Materials and Methods
Data for this study were obtained from publicly funded maternal child health program clinic records submitted to the CDC Pregnancy Nutrition Surveillance System between 1990 and 1993. Ninety-six percent of the records were from women attending the Special Supplemental Nutrition Program for Women, Infants, and Children. We included all surveillance records on singleton pregnant women from 11 states in the United States who entered the maternal child health program between 1 and 36 weeks' gestation and who delivered a liveborn infant at 26 to 42 weeks' gestation (n = 407,416). States were selected on the basis of participation in the Pregnancy Nutrition Surveillance System during this period and data quality (gestational age, birth weight, maternal smoking, and hemoglobin or hematocrit data were normally distributed and within expected ranges). We excluded records with missing data on infant birth weight (n = 8899), infant sex (n = 20,299), maternal smoking status (n = 51,232), and maternal hemoglobin level and hematocrit (n = 44,863). These exclusions reduced our sample size to 282,123.
We excluded records with incompatible birth-weight-for-gestational age data by comparing the recorded birth weight with a sex-, race-, altitude-, and gestational-age–specific fetal growth reference developed at the CDC (Yip R, McLaren N. Optimal birth weight distribution at specific gestational age of the United States infants: The CDC intrauterine growth standard [Atlanta: Division of Nutrition, CDC, 1992. Unpublished document]). If the birth weight was more than three standard deviations (SD) from the reference median birth weight, we considered the birth weight or gestational age on the record to be erroneous and excluded the record from our study (n = 32,354). Table 1 shows birth weight ranges within three SDs of the reference median. After excluding records with probable errors in birth weight or gestational age, the final sample size was 249,769 pregnant women (61% of 407,416). Our reduced sample was similar to our original sample in terms of maternal age, race/ethnicity, marital status, gestational length, and week of program entry.
Hemoglobin status during pregnancy was determined by hemoglobin level or hematocrit at the woman's first prenatal visit to the program. Gestational age at hemoglobin measurement and delivery were based on the date of the woman's last menstrual period, reported at the woman's initial prenatal visit. Hematocrit values were converted to hemoglobin values (g/L) by using the following formula: hemoglobin = (hematocrit/2.97) × 10. The adjustment factor of 2.97 was based on all records on the Pregnancy Nutrition Surveillance System with both hemoglobin level and hematocrit reported from the same clinic visit (n = 145,886).
Maternal hemoglobin was adjusted for altitude and maternal smoking according to CDC criteria.15 We then calculated a hemoglobin-for-gestational age Z score (hemoglobin Z score) by comparing a woman's hemoglobin value with a gestational-age-specific hemoglobin reference value.14 Adjustment was made by week of gestation [hemoglobin Z score = (measured hemoglobin adjusted for smoking status and altitude–reference median hemoglobin for gestational week of measurement) ÷ SD of reference hemoglobin distribution]. A hemoglobin Z score of −1.0 to 1.0 was used as the referent, and Z scores less than −3.0 were defined as very low hemoglobin level (moderate-to-severe anemia), −3.0 to less than −2.0 as low hemoglobin level (mild anemia), −2.0 to less than −1.0 as low normal hemoglobin level, over 1.0 to 2.0 as high normal hemoglobin level, greater than 2.0 to 3.0 as high hemoglobin level, and greater than 3.0 as very high hemoglobin level. Hemoglobin equivalents of selected hemoglobin-for-gestational age Z scores are shown in Table 2.
Preterm birth was defined as less than 37 completed weeks' gestation, and SGA birth was defined as birth weight below the tenth percentile of a United States national sex- and gestational-age–specific reference for fetal growth.16 We used multiple logistic regression in SAS (SAS Institute, Inc., Cary, NC) to calculate odds ratios (ORs) and 95% confidence intervals (CIs). All covariates were tested for confounding and effect modification. The covariates tested were maternal age, maternal education, marital status, maternal race/ethnicity, pregnancy weight gain, prepregnancy body mass index (BMI) (weight in kilograms/height in meters squared), and maternal use of cigarettes. In addition, because previous studies have suggested differential effects by race, separate models were constructed to examine the association between maternal anemia and birth outcome in white women compared with black women.4,17 The final sample included Pregnancy Nutrition Surveillance System records from the ten states that contained data on all covariates (n = 173,031 [70% of 249,769]). Associations between maternal hemoglobin level and birth outcome in each state were very similar to associations in the combined data set. Furthermore, unadjusted associations between maternal hemoglobin level and birth outcome in the full sample (n = 249,769) were very similar to unadjusted associations in the final sample (n = 173,031). We present results of the adjusted models from the final multistate sample.
