Maternal anemia and risk of adverse birth and health ...

Maternal anemia and risk of adverse birth and health outcomes in low- and middle-income countries: systematic review and meta-analysis1,2 Md Mizanur Rahman,3,4* Sarah Krull Abe,3 Md Shafiur Rahman,3 Mikiko Kanda,3 Saki Narita,3 Ver Bilano,3 Erika Ota,5 Stuart Gilmour,3 and Kenji Shibuya3 3

Department of Global Health Policy, The University of Tokyo, Tokyo, Japan; 4Department of Population Science and Human Resource Development, University of Rajshahi, Rajshahi, Bangladesh; and 5Department of Health Policy, National Centre for Child Health and Development, Tokyo, Japan

ABSTRACT Background: Anemia is a leading cause of maternal deaths and adverse pregnancy outcomes in developing countries. Objectives: We conducted a systematic review and meta-analysis to estimate the pooled prevalence of anemia, the association between maternal anemia and pregnancy outcomes, and the populationattributable fraction (PAF) of these outcomes that are due to anemia in low- and middle-income countries. Design: PubMed, EMBASE, CINAHL, and the British Nursing Index were searched from inception to May 2015 to identify cohort studies of the association between maternal anemia and pregnancy outcomes. The anemic group was defined as having hemoglobin concentrations ,10 or ,11 g/dL or hematocrit values ,33% or ,34% depending on the study. A metaregression and stratified analysis were performed to assess the effects of study and participant characteristics on adverse pregnancy risk. The pooled prevalence of anemia in pregnant women by region and country-income category was calculated with the use of a random-effects metaanalysis. Results: Of 8182 articles reviewed, 29 studies were included in the systematic review, and 26 studies were included in the meta-analysis. Overall, 42.7% (95% CI: 37.0%, 48.4%) of women experienced anemia during pregnancy in low- and middle-income countries. There were significantly higher risks of low birth weight (RR: 1.31; 95% CI: 1.13, 1.51), preterm birth (RR: 1.63; 95% CI: 1.33, 2.01), perinatal mortality (RR: 1.51; 95% CI: 1.30, 1.76), and neonatal mortality (RR: 2.72; 95% CI: 1.19, 6.25) in pregnant women with anemia. South Asian, African, and low-income countries had a higher pooled anemia prevalence than did other Asian and uppermiddle-income countries. Overall, in low- and middle-income countries, 12% of low birth weight, 19% of preterm births, and 18% of perinatal mortality were attributable to maternal anemia. The proportion of adverse pregnancy outcomes attributable to anemia was higher in low-income countries and in the South Asian region. Conclusion: Maternal anemia remains a significant health problem in low- and middle-income countries. Am J Clin Nutr doi: 10.3945/ajcn.115.107896. Keywords: birth and health outcomes, low- and middle-income countries, maternal anemia, meta-analysis, population-attributable fraction

INTRODUCTION

Anemia remains a significant health problem globally, accounting for 60,534 deaths and 3.4% of global disability-adjusted life years (DALYs) in 2010 in women aged 15–49 y (1). The majority of DALYs that are due to anemia occur in low-income countries, particularly in South Asia (5.7% of DALYs in women) and Sub-Saharan Africa (3.9% of DALYs in women) (1). In highincome countries, 16% of women and 22% of pregnant women had anemia in 2011 (2). Rates of anemia are highest in lowincome countries, especially in Central and West Africa (48% of reproductive-age women and 56% of pregnant women) and in South Asia (47% of reproductive-age women and 52% of pregnant women) (2). Despite achievements in maternal and child health-related programs over the past decade (3–6), anemia remains a key health problem in pregnant women in low- and middle-income countries (2, 7, 8). The principal causes of anemia are poor nutrition (iron, folic acid, and vitamin deficiencies), infectious diseases such as malaria, and untreated genetic hemoglobin disorders (7–12). Anemia during pregnancy may cause low birth weight, preterm birth, and perinatal, neonatal and maternal mortality (13, 14), although findings on these risks have not been consistent, and systematic reviews are lacking for low- and middle-income countries. In the most-comprehensive review currently available, Haider et al. (13) compared risk of low birth weight and preterm birth in low- or middle-income countries combined with high-income countries. However, the review did not stratify results by country-income categories or regions for 1 Supported in part by the Japan Ministry of Health, Labour and Welfare (grant H25-chikyukibo-ippan-007), the Japan Agency for Medical Research and Development, Japan (grant 27300101), and the WHO (grant HQHWA1208014). 2 Supplemental Protocol, Supplemental Figures 1–8, and Supplemental Tables 1–10 are available from the “Online Supporting Material” link in the online posting of the article and from the same link in the online table of contents at http://ajcn.nutrition.org. *To whom correspondence should be addressed. E-mail: mizanur_rub@ yahoo.com. Received January 27, 2015. Accepted for publication November 30, 2015. doi: 10.3945/ajcn.115.107896.

