7, 8–11, or ≥12 y), occupation (business/professional, skilled formal, skilled informal, unskilled, or unemployed), number of household assets (0–1, 2–3, or 4–5), marital status (never married, not currently married, or married/cohabiting), clinical site (Amtullabai or Sinza), and amount spent on food [<3128TZS$ (equivalent to 2 US$ based on the historical exchange rate on 1 Jan 2012) or ≥3128TZS$] (35). BMI (in kg/m²; <18.5, 18.5–24.99, or ≥25), consumption of meat (<75 or ≥75 g/wk), season of enrollment [December–March (long rains), April–May (harvest), June–September (postharvest), or October–November (short rains)], compliance rate (<80% or ≥80%), mean corpuscular volume (<80 or ≥80 fL), low birth weight delivery in previous pregnancy (yes or no), history of obstetric hemorrhage in current pregnancy (yes or no), multiple gestation (yes or no), and baseline categories of hemoglobin (<110 or ≥110 g/L), ferritin (≤12 or >12 μg/L), sTfR (>4.4 or ≤4.4 mg/L), ZPP (>70 or ≤70 mmol/L), or hepcidin (≤1 or >1 μg/L) were also considered. Most of the participants had hepcidin concentrations below the cutoff for deficiency, 13.3 μg/L, and the hepcidin concentration was dichotomized at the median, 1 μg/L, to obtain categories that varied. To obtain compliance the number of days supplement pills were known to be taken, based on a pill count, was calculated as a proportion of the time between enrollment and delivery.

The sample size of 600 was calculated to detect an 11% change in ferritin concentration at 80% statistical power, assuming a mean concentration of 20 μg/L and SD of 10 μg/L. The study would also have sufficient power to detect changes in biomarker concentrations up to 8% and 12% for sTfR and hepcidin, respectively. Covariates with missing data were retained in the analysis using the missing indicator method (36). P values were 2-sided, and significance was set at <0.05. Statistical analyses were conducted with SAS v9.2 (SAS Institute Inc). Values presented in the text are medians ± IQRs, means (95% CI), mean ± SE, and RR.

Participants gave written informed consent at enrollment. Ethical approval for the study was obtained from the institutional review boards of the Harvard T.H. Chan School of Public Health and Muhimbili University of Health and Allied Sciences, and regulatory approval from the Tanzanian National Institute for Medical Research, and the Tanzanian Food and Drug Administration.

Results

The study included 600 HIV-negative pregnant women iron deficient at the time of commencing antenatal care. The mean age ± SD of the women was 24 ± 4 y. They were enrolled at a mean gestational age of 20 ± 4 wk and received iron supplementation for 4.66 ± 1.28 mo, until delivery. Fifty-three percent (n = 317) were primigravida, and the rest were secundigravida (Table 1). Use of malaria prevention measures was predominant, especially bed nets. Most were in the late convalescence stage with elevated AGP and normal CRP concentration.

TABLE 1.

Basic sociodemographic and clinical characteristics of iron-deficient Tanzanian pregnant women who received iron supplementation¹

Characteristics	Values
Age, y	24.2 ± 4.0
18–20	120 (20)
21–25	282 (37)
>25	198 (33)
Gestational age at enrollment, wk	20.1 ± 3.8
4–13	34 (6)
>13–20	273 (45)
>20–27	293 (49)
Gravidity
Primigravida	317 (53)
Secundigravida	283
BMI, kg/m2	24.2 ± 3.9
<18.5	19 (3)
18.5–25	370 (62)
>25	206 (35)
Education, y
0–7	337 (56)
>7–11	161 (27)
>11	100 (17)
Marital status
Married or cohabiting	475 (81)
Never married	113 (19)
Widowed or divorced	2 (0.3)
Occupation
Unemployed	288 (48)
Unskilled	191 (32)
Skilled informal	7 (1)
Skilled formal	51 (9)
Business/professional	61 (10)
Meat consumption, g/wk
<75	129 (22)
≥75	462 (78)
Number of household assets2
0–5	34 (6)
6–8	263 (44)
9–10	298 (50)
Hemoglobinopathy
Suggestive of thalassemia	48 (8)
Not suggestive	552 (92)
Malaria prevention measure
Any malaria prevention measure	557 (93)
Bed nets	539 (90)
Insecticide-treated bed nets	369 (62)
Fumigation	175 (29)
Mosquito-repellant coil	36 (6)
Inflammation3
None	18 (3)
Incubating	11 (2)
Early convalescence	152 (25)
Late convalescence	285 (48)
Compliance with iron supplementation
<80%	361 (60)
≥80%	239 (40)
Duration of use of iron supplements, d
<90	279 (46)
≥90	321 (54)
Multiple gestation
Singleton pregnancy	583 (99)
Twin gestation	8 (1)