Table 3 shows maternal and infant characteristics by trimester of entry into the maternal-child health program, the trimester during which maternal hemoglobin level was measured. Because of the large sample size in each group, even small differences were statistically significant. Thus, the magnitude of the relative differences rather than their statistical significance was examined. Relative differences of more than 25% were considered noteworthy.
Women who entered the maternal-child health program during the first, second, and third trimester of pregnancy were similar in terms of age, education, marital status, pregnancy weight gain, prepregnancy BMI, and reported use of cigarettes during pregnancy. Second- and third-trimester visits included a slightly higher proportion of black women compared with first-trimester visits, but the relative difference was less than 25%.
The prevalence of low normal hemoglobin level, mild anemia, and moderate-to-severe anemia was higher among women who entered the maternal-child health program during the third trimester of pregnancy (Table 3). The prevalence of high and very high hemoglobin level was lower among women who entered during the third trimester. As expected, preterm birth was lower among women during the third trimester because these women had less opportunity to deliver before term. The rate of SGA increased very slightly from the first to the third trimester.
We found an increased risk of preterm birth among women with a low hemoglobin level during the first and second trimesters of pregnancy; this increased risk did not occur among those with a high hemoglobin level. Among women with a first-trimester hemoglobin measurement, the unadjusted rate of preterm birth was 8.0% among those with hemoglobin levels in the normal range (reference) and 11.1%, 11.8%, and 15.0% among women with low normal hemoglobin level, mild anemia, and moderate-to-severe anemia, respectively (data not shown). Among women in whom hemoglobin was measured during the second trimester, the corresponding rate of preterm birth was 8.3% among women with hemoglobin levels in the normal range and 10.3%, 13.4%, and 16.5% among women with low normal hemoglobin, mild, and moderate-to-severe anemia, respectively (data not shown). These risks persisted after adjustment for covariates (Table 4). Compared with first- and second-trimester women with hemoglobin levels in the normal range, women with low normal hemoglobin levels had a 10–30% increased risk of preterm birth, those women with mild anemia had a 30–40% increased risk, and those with moderate-to-severe anemia had a nearly 70% increased risk of preterm birth (Table 4). Low maternal hemoglobin level during the third trimester was not associated with increased risk of preterm birth.
In contrast, the risk of SGA was increased with high, but not low, maternal hemoglobin level during the first and second trimesters. The unadjusted rates of SGA among women with a normal (reference), high normal, high, and very high hemoglobin level during the first trimester were 11.6%, 10.9%, 12.7%, and 13.1%, respectively (data not shown). The rates were 12.5%, 12.2%, 14.0%, and 16.9% among women with a normal, high normal, high, and very high hemoglobin level during the second trimester, respectively (data not shown). Again, adjusted ORs confirmed these findings (Table 4). Compared with first- and second-trimester women with hemoglobin levels in the normal range, women with high hemoglobin levels had a nearly 30–40% increased risk of delivering an SGA infant; women with very high hemoglobin levels during the second trimester had the highest odds of an SGA birth (OR = 1.79). There was little association between maternal hemoglobin during the third trimester and SGA risk.
The pattern and magnitude of association between maternal anemia and preterm birth and SGA were similar for white and black women (data not shown), with the exception of slightly higher odds for preterm birth among black women with moderate-to-severe anemia during the second trimester (OR 1.65; 95% CI 1.19, 2.30 for black women and OR 1.20; 95% CI 0.64, 2.25 for white women).
We found that maternal anemia during pregnancy was associated with preterm birth but not with delivery of an SGA infant. Moreover, high maternal hemoglobin level was associated with SGA but not with preterm birth. We observed the strongest increased risk of preterm birth among women with more severe anemia (hemoglobin Z score less than −3.0), which corresponds to an approximate hemoglobin level less than 90 g/L to 100 g/L, depending on week of gestation (Table 2). The strongest risk of SGA was found among women with a hemoglobin Z score greater than 3.0 during the second trimester; this score in general, corresponds to a hemoglobin level greater than 144 g/L. We observed smaller increases in risk among women with more modest decreases or elevations in hemoglobin.