Am J Clin Nutr doi: 10.3945/ajcn.115.107896. Printed in USA. Ó 2016 American Society for Nutrition

1 of 10

2 of 10

RAHMAN ET AL.

small-for-gestational-age and preterm births. Previous metaanalyses have not comprehensively studied the association between maternal anemia and adverse pregnancy outcomes by both geographic region and national income category despite the wide variation in anemia burden within and between regions and national income categories (1, 2). To our knowledge, no previous study has estimated the population-attributable fraction (PAF) of adverse pregnancy outcomes for maternal anemia. An understanding of these outcomes, the current trends in maternal anemia, and the association of maternal anemia with adverse pregnancy outcomes at the regional level and stratified by income is essential to inform policies and program development to prevent maternal anemia and improve maternal and child health outcomes. In this study, we aimed to conduct a systematic review with a meta-analysis of published cohort studies of low birth weight, preterm birth, small for gestational age, perinatal mortality, neonatal mortality, gestational diabetes, preeclampsia, and mode of delivery according to maternal anemia status in low- and middle-income countries. To assess the role of maternal anemia at the population level, we estimated the PAF for selected adverse pregnancy outcomes for maternal anemia. In addition, we estimated the pooled prevalence of maternal anemia by geographic region, national income category, and year with the use of available Demographic and Health Survey data.

as subjects. We included studies that examined maternal hemoglobin, hematocrit, or anemia status measured in the first or second trimester during pregnancy and pregnancy and perinatal outcomes. Anemia was defined as the exposure variable with hemoglobin concentrations ,11 g/dL or hematocrit ,33% (18). We included studies that reported any hemoglobin or hematocrit cutoffs. Birth and health outcomes including preterm delivery (defined as a birth before 37 wk of gestation), low birth weight (defined as weight ,2500 g), small for gestational age (defined as birth weight below the sex-specific 10th percentile of the gestational age), perinatal mortality [defined as deaths including death of a fetus .22 wk of gestation (stillbirth)], early neonatal mortality (,7 d of life), neonatal mortality (defined as death of a neonate in the first month of life), gestational diabetes, preeclampsia, and cesarean delivery were included in our studies. We excluded cross-sectional and case-control studies because these trials do not allow for the assessment of the temporal association between exposure and outcome. Studies that considered high-risk subjects with HIV, AIDS, heart disease, or diabetes at baseline were not included in our review. A small sample size may introduce bias in the estimation of an effect size; therefore, we excluded studies if they recruited ,100 subjects (19). Details of the inclusion and exclusion criteria and definitions of exposure and outcomes variables are presented in the Supplemental Protocol.

METHODS

Data extraction and quality assessment

Search strategy This review was undertaken according to the protocol established in the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines (15). With the help of a librarian, we searched PubMed, EMBASE, CINAHL, and British Nursing Index databases (Supplemental Tables 1–4) for studies concerning maternal anemia and risk of pregnancy and maternal and newborn health outcomes published between 1966 and February 2014 initially and updated in May 2015. Our search strategies consisted of a combination of free-text words, words in titles and abstracts, and Medical Subject Headings for exposure, participants, and study designs. The detailed search strategies and initial and updated search results for PubMed, EMBASE, CINAHL, and the British Nursing Index are presented in Supplemental Tables 1–4. Additional eligible studies on anemia and birth or health outcomes were sought by reviewing the reference lists of identified articles and searching relevant journals related to our search topic. We did not apply any language restrictions during the search. We defined lowand middle-income countries according to World Bank criteria in 2013 (16). Study selection In the screening and article selection, we followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses flowchart (17). Two reviewers (MMR and MSR) independently screened titles and abstracts and critically reviewed the full texts of all selected studies on the basis of the inclusion and exclusion criteria. Studies were included if they were cohorts (prospective or retrospective) with pregnant women aged $15 y

Two authors (MMR and MSR) independently extracted data on the country and year of the study, the study design, participants, exposures, the time of exposure assessment, outcomes, confounders, and measures of an association. We did not find any foreign-language articles that needed to be translated into English. We resolved any inconsistency through a consensus process. We used a specific checklist to assess the methodologic quality of all included cohort studies with the use of the Newcastle–Ottawa Scale recommended by Wells et al. (20). Data analyses RR was used as the common outcome measure in observational studies. We converted ORs into RRs according to the proposed methodology of Zhang (21, 22) when the incidence of an outcome was common (.10%) in the study population. If the RR or OR was unavailable, we estimated the unadjusted RR and 95% CI from raw data. We used fixed-effect (Mantel-Haenszel method) (23) or random-effect (DerSimonian-Laird method) (24) models for calculating summary estimates for the effects of maternal anemia with the model choice made on the basis of heterogeneity (I2 statistic) assessments. An I2 value refers to the percentage of total variation across studies that was due to between-study heterogeneity. Fixed-effects models were performed for I2 #25%, and random-effects models were performed for I2 .25% (25). Prediction intervals were estimated on the basis of t when the random model was used because of the presence of heterogeneity and a minimum of 3 studies. We presented summary estimates according to WHO thresholds for anemia and according to the definitions used in the original studies separately. Publication bias and biases related to a small sample size and reporting bias were assessed with the use of the regression

MATERNAL ANEMIA AND PREGNANCY OUTCOMES

asymmetry test of Egger (26). In addition, we performed trimand-fill procedures to further evaluate possible effects of publication bias in the meta-analyses (27). We investigated sources of heterogeneity through subgroup and metaregression analyses (28) according to the study design (prospective compared with retrospective), sample size above or below the median observed sample size (#800 compared with .800), confounding factors (adjusted compared with unadjusted), country-income category, study location (South Asia (Bangladesh, India, Nepal, Pakistan, India, and Sri Lanka), EastWest Asia (China, Malaysia, Iran, and Turkey), or Africa and South America (Malawi, Tanzania, Ghana, and Peru), and mean maternal age above or below the median from all studies (,26 compared with $26 y).We undertook sensitivity analyses to determine differences in summary effects by dropping a small number of studies that we defined as highly influential on the basis of variance and weight estimates from the meta-analysis. We calculated the PAF for birth outcomes that were due to anemia with the use of estimates obtained from the meta-analysis on the assumption that attributable risk arises from any association and not only a causal relation. The PAF was calculated on the basis of the pooled RR of pregnancy outcomes and the proportion (P) of maternal anemia during pregnancy as follows: PAF ¼

P ðRR 2 1Þ ½1 þ P ðRR 2 1Þ

ð1Þ

The pooled prevalence of anemia by region and countryincome level was estimated with the use of a random-effects meta-analysis and available Demographic and Health Survey data sets across 23 developing countries. We used the Freeman-Tukey transformation method to estimate the pooled prevalence of anemia (29). We used Stata version 12.1/MP software (StataCorp LP) for all analyses.