¹Values are means ± SDs and n (%). AGP, α₁-acid glycoprotein; CRP, C-reactive protein.

²Number of household assets was computed from a simple list of assets owned by a participant: car, generator, bike, sofa, television, radio, refrigerator, fan, electricity, and potable water.

³Participants were categorized into 4 groups based on concentrations of CRP and AGP: reference/no inflammation (CRP <5 μg/mL, AGP <1.0 g/L); incubating (elevated CRP only, ≥5 μg/mL); early convalescence (elevated CRP and elevated AGP, ≥1 g/L); and late convalescence (elevated AGP only).

Most of the participants were iron deficient (n = 557; 92.8%) at enrollment into the study and had mild to moderate anemia with hemoglobin concentrations of 85–110 g/L (n = 423; 70.5%). Only 12 participants (2%) had severe anemia at baseline (Table 2). At baseline, the median ± IQR concentrations of the hematologic biomarkers were 101 ± 19 g/L for hemoglobin, 5.9 ± 6.3 μg/L for ferritin, 3.4 ± 2.9 g/L for sTfR, 77 ± 78 mmol/L for ZPP, and 1.0 ± 0.6 μg/L for hepcidin. Concentrations of hepcidin were correlated with those of CRP and AGP [r = 0.26 (P < 0.001) and r = 0.25 (P < 0.001) at baseline, respectively, and r = 0.24 (P < 0.001) and r = 0.19 (P = 0.007) at delivery, respectively]. There were significant improvements in maternal biomarker concentrations between baseline and delivery, with increases in hemoglobin concentration (15.2 g/L; 95% CI: 13.2, 17.2 g/L), serum ferritin (51.6 μg/L; 95% CI: 49.5, 58.8 μg/L), and serum hepcidin (14.0 μg/L; 95% CI: 12.4, 15.6 μg/L) and decreases in sTfR (−1.7 mg/L; 95% CI: −2.0, −1.3 mg/L) and ZPP (−17.8 mmol/L; 95% CI: −32.1, 3.5 mmol/L). The proportions of participants with low hemoglobin, ferritin, and hepcidin were 73%, 93%, and 99%, respectively, at baseline and 34%, 12%, and 46%, respectively, at delivery. For all biomarkers, the differences in biomarker concentrations were significantly greater among those who had poorer iron status at baseline than those who were replete.

TABLE 2.

Hematologic biomarker concentrations of iron-deficient Tanzanian pregnant women who received iron supplementation at baseline and delivery¹

