IS THERE A CAUSAL RELATIONSHIP? 591S Limitations of the data reviewed: all studies As is the case in nonpregnant adults, anemia in pregnant women does not result solely from lack of dietary iron. Other causes of anemia include the following: hookworm infection; malaria; schistosomiasis; recent or current infections; chronic inﬂammation; hereditary anemias; and other nutritional deﬁ- ciencies, particularly of folic acid or vitamin B-12. The im- portance of these other causes of anemia varies from popula- tion to population. Some of these causes of anemia are also FIGURE 1 General approach to the review of the literature. FGR,fetal growth retardation. independently associated with birth outcomes. The differential increases in plasma volume and red cell mass that are characteristic of pregnancy make interpretation of hemoglobin values challenging. The ﬁrst problem is thatLBW, prematurity, fetal growth and perinatal mortality. In plasma volume expansion, with its corresponding fall in he-addition, the Cochrane Reviews on routine iron (Mahomed moglobin concentration, obscures the usual relationship be-1998b) and folate (Mahomed 1998a) supplementation during tween iron deﬁciency and low hemoglobin values. It alsopregnancy were consulted; the studies cited there, which date makes it difﬁcult to interpret the plasma-based indicators ofback to 1955, were also reviewed. iron deﬁciency (e.g., ferritin), which are also diluted by plasma Despite the relatively comprehensive nature of this search volume expansion during pregnancy. The second problem isstrategy, some limitations are nonetheless present. It did not that plasma volume and red cell mass change throughoutinclude a speciﬁc search for each possible cause of maternal pregnancy. There is little consistency in the point at which Downloaded from jn.nutrition.org by guest on February 21, 2012anemia and the outcomes of interest. It was assumed that most maternal hemoglobin concentration was assessed during preg-of these would be picked up with the search terms “anemia” nancy; some investigators assessed this early in pregnancy,and “hemoglobin” and by the more speciﬁc attention to folate some later in pregnancy and others only at delivery. Moredeﬁciency, the next most common nutritional cause of mater- confusing still are the papers that reported an associationnal anemia after iron deﬁciency. between the lowest maternal hemoglobin value and some The studies obtained were grouped into two broad catego- outcome but did not reveal when this lowest value was ob-ries, i.e., those studies suitable for establishing whether there is tained, making it impossible to correct for the gestational agean association between maternal anemia and birth outcomes at which the measurement was made.and those studies suitable for establishing whether this associ- The association between anemia and birth outcomes mayation is causal. The ﬁrst group included observational studies be stronger if the anemia occurs at one time during pregnancyas well as intervention trials that either did not meet usual rather than at another time. This is because of differences incriteria for causal inference (e.g., random assignment of sub- the rates of fetal growth and development during gestation.jects to treatment groups, blind assessment of outcomes) or Similarly, the effectiveness of treatments may vary dependingwere analyzed outside the framework of the intervention. The on when and for how long they are offered (G. H. Beaton andsummaries of these studies are not included here. The second G. P. McCabe, unpublished, 2000). Finally, any effectivegroup consisted of interventions that were designed to elimi- treatment for anemia will reduce the association betweennate maternal anemia, usually with the provision of iron or preexisting anemia and birth outcomes in observational stud-folic acid supplements or both and in which relevant birth ies, and women probably received supplemental iron in manyoutcomes were assessed (Table 1). of these investigations. FIGURE 2 Detailed conceptual framework used to guide interpretation of the literature. Hb, hemoglobin con- centration.
592S SUPPLEMENT TABLE 1 Effects of interventions to alleviate of iron deﬁciency, iron deﬁciency anemia or folate deﬁciency on size at birth and duration of gestation1 Effect on perinatal Effect on mortality; Design Effect on duration fetal HbStudy site; authors Intervention Subjects concerns size at birth of gestation concentrationScotland Iron, 12 or 115 mg/ n ϭ 173; Hb Ͼ100 False negative (not No difference in n/a n/a (Aberdeen); d from 20 to 36 g/L at ﬁrst visit anemic at outset, BW among the (Paintin et al. wk of gestation high mean BW) treatment groups 1966) or a placebo (mean BW about 3.3 kg)Nigeria (Ibadan); Folic acid, 5 mg/d n ϭ 75, but only 54 Confounding (non- No effect of the n/a n/a (Fleming et al. or placebo completed the random treatment on BW 1968) (alternate study; PCV assignment), bias (mean ϭ 3.0 kg) assignment) Ն27% at 26 wk (high dropout gestation rate), false negative (small sample size)England (Liverpool); Ferrous gluconate, n ϭ 698 Confounding (non- Women with n/a n/a Downloaded from jn.nutrition.org by guest on February 21, 2012 (Rae and Robb 200 mg/d, or iron randomized by random megaloblastic 1970) ϩ folic acid, 5 day of clinic assignment); false (not mg/d attendance, but negative (anemia normoblastic) all women seen not corrected by anemia (Hb Ͻ109 in the 1st either treatment) g/L) tended to trimester were have babies with assigned for Fe a lower BW than ϩ folic acid those who were group never anemic (not signiﬁcant)South Africa; Iron, 200 mg/d; iron n ϭ 183 Bantu Need for No difference in Preterm deﬁned n/a (Baumslag et al. ϩ folic acid, 5 (after 28 wk supplements as BW among the as BW Ͻ5 lb 1970) mg/d; iron ϩ folic gestation) and well as white subjects; acid ϩ vitamin 172 whites (after hematologic excess of babies B-12 50 g/d 24 wk gestation); response to them Ͻ5 lb among the initial Hb not were not Bantu given iron reported reported only but can’t distinguish between term and preterm LBWEngland (London); Ferrous sulfate, 200 n ϭ 643, mean Hb False negative (not No difference n/a n/a (Fletcher et al. mg/d; iron ϩ folic at booking about anemic, high between 1971) acid, 5 mg/d 130 g/L mean BW) treatment groups (mean BW ϭ 3.3 kg)India (Hyderabad); Iron, 60 mg/d; iron n ϭ 200 at 20–24 Bias (high dropout BW 