RESULTS

Our search identified 8182 records from inception to May 2015 of which 7987 records remained after the removal of duplicates (Figure 1). On the basis of title and abstract screening, 99 records were considered potentially eligible from databases. An additional 6 articles were identified from reference lists and hand searches. In total, 105 full-text articles were reviewed. In this full-text screening, 74 articles were further excluded because of small sample sizes (,100 subjects), different study designs (case-control, cross-sectional, and secondary data analyses), nonresearch materials, hemoglobin measured only at the third trimester during pregnancy, and high-risk populations, which left 31 articles (Figure 1). Two articles were based on overlapping data from the same cohort in China (30, 31). In addition, 2 articles used the same data source in Pakistan (32, 33). To avoid the duplicate inclusion of data, we merged outcomes and treated each of these pairs of studies as one study (which left 29 studies in the systematic review). Of these 29 studies, 3 studies were dropped from the meta-analysis because the articles did not assess health outcomes according to anemia thresholds (34–36). This exclusion left 26 studies for the metaanalysis.

3 of 10

Study characteristics Of 29 studies, 12 studies were conducted in South Asia, 13 studies were conducted in East-West Asia, and 4 studies were conducted in the African and South American regions (Supplemental Table 5). Twenty-four articles included prospective cohorts, and 5 articles were retrospective cohort studies. The selected studies were published between 1994 and 2014. The number of subjects per study ranged from 253 to 399,274 with a total of w0.72 million pregnant women with a mean age that ranged from 20 to 30 y. In the 29 studies, 18 studies reported low-birth-weight, 15 studies reported preterm birth, 12 studies reported perinatal mortality, 5 studies reported small-forgestational-age, 3 studies reported gestational diabetes, 4 studies reported preeclampsia, 2 studies reported neonatal mortality, 2 studies reported cesarean delivery, and one study reported stillbirth outcomes (Supplemental Table 5). All studies were of high quality (Supplemental Table 6).

Pooled and sensitivity analysis The pooled RR, publication bias, and trim-and-fill estimates across all studies are presented in Table 1. Risk of low birth weight was significantly higher in anemic pregnant women during the first or second trimester than in the nonanemic group (RR: 1.31; 95% CI: 1.13, 1.51; I2 = 66%; 17 studies). We showed significantly greater risk of preterm birth (RR: 1.63; 95% CI: 1.33, 2.01; I2 = 88%; 13 studies), perinatal mortality (RR: 1.51; 95% CI: 1.30, 1.76; I2 = 0%; 12 studies), and neonatal mortality (RR: 2.72; 95% CI: 1.19, 6.25; I2 = 0%; 2 studies) in our study. However, when we calculated 95% prediction intervals, the associations become insignificant between anemia and risk of low birth weight, preterm birth, and small for gestational age (Supplemental Figure 1). Anemia during the first or second trimester was also not significantly associated with small for gestational age (RR: 0.87; 95% CI: 0.63, 1.20; I2 = 95%; 5 studies), gestational diabetes (RR: 1.02; 95% CI: 0.86, 1.21; I2 = 20%; 2 studies), preeclampsia (RR: 2.66; 95% CI: 0.61, 11.52: I2 = not applicable; one study), or cesarean delivery (RR: 1.68; 95% CI: 0.76, 3.72; I2 = not applicable; one study). Our narrative review showed mixed results for maternal anemia and preeclampsia. One study indicated that a hemoglobin concentration .13.2 g/dL during the first trimester was significantly associated with preeclampsia (OR: 1.73; 95% CI: 1.07, 2.81). However, 2 other studies showed that low concentrations of hemoglobin during pregnancy were associated with preeclampsia (P , 0.01) (Supplemental Table 7) (34, 36). To account for any form of publication bias, we performed a sensitivity analysis with the use of the trim-and-fill method and showed a negligible effect on the results from the inclusion of possible imputed negative or small sample-size studies (Table 1). In the sensitivity analysis, low birth weight, preterm birth, and perinatal mortality remained risks in anemic women after highly influential studies were dropped (Supplemental Figures 2–4). In addition, we calculated pooled estimates of birth and health outcomes according to the definition of anemia on the basis of individual study definitions of anemia and showed almost similar results when we considered only WHO thresholds for anemia (Supplemental Figures 5–7). Furthermore, we calculated another sensitivity analysis of pregnancy outcomes after

4 of 10

RAHMAN ET AL.

FIGURE 1 Preferred Reporting Items for Systematic Reviews and Meta-Analyses flowchart for the selection of studies.

dropping studies that reporting ORs, and pooled estimates of low birth weight, preterm birth, and perinatal mortality showed similar results to those in the analysis with the OR conversion (Supplemental Table 8). Stratified analyses The study showed moderate heterogeneity in the low-birthweight outcome and severe heterogeneity in the preterm birth and small-for-gestational-age outcomes. To examine these heterogeneities, we conducted stratified analyses according to study designs, sample sizes, confounding adjustments, country-income categories, study locations, and maternal ages shown in Table 2 and Supplemental Table 9. The RR differed according to the subgroup analysis by country-income category. Risks of low

birth weight and preterm delivery were substantially higher in low-income countries than in upper-middle-income countries. Stratification by geographic region revealed increased risk of low birth weight and preterm delivery in anemic pregnant women in South Asia than in East-West Asia and the African and South American regions. However, the result was not statistically significant (P . 0.05). Prevalence of anemia Figure 2 presents the random-effects estimate for maternal anemia during pregnancy by country-income category. Prevalence was estimated from 28 recent surveys and 25 countries with a pooled prevalence of 42.7% (95% CI: 37.0%, 48.4%) in low- and middle-income countries. There were slight differences

