Hemoglobin Concentration and Pregnancy Outcomes: A Systematic Review and Meta-Analysis

Objective. To conduct a systematic review and meta-analysis of hemoglobin effect on the pregnancy outcomes. Methods. We searched MEDLINE and SCOPUS from January 1, 1990 to April 10, 2011. Observational studies addressing association between hemoglobin and adverse pregnancy outcomes were selected. Two reviewers independently extracted data. A mixed logistic regression was applied to assess the effects of hemoglobin on preterm birth, low birth weight, and small for gestational age. Results. Seventeen studies were included in poolings. Hemoglobin below 11 g/dL was, respectively, 1.10 (95% CI: 1.02–1.19), 1.17 (95% CI: 1.03–1.32), and 1.14 (95% CI: 1.05–1.24) times higher risk of preterm birth, low birth weight, and small for gestational age than normal hemoglobin in the first trimester. In the third trimester, hemoglobin below 11 g/dL was 1.30 (95% CI: 1.08–1.58) times higher risk of low birth weight. Hemoglobin above 14 g/dL in third trimester decreased the risk of preterm term with ORs of 0.50 (95% CI: 0.26–0.97), but it might be affected by publication bias. Conclusions. Our review suggests that hemoglobin below 11 g/dl increases the risk of preterm birth, low birth weight, and small gestational age in the first trimester and the risk of low birth weight in the third trimester.


Introduction
Anemia has been claimed to be the most common nutritional disorder in pregnancy across the world [1,2].The worldwide prevalence is estimated at 41.8% (95% CI: 39.9-43.8)[2] and is more common in African (57.1%, 95% CI: 52.8-61.3)pregnant women.The prevalence, however, depends on the definition of anemia, in which two definitions are commonly used, that is, the Centers for Disease Control and Prevention (CDC) and the World Health Organization (WHO) definitions [3,4].Adverse pregnancy outcomes thought to be affected by anemia include maternal mortality, perinatal mortality, preterm birth (PTB), low birth weight (LBW), and small for gestational age (SGA).Previous research has demonstrated a strong association between severe anemia and maternal mortality [5], but the risk of maternal mortality in pregnant women with moderate anemia (i.e., hemoglobin concentration of 40-80 g/dL) was inconclusive.The impact of anemia on other adverse pregnancy outcomes (e.g., PTB, LBW and SGA) is controversial.Some studies found significant associations [6][7][8][9][10][11][12], while other studies did not [13][14][15][16]; this is likely the result of different studies using different criteria or cutoff thresholds for defining anemia.In another way, some studies reported the association between high hemoglobin concentration and adverse pregnancy outcomes [17,18].Although a previous systematic review published in 2000 [19] reported maternal anemia in early pregnancy (<20 gestational week) increased the risk of PTB but not for LBW and SGA.Most of the included studies in this review were from developed countries, except 1 study was from African [20], and 2 studies were from China [9,21].Anemic pregnancy was more prevalent in developing countries; thus the effects of anemia on pregnancy outcomes in developing countries were still in question.We therefore conducted a systematic review to determine the association of hemoglobin concentration and adverse pregnancy outcomes including PTB, LBW, and SGA in each trimester of pregnancy.

Identification of Studies.
Studies were identified from MEDLINE and SCOPUS from January 1, 1990 to April 10, 2011.Reference lists from selected articles, narrative reviews, and systematic reviews were also reviewed to find relevant articles that were not identified by the initial search strategies.The following search terms were used: pregnancy, pregnant women, hemoglobin/haemoglobin, anemia/anaemia, hematologic/haematologic parameter, mortality, preterm birth/ delivery, low birth weight, and small for gestational age.Search strategies are clearly described in Appendix 1 in the Supplementary Material available online at http://dx.doi.org/10.1155/2013/769057.Only human studies published in English were considered.