200–300 g No difference No difference (Iyengar 1971) ϩ folic acid, 100 wk gestation, but rate) higher with 200 between between or 200 or 300 only 114 or 300 g folic treatment treatment g/d completed the acid than none or groups groups trial; Hb Ͼ85 g/L 100 g (P Ͻ 0.05)Australia; (Fleming Ferrous sulfate, 60 n ϭ 146 with Hb False negative (not Placebo, 3.476 kg Premature n/a et al. 1974) mg/d; folic acid, Ͼ10/dL at 20 wk anemic; high (n ϭ 17); Fe, deliveries were 5 mg/d, both or gestation mean BW and 3.310 kg (n ϭ excluded placebo low statistical 21); folic acid, power) 3.278 kg (n ϭ 15); both, 3.395 (n ϭ 20) (NS for main effects of Fe or folic acid)India (Delhi and Ferrous fumarate, n ϭ 647, stratiﬁed False negative [Fe No overall effect of n/a n/a Vellore); (Sood et none, 30, 60, 120 by initial Hb (all doses of 120 mg/ hematinics on al. 1975) or 240 mg/d; Ͼ50 g/L); d or less did not BW (data with folic acid, 5 treatment started eliminate anemia available for only mg/d, and B-12, at 22 wk (but even 30 mg 47% of subjects; 100 g every 2 gestation and of Fe with folic low mean BW, wk continued for 10– acid and B-12 2.7 kg; n ϭ 33– 12 wk produced ﬁnal Hb 56/group); 71 g values Ͼ100 g/L) difference in BW low statistical between 120 mg power]; bias (high Fe and controls dropout rate for (not signiﬁcant) BW)
IS THERE A CAUSAL RELATIONSHIP? 593S TABLE 1 (continued) Effects of interventions to alleviate of iron deﬁciency, iron deﬁciency anemia or folate deﬁciency on size at birth and duration of gestation1 Effect on perinatal Effect on mortality; Design Effect on duration fetal HbStudy site; authors Intervention Subjects concerns size at birth of gestation concentrationIndia (Hyderabad); Ferrous fumarate, n ϭ 282 with Hb Confounding (non- No treatment, n/a No effect of Fe (Iyengar and 60 mg/d alone or Ͼ85 g/L at 20–28 random 2.567 kg (30.8% or vitamins on Rajalakshmi with folic acid, wk gestation; assignment); LBW); Fe only, infant Hb at 3 1975) 0.5 mg/d subjects source of 2.650 kg (30.2% mo of age (n (alternate matched for controls not LBW); Fe ϩ folic ϭ assignment) height and parity speciﬁed acid, 2.890 kg 31–53/group) (15.5% LBW) (P Ͻ 0.001)England; (Trigg et Ferrous sulfate, 50 n ϭ 76 Fe alone; n False negative (not No effect of folic n/a n/a al. 1976) mg/d or ferrous ϭ 82 Fe ϩ folic anemic; high acid on BW in sulfate ϩ folic acid mean BW) Fe-supplemented acid, 0.05 mg/d women (mean BW ϭ 3.4 kg) Downloaded from jn.nutrition.org by guest on February 21, 2012England; (Taylor et Ferrous sulphate, n ϭ 48 randomly No placebo- No effect of Fe No difference n/a al. 1982) 350 mg/d ϩ 350 assigned to controlled group, treatment with between Fe g/d folic acid or receive either false negative BW (mean ϭ 3.5 supplementation no supplement iron or no (those randomly kg) and no treatment assigned were supplementation not anemic), in duration of excluded 3 gestation, but subjects who had excluded 3 premature subjects deliveries because of premature birthsFinland; (Romslo et Ferrous sulfate, 200 n ϭ 45 healthy False negative (not No effect of iron on n/a n/a al. 1983) mg/d or placebo women who anemic; high BW (mean ϭ 3.5 delivered mean BW) kg) singleton infants at termFrance; (Tchernia et Iron or placebo n ϭ 203 for study 1 False negative (not BW (P Ͻ 0.05) and Serum folate n/a al. 1983) (study 1); iron or (n ϭ 155 with Hb anemic, low birth length (P Ͻ values were iron ϩ folic acid Ͼ110 g/L who power); bias 0.001) were lower (P Ͻ 0.01) (study 3) were randomly (assignment to higher in infants among mothers assigned to treatment group of mothers who of preterm (Յ39 treatment) and n not speciﬁed received Fe ϩ wk) infants; ϭ 200 for study (study 3) folic acid length of 3 compared with gestation was those who longer (P Ͻ received Fe alone 0.025) among women with higher (Ͼ200 g/L) folate; length of gestation was longer (P Ͻ 0.001) among women treated with Fe ϩ folic acid than with Fe aloneFrance; (Zittoun et Fe sulfate, 105 mg n ϭ 203 at 28 wk; False negative No association of No association of n/a al. 1983) elemental Fe/d ϩ if Hb Ͻ110 g/L, (those randomly Fe treatment with Fe treatment 500 mg ascorbic treated; assigned were BW (mean ϭ 3.3 with length of acid otherwise not anemic; high Ϯ 0.5 kg) or gestation randomly mean BW) length of assigned to gestation or fetal treatment or Fe status placeboNigeria (Zaria); Clorquine ϩ n ϭ 200, Ͻ24 wk False negative (not No difference n/a No differences (Fleming et al. proguanil and gestation anemic) among the among the 1986) either ferrous groups (mean groups; 10.5% sulfate, 60 mg/d, BW ϭ 2.85 kg) perinatal folic acid, 1 mg/ mortality d, or both among n ϭ 152 with known outcome
594S SUPPLEMENT TABLE 1 (continued) Effects of interventions to alleviate of iron deﬁciency, iron deﬁciency anemia or folate deﬁciency on size at birth and duration of gestation1 Effect on perinatal Effect on mortality; Design Effect on duration fetal HbStudy site; authors Intervention Subjects concerns size at birth of gestation concentrationFrance; (de Benaze Ferrous betainate, n ϭ 191 pregnant False negative (not n/a No difference n/a et al. 1989) 45 mg elemental women beginning anemic) between the Fe/d in a divided at 22 wk groups in dose, or placebo gestation and duration of continuing until 2 gestation mo postpartum; initial Hb ϭ 125 g/L, serum ferritin ϭ 60 g/LFinland; (Hemminki Elemental Fe, 100 n ϭ 1451 routine False negative (not No difference No difference Neonatal and Rimpela¨ mg/d (routine); supplementation, anemic, high between the 2 Fe between the 2 mortality was 1991) slow release n ϭ 1461 mean BW) supplementation Fe signiﬁcantly ferrous sulfate, selective regimens in BW supplementation higher in the Downloaded from jn.nutrition.org by guest on February 21, 2012 50 mg 2 times/d supplementation; (mean 3.6 kg) regimens in routine (7.5/ (selective) Hb Ͼ110 g/L duration of 1000) than gestation the selective (2.2/1000) groupsIndia (Varanasi); Ferrous sulfate, 60 n ϭ 418 randomly Bias (high dropout BW was higher n/a Reduced (Agarwal et al. mg/d ϩ folate, assigned by rates) and false (2.88 kg) in the neonatal 1991) 500 g/d subcenters (n ϭ positive treatment than in death rates in 6) to supplement (randomization the control (2.59 the Fe- or placebo; only by subcenters kg) group (P Ͻ supplemented 137 of 215 in the with analysis by 0.001); LBW group (P Ͻ treated group individuals) reduced from 0.04) and 