	Baseline		Delivery		Change, baseline to delivery		P value
Biomarkers	n	Values	n	Values	n	Values	Within-group change from baseline to delivery2	Between-group difference from baseline to delivery3
Hemoglobin
All	599	101 ± 19	403	117 ± 19	402	15.2 (13.2, 17.2)	<0.0001
Normal, ≥110 g/L	165	118 ± 8	111	124 ± 22	111	6.4 (3.0, 9.7)	<0.0001	<0.0001
Deficient, <110 g/L	434	97 ± 14	291	115 ± 24	291	18.5 (16.1, 20.9)	<0.0001
% Deficient	599	72.5	403	34.0			<0.0001
Ferritin4,5
All	600	5.9 ± 6.3	381	47.5 ± 58.4	381	51.6 (49.5, 58.8)	<0.0001
Normal, >12 μg/L	44	26.1 ± 11.5	29	58.0 ± 75.4	29	38.2 (21.8, 54.6)	<0.0001	<0.0001
Deficient, ≤12 μg/L	556	5.4 ± 5.7	352	46.8 ± 58.2	352	52.7 (47.9, 57.5)	<0.0001
% Deficient	600	92.8	381	12.1			<0.0001
ZPP
All	278	77.0 ± 78.0	138	62.0 ± 43.0	96	−17.8 (−32.1, 3.5)	0.02
Normal, ≤70 mmol/L	114	52.0 ± 17.0	42	52.0 ± 22.0	42	5.6 (−2.9, 14.1)	0.71	0.001
High,6 >70 mmol/L	164	111.0 ± 95.0	54	78.0 ± 56.0	54	−36.0 (−59.9, 12.5)	0.004
% High	278	59.0	138	41.3			0.0007
Hepcidin								NA
All	404	1.0 ± 0.6	333	12.3 ± 17.8	319	14.0 (12.4, 15.6)	<0.0001
Deficient,7 ≤13.3 μg/L	403	1.0 ± 0.6	318	12.5 ± 17.9	318	14.1 (12.5, 15.7)	<0.0001
% Deficient	404	99.8	333	45.7			<0.0001
sTfR
All	395	3.4 ± 2.9	319	1.6 ± 2.5	304	−1.7 (−2.0, −1.3)	<0.0001
Normal, ≤4.4 mg/L	266	2.5 ± 1.8	272	1.6 ± 2.6	205	−0.2 (−0.5, 0.1)	0.02	<0.0001
High,6 >4.4 mg/L	129	6.0 ± 2.5	99	1.5 ± 3.2	99	−4.6 (−5.4, −3.8)	<0.0001
% High	395	32.7	319	14.7			<0.0001

¹Values are median ± IQR and mean change (95% CI). AGP, α₁-acid glycoprotein; CRP, C-reactive protein; NA, not applicable; sTfR, serum soluble transferrin receptor; ZPP, zinc protoporphyrin.

²P values for within-group change compare the baseline and delivery biomarker concentrations and are derived from the Wilcoxon's rank-sum test, which does not require normally distributed data. Only hemoglobin concentration was normally distributed.

³P values for between-group change compare the quantitative change in biomarker concentrations among those with normal and deficient baseline concentrations using Oldham’s test. The null hypothesis of no baseline effect was tested. The SDs of the measurement errors are assumed to be equal at baseline and delivery (21, 22).

⁴Participants were enrolled based on a rapid ferritin screening test that identified participants with ferritin <10 μg/L. In 7.2% of cases, the rapid test incorrectly classified participants, when compared with the serum testing later conducted in the research laboratory from same-day blood samples. The participants were followed up nonetheless and were not excluded from subsequent analysis. Hemoglobin was not a screening criterion.

⁵Participants were categorized into 4 groups: reference/no inflammation (CRP <5 μg/mL, AGP <1.0 g/L); incubating (elevated CRP only, ≥5 μg/mL); early convalescence (elevated CRP, elevated AGP, ≥1 g/L); and late convalescence (elevated AGP only). We corrected ferritin for inflammation by multiplying values by the ratio of the median ferritin concentration for the reference over the median ferritin concentration for each group (17).

⁶High concentrations of ZPP represent poor iron-deficient erythropoiesis. High concentrations of sTfR represent depletion of blood iron.

⁷Hepcidin concentration ≤13.3 μg/L corresponds to deficient iron status. Almost all participants in our study belonged to this category at baseline.

The median ± IQR hemoglobin concentration was 117 ± 19 g/L at the time of delivery (Table 3). Delivery occurred before term in 14% of the participants. The mean birth weight of neonates was 3180 ± 483 g, with 10% low birth weight and 12% small for gestational age. There was 4% perinatal mortality and 4% early infant deaths.

TABLE 3.