5 of 10

MATERNAL ANEMIA AND PREGNANCY OUTCOMES TABLE 1 Summary, publication bias, and trim-and-fill estimates Trim-and-fill estimates1

Summary estimates

Characteristic Low birth weight Preterm birth Small for gestational age Perinatal mortality Neonatal mortality Gestational diabetes Preeclampsia Cesarean delivery

Studies, n 17 13 5 12 2 2 1 1

RR (95% CI) 1.31 1.63 0.87 1.51 2.72 1.02 2.66 1.68

(1.13, (1.33, (0.63, (1.30, (1.19, (0.86, (0.61, (0.76,

1.51)2 2.01)2 1.20)2 1.76)3 6.25)3 1.21)3 11.52)4 3.72)4

Heterogeneity index

P-bias test

Missing studies, n

65.7 88.2 95.0 0.0 0.0 19.8 NA5 NA

0.03 0.05 0.41 ,0.001 0.72 0.16 NA NA

4 0 0 5 0 0 0 0

RR (95% CI) 1.18 1.63 0.87 1.43 2.72 1.02 2.66 1.68

(1.02, (1.33, (0.63, (1.24, (1.19, (0.86, (0.61, (0.76,

1.37) 2.01) 1.20) 1.65) 6.25) 1.21)3 11.52) 3.72)

1 Trim-and-fill method simulated studies that were likely to be missing from the literature because of publication or other forms of bias. Trim-and-fill RRs estimate what the pooled RRs would be if the missing studies were included in the analysis. 2 On the basis of random-effects methods. 3 On the basis of fixed-effects methods. 4 No pooling method was used because there was only a single study. 5 NA, not applicable.

between the pooled prevalence of anemia in pregnant women in country-income categories. The East-West Asia region displayed a lower pooled prevalence (39.9%, 95% CI: 27.3%, 53.2%) of anemia than that of the South Asian region (48.6%; 95% CI: 44.7%, 52.4%) and African and South American regions (43.5%; 95% CI: 36.8%, 50.3%) (Supplemental Figure 8). Role of anemia The PAFs for selected adverse pregnancy outcomes that were attributable to maternal anemia during pregnancy are presented in Table 3. The prevalence of anemia with data sources used in the PAF calculations is presented in Supplemental Table 10. Overall, 12% of low birth weight, 19% of preterm birth, and 18% of perinatal mortality were attributable to maternal anemia during pregnancy in low- and middle-income countries. There was a wide difference in the PAF of pregnancy outcomes across geographic regions and country-income levels. In low-income countries, 25% of low birth weight, 44% of preterm births, and 21% of perinatal mortality were attributable to anemia during pregnancy. However, the respective PAFs were substantially smaller in lower-middle- and upper-middle-income countries. In low-income and lower-middle-income countries, there was a relatively higher anemia-attributable proportion of adverse low birth weight in Pakistan and Bangladesh than in Ghana and India. The highest anemia-attributable proportion of preterm birth was observed in Pakistan (54%) and followed by India (27%) and Iran (18%). DISCUSSION

To date, most knowledge relating to the birth and health consequences of maternal anemia has come from cross-sectional, case-control, and cohort studies. This systematic review and meta-analysis summarizes these associations by region and country-income category in low- and middle-income settings with the use of high-quality cohort studies. This study also assesses the proportion of low birth weight, preterm birth, and perinatal mortality that were attributable to maternal anemia.

Significant, positive associations were shown between anemia in the first or second trimester and low birth weight, preterm birth, and perinatal and neonatal mortality. However, no association was shown between maternal anemia and risk of small for gestational age, gestational diabetes, preeclampsia, and cesarean delivery. Maternal anemia was shown to be associated with a significant proportion of pregnancy outcomes in low-income countries with this proportion declining with increasing national income and varying substantially between countries and geographic regions. Maternal anemia remains one of the most-serious health problems in low-income countries despite the high priority of maternal and child health programs. Our study showed nearly one-half (42.7%) of pregnant women were anemic in low- and middle-income countries and the prevalence of anemia varied by the country economic profile (45.4% in low-income, 39.8% in lower-middle-income, and 37.1% in upper-middle-income countries). Our study also identified substantial regional differences in the prevalence of anemia. Consistent with a recent multicountry study (2), there was a higher prevalence of maternal anemia in the South Asian region and African and South American regions (48.6% and 43.5%, respectively) than in East-West Asia (39.9%). Our study showed that the prevalence of anemia during pregnancy has remained almost unchanged in low-income countries since 2000. The main reason for this stable and high prevalence of maternal anemia during pregnancy in low-income countries, especially in African and Asian regions, may be because of the high prevalence of malaria and poor nutrition including underweight and iron deficiency (10–12, 37). Consistent with previous studies (13, 38), women with anemia in the first or second trimester had a significantly greater risk of low birth weight, preterm birth, and perinatal and neonatal mortality. Sensitivity analyses confirmed a similar association after publication bias was accounted for or a small number of highly influential studies were dropped. In subgroup analyses, we showed an association between the hemoglobin concentration and risk of birth and health outcomes in low-income countries compared with in lower-middle-income or upper-middle-income countries. We showed substantial heterogeneity in low-birthweight, preterm birth, and small-for-gestational-age outcomes.