Study Selection.
Eligibility assessment was performed by one reviewer using the following inclusion criteria.Any observational study (i.e., case control or cohort) performed in singleton pregnancy that assessed the association between hemoglobin concentration and any adverse pregnancy outcomes (i.e., still births, neonatal mortality, perinatal mortality, LBW, PTB, and SGA) reported gestational age at the time of hemoglobin testing and had sufficient data to allow calculation of the odds ratio and 95% confidence interval for dichotomous outcomes, number of subjects, mean and standard deviation according to hemoglobin concentration groups for continuous outcomes.

Data Extraction.
The data extraction was performed independently by two reviewers.Data on study characteristics (i.e., maternal age, ethnicity, gestational age at first antenatal visit, gestational age at delivery, parity, number of antenatal care visits, and smoking), mean and standard deviation of continuous outcomes, and frequencies of crosstabulation between each hemoglobin group and outcome for categorical data were extracted.Any disagreement was resolved by consensus between the two reviewers.If no agreement could be reached, it was adjudicated by a third reviewer.

Risk of Bias Assessment.
A risk of bias assessment was independently performed by 2 reviewers.The tool use was modified from meta-analysis for genetic association studies using epidemiological part [22].Four domains were assessed, which were representativeness of subjects, ascertainment of outcomes, ascertainment of exposures, and confounding bias.Disagreements between the two reviewers were solved by the third author.

Outcomes of Interest.
The outcomes of interests were PTB, LBW, and SGA.Briefly, LBW was defined as a newborn with weight at birth of less than 2500 g.PTB was defined as a neonate born before 37 weeks gestational age (the 259th day).SGA was a newborn whose birth weight was below the 10th percentile for gestational age.

Statistical Analysis.
Characteristics of all included studies were described including study design, number of participants, hemoglobin cutoff, trimester, and pregnancy outcomes.Hemoglobin concentration was categorized as <9, <10, <11, 11-13.9, and ≥14 g/dL in mutually exclusive fashion for each study.To assess hemoglobin effects, data were pooled separately according to the trimester and pregnancy outcomes using the cutoff of 11-13.9g/dL as the reference group.To compare pregnancy outcomes between multiple hemoglobin cutoffs in the same time, summary data of hemoglobin cutoffs and outcome groups were then expanded to individual patient data using the expand command in STATA.A mixed logit model with a random intercept (i.e., to account for between-study variation) was applied to assess hemoglobin effects on pregnancy outcome.The estimated pooled odds ratio (OR) along with 95% confidence interval (CI) was estimated by exponential logit coefficients.A degree of heterogeneity was estimated using multivariate metaanalysis method [23].All analyses were performed using STATA version 12.The statistical significance was set to twosided  < 0.05 for all analyses.

Results
Eighty-five potentially relevant articles were identified, of which 65 studies were excluded, leaving 20 studies [7, 9-12, 15-18, 24-34] for data extraction.The reasons for exclusion are described in Figure 1.Characteristics of all included studies are summarized in Table 1.Briefly, 50% percent of included studies were prospective cohorts.Mean age ranged from 16.1 to 30.6 years.Most of the eligible studies included pregnant women from Asian populations (12 studies), followed by European (4 studies), Africans (2 studies), and North Americans (2 studies).PTB, LBW, and SGA were outcomes of interests in 15, 14, and 6 studies, respectively.Still births, neonatal deaths, and perinatal deaths were less frequently reported only in 5, 3, and 4 studies, respectively, and we therefore did not pool these outcomes.The number of studies available for analyses according to trimester and outcome is shown in Figure 1.
Risk of bias was performed independently by the two reviewers with the total agreement rate of 98.75% with the kappa statistic of 0.946 ( < 0.001).For those items where there were disagreements, the senior reviewer (AT) had performed risk of bias and made decisions.Results of assessments for individual studies were described in Supplementary Table 1 .The highest quality was in the domain  of ascertainment of outcome (100.0%low risk) followed by representativeness of subjects (95.0%), while the lowest was ascertainment of hemoglobin test or definition (70.0%).
The Egger test was applied and suggested no evidence of small study effect with coefficients of −0.019 ( = 0.322), 0.140 ( = 0.513), 0.076 ( = 0.725), and 0.251 ( = 0.474) for the cutoffs below 9, 10, 11, and above 14 g/dL, respectively.Pooling studies in developing countries and in cohort studies did not change much results (data were not shown).