Maternal and child health outcomes of iron-deficient Tanzanian pregnant women who received iron supplementation¹

Outcome	Values
Birth weight	3180 ± 483
Low, <2500 g	54 (10)
Normal, ≥2500 g	503 (90)
Small for gestational age	64 (12)
Gestational age at delivery, wk	39 ± 2.9
Preterm	81 (14)
Full-term	510 (86)
Perinatal mortality2
Died	22 (4)
Survived	570 (96)
Neonatal mortality3
Died	23 (4)
Survived	560 (96)
Delivery hemoglobin, g/L	117 ± 19
Delivery anemia, <110	137 (34)
Moderate to severe anemia, <100	68 (17)

¹Values are mean ± SD or n (%). Based on the International Fetal and Newborn Growth Consortium for the 21st Century standard (25).

²Perinatal mortality included stillbirths and neonatal deaths in the first week of life.

³Neonatal mortality included infant deaths in the first 6 wk of life.

To evaluate the clinical response to iron supplementation, we compared the association of the quantitative change in the different biomarkers and the risk of birth outcomes (Table 4). For every 10-g/L increase in hemoglobin concentration, the risk of perinatal mortality decreased by 24% (RR = 0.76; 95% CI: 0.59, 0.98) and the risk of early infant death decreased by 23% (RR = 0.77; 95% CI: 0.60, 0.99). For every 10-μg/L increase in ferritin between baseline and delivery, hemoglobin concentration at delivery increased by 0.03 g/L and the risk of anemia reduced by 5% (RR = 0.95; 95% CI: 0.90, 0.99). We did not find associations with any of the other birth outcomes evaluated.

TABLE 4.

Association of change in hematologic biomarker concentrations with maternal and child health outcomes of iron-deficient Tanzanian pregnant women who received iron supplementation¹

	Change in hemoglobin, per 10-g/L increase			Change in ferritin, per 10-μg/L increase2
Outcome	Values3	Unadjusted RR (95% CI)4,5	Adjusted RR4–6	Values3	Unadjusted RR (95% CI)4,5	Adjusted RR4–6
Low birth weight, RR	39/390	0.99 (0.88, 1.15)	1.00 (0.85, 1.17)	35/370	1.01 (0.94, 1.07)	1.02 (0.96, 1.09)
Small for gestational age, RR	44/390	0.90 (0.79, 1.04)	0.90 (0.78, 1.04)	36/370	1.01 (0.95, 1.08)	1.03 (0.96, 1.10)
Birth weight, g	3179 ± 475	−0.6 (−22.9, 21.7)	6.1 (−18.8, 31.0)	3196 ± 465	−9.0 (−19.1, 1.01)	−5.1 (−16.3, 6.10)
Preterm, RR	63/403	1.02 (0.92, 1.14)	1.01 (0.88, 1.15)	56/380	0.99 (0.94, 1.04)	0.97 (0.91, 1.03)
Gestational age at delivery, wk	38.8 ± 2.9	−0.07 (−0.20, 0.06)	−0.04 (−0.18, 0.10)	38.8 ± 2.9	0.02 (−0.04, 0.07)	0.05 (−0.02, 0.11)
Mortality, RR
Perinatal	16/402	0.79 (0.63, 0.99)*	0.76 (0.59, 0.99)*	14/381	0.98 (0.87, 1.10)	0.96 (0.84, 1.09)
Neonatal	17/395	0.78 (0.63, 0.97)*	0.77 (0.60, 0.99)	15/377	0.99 (0.88, 1.10)	0.99 (0.88, 1.11)
Anemia, RR
Delivery	137/402	0.68 (0.63, 0.73)*	0.64 (0.59, 0.69)*	119/355	0.93 (0.90, 0.97)	0.95 (0.91, 0.99)*
Moderate to severe	68/402	0.61 (0.55, 0.67)*	0.55 (0.49, 0.62)*	57/355	0.92 (0.86, 0.98)	0.92 (0.86, 0.99)*
Delivery hemoglobin, g/L	116 ± 20	7.2 (6.6, 7.8)*	7.2 (6.6, 7.9)*	116 ± 19	0.06 (0.01, 0.09)*	0.03 (−0.01, 0.07)

¹*, significant at P < 0.05. AGP, α₁-acid glycoprotein; CRP, C-reactive protein.