Our finding of no association between third-trimester anemia and preterm birth is consistent with those of previous studies.9,11,18 Scholl and Hediger18 suggest that the lack of association between third-trimester anemia and preterm birth may be attributable to the normal expansion of maternal plasma that occurs in pregnancy, which results in a physiologic anemia that is more difficult to differentiate from iron-deficiency anemia later in pregnancy and thus lowers the predictive value of anemia tests during the third trimester to detect true iron-deficiency anemia. We cannot definitively explain the lack of association between high hemoglobin level during the third trimester and SGA, but it is possible that women who entered the maternal-child health program during the third trimester were more likely overall to deliver an SGA infant and thus had multiple and competing risks for SGA.
Previous studies have suggested that at the same low hemoglobin level, white women have an increased risk of adverse pregnancy outcome compared with black women, suggesting that the lower hemoglobin or hematocrit levels observed in black women19,20 may represent a normal physiologic difference between racial groups.4,17 In our study, maternal hemoglobin level was an average of 6.9 g/L lower among black women compared with white women; however, we found little difference in risk of preterm birth or SGA among white and black women at the same hemoglobin level. The one exception of a slightly higher risk for preterm birth among black women with moderate anemia during the second trimester was in the opposite direction than would be expected if lower hemoglobin level in black women were physiologically normal.
Studies suggest that iron deficiency anemia influences preterm delivery.11,12 Scholl et al,12 in a US study comparing risk of adverse pregnancy outcome among women with iron-deficiency anemia, anemia, or anemia from causes other than iron deficiency, found that only iron-deficiency anemia during the first and second trimesters of pregnancy increased a woman's risk for preterm delivery and delivering an LBW infant. The results of the study by Scholl et al12 provide stronger evidence that iron-deficiency anemia is associated with preterm birth. The finding of Lu et al11 that maternal anemia was not associated with preterm delivery among iron- and folate-supplemented women provides additional evidence that anemia from causes other than iron deficiency is not associated with preterm delivery. Despite these findings, the causal relation and possible mechanism by which iron deficiency or iron-deficiency anemia is associated with preterm delivery are not yet established.21 The postulated risks associated with iron deficiency relate to impaired transport of hemoglobin and thus oxygen to the uterus, placenta, and fetus.2 The observed association between maternal anemia and preterm delivery reported here and by others may not indicate a causal association but may be due to underlying maternal or fetal conditions that produce both the anemia (for example, through increased bleeding) and preterm delivery. Nevertheless, our findings indicate that low maternal hemoglobin level represents an important indicator of a complication of pregnancy that can adversely affect gestational length and thus the subsequent health and survival of the infant.
A high hemoglobin level during pregnancy may indicate a failure in plasma volume expansion. During normal pregnancy, plasma volume expands by an average of 50%; and failure of such expansion to occur results in hemoconcentration or higher-than-expected hemoglobin values.22 Lack of normal plasma volume expansion occurs in pregnancy-induced hypertension and preeclampsia, conditions associated with poor fetal growth as a result of poor placental–fetal perfusion.22 Thus, the most plausible explanation for the association of high hemoglobin level and SGA is pregnancy-induced hypertension.
Our study design had several strengths. We adjusted measured hemoglobin to take into account maternal smoking status and altitude. Because cigarette smoking and residency at high altitude increase hemoglobin and hematocrit, adjustment is necessary to avoid bias in risk estimates. We also used appropriate-for-gestational-age–specific criteria to define anemia and high hemoglobin level, which most previous studies failed to do. Calculation of the hemoglobin-for-gestational-age Z score enabled us to take into account the normal change in hemoglobin levels over pregnancy in response to increased plasma and erythrocyte volume.22
Use of data from a large, multistate surveillance system enabled us to examine the entire spectrum of hemoglobin status to refine the level at which risk is increased. Our method of comparing levels of low maternal hemoglobin to a referent level that did not include high hemoglobin level enabled us to capture the risk associated with severe anemia. Higher hemoglobin cutoff values used to define anemia in previous studies may have masked the stronger association with more severe anemia because all women with hemoglobin levels below the cutoff value were combined into one anemic group.12 In our study, we found a weaker association between maternal anemia and preterm delivery when we used CDC anemia criteria to classify women as anemic (hemoglobin level less than 110 g/L during the first and third trimester and less than 105 g/L during the second trimester15) and included all nonanemic women in our referent group (OR 1.31; 95% CI 1.19, 1.45 among first-trimester women and OR 1.28; 95% CI 1.18, 1.38 among second-trimester women). Only Lieberman et al7,8 reported an increased risk of preterm birth using a relatively high cutoff (hematocrit less than 38%); however, they used hematocrit at delivery without taking into account the normal increase in hematocrit at the end of pregnancy, thus producing a spurious positive effect.9 Lu et al11 examined the spectrum of hematocrit values and found no association between anemia and birth outcome after adjusting for potential confounders; however, their study was based on iron- and folate-supplemented pregnant women, suggesting that the low hematocrits in their study population may have been due to causes other than iron-deficiency anemia.