123 of 203 in 37.9% in the the control group controls to 20.4% completed the (P Ͻ 0.05); later trial; initial Hb start of 10.1–109 g/L supplementation (20–40 wk vs. 16– 19 wk) associated with higher LBW (23.1 vs. 12.1%, respectively)Denmark; Ferrous iron, 100 n ϭ 52 randomly False negative (not No difference No difference n/a (Thomses et al. mg/d ϩ 18 mg/g assigned as anemic), no between the 2 Fe between the 2 1993) entered clinic to control group, supplementation Fe multivitamin excluded those regimens in BW supplementation containing either with premature (mean ϭ 3.5 kg) regimens in 100 mg (n ϭ 22) births duration of or 18 mg (n ϭ gestation, but 21) ferrous iron excluded 3 subjects because of premature birthsGambia; Ferrous sulfate, 200 n ϭ 550 False negative No statistically n/a n/a (Menendez et al. mg/d (60 mg multigravidas (52% of treated signiﬁcant 1994) elemental Fe) or with PCV Ͼ25% group still difference in BW placebo; 5 mg randomly anemic after (56 g) or %LBW folic acid weekly assigned by delivery) between the compound of treatment groups; residence; 3.103 kg and 3% double-blind; LBW for the initial Hb 100– supplemented 101 g/L and 3.047 kg and 5% LBW for the placebo group; among the women who took Ͼ80 Fe tablets, BW was 96 g higher in the supplemented group (P ϭ 0.04)
IS THERE A CAUSAL RELATIONSHIP? 595S TABLE 1 (continued) Effects of interventions to alleviate of iron deﬁciency, iron deﬁciency anemia or folate deﬁciency on size at birth and duration of gestation1 Effect on perinatal Effect on mortality; Design Effect on duration fetal HbStudy site; authors Intervention Subjects concerns size at birth of gestation concentrationDenmark; (Milman Ferrous fumarate, n ϭ 135 randomly False negative (not No signiﬁcant n/a n/a et al. 1994) 200 mg/d or assigned to anemic), bias differences placebo receive placebo (9% exclusion among the (n ϭ 57) or after treated and ferrous fumarate randomization— control in total (n ϭ 63), double- primarily in duration of blind placebo group) gestation or mean birth weightIndia (Tamil Nadu); High risk (Fe ϩ n ϭ 12 subcenters, False negative (low No signiﬁcant n/a n/a (Srinivasan et al. folic acid if not assigned statistical power) differences 1995) anemic, double randomly within among the dose if anemic, 4 primary health treatments in BW Downloaded from jn.nutrition.org by guest on February 21, 2012 parenteral Fe if centers; initial Hb or %LBW (poor Hb Ͻ80 g/L), 93–100 g/L at 34 ascertainment of usual care (Tamil wk of gestation BW) Nadu Government) (Fe ϩ folic acid, 100 doses regardless of Hb value) or control (government program without services of midwives)Niger; (Preziosi et Ferrous betainate, n ϭ 197 at 28 wk False negative No difference n/a n/a al. 1997) 100 mg gestation; Ͼ65% (42% of the between elemental Fe/d anemic (Hb Ͻ110 treated group still treatment groups g/L) at 6 mo anemic at in BW (mean ϭ gestation delivery) 3.0 kg); birth length was longer in Fe- supplemented group (P Ͻ 0.05) 1 Abbreviations used: BW, birth weight; LBW, low birthweight; Hb, hemoglobin; n/a, not available; NS, not signiﬁcant; OR, odds ratio; PCV, packedcell volume; RR, relative risk. The relevant literature on this topic includes many older subjects when the unit of randomization was, for example, thestudies in which investigators did not distinguish between village and not the individual woman) were not often ainfants who were small for their gestational age and those who problem in this literature and therefore are not considered inwere born prematurely; both were included in the group la- detail.beled LBW. Some investigators solved this problem by re- To eliminate confounding and bias, random assignment tostricting their sample to term births. This strategy removes treatment, double-blind assessment of outcomes and a placebopreterm babies from the LBW group but also makes it impos- in the control group are normally used. Some of the oldersible to evaluate the effect of treatment on the duration of studies did not provide details on all of these procedures andgestation. may not have included them. To be able to attribute the positive outcomes to theLimitations of the data reviewed: intervention studies elimination of anemia, iron deﬁciency or both, these factors For the intervention studies to demonstrate a causal rela- must, in fact, be eliminated. This requires that the subjectstionship between correction of maternal anemia and an in- be offered an adequate dose of the target hematinic (e.g.,crease in birth weight, a number of conditions must be met. iron, folic acid, vitamin B-12, blood transfusions) and thatFor the purpose of this review, these factors fall into three they take the dose assigned for a sufﬁcient period. Unfor-broad categories, i.e., those that eliminate confounding and tunately, sometimes the doses of iron (and other hema-bias, those that permit one to attribute the effect observed to tinics) used in these studies were ineffective in correctingthe elimination of anemia, iron deﬁciency or both, and those maternal anemia, possibly because iron deﬁciency was notthat eliminate false-negative ﬁndings. False-positive ﬁndings the sole or even the primary cause of the anemia. In some(such as those that come from analyzing the data by individual cases, the investigators acknowledged that the dose of iron
596S SUPPLEMENTthat they used was too low; in other cases, women did not 1986) and data from the North West Thames region in thetake a sufﬁcient number of the pills provided. United Kingdom (Ͼ150,000 births) (Steer et al. 1995). To eliminate false-negative ﬁndings, subjects must have the It is likely that the causes of small size at birth differ at thepotential to respond to the treatment offered and there must two ends of the range of maternal hemoglobin concentrations.be a statistically adequate sample size to be able to detect this High hemoglobin values may reﬂect poor plasma volume ex-response. First, this means that subjects had to have a cause of pansion, which is itself associated with impaired fetal growthLBW that could be corrected by receipt of a hematinic such as (Duffus et al. 1971, Gibson 1973), or other pathological con-iron or folic acid. Those experiments conducted in women ditions (Yip 2000). Low (100 –110 g/L) hemoglobin values inwhose anemia was not caused by, for example, iron or folic late pregnancy probably reﬂect changes in plasma volumeacid