²Ferritin concentrations were scaled by 10 μg/L for better interpretability.

³Values in the column are number of events/number of observations (n/N) or means ± SDs.

⁴Values in the column are RRs or difference measures, with 95% CIs.

⁵Binomial regression models were estimated to obtain the RR of critical pregnancy outcomes. RR > 1 implies the pregnancy outcome is more likely to occur as the biomarker concentration increases by every 10 μg/L for ferritin. RR < 1 implies the pregnancy outcome is less likely as the biomarker concentration increases. Continuous outcomes were modeled using robust regression (27).

⁶Multivariate estimates were adjusted for age (18–20, 21–25, or >25 y); parity (0 or 1); gestational age at enrollment (4–13, 14–20, or 21–27 wk); years of formal education (0–7, 8–11, or ≥12 y); BMI (in kg/m²; <18.5, 18.5–25, or >25); meat consumption (<75 or ≥75 g/wk); season of enrollment [December–March (long rains), April–May (harvest), June–September (postharvest), or October–November (short rains)]; supplement compliance rate (≥80% or <80%); multiple gestation (yes or no); and baseline lab findings: hemoglobinopathy (reduced hemoglobin A₂/elevated hemoglobin F or not), inflammation status (normal CRP and AGP, elevated CRP only, CRP and AGP elevated, or elevated AGP), use of malaria prevention measures (bed nets, other measures, or none). Variables were included in the model if they were known to be associated with iron status and pregnancy outcomes, or were selected from univariate associations with the change in hemoglobin and ferritin concentration (P < 0.2).

To identify participants who were most likely to benefit from supplementation, we examined the baseline and follow-up covariates that may predict the risk of delivery anemia, delivery iron deficiency, and the change in hemoglobin and ferritin concentrations (Table 5). The risk of maternal anemia at delivery despite supplementation was predicted by baseline anemia (RR = 2.11; 95% CI: 1.39, 3.18). Participants who were 26–35 y old had a significantly lower risk of delivery iron deficiency than those who were younger (RR = 0.44; 95% CI: 0.20, 0.99). Improvements in ferritin concentration by up to 10 μg/L were more likely to be observed in participants who took iron supplements for up to 90 d (RR = 1.41; 95% CI: 1.13, 1.76).

TABLE 5.

Predictors of maternal delivery anemia and iron deficiency of iron-deficient Tanzanian pregnant women who received iron supplementation¹