Our examination of LBW-associated outcomes preterm delivery and SGA enabled us to identify the specific outcomes associated with low and high maternal hemoglobin level. Consistent with previous studies,4,6,13 our early analyses for this study detected an increased risk of LBW with both low and high maternal hemoglobin. However, when we examined separately the association between hemoglobin and the birth weight–associated outcomes, a distinct picture emerged: Low hemoglobin level was specific for preterm birth and high hemoglobin level was specific for SGA. The different pattern of association with the outcomes may shed some light on the pathophysiologic mechanisms underlying the observed associations.
A final strength of our study is that our data set provided information on potential confounders of the association between maternal hemoglobin level and birth outcome. Not all previous studies adjusted for potential confounders.4,6 Of studies that performed adjusted analyses, not all were able to adjust for important potential confounders, such as maternal smoking and undernutrition.9,13
Our study also had limitations. Because the Pregnancy Nutrition Surveillance System data include only one hemoglobin measurement for each pregnant woman, we do not know whether a woman's anemia status at program entry differed from her status at an earlier or later stage of the pregnancy. Nevertheless, our analysis compared a women's hemoglobin level at a specific stage of pregnancy with those of other women in whom hemoglobin was measured at the same gestational stage. Random error in measurement of hemoglobin and hematocrit may have occurred more in our study than in a controlled study because we used hemoglobin data collected across clinics by multiple professionals using capillary sampling techniques. Random error may have diminished the strength of the observed ORs. Furthermore, one of the benefits of the Special Supplemental Program for Women, Infants, and Children is that women are given iron-fortified cereals and vitamin C–containing juices as part of their food package. In addition, women who screen positive for anemia are referred to their practitioner for treatment. Thus, program interventions may have diminished the strength of the association observed between anemia and preterm delivery. Data on pregnancy complications (eg, first- and second-trimester bleeding) and the cause of preterm birth (induced or spontaneous) were lacking. Data on complications may have shed some light on the proposed mechanisms in our study.
A final limitation of our study is that data on maternal iron status are not available on the Pregnancy Nutrition Surveillance Data set. Therefore, we could not verify whether the observed anemia was related to iron deficiency or some other cause. Although anemia stems from many causes, iron deficiency is by far the most important cause, in part because the very high iron requirement of pregnancy is difficult to meet. Therefore, it is reasonable to assume that the majority of the moderate anemia cases in our study were due to iron deficiency. Furthermore, our finding of a stronger increased risk of preterm delivery with more severe anemia suggests an association with iron-deficiency anemia because decreasing hemoglobin values increases the predictive value of anemia for iron deficiency.20
Iron supplementation during pregnancy is a routine practice intended to prevent iron-deficiency anemia. Our study suggests that iron supplementation may prevent preterm birth, if the observed association can be established as causal. Results of one trial that examined whether iron supplementation reduced the occurrence of preterm delivery were inconclusive.23 Because standard clinical practice is to provide iron supplementation to all pregnant women, it would not be ethical to withhold iron supplementation from a control study group of pregnant women without evidence that the group was iron-replete. Nevertheless, in the absence of clinical trial evidence of the benefit of iron supplementation on birth outcome, the high risk of iron deficiency among pregnant women and the demonstration that iron supplementation reduces the prevalence of iron-deficiency anemia in the absence of adverse effects are sufficient to justify the practice of routine supplementation during pregnancy.24
Evidence of a causal relation in the observed association between high maternal hemoglobin level and SGA is lacking. Iron supplementation cannot increase the hemoglobin level beyond what is optimal for a given person; it thus cannot be regarded as one of the causes of high hemoglobin levels. It is most likely that both elevated hemoglobin levels and SGA are the result of common disorders of pregnancy. We agree with other researchers11,17 that high hemoglobin level during pregnancy deserves more attention than it currently receives in clinical practice.
1. Centers for Disease Control and Prevention. Pregnancy nutrition surveillance, 1996 full report. Atlanta: US Department of Health and Human Services, 1998.