deﬁciency, cannot be expected to respond to supplemen- (Whittaker et al. 1996). Only hemoglobin values Ͻ100 g/L aretation with these substances with either a reduction in anemia likely to reﬂect inadequate maternal nutritional status withor an increase in birth weight. Second, and similarly, the mean respect to iron, folic acid and other nutrients. The speciﬁcbirth weight in the treated population had to be sufﬁciently cause of the low maternal hemoglobin values remains un-low so that it could be expected to rise if the therapy were known in most available studies. It is noteworthy that theeffective. Mean birth weight of populations (high end of the U-shaped relationship is more apparent in studies that usedistribution: 3.5 kg) has long been known to be somewhat “lowest hemoglobin” than in those that control for the stage ofbelow the range of birth weights that are associated with gestation (Scanlon et al. 2000) or include data only fromminimal infant mortality (low end of the distribution: 3.5 kg) women very early in pregnancy, when changes in plasma(Hytten and Leitch 1971). The standard deviation around volume are minimal (Zhou et al. 1998). Thus, it is possiblethese means is usually ϳ0.5 kg. For this review, study popu- that this shape is spurious.lations in which the mean birth weights in the control group The large studies permit assessment of the maternal hemo-was Ն3.3 kg were not considered to have the potential to Downloaded from jn.nutrition.org by guest on February 21, 2012 globin values associated with the best birth outcomes. In therespond to treatment. Third, iron or folic acid deﬁciency must high risk population studied as part of the National Collabo-be the factor limiting birth weight so that correcting anemia rative Perinatal Project (Garn et al. 1981a), the LBW rate wascaused by these deﬁciencies will permit birth weight to rise. minimal at maternal hemoglobin values of 105–125 g/L inThere are numerous examples in which this condition proba- Caucasian women. In the Cardiff Births Survey (Murphy et al.bly was not met. It is especially likely to have been the case 1986), LBW was minimal when the maternal hemoglobinwhen the population’s mean birth weight was low. value at booking was 104 –132 g/L, regardless of whether Finally, a statistically adequate sample size to ascertain booking was before 13 wk gestational age or 13–19 or 20 –24whether iron or folic acid improved maternal hematologic wk gestational age. In the recent data from the United King-status is much lower than that needed to ascertain whether dom (Steer et al. 1995), birth weight was highest at maternalbirth weight or the duration of gestation has increased or hemoglobin values of 86 –95 g/L; LBW rates were lowest atperinatal mortality has decreased. For example, it is often maternal hemoglobin values of 96 –105 g/L. Both of thesepossible to see a hematologic response to iron treatment with hemoglobin values are below the cut-off value for anemia in50 women in each treatment group, but at least 250 women in pregnant women by current WHO criteria (i.e., 110 g/L).each treatment group would be required to detect a 100-g Interpreting the data in this report is not straightforward,difference in birth weight and even more subjects to detect an however, because the hemoglobin values used were deter-effect on mortality. Therefore, it is not surprising that many of mined at various times during gestation. The only data avail-the studies reviewed were able to detect an improvement in able for African-American women come from the Nationalhematologic values but still lacked sufﬁcient statistical power Collaborative Perinatal Project and show a minimal rate ofto detect an effect on these birth outcomes if such an effect LBW at lowest maternal hemoglobin values of 85–95 g/Lhad been present. (Garn et al. 1981a). In a recent study of Chinese women Overall, it is noteworthy that the effects of failing to treat (Zhou et al. 1998), the minimum risk of LBW occurred atanemia successfully or to eliminate the sources of false-nega- hemoglobin values of 110 –119 g/L, but these values weretive results that are listed above is to bias ﬁndings toward the determined very early in pregnancy (4 – 8 wk of gestation), atnull. That is, investigations with one or more of these prob- a time of minimal expansion of plasma volume.lems are likely to ﬁnd that iron or folic acid did not improve This same U-shaped pattern was also observed for thebirth outcomes when this might not have been true if a more association between maternal hemoglobin concentration andadequate experimental design for this purpose had been used. duration of gestation as well as for the association between maternal hemoglobin and neonatal mortality. The group ofEvidence for an association between iron deﬁciency, iron- studies from which this assessment can be made is much moredeﬁciency anemia, or anemia and birth outcomes limited than that for birth weight. Many of the latter studies either did not report the duration of gestation or restricted There is ample evidence from observational studies, both their sample speciﬁcally to mothers of term infants. Minimallarge and small, that there is an association between maternal rates of prematurity occurred at maternal hemoglobin values ofanemia (as deﬁned by hemoglobin concentration) and size at 115–125 g/L in Caucasian women in the National Collabora-birth, duration of gestation, and neonatal or perinatal mortal- tive Perinatal Project (Garn et al. 1981a). In the recent dataity. from the United Kingdom (Steer et al. 1995), the lowest rate In its broadest form, this association is U-shaped, i.e., the of preterm birth occurred at maternal hemoglobin concentra-proportion of LBW infants rises (and the mean birth weight tions of 96 –105 g/L, also below the current cut-off value fordrops) when maternal hemoglobin values are either at the low maternal anemia. Prematurity was minimal at the lowest ma-or high end of the range. This association is most obvious in ternal hemoglobin values of 105–115 g/L in African-Americanthe three largest data sets examined, namely, the National women in the National Collaborative Perinatal Project (GarnCollaborative Perinatal Project from the United States (nearly et al. 1981a). When the duration of gestation was controlled60,000 births) (Garn et al. 1981a), the Cardiff Births Survey for, the minimum risk of preterm birth occurred above thefrom the United Kingdom (ϳ55,000 births) (Murphy et al. cut-off value for anemia in a smaller cohort of Chinese women