	Delivery anemia (Hemoglobin <110 g/L)		≥10-g/L Change in hemoglobin		Delivery iron deficiency (Ferritin ≤ 12 μg/L)		≥10-μg/L Change in ferritin
Baseline characteristic	Univariate RR (95% CI)2	Multivariate RR2–4	Univariate RR (95% CI)2	Multivariate RR (95% CI)2–4	Univariate RR (95% CI)2	Multivariate RR (95% CI)2–4	Univariate RR (95% CI)2	Multivariate RR (95% CI)2–4
Age, y
18–20	1.00		1.00		1.00	1.00	1.00	1.00
21–25	1.20 (0.80, 1.80)		0.82 (0.66, 0.97)*		0.52 (0.25, 1.07)	0.44 (0.20, 0.99)*	1.16 (0.92, 1.47)	1.17 (0.85, 1.62)
>25	1.21 (0.70, 1.84)		0.90 (0.74, 1.09)		0.89 (0.46, 1.74)	0.68 (0.31, 1.52)	1.34 (1.06, 1.69)	1.32 (0.93, 1.86)
Gravidity
Primigravida	1.00		1.00		1.00	1.00	1.00	1.00
Secundigravida	0.99 (0.75, 1.30)		0.99 (0.85, 1.16)		1.60 (0.92, 2.79)	1.77 (0.93, 3.38)	1.03 (0.89, 1.20)	0.94 (0.74, 1.19)
Baseline hemoglobin, g/L
Anemic, ≤110	2.11 (1.39, 3.18)*	2.16 (1.34, 3.47)*	1.37 (1.11, 1.69)	1.34 (0.80, 2.25)	1.14 (0.60, 2.15)		1.09 (0.92, 1.31)
Normal, >110	1.00	1.00	1.00	1.00	1.00		1.00
Baseline ferritin, μg/L
Deficient, ≤12	0.99 (0.60, 1.63)	0.99 (0.53, 1.87)	1.70 (1.08, 2.68)	1.64 (0.58, 4.61)	1.85 (0.47, 7.28)	1.72 (0.41, 7.16)	1.27 (0.89, 1.83)	1.26 (0.78, 2.03)
Normal, >12	1.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00
Baseline sTfR, mg/L
High, >4.4	0.99 (0.73, 1.34)		1.33 (1.13, 1.55)*	1.14 (071, 1.84)	0.71 (0.37, 1.36)		1.10 (0.98, 1.26)
Normal, ≤4.4	1.00		1.00	1.00	1.00		1.00
Baseline ZPP, mmol /L
High, >70	1.28 (0.85, 1.93)		1.36 (1.03, 1.80)*	1.17 (0.73, 1.89)	0.99 (0.50, 1.96)		1.00 (0.79, 1.27)
Normal, ≤70	1.00		1.00		1.00		1.00
Baseline hepcidin,5 μg/L
≤1	0.93 (0.70, 1.25)		1.08 (0.91, 1.27)		0.74 (0.41, 1.32)		1.04 (0.92, 1.18)
>1	1.00		1.00		1.00		1.00
BMI, kg/m2
<18.5	1.34 (0.73, 2.47)	1.34 (0.58, 3.11)	0.93 (0.60, 1.44)	0.69 (0.21, 2.28)	**		0.89 (0.55, 1.45)
18.5–25	1.00	1.00	1.00	1.00	1.00		1.00
>25	0.93 (0.69, 1.26)	1.08 (0.74, 1.56)	0.80 (0.67, 0.96)*	0.96 (0.62, 1.49)	1.07 (0.61, 1.86)		0.98 (0.83, 1.15)
Number of household assets
0–5	0.84 (0.44, 1.64)	0.89 (0.52, 1.52)	0.88 (0.63, 1.23)	1.11 (0.54, 2.31)	0.71 (0.18, 2.86)		1.15 (0.86, 1.53)
6–8	1.00	1.00	1.00	1.00	1.00		1.00
9–10	1.07 (0.81, 1.42)	0.85 (0.65, 1.12)	0.82 (0.70, 0.97)*	0.75 (0.49, 1.15)	1.15 (0.66, 1.99)		0.94 (0.80, 1.10)
Meat intake, g/wk
<75	1.00		1.00		1.00		1.00
≥75	1.00 (0.99, 1.01)		1.00 (0.99, 1.01)		1.01 (0.99, 1.02)		0.99 (0.99, 1.01)
Compliance,5 %
<80	1.00		1.00		1.00		1.00
≥80	0.79 (0.60, 1.06)		1.10 (0.63, 1.92)		1.21 (1.03, 1.41)*		1.43 (1.24, 1.66)*
Duration of use, d
<90	1.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00
≥90	0.82 (0.62, 1.08)	0.83 (0.59, 1.17)	0.96 (0.56, 1.65)	1.19 (0.79, 1.79)	1.22 (1.04, 1.43)*	0.91 (0.51, 1.61)	1.43 (1.23, 1.67)	1.41 (1.13, 1.76)*
Malaria prevention
Any measure	1.29 (0.71, 2.39)		0.87 (0.68, 1.12)		0.79 (0.31, 2.02)		1.14 (0.82, 1.58)
Bed nets	1.03 (0.64, 1.68)		0.91 (0.71, 1.17)		0.65 (0.30, 1.42)		1.17 (0.87, 1.58)
None	1.00		1.00		1.00		1.00
Multiple gestation
Singleton	1.00	1.00	1.00		1.00		1.00
Twin gestation	1.98 (1.11, 3.55)*	1.94 (0.70, 5.36)	0.27 (0.05, 1.62)		**		0..47 (0.14, 1.55)
Hemoglobinopathy
Suggests thalassemia	1.64 (1.15, 2.33)*	1.64 (0.99, 2.72)	0.69 (0.47, 1.03)	0.81 (0.44, 1.49)	1.01 (0.39, 2.65)	1.12 (0.40, 3.16)	0.95 (0.71, 1.27)	1.02 (0.67, 1.55)
Not suggestive	1.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00

¹*, significant at P <0.05; **, too few events and value not available. AGP, α1-acid glycoprotein; CRP, C-reactive protein; sTfR, serum soluble transferrin receptor; ZPP, zinc protoporphyrin.