3.3.
Small for Gestational Age.Effects of low hemoglobin concentration in the first trimester on SGA were assessed in 6 studies [9,12,17,18,27,28] with the sample size totaling 94, 280 women (Supplementary Table 4).The mixed logistic model suggested that hemoglobin concentrations below 10 and 11 g/dL increased the risk of SGA by 26% (OR: 1.26, 95% CI: 1.09-1.45)and 14% (OR: 1.14, 95% CI: 1.05-1.24),respectively (Table 2 and Figure 2).The hemoglobin effects were mildly heterogeneous for both cutoffs with the  2 of 1% and 2%.There was no evidence of small study effects as suggested by the Egger test with coefficients of −0.124 ( = 0.402) and 0.229 ( = 0.645), respectively.Pooling effects in 4 studies conducted in developing countries did not change much results (data were not shown).

Discussion
We have performed a systematic review and meta-analysis to assess effects of hemoglobin concentration on pregnancy outcomes according to trimesters.Our results suggest that lower hemoglobin concentration is associated with a higher risk of poor pregnancy outcomes in both first and third trimesters.
The risk of PTB, LBW, and small gestational age was approximately 10-17% and 26-57% higher in pregnant women who had a hemoglobin concentration below 10 and 11 g/dL in the first trimester, respectively.In the third trimester, hemoglobin below 11 g/dL increases the risk of LBW by 30% but not for preterm term.Hemoglobin below 10 g/dL in the third trimester also increases the risk of PTB and LBW by 2.6 and 3.6 times, respectively.Hemoglobin ≥14 g/dL did not increase risk in any trimester of pregnancy but conversely reduced a risk of PTB by 50%.
Our results confirm the findings from a previous metaanalysis [19] showing that low hemoglobin concentration in early pregnancy (<20 weeks gestation) was associated with PTB.In addition, we also found that low hemoglobin concentration in the first trimester was a risk of LBW and SGA, which has not been reported in the previous review.These findings may be explained by reduced oxygen transportation from the mother to fetus and may reflect inadequate iron reserves during early pregnancy.Subgroup analysis did not show any clear differences in the effects of low hemoglobin concentration between developing and developed countries.The magnitude of the effect in developing countries was similar to results of other previous studies [35][36][37].Pooling studies based on cohorts only did not change the results when compared to pooling all studies with cohorts and case controls.
Our study has a number of strengths.We assessed effects of various hemoglobin cutoffs on pregnancy outcomes stratified by trimesters.The use of hemoglobin cutoff in individual studies varied from 7 to 16 g/dL, and one study had more than one cutoff.We therefore applied a mixed logistic model in order to simultaneously assess hemoglobin effects without inflating type one error.Between-study variations were also taken into account in the mixed logistic model.A degree of heterogeneity was also estimated using multivariate metaanalysis method.
There are, however, some limitations and caveats with our study.Because of the varying cutoffs used in different studies, the only way to reasonably pool data was to expand the summary data to individual level data and then pool based on common thresholds; thus some subjects were excluded if studies used a reference cutoff lower than 11 g/dL.It is also possible that our results are confounded by other factors in which analysis based on summary data could not adjust for.A meta-analysis of individual patient data should be conducted to calibrate hemoglobin cutoff with adjusting for confounders; that is, anemia is a marker of general poor health in developing countries but not in developed countries.Our review also excluded non-English articles due to limitation in translation issue.

Conclusion
Our review suggests that hemoglobin below 11 g/dL increases the risk of LBW in both first and third trimesters, PTB and small gestational age in the first trimester.Conversely, hemoglobin 14 g/dL or higher can conversely reduce the risk of PTB in the third trimester, but the result might be affected by publication bias.

Figure 3 :
Figure 3: Pooling effects of hemoglobin concentration on pregnancy outcomes in the third trimester.

Table 1 :
Characteristics of included studies.

Table 2 :
Association between hemoglobin concentration and pregnancy outcomes.