IS THERE A CAUSAL RELATIONSHIP? 597S TABLE 2 Relative and attributable risk of low birth weight according to severity and type of maternal anemia during pregnancy1,2 Relative risk (attributable risk) of LBW compared to mild or no anemiaStudy site; authors Moderate anemia Severe anemia (usually Յ 80 g/L) Iron deﬁciency anemiaKashmir; (Verma and Dhar 1976) 2.13 (53%) 6.33 (84%) —United States; (Garn et al. 1981b) — 1.55 (36%) for whites, 1.0 for blacks (lowest — Hb midpoint of 80 g/L compared with 110 g/L)Nigeria (Zaria); (Lister et al. 1985) 1.71 (42%) — —Papua New Guinea; (Brabin et al. — At booking: 5.91 (83%) for primiparas, 1.42 6.0 (OR) for primiparas early in 1990) (42%) for multiparas; at delivery: 2.38 pregnancy; not signiﬁcant (57%) for primiparas, 1.94 (48%) for for multiparas or iron multiparas deﬁciency late in pregnancyUnited States; (Scholl et al. 1992) — — 3.10 (adjusted OR)India (Pune); (Hirve and Ganatra — 1.53 (34.5%) — 1994)India (Varanasi); (Swain et al. 1994) 2.22 (55%) — —Italy; (Spinillo et al. 1994) — 5.05 (adjusted OR for SGA speciﬁcally) —Brazil; (Rondo et al. 1995) 0.76 — —England; (Steer et al. 1995) 0.76 (lowest Hb Յ 105 g/L 2.44 [lowest Hb Յ 85 g/L compared with — Downloaded from jn.nutrition.org by guest on February 21, 2012 compared with 106–125 Hb 96–105 g/L (lowest LBW rate)] (59%) g/L)Ghana; (Onadeko et al. 1996) 1.07 (6.3%) — —Papua New Guinea; (Brabin and Goroka (non-malarious): Goroka (nonmalarious): 1.65 primiparas, 5.0 — Piper 1997) 1.35 primiparas, 1.15 multiparas; Madang (malarious), 1.59 multiparas; Madang primiparas, 1.74 multiparas (malarious): 1.54 primiparas, 1.14 multiparasChina; (Zhou et al. 1998) 2.96 (but only 0.99 for — — SGA) 1 Relative risk calculated as LBW rate in anemic women/LBW rate in nonanemic women; attributable risk: (LBW rate in anemic women Ϫ LBWrate in nonanemic women)/LBW rate in anemic women). 2 Abbreviations used: Hb, hemoglobin; LBW, low birth weight; OR, odds ratio; SGA, small-for-gestational age.(hemoglobin values of 110 –119 g/L at 4 – 8 wk of pregnancy) eliminated any alternative explanations for this association,(Zhou et al. 1998) and also in a very large cohort of American which is an important failing because confounding might bewomen (Scanlon et al. 2000). expected. Data from the National Collaborative Perinatal Project The relative risk of delivering a LBW baby when theshowed that fetal death was minimal at maternal hemoglobin mother has moderate or severe anemia or iron-deﬁciency ane-values of 95–105 g/L for Caucasians and 85–95 g/L for African- mia is provided in a few of the studies reviewed and wasAmericans (Garn et al. 1981a). Perinatal mortality rate was calculated, where possible, from data included in others (Tableminimal at maternal hemoglobin values of 104 –132 g/L in the 2). It is difﬁcult to compare these results across studies becausedata from the Cardiff Births Survey (Murphy et al. 1986). As the reference group was deﬁned in various ways. Comparedwas the case for birth weight and duration of gestation, some with no or mild anemia, moderate anemia had a relative riskof these values are below the current cut-off value for anemia. of LBW of 0.76 –2.96 and severe anemia had a relative risk of An association between maternal hemoglobin concentra- LBW of 1– 6.33 in the studies reviewed. Only two studies weretion and birth weight was most likely to be detected in studies, identiﬁed in which the authors considered iron-deﬁciencyusually with a small sample size, that were conducted in anemia speciﬁcally. In the United States, the adjusted oddspopulations with lower maternal hemoglobin concentrations ratio for LBW was 3.10 (Scholl et al. 1992). In Papua Newand lower birth weight. Even when birth weights were higher, Guinea, the odds ratio for LBW was 6.0 for primiparas whena speciﬁc association between iron-deﬁciency anemia (i.e., low iron-deﬁciency anemia was recorded early in pregnancy; therehemoglobin combined with low serum ferritin concentration) was no excess probability for multiparas or for anemia late inand birth weight, preterm birth or both could be detected pregnancy (Brabin et al. 1990). These data also were used to(Nemet et al. 1986, Scholl et al. 1992, Singla et al. 1997). ´ calculate the attributable risk (Table 2). For moderate anemia, The effect of the severity of anemia on birth outcomes the attributable risk was 42–55%; for severe anemia, it wascould be examined only in studies that did not eliminate 34.5– 83%. One group calculated the proportion of LBW thatwomen with severe anemia (usually deﬁned as hemoglobin could be attributed to maternal anemia (the population-attrib-values Ͻ80 g/L). These studies (Bhargava et al. 1989, Duthie utable risk). With data from Papua New Guinea, Brabin andet al. 1991, Msolla and Kinabo 1997, Singla et al. 1997, Verma Piper (1997) calculated that, if the relationship were causal,and Dhar 1976) all report either a strong statistical association severe (Ͻ70 g/L) maternal anemia was responsible for Ͻ10%between the lowest maternal hemoglobin values and low birth of the LBW; in comparison, malaria was responsible for 40% ofweight or a difference between 200 and 400 g in birth weight the LBW.between women with hemoglobin values Ͻ80 g/L and those The relative risk of delivering a preterm baby when thewith higher values (Ͼ100 g/L). None of these investigations mother has moderate or severe anemia or iron-deﬁciency ane-