6 of 10

RAHMAN ET AL. TABLE 2 Stratified analysis of pooled RRs of low birth weight, small for gestational age, perinatal mortality, and preterm birth for anemic pregnant women1 P Characteristic Low birth weight Study design Prospective Retrospective Confounding factors Adjusted Unadjusted Country-income category Low income Lower-middle income Upper-middle income Geographic region South Asia East-West Asia Africa and South America Preterm birth Study design Prospective Retrospective Confounding factors Adjusted Unadjusted Country-income category Low income Lower-middle income Upper-middle income Geographic region South Asia East-West Asia Africa and South America Small for gestational age Study design Prospective Retrospective Confounding factors Adjusted Unadjusted Geographic region South Asia East-West Asia Africa and South America Perinatal mortality Study design Prospective Retrospective Confounding factors Adjusted Unadjusted Country-income category Low income Lower-middle income Upper-middle income Geographic region South Asia East-West Asia Africa and South America 1 2

Pooled RR (95% CI)

Heterogeneity

Metaregression2

1.41 (1.18, 1.68) 0.95 (0.60, 1.50)

,0.001 0.03

0.09

1.28 (1.03, 1.59) 1.34 (1.13, 1.59)

,0.001 0.17

0.64

1.72 (1.32, 2.25) 1.12 (0.94, 1.33) 1.27 (0.89, 1.79)

0.09 0.12 0.02

0.05

1.36 (1.11, 1.66) 1.27 (0.89, 1.79) 1.32 (0.76, 2.29)

,0.001 0.02 NA

0.91

1.94 (1.31, 2.89) 1.18 (1.09, 1.27)

,0.001 0.24

0.01

1.63 (1.16, 2.31) 1.70 (1.17, 2.46)

,0.001 ,0.001

0.88

2.73 (1.29, 5.79) 1.36 (0.97, 1.91) 1.20 (1.00, 1.44)

,0.001 0.02 0.04

0.01

2.03 (1.23, 3.36) 1.20 (1.00, 1.44) 1.24 (1.12, 1.37)

,0.001 0.04 NA

0.12

0.95 (0.65, 1.40) 0.85 (0.57, 1.25)

0.73 ,0.001

0.80

1.00 (0.84, 1.19) 0.65 (0.60, 0.72)

0.15 NA

0.01

0.84 (0.38, 1.88) 1.00 (0.82, 1.22) 0.65 (0.60, 0.72)

NA 0.09 NA

0.05

1.67 (1.30, 2.14) 1.43 (1.18, 1.73)

0.73 0.68

0.34

1.72 (1.28, 2.32) 1.45 (1.22, 1.73)

0.79 0.64

0.33

1.61 (1.16, 2.23) 1.44 (1.18, 1.76) 1.63 (1.18, 2.25)

0.33 0.87 0.79

0.75

2.05 (1.18, 3.55) 1.63 (1.18, 2.25) 1.43 (1.20, 1.72)

0.75 0.79 0.44

0.43

NA, not applicable. Metaregression P values represent a test of the entire characteristic.


7 of 10

FIGURE 2 Random-effects meta-analysis pooled prevalence estimates during 2010–2013 by country-income categories. Open diamonds represent pooled prevalence (95% CIs). Small filled diamonds represent the prevalence for each survey, and black bars denote 95% CIs. ES, prevalence; I^2, percentage of variation attributable to heterogeneity; n, total number of anemic women during pregnancy; N*, total number of pregnant women.

When the substantial heterogeneity in 95% prediction intervals was accounted for, the results indicated that the association between anemia and risks of low birth weight, preterm birth, and small for gestational age became insignificant. These results do not necessarily indicate that there is no impact of maternal anemia on birth outcomes. However, the results do indicate that there is still substantial uncertainty about the significance of the association. Our findings expand significantly on the recent metaanalysis of Haider et al. (13), which collapsed findings across countries and accessed only limited information on birth outcomes. We compared risks of preterm birth and low birth weight separately for low- or middle-income countries rather than conducting a combined comparison against high-income countries. In addition, we presented information on pregnancy outcomes by geographic regions and country-income categories in recognition of the substantially differing patterns of prevalence and birth outcomes in these country categories. Our study showed a con-

sistent regional variation in risks of low birth weight, preterm birth, and perinatal mortality. Highest risk of perinatal mortality attributable to maternal anemia was shown in Ghana, Pakistan, India, and Malawi. Greater risk of low birth weight was also observed in Pakistan, Bangladesh, and Ghana. In some low-income and lower-middle-income countries, the control of infectious diseases such as HIV, AIDS, and malaria has not yet been achieved, and health-service delivery, access, and effective coverage and access to affordable care are limited (11, 39–45). In previous studies anemia, malnutrition, and malaria during pregnancy were shown to be significant risks to both maternal and neonatal health (2, 8, 11, 37). However, many lowincome countries are facing challenges in implementing immunization, malaria control, and nutrition support programs (5, 46, 47). The war in Afghanistan and internal conflict in Pakistan targeted female health workers, and thus, many parts of these areas are severely affected by workforce-related barriers to the

8 of 10

RAHMAN ET AL.