2. United States Preventive Services Task Force. Routine iron supplementation during pregnancy. JAMA 1993;270:2846–54.
3. United States Preventive Services Task Force. Screening for iron deficiency anemia-including iron prophylaxis. In: Guide to clinical preventive services. 2nd ed. Alexandria, Virginia: International Medical Publishing; 1996:231–46.
4. Garn SM, Ridella SA, Petzold AS, Falkner F. Maternal hematological levels and pregnancy outcomes, Semin Perinatol 1981;5:155–62.
5. Kim I, Hungerford DW, Yip R, Kuester SA, Zyrkowski C, Trow-bridge FL. Pregnancy nutrition surveillance system—United States, 1979–1990. MMWR CDC Surveill Summ 1992;41:26–41.
6. Murphy JF, O'Riordan J, Newcombe RG, Coles EC, Pearson JF. Relation of haemoglobin levels in first and second trimesters to outcome of pregnancy. Lancet 1986;1:992–5.
7. Lieberman E, Ryan KJ, Monson RR, Schoenbaum SC. Risk factors accounting for racial differences in the rate of premature birth. N Engl J Med 1987;317:743–8.
8. Lieberman E, Ryan KJ, Monsen RR, Schoenbaum SC. Association of maternal hematocrit with premature labor. Am J Obstet Gynecol 1988;159:107–14.
9. Klebanoff MA, Shiono PH, Berendes HW, Rhoads GG. Facts and artifacts about anemia and preterm delivery. JAMA 1989;262:511–5.
10. Klebanoff MA, Shiono PH, Selby JV, Trachtenberg AI, Graubard BI. Anemia and spontaneous preterm delivery. Am J Obstet Gynecol 1991;164:59–63.
11. Lu ZM, Goldenberg RL, Cliver SP, Cutter G, Blankson ML. The relationship between maternal hematocrit and pregnancy outcome. Obstet Gynecol 1991;71:190–4.
12. Scholl TO, Hediger ML, Fischer RL, Shearer JW. Anemia versus iron deficiency: Increased risk of preterm delivery in a prospective study. Am J Clin Nutr 1992;55:985–8.
13. Steer P, Alam MA, Wadsworth J, Welch A. Relation between maternal haemoglobin concentration and birth weight in different ethnic groups. BMJ 1995;310:489–91.
14. CDC criteria for anemia in children and childbearing-aged women. MMWR Morb Mortal Wkly Rep 1989;38:400–4.
15. Yip R, Parvanta I, Cogswell ME, McDonnell SM, Bowman BA, Grummer-Strawn LM, et al. Recommendations to prevent and control iron deficiency in the United States. MMWR Morb Mortal Wkly Rep 1998;47(RR-3):1–29.
16. Alexander GR, Himes JH, Kaufman RB, Mor J, Kogan M. A United States national reference for fetal growth. Obstet Gynecol 1996;87:163–8.
17. Blankson ML, Goldenberg RL, Cutter G, Cliver SP. The relationship between maternal hematocrit and pregnancy outcome: Black-white differences. J Natl Med Assoc 1993;85:130–4.
18. Scholl T, Hediger ML. Anemia and iron-deficiency anemia: Compilation of data on pregnancy outcome. Am J Clin Nutr 1994; 59(Suppl):492S–501S.
19. Perry GS, Byers T, Yip R, Margen S. Iron nutrition does not account for the hemoglobin differences between blacks and whites. J Nutr 1992;122:1417–24.
20. Johnson-Spear MA, Yip R. Hemoglobin difference between black and white women with comparable iron status: Justification for race-specific anemia criteria. Am J Clin Nutr 1994;60:117–21.
21. Allen L. Pregnancy and iron deficiency: Unresolved issues. Nutr Rev 1997;55:91–101.
22. Blackburn ST, Loper DL. Maternal, fetal, and neonatal physiology. A clinical perspective. Philadelphia: WB Saunders, 1992:159–200.
23. Hemminki E, Rimpelä U. A randomized comparison of routine versus selective iron supplementation during pregnancy. J Am Coll Nutr 1991;10:3–10.
24. Binkin NJ, Yip R. When is anemia screening of value in detecting iron deficiency? In: Hercberg S, Galan P, Dupin H, eds. Recent knowledge of iron and folate deficiencies in the world. Vol. 197. Paris: L'Institut National de la Santé et de la Recherche Médicale, 1990:137–45.