598S SUPPLEMENT TABLE 3 Relative and attributable risk of preterm birth according to severity and type of maternal anemia during pregnancy1,2 Relative risk (attributable risk) of preterm birth compared to mild or no anemia Severe anemia (usually Hb Յ Iron deﬁciencyStudy site; authors Moderate anemia 80 g/L) anemiaUnited States (various locations); (Garn — 1.43 (30%) for whites, 1.10 — et al. 1981b) (9%) for blacks (lowest Hb midpoint of 80 g/L compared with 110 g/L)Germany; (Goepel et al. 1988) 1.30 (23%) — —United States (Boston); (Lieberman et 1.90 (47%) (hematocrit Յ — — al. 1988) 34% compared to all higher values)United States (various locations); 0.6–1.6 for black women and — — (Klebanoff et al. 1989) 0.7–2.1 for white women depending on the duration of pregnancy (higher earlier and lower later in pregnancy)Papua New Guinea; (Brabin et al. — “Not signiﬁcantly increased” (at — Downloaded from jn.nutrition.org by guest on February 21, 2012 1990) booking: 1.89 for primiparas and 0.55 for multiparas; at delivery: 1.08 for primiparas and 0.43 for multiparas) (Hb Ͻ80 g/L compared to all higher values)Japan; (Fukushima and Wantabe 1991) 3.19 (67%) — —United States; (Scholl et al. 1992) — — 2.66 (adjusted OR)England; (Steer et al. 1995) 0.84 (lowest Hb Յ 105 g/L 2.46 [lowest Hb Յ 85 g/L — compared with 106–125 compared with Hb 96–105 g/L) g/L (lowest LBW rate)] (59%)Wales (Cardiff Births Survey); (Meis et 1.23 (adjusted OR of Hb — — al. 1995) Ͻ104 g/L compared with Hb 118–132 g/L)China; (Zhou et al. 1998) 2.07 — —Egypt; (Afrafa et al. 1998) 2.63 (adjusted OR) in early 4.01 (adjusted OR) in early — pregnancy, 2.03 (adjusted pregnancy (Ͻ90 g/L) OR) in the 3rd trimesterPapua New Guinea; (Allen et al. 1998) 0.64 (OR) — —United States (Los Angeles); (Siega-Riz 1.83 (adjusted OR for anemia — — et al. 1998) at 28–32 wk gestational age) 1 Relative risk calculated as preterm rate in anemic women/preterm rate in nonanemic women; attributable risk: (preterm rate in anemic womenϪ preterm rate in nonanemic women)/preterm rate in anemic women). 2 Abbreviations used: Hb, hemoglobin; OR, odds ratio.mia is also provided in a few of the studies reviewed and was than there are observational studies that report associationscalculated from data included in others (Table 3). These data between these factors. Although for the most part, the trialshave the same limitations as described above for LBW. On the listed in Table 1 were randomized and blind, often they didwhole, the relative risks of preterm birth were lower than those not meet the other criteria described above for demonstratingfor LBW. Compared with no or mild anemia, moderate anemia an effect of iron or folic acid supplementation on birth weighthad a relative risk of preterm birth of 0.6 –3.2 and severe or duration of gestation. The possibility of false-negative re-anemia had a relative risk of preterm of 0.55– 4.01 in the sults was particularly high because most studies were con-studies reviewed. The adjusted odds ratio of preterm birth was ducted in populations with adequate values for initial hemo-2.66 for iron-deﬁciency anemia in the one study in which this globin and birth weight. Therefore, these subjects had littlewas determined (Scholl et al. 1992). These data also were used potential to respond to supplementation with increases into calculate the attributable risk (Table 3). For moderate birth weight or duration of gestation. Paintin and coworkersanemia, the attributable risk was 23– 67%; for severe anemia it (1966) even commented that “the range of hemoglobin con-was 9 –30%. centrations at the 20th week was mainly due to factors other than iron deﬁciency.” Some of these studies provided infor-Evidence that iron deﬁciency, iron-deﬁciency anemia, or mation on iron deﬁciency late in pregnancy. In general, feweranemia causes poor birth outcomes than half of the subjects were iron deﬁcient even if their hemoglobin concentrations had dropped into the anemic Controlled experiments are necessary to examine whether range by this time.there is a causal relationship between maternal anemia and The research trials in which women have been supple-poor birth outcomes; there are many fewer such experiments mented with iron or folic acid have been reviewed several
IS THERE A CAUSAL RELATIONSHIP? 599Stimes in recent years (Mahomed 1998a and 1998b, Scholl and In summary, only one intervention trial was identiﬁed thatReilly 2000, U.S. Preventive Services Task Force 1993). The was without major design defects and provided evidence of aU.S. Task Force review concluded: “Although iron supple- statistically signiﬁcant positive effect of iron supplementationmentation can improve maternal hematologic indexes, con- on birth weight, and that evidence was provided only for atrolled clinical trials . . . have failed to demonstrate that iron subgroup of the subjects. No such positive ﬁndings were iden-supplementation or changes in hematologic indexes actually tiﬁed in trials conducted in nonanemic populations. Impor-improve clinical outcomes for the mother or newborn.” The tantly, no intervention trials were identiﬁed that providedresults of the two recent Cochrane Reviews were similar. For evidence of a negative effect of iron supplementation on birthiron supplementation, the author said: “. . . There is very little weight.information regarding the effect if any on any substantivemeasures of either maternal or fetal outcome . . .” (Mahomed Summary and conclusions1998b). For folate supplementation, the other major nutri-tional cause of anemia during pregnancy, the author said: In populations in which the rate of iron or folate deﬁciency“. . . No advantage of routine folate supplementation was de- is low among nonpregnant women, the primary cause of ane-tected in terms of . . . preterm delivery. There is a nonsigniﬁ- mia during pregnancy is likely to be plasma volume expansion,cant reduction in the incidence of low birth weight associated and this anemia is not associated with negative birth out-with folate supplementation” (Mahomed 1998a). No addi- comes.tional studies were identiﬁed for the present review that would Maternal hemoglobin values during pregnancy are associ-change these conclusions. ated with birth weight and preterm birth in a U-shaped rela- However, caution is warranted in interpreting these results tionship with high rates of babies who are small, early or both,because, relative to ascertaining an effect on birth outcomes, at low and high concentrations of maternal hemoglobin. How- Downloaded from jn.nutrition.org by guest on February 21, 2012the design problems characteristic of these studies tend to bias ever, some of this association may result from using “lowestthem toward null ﬁndings. Furthermore, these null ﬁndings hemoglobin” rather than a hemoglobin value controlled forcontrast strongly with the expectation of a causal relationship, the stage of pregnancy. A similar U-shaped relationship isalbeit a complicated one, derived from the large body of likely to be present between maternal hemoglobin concentra-observational data on this subject. Although the 23 studies tion and neonatal or perinatal mortality, but the data tolisted in Table 1 include many that