TABLE 3 LBW, PTB, and PNM attributed to maternal anemia1 LBW Country 2

Overall Country-income category2 Low-income Lower middle-income Upper middle-income Region2 South Asia East-West Asia Africa and South America Country specific3 Bangladesh Pakistan India China Nepal Iran Peru Ghana Malawi Tanzania

PTB

PNM

Prevalence, %

RR (95% CI)

PAF, %

RR (95% CI)

PAF, %

RR (95% CI)

PAF, %

42.7

1.31 (1.13, 1.51)

12.1

1.63 (1.33, 2.01)

19.0

1.51 (1.30, 1.76)

17.9

45.4 39.8 37.1

1.72 (1.32, 2.25) 1.12 (0.94, 1.33) 1.27 (0.89, 1.79)

24.6 4.6 9.1

2.73 (1.29, 5.79) 1.36 (0.97, 1.91) 1.20 (1.00, 1.44)

44.0 12.5 4.9

1.60 (1.15, 2.23) 1.44 (1.18, 1.76) 1.63 (1.18, 2.25)

21.4 14.9 18.9

48.6 39.9 43.5

1.36 (1.11, 1.66) 1.27 (0.89, 1.79) 1.32 (0.76, 2.29)

14.9 9.8 12.5

2.03 (1.23, 3.36) 1.20 (1.00, 1.44) 1.24 (1.12, 1.37)

33.4 5.3 9.5

2.05 (1.18, 3.55) 1.63 (1.18, 2.25) 1.43 (1.19, 1.72)

38.8 20.1 15.8

49.6 40.0 59.0 28.9 47.6 14.0 28.8 70.0 37.5 52.7

1.80 2.10 1.13 1.15 1.24 2.00

28.6 30.6 7.1 5.5 10.3 12.3 — 18.3 — —

— (3.04, 5.13) (0.88, 3.04) (1.01 1.28) (0.62, 1.39) (1.33, 5.10) (1.12, 1.37) — — —

— 54.1 27.1 0.9 23.5 18.4 6.5 — — —

— 2.25 (1.14, 4.44) 1.70 (0.66, 4.38) 1.63 (1.18, 2.25) — — 1.40 (1.14, 1.73) 2.16 (0.74, 6.30) 2.00 (1.12, 3.57) 1.14 (0.70, 1.87)

— 42.4 29.2 15.4 — — 10.3 44.8 27.5 6.9

(1.18, (1.56, (0.92, (0.70, (0.97, (1.08, — 1.32 (0.76, — —

2.27) 2.82) 1.38) 1.89) 1.58) 3.70) 2.29)

3.95 1.63 1.14 0.93 2.61 1.24

1

LBW, low birth weight; PAF, population-attributable fraction; PNM, perinatal mortality; PTB, preterm birth. Pooled prevalence of anemia by meta-analysis with the use of Demographic and Health Survey most-recent data from 2010 to 2013 (more-detailed information is shown in Supplemental Table 10 and Supplemental Figure 8). 3 Country-specific prevalence of anemia data were mainly from most-recent Demographic and Health Survey data and other representative data (Supplemental Table 10). 2

resolution of their maternal and child health issues (5). Cost is another barrier to accessing health services in low-income countries (48, 49), and many poor households may avoid consulting doctors during pregnancy to minimize financial risks associated with high-treatment costs. Consequently, these women may be unaware of their nutritional status during pregnancy. Service delivery, effective coverage, and access and affordable care during pregnancy can be ensured by introducing universal health coverage plans (43, 45, 48, 49). For example, in Ghana, Indonesia, Uganda, and China, after health insurance was introduced, the burden of treatment costs sharply decreased and access to care increased (48, 50). Ensuring access to comprehensive, integrated primary care and maternal and child health services through better health-financing methods will help to ensure that women understand their nutritional status and are able to act earlier in pregnancy to minimize worst risks associated with maternal anemia. Our study had several strengths. We used comprehensive search techniques and validated systematic review methods, followed a predesigned protocol, and observed the Meta-analysis of Observational Studies in Epidemiology (15) and Preferred Reporting Items for Systematic Reviews and Meta-Analyses (17) guidelines, which strengthened the review quality and conclusions. We investigated the possible association between maternal anemia and birth and health outcomes by region, country-income category, and specific countries. In the meta-analysis, appropriate statistical techniques were used to estimate the pooled prevalence, RR, and presence of bias. Despite these strengths, limitations of this systematic review and meta-analysis must be considered. We included studies from

low- and middle-income countries, and thus, the results of review are not applicable to high-income countries. We only included studies that measured hemoglobin during the first or second trimester and may have overlooked some studies that addressed the effect of anemia in the third trimester. However, a recent meta-analysis suggested that anemia during the third trimester is not a potential risk factor for adverse birth outcomes, and its exclusion was unlikely to have biased this review (8). We defined anemia on the basis of WHO standard thresholds based on hemoglobin [anemic: ,10–11 g hemoglobin/dL or hematocrit ,30–34%; nonanemic: .11 g hemoglobin/dL) (18), but some studies did not use these categorizations. Different definitions and categorizations can lead to variations in RRs or ORs even within a single data set. However, in our systematic review, almost all studies used WHO cutoffs except for 5 studies from Ghana, Bangladesh, India, China, and Turkey. We performed a pooled analysis separately in which thresholds proposed by the WHO and other thresholds according to the definitions of the original studies. Despite different definitions, there was no significant difference in pooled estimates between WHO thresholds and others. This stable pooled estimate may have been because only 5 studies used different cutoffs to those recommended by the WHO. We also had to use estimated RRs for 7 studies that reported ORs, which were converted to RRs for the meta-analysis. There was risk that the variance of the derived RRs could have been underestimated in the proposed conversion methodology of Zhang (15, 21). However, we performed a sensitivity analysis that excluded the affected studies and showed negligible effect on the results. Finally, we did not include gray literature, which may have contained smaller null-result studies that were not