²Binomial regression models were estimated to evaluate the association of delivery anemia and iron deficiency and the change in hemoglobin and ferritin (scaled by 10 μg/L) with a set of covariates. RR > 1 implies the follow-up outcome is more likely to occur compared to the reference. RR < 1 implies the outcome is less likely to occur.

³Variables considered for inclusion in the model include age (18–20, 21–25, or >25 y); parity (0 or 1); gestational age at enrollment (4–13, 14–20, or 21–27 wk); years of formal education (0–7, 8–11, or ≥12 y); occupation (business/professional, skilled formal, informal skilled and unskilled, or unemployed); marital status (married/cohabiting or not currently married); clinical site; amount spent on food; BMI (in kg/m²; <18.5, 18.5–25, or > 25); meat consumption (<75 or ≥75 g/wk); season of enrollment [December–March (long rains), April–May (harvest), June–September (postharvest), or October–November (short rains)]; compliance to regimen (≥80% or <80%); duration of use (≥90 or <90 d); low birth weight delivery in previous pregnancy (yes or no); history of obstetric hemorrhage in current pregnancy (yes or no); multiple gestation (yes or no); antenatal urinary tract infection (yes or no); hypertension (yes or no); and baseline lab findings: mean corpuscular volume (<80 or ≥80 fL), mean corpuscular hemoglobin (<27 or ≥27 pg), hemoglobinopathy (reduced hemoglobin A₂/elevated hemoglobin F or not), inflammatory biomarkers (normal CRP and AGP, elevated CRP only, CRP and AGP elevated, or elevated AGP).

⁴The significance level for entry into the multivariate model was P < 0.2 from univariate analyses.

⁵Most of the participants in our study had hepcidin concentrations below the cutoff for deficiency, 13.3 μg/L. We therefore dichotomized by the median hepcidin concentration, 1 μg/L, to obtain categories that varied for this analysis. Compliance and duration of use of supplements are correlated, and only duration of use of supplements was used in the multivariate models because it has the greater effect on nutritional status than most indicators (5).

In sensitivity analysis, we excluded participants with twin deliveries, and assessed the risk of adverse clinical outcomes, and the results did not change materially.

Discussion

We evaluated the clinical and hematologic response to iron supplementation among iron-deficient pregnant women and found significant improvements in the concentrations of hematologic status biomarkers. There was a significantly greater response in women with poorer baseline hematologic status. Increase in hemoglobin concentration was associated with a reduced risk of perinatal and early infant mortality.

Our study findings corroborate the evidence that iron supplementation effectively improves maternal hematologic status. Previous meta-analyses of iron supplementation studies have reported widely varying improvements in hemoglobin concentration of 4.59 g/L (3) and 5.1 g/L (37) during pregnancy. There was an increase of 15.2 g/L among participants who were anemic and iron deficient at enrollment in our study, whereas a placebo-controlled trial among iron-replete women in the same setting observed 1.3 g/L in the supplementation arm (15). The role of baseline hematologic status in modifying the response to iron supplementation was previously unclear. The greater magnitude of response to iron supplementation in those with poorer baseline hematologic status may be due to increases in erythropoietin secretion (13, 38). Because iron supplementation is safe in iron-deficient and replete pregnant women (15, 39), universal supplementation, as recommended in numerous countries, is clinically appropriate and ethically justified as it improves the condition of those who are worse off and eliminates the need for costly tests to improve detection of poor iron status in the presence of inflammation, yet places no substantial burden on pregnant women in malaria-endemic settings with good uptake of malaria prevention measures. Efforts to improve coverage should be strengthened.