are randomized, placebo establish this association remain insufﬁcient.controlled, and double blind, none was free of possible bias. The relative risk of LBW that results from moderate orSome trials had multiple problems with design and interpre- severe anemia is inconsistent; nonetheless, it is generallytation. Among these 23 intervention trials, there was 1 with higher than the also inconsistent relative risk of preterm birthfalse-positive bias, 19 with false-negative bias and 6 with that results from these conditions.possible bias of unknown direction; confounding was a prob- Severe maternal anemia (Ͻ80 g/L) is associated with birthlem in 3 studies, and 1 had insufﬁcient information to evaluate weight values that are 200 – 400 g lower than in women withthe possibility of bias and confounding. higher (Ͼ100 g/L) hemoglobin values, but researchers gener- It is perhaps instructive to examine in more detail those few ally have not excluded other factors that might also haveexperimental studies that were conducted in populations in contributed to both LBW and the severity of the anemia.which anemia was common and iron deﬁciency was a likely Supplementation of anemic or nonanemic pregnant womencause of this anemia (Agarwal et al. 1991, Menendez et al. with iron, folic acid or both does not appear to increase birth1994, Preziosi et al. 1997, Sood et al. 1975, Srinivasan et al. weight or the duration of gestation, but the intervention trials1995). There was a wide range in the size of the effect of iron on which this conclusion is based generally suffered fromsupplementation on birth weight reported in these investiga- design problems that would tend to produce false-negativetions, i.e., from 0 to 290 g. The study with the largest effect ﬁndings.(Agarwal et al. 1991) was the only one reviewed with the In a number of studies, maximal values for birth weight andpossibility of false-positive ﬁndings. In addition, the results of minimal values for preterm birth occurred at maternal hemo-this study may be biased because information on birth weight globin values (all uncontrolled for the stage of gestation)was available only for a limited number of the subjects. The below current cut-off values for anemia during pregnancy.observed effect (71 g, nonsigniﬁcant) may have been under-estimated in an older study (Sood et al. 1975) in which the Implications for researchdose of iron given also was insufﬁcient to cure the subjects’anemia. However, bias is also a possibility in this investigation Effort should be directed toward using the available obser-because such a small proportion of the subjects provided data vational data to estimate the risk of LBW and preterm birthon birth weight. In a study with a superior design (Menendez that is attributable to iron-deﬁciency anemia as distinct fromet al. 1994), the overall effect (56 g) was not statistically anemia from other causes. This requires studies in which ironsigniﬁcant, but the effect of iron supplementation on birth deﬁciency was ascertained by some method in addition toweight in a subgroup of women who took more of the iron pills maternal hemoglobin concentration. Because data in many ofwas greater (96 g) and statistically signiﬁcant. This ﬁnding and the published papers were not presented in a way that wouldthe fact that supplementation did not correct the subjects’ permit this calculation to be made, access to the original data,anemia suggest that the overall effect on birth weight may which was not possible for the present review, will be requiredhave been underestimated. The remaining studies showed no to estimate this risk.difference in birth weight between the treatment groups and Priority should be given to conducting studies of iron andsuffered from low statistical power (Srinivasan et al. 1995) and folate supplementation during pregnancy that meet the crite-failure to eliminate anemia (Preziosi et al. 1997), both causes ria for demonstrating a positive effect of supplementation onof false-negative results. These results suggest that adequate birth outcomes, should such an effect exist. In particular, thisiron supplementation could increase birth weight by 100 g at means studying a population in which the mean birth weightthe most, an effect that would not be inconsequential if it is Ͻ3.3 kg, treating all women to eliminate other causes ofcould be substantiated. LBW or preterm birth, selecting women with iron-deﬁciency
600S SUPPLEMENTanemia for iron supplementation (or folate deﬁciency for folic the value of serum ferritin during pregnancy. Gynecol. Obstet. Investig. 26: 265–273.acid supplementation), and including a sufﬁcient number of Hemminki, E. & Rimpela, U. (1991) A randomized comparison of routine versus ¨subjects for adequate statistical power. selective iron supplementation during pregnancy. J. Am. Coll. Nutr. 10: 3–10. Hirve, S. S. & Ganatra, B. R. (1994) Determinants of low birth weight: a community based prospective cohort study. Indian Pediatr. 31: 1221–1225.Implications for public health Hytten, F. & Leitch, I. (1971) The Physiology of Human Pregnancy, vol. 2. Blackwell Scientiﬁc Publications, Oxford, UK. Consideration should be given to lowering the hemoglobin Iyengar, L. (1971) Folic acid requirements of Indian pregnant women. Am. J. Obstet. Gynecol. 111: 13–16.cut-off value for anemia during pregnancy because optimal Iyengar, L. & Rajalakshmi, K. (1975) Effect of folic acid supplement on birthbirth outcomes may be achieved at hemoglobin values in the weights of infants. Am. J. Obstet. Gynecol. 122: 332–336.range currently designated as anemic. Klebanoff, M. A., Shiono, P. H., Berendes, H. W. & Rhoads, G. G. (1989) Facts Although there may be other reasons to offer women sup- and artifacts about anemia and preterm delivery. J. Am. Med. Assoc. 262: 511–515.plemental iron during pregnancy, the currently available evi- Lieberman, E., Ryan, K. J., Monson, R. R. & Schoenbaum, S. C. (1988) As-dence from studies with designs appropriate to establish a sociation of maternal hematocrit with premature labor. Am. J. Obstet. Gy-causal relationship is insufﬁcient to support or reject this necol. 159: 107–114. Lister, U. G., Rossiter, C. E. & Chong, H. (1985) 11. Perinatal mortality. Br. J.practice for the speciﬁc purposes of raising birth weight or Obstet. Gynæcol. (suppl.) 