accepted for publication, but we adjusted for this publication bias with the use of the trim-and-fill method and showed little effect on our results from the inclusion of possible imputed negative studies. In conclusion, maternal anemia is a risk factor for adverse birth and perinatal health outcomes in low-and middle-income countries. Significantly higher risk of low birth weight and preterm birth were observed in low-income countries and in South Asian countries despite the greater priority and larger investment on maternal and child health programs in recent decades. On the basis of our review findings, we advocate that greater attention be placed on the impact of maternal and child health programs on anemia-related outcomes in low-income countries. With the management of maternal anemia during pregnancy in low-income countries, a substantial proportion of anemia-related adverse pregnancy outcomes could be avoided. We thank Miwako Segawa for her help with the electronic search strategies and retrieval of articles. The authors’ responsibilities were as follows—MMR: contributed to the statistical analysis, interpretation of the data, and wrote the first draft of the manuscript; MMR and MSR: performed the literature screening and data extraction; MMR, SKA, MK, and SN: collaborated on the quality assessment; MMR, EO, and KS: conceived of the study; EO: consulted on the quality assessment; SKA: collaborated on the writing of the first draft of the manuscript; SKA and VB: checked for consistency of the study; EO, SG, and KS: revised the manuscript critically for important intellectual content; and all authors: reviewed and approved the final manuscript. None of the authors reported a conflict of interest related to the study.

REFERENCES 1. Institute for Health Metrics and Evaluation HDN, The World Bank. The global burden of disease (GBD) compare [Internet]. 2013 ed. Seattle, WA: IHME, 2013 [cited 2014 Jun 24]. Available from: http:// vizhub.healthdata.org/gbd-compare/. 2. Stevens GA, Finucane MM, De-Regil LM, Paciorek CJ, Flaxman SR, Branca F, Peña-Rosas JP, Bhutta ZA, Ezzati M. Global, regional, and national trends in haemoglobin concentration and prevalence of total and severe anaemia in children and pregnant and non-pregnant women for 1995–2011: a systematic analysis of population-representative data. Lancet Glob Health 2013;1:e16–25. 3. Acuin CS, Khor GL, Liabsuetrakul T, Achadi EL, Htay TT, Firestone R, Bhutta ZA. Maternal, neonatal, and child health in southeast Asia: towards greater regional collaboration. Lancet 2011;377:516–25. 4. Amouzou A, Habi O, Bensaid K, Niger Countdown Case Study Working Group. Reduction in child mortality in Niger: a countdown to 2015 country case study. Lancet 2012;380:1169–78. 5. Bhutta ZA, Hafeez A, Rizvi A, Ali N, Khan A, Ahmad F, Bhutta S, Hazir T, Zaidi A, Jafarey SN. Reproductive, maternal, newborn, and child health in Pakistan: challenges and opportunities. Lancet 2013;381:2207–18. 6. Chowdhury AM, Bhuiya A, Chowdhury ME, Rasheed S, Hussain Z, Chen LC. The Bangladesh paradox: exceptional health achievement despite economic poverty. Lancet 2013;382:1734–45. 7. Black RE, Victora CG, Walker SP, Bhutta ZA, Christian P, de Onis M, Ezzati M, Grantham-McGregor S, Katz J, Martorell R. Maternal and child undernutrition and overweight in low-income and middle-income countries. Lancet 2013;382:427–51. 8. Branca F, Mahy L, Mustafa TS. The lack of progress in reducing anaemia among women: the inconvenient truth. Bull World Health Organ 2014;92:231. 9. Balarajan Y, Ramakrishnan U, Ozaltin E, Shankar AH, Subramanian SV. Anaemia in low-income and middle-income countries. Lancet 2011;378:2123–35. 10. Ouma P, Van Eijk AM, Hamel MJ, Parise M, Ayisi JG, Otieno K, Kager PA, Slutsker L. Malaria and anaemia among pregnant women at first antenatal clinic visit in Kisumu, western Kenya. Trop Med Int Health 2007;12:1515–23.