Iron supplementation improves fetal and neonatal health and survival (40). Fetal iron status and ferritin concentrations are known to be determined to a large extent by maternal iron status and to influence fetal growth and survival in the first 24 mo of life (1, 38). Improvements in iron status translated to reduced risk of perinatal mortality among participants in our study, but did not influence other outcomes. A 19% reduction in risk of low birth weight has been previously reported, with a linear increase in birth weight corresponding to increased hemoglobin concentration (3). Reduced risk of preterm birth, and early neonatal mortality have also been reported in developing country setting following iron supplementation (41). Most participants in our study enrolled after 20 wk of gestation, and it is possible that the duration of supplementation was not long enough to alter the risk of the other outcomes in this population. Although iron absorption increases substantially after the 20th week of gestation (13), the effect of reduced fetal growth from iron deficiency in the first trimester persists despite subsequent supplementation (42). Further, most of the participants in our study had only mild anemia at baseline, and it is not clear whether the treatment of mild maternal anemia leads to significant improvement in child outcomes (43).

Targeted iron supplementation, as an alternative to universal supplementation, may be cost-effective (44) but is limited by the suitability of available biomarkers to detect those who may reap the most benefit. Plasma ferritin is regarded as the best representation of body iron stores (45). Its concentrations are, however, influenced by inflammation, with up to 5-fold increases, and the use of inflammatory biomarkers to improve reliability has been recommended. We enrolled participants into the study based on rapid ferritin screening tests and observed remarkable response to supplementation in blood concentrations of all 5 biomarkers assessed in our study regardless of baseline concentrations, with the greatest magnitude in ferritin and hemoglobin. Targeted supplementation may also be useful in a setting of universal supplementation to fill gaps in coverage. Women identified using rapid ferritin tests may be more closely followed-up while in antenatal care using adherence improvement strategies for improved results.

We explored the relation of several covariates with the concentrations of hemoglobin and ferritin at delivery, as proxies for effective supplementation. Despite supplementation, pregnant women that were anemic at baseline had 2.17 times greater risk of delivery anemia than others, suggesting a potential role for preconceptional supplementation (46). Improvements in ferritin concentration were more likely to be observed in participants who took iron supplements for up to 90 d in our study. A recent clinical trial in China found a similar dose-related relation between the use of supplements and maternal iron status (47). Age <25 y, poor compliance to supplements, and twin gestation are related to the effectiveness of supplementation. Identifying participants with these factors for closer follow-up may be helpful in clinical or program settings, to reduce the burden of iron deficiency anemia and related adverse outcomes. The possession of certain genetic variants may be associated with better response to iron supplementation (48). Hemoglobinopathies, such as thalassemias, are important causes of nonnutritional anemia for which supplementation with iron may be unnecessary and ineffective.

Approximately 48% of participants in our study were in the late convalescence stage at enrollment, with high AGP concentrations and low CRP concentrations. The predominance of late convalescence may relate to poor nutritional status of the women (45). This study included women in their first and second pregnancies in malaria-endemic settings, limiting the generalizeability of the findings to this group of women. Compliance was estimated based on confirmed pill counts and probably underestimated the true picture. Compliance to supplements was 77% among participants for whom pill count information was complete. We are unaware of any previous study that assessed the response to supplementation based on concentrations of hemoglobin, ferritin, ZPP, sTfR, and hepcidin in a cohort of Tanzanian pregnant women and carefully assessed the influence of baseline hematologic status on the response of the biomarkers.

Conclusion

Iron supplementation during pregnancy may effectively improve maternal hematologic status and improve perinatal and newborn survival in a setting of good malaria prevention. Biomarkers demonstrate response to iron supplementation, and rapid screening tests may be helpful for more intensive targeting of supplementation for the iron deficient and anemic to close the gaps in coverage and reduce the burden of anemia in developing countries. Efforts to initiate iron supplementation earlier in pregnancy, and possibly prepregnancy, and increase compliance are likely to improve the effectiveness of supplementation.