5: 86 –99.lowering the rate of preterm birth. Mahomed K. (1998a) Routine folate supplementation in pregnancy (Cochrane Review). In: The Cochrane Library Update Software, Oxford, UK. Mahomed K. (1998b) Routine iron supplementation during pregnancy (Co- ACKNOWLEDGMENTS chrane Review). In: The Cochrane Review Update Software, Oxford, UK. Meis, P. J., Michielutte, R., Peters, T. J., Wells, H. B., Sands, R. E., Coles, E. C. The author thanks Jean Pierre Habicht and, especially, Mary E. & Johns, K. A. (1995) Factors associated with preterm birth in Cardiff,Cogswell, for their thoughtful and helpful comments on this paper Wales I. Univariable and multivariable analysis. Am. J. Obstet. Gynecol. 173: Downloaded from jn.nutrition.org by guest on February 21, 2012before, during and after its presentation. A number of their ideas are 590 –596. Menendez, C., Todd, J., Alonso, P. L., Francis, N., Lulat, S., Ceesay, S., M’Boge,included here. In addition, the constructive criticism provided by B. & Greenwood, B. M. (1994) The effects of iron supplementation duringLaurence Grummer-Strawn is gratefully acknowledged. pregnancy, given by traditional birth attendants, on the prevalence of anaemia and malaria. Trans. R. Soc. Trop. Med. Hyg. 88: 590 –593. Milman, N., Agger, A. O. & Nielsen, O. J. (1994) Iron status markers and serum LITERATURE CITED erythropoietin in 120 mothers and newborn infants: effect of iron supplemen- tation in normal pregnancy. Acta Obstet. Gynecol. Scand. 73: 200 –204.Afrafa, M., Abou-Zied, H., Attia, A. F. & Youssof, M. (1998) Maternal hemo- Msolla, M. J. & Kinabo, J. L. (1997) Prevalence of anaemia in pregnant women globin and premature child delivery. Rev. Sante Mediterr. Orient. 4: 480 – 486. ´ ´ during the last trimester. Int. J. Food Sci. Nutr. 48: 265–270.Agarwal, K. N., Agarwal, D. K. & Mishra, K. P. (1991) Impact of anaemia Murphy, J. F., O’Riordan, J., Newcombe, R. G., Coles, E. C. & Pearson, J. F. prophylaxis in pregnancy on maternal haemoglobin, serum ferritin & birth (1986) Relation of haemoglobin levels in ﬁrst and second trimesters to weight. Indian J. Med. Res. 94: 277–280. outcome of pregnancy. Lancet i: 992–994.Allen, S. J., Raiko, A., O’Donnell, A., Alexander, N.D.E. & Clegg, J. B. (1998) Nemet, K., Andrassy, K., Bognar, K., Czappan, P., Stuber, A. & Simonovits, I. ´ ´ ´ ´ Causes of preterm delivery and intrauterine growth retardation in a malaria (1986) Relationship between maternal and infant iron stores: 1. Full term endemic region of Papua New Guinea. Arch. Dis. Child. 79: F135–F140. infants. Haematologia 19: 197–205.Baumslag, N., Edelstein, T. & Metz, J. (1970) Reduction of incidence of Onadeko, M. O., Avokey, F. & Lawoyin, T. O. (1996) Observations on stillbirths, prematurity by folic acid supplementation in pregnancy. Br. Med. J. 1: 16 –17. birthweight and maternal haemoglobin in teenage pregnancy in Ibadan, Ni-Bhargava, M., Kuman, R., Iyer, P. U., Ramji, S., Kapani, V. & Rhargava, S. K. geria. Afr. J. Med. Sci. 25: 81– 86. (1989) Effect of maternal anaemia and iron depletion on foetal iron stores, Paintin, D. B., Thomson, A. M. & Hytten, F. E. (1966) Iron and the haemoglobin birthweight and gestation. Acta Pædiatr. Scand. 78: 321–322. level in pregnancy. J. Obstet. Gynaecol. Br. Commonw. 73: 181–190.Brabin, B., Ginny, M., Sapau, J., Galme, K. & Paino, J. (1990) Consequences Preziosi, P., Prual, A., Galan, P., Daouda, H., Boureima, H. & Hercberg, S. (1997) of maternal anaemia on outcome of pregnancy in a malaria endemic area in Effect of iron supplementation on the iron status of pregnant women: conse- Papua New Guinea. Ann. Trop. Med. Parasitol. 84: 11–24. quences for newborns. Am. J. Clin. Nutr. 66: 1178 –1182.Brabin, B. & Piper, C. (1997) Anaemia- and malaria-attributable low birth- Rae, P. G. & Robb, P. M. (1970) Megaloblastic anemia of pregnancy: a clinical weight in two populations in Papua New Guinea. Ann. Hum. Biol. 24: 547– and laboratory study with particular reference to the total and labile serum 555. folate levels. J. Clin. Pathol. 23: 379 –391.de Benaze, C., Galan, P., Wainer, R. & Hercberg, S. (1989) Prevention de ´ Romslo, I., Haram, K., Sagen, N. & Augensen, K. (1983) Iron requirement in l’anemie ferripreive au cours de la grossesse par une supplementation mar- ´ ´ normal pregnancy as assessed by serum ferritin, serum transferrin saturation tiale precoce: un essai controle. Rev. Epidemiol. Sante Publique 37: 109 –118. ´ ˆ ´ and erythrocyte protoporphyrin determinations. Br. J. Obstet. Gynæcol. 90:Duffus, G. M., MacGillivray, I. & Dennis, K. J. (1971) The relationship between 101–107. baby weight and changes in maternal weight, total body water, plasma Rondo, P.H.C., Abbott, R., Rodrigues, L. C. & Tompkins, A. M. (1995) Vitamin volume, electrolytes and proteins and urinary oestriol excretion. J. Obstet. A, folate, and iron concentrations in cord and maternal blood of intra-uterine Gynaecol. Br. Commonw. 78: 97–104. growth retarded and appropriate birth weight babies. Eur. J. Clin. Nutr. 49:Duthie, S. J., King, P. A., To, W. K., Lopes, A. & Ma, H. K. (1991) A case controlled study of pregnancy complicated by severe maternal anemia. Aust. 391–399. N. Z. J. Obstet. Gynaecol. 31: 125–127. Scanlon, K. S., Yip, R., Schieve, L. A. & Cogswell, M. E. (2000) High and lowFleming, A. F., Chatoura, G.B.S., Harrison, K. A., Briggs, N. D. & Dunn, D. T. hemoglobin levels during pregnancy: differential risks for preterm birth and (1986) The prevention of anaemia in pregnancy in primigravidae in the small for gestational age. Obstet. Gynecol. 96: 741–748. guinea savanna of Nigeria. Ann. Trop. Med. Parasitol. 80: 211–233. Scholl, T. O., Hediger, M. L., Fischer, R. L. & Shearer, J. W. (1992) Anemia vsFleming, A. F., Henrickse, J. P. d. V. & Allan, N. C. (1968) The prevention of iron deﬁciency: increased risk of preterm delivery in a prospective study. megaloblastic anaemia in pregnancy in Nigeria. J. Obstet. Gynaecol. Br. Am. J. Clin. Nutr. 55: 985–988. Commonw. 75: 425– 432. Scholl, T. O. & Reilly, T. (2000) Anemia, iron and pregnancy outcome. J. Nutr.Fleming, A. F., Martin, J. D., Hahnel, R. & Westlake, A. J. (1974) Effects of iron 130: 443S– 447S. and folic acid antenatal supplements on maternal haematology and fetal Siega-Riz, A. M., Adair, L. S. & Hobel, C. J. (1998) Maternal hematologic wellbeing. Med. J. Aust. 2: 429 – 436. changes during pregnancy and the effect of iron status on preterm delivery inFletcher, J., Gurr, A., Fellingham, F. R