9 of 10

11. Desai M, ter Kuile FO, Nosten F, McGready R, Asamoa K, Brabin B, Newman RD. Epidemiology and burden of malaria in pregnancy. Lancet Infect Dis 2007;7:93–104. 12. Ouédraogo S, Koura GK, Bodeau-Livinec F, Accrombessi MM, Massougbodji A, Cot M. Maternal anemia in pregnancy: assessing the effect of routine preventive measures in a malaria-endemic area. Am J Trop Med Hyg 2013;88:292–300. 13. Haider BA, Olofin I, Wang M, Spiegelman D, Ezzati M, Fawzi WW. Anaemia, prenatal iron use, and risk of adverse pregnancy outcomes: systematic review and meta-analysis. BMJ 2013;346:f3443. 14. Rasmussen K. Is there a causal relationship between iron deficiency or iron-deficiency anemia and weight at birth, length of gestation and perinatal mortality? J Nutr 2001;131:590S–601S; discussion 601S-603S. 15. Stroup DF, Berlin JA, Morton SC, Olkin I, Williamson GD, Rennie D, Moher D, Becker BJ, Sipe TA, Thacker SB. Meta-analysis of observational studies in epidemiology: a proposal for reporting. Metaanalysis Of Observational Studies in Epidemiology (MOOSE) group. JAMA 2000;283:2008–12. 16. World Bank. World Bank list of economies [Internet]. 2014 [cited 2014 Feb 24]. Available from: http://data.worldbank.org/about/countryclassifications/country-and-lending-groups. 17. Moher D, Liberati A, Tetzlaff J, Altman DG, The PRISMA Group. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Med 2009;6:e1000097. 18. World Health Organization. Vitamin and Mineral Nutrition Information System: haemoglobin concentrations for the diagnosis of anaemia and assessment of severity [Internet]. WHO reference number: WHO/NMH/NHD/MNM/11.1. Geneva (Switzerland): World Health Organization; 2011. [cited 2014 Feb 24]. Available from: http://www. who.int/vmnis/indicators/haemoglobin. 19. Raudys SJ, Jain AK. Small sample size effects in statistical pattern recognition: Recommendations for practitioners. IEEE Trans Pattern Anal Mach Intell 1991;13:252–64. 20. Wells G, Shea B, O’connell D, Peterson J, Welch V, Losos M, Tugwell P. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses. The World Bank [Internet]. 2014 [cited 2014 Feb 25]. Available from: http://www.ohri.ca/programs/ clinical_epidemiology/nos_manual.pdf. 21. Zhang J, Yu KF. What’s the relative risk? A method of correcting the odds ratio in cohort studies of common outcomes. JAMA 1998;280:1690–1. 22. Willi C, Bodenmann P, Ghali WA, Faris PD, Cornuz J. Active smoking and the risk of type 2 diabetes: a systematic review and meta-analysis. JAMA 2007;298:2654–64. 23. Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst 1959;22: 719–48. 24. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;7:177–88. 25. Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ 2003;327:557–60. 26. Egger M, Smith GD, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997;315:629–34. 27. Duval S, Tweedie R. Trim and fill: A simple funnel-plot–based method of testing and adjusting for publication bias in meta-analysis. Biometrics 2000;56:455–63. 28. Thompson SG, Higgins JPT. How should meta-regression analyses be undertaken and interpreted? Stat Med 2002;21:1559–73. 29. Freeman MF, Tukey JW. Transformations related to the angular and the square root. Ann Math Stat 1950;21:607–11. 30. Zhang Q, Ananth CV, Li Z, Smulian JC. Maternal anaemia and preterm birth: a prospective cohort study. Int J Epidemiol 2009;38:1380–9. 31. Zhang Q, Ananth CV, Rhoads GG, Li Z. The impact of maternal anemia on perinatal mortality: a population-based, prospective cohort study in China. Ann Epidemiol 2009;19:793–9. 32. Lone FW, Qureshi RN, Emanuel F. Maternal anaemia and its impact on perinatal outcome. Trop Med Int Health 2004;9:486–90. 33. Lone FW, Qureshi RN, Emmanuel F. Maternal anaemia and its impact on perinatal outcome in a tertiary care hospital in Pakistan. East Mediterr Health J 2004;10:801–7. 34. Khoigani MG, Goli S,, Hasanzadeh A. The relationship of hemoglobin and hematocrit in the first and second half of pregnancy with pregnancy outcome. Iran J Nurs Midwifery Res 2012;17:S165–70. 35. Goshtasebi A, Moghaddam Banaem L. High maternal hemoglobin concentration and preeclampsia. Int J Fertil Steril 2012;6:41.

10 of 10

RAHMAN ET AL.

36. Muniswaran G, Suharjono H, Soelar S, Karalasingam S, Jeganathan R. Severity of anemia and its obstetric implications. Int J Gynaecol Obstet Hoboken, NJ: Wiley-Blackwell; 2014. p. 156–7. 37. Stoltzfus RJ. Iron deficiency: global prevalence and consequences. Food Nutr Bull 2003;24:S99–103. 38. Sukrat B, Wilasrusmee C, Siribumrungwong B, McEvoy M, Okascharoen C, Attia J, Thakkinstian A. Hemoglobin concentration and pregnancy outcomes: a systematic review and meta-analysis. Biomed Res Int 2013; 2013:769057. 39. Kumar AK, Chen LC, Choudhury M, Ganju S, Mahajan V, Sinha A, Sen A. Financing health care for all: challenges and opportunities. Lancet 2011;377:668–79. 40. Rahman MM, Gilmour S, Saito E, Sultana P, Shibuya K. Health-related financial catastrophe, inequality and chronic illness in Bangladesh. PLoS One 2013;8:e56873. 41. Xu K, Evans DB, Kawabata K, Zeramdini R, Klavus J, Murray CJ. Household catastrophic health expenditure: a multicountry analysis. Lancet 2003;362:111–7. 42. Nishtar S, Boerma T, Amjad S, Alam AY, Khalid F, ul Haq I, Mirza YA. Pakistan’s health system: performance and prospects after the 18th Constitutional Amendment. Lancet 2013;381:2193–206. 43. Boerma T, AbouZahr C, Evans D, Evans T. Monitoring intervention coverage in the context of universal health coverage. PLoS Med 2014; 11:e1001728.

44. de Andrade LO, Filho AP, Solar O, Rigoli F, de Salazar LM, Serrate PC, Ribeiro KG, Koller TS, Cruz FN, Atun R. Social determinants of health, universal health coverage, and sustainable development: case studies from Latin American countries. Lancet 2015;385:1343–51. 45. Marten R, McIntyre D, Travassos C, Shishkin S, Longde W, Reddy S, Vega J. An assessment of progress towards universal health coverage in Brazil, Russia, India, China, and South Africa (BRICS). Lancet 2014; 384:2164–71. 46. Masud T, Navaratneb KV. The expanded program on immunization: recommendations for improving performance. Washington (DC): The International Bank for Reconstruction and Development/The World Bank; 2012. 47. Ghanchi NK, Shakoor S, Thaver AM, Khan MS, Janjua A, Beg MA. Current situation and challenges in implementing malaria control strategies in Pakistan. Criti Rev Microbiol 2014 Oct 30 (Epub ahead of print; DOI: 10.3109/1040841X.2014.973368). 48. Lagomarsino G, Garabrant A, Adyas A, Muga R, Otoo N. Moving towards universal health coverage: health insurance reforms in nine developing countries in Africa and Asia. Lancet 2012;380:933–43. 49. Sachs JD. Achieving universal health coverage in low-income settings. Lancet 2012;380:944–7. 50. Meng Q, Xu L, Zhang Y, Qian J, Cai M, Xin Y, Gao J, Xu K, Boerma J, Barber SL. Trends in access to health services and financial protection in China between 2003 and 2011: a cross-sectional study. Lancet 2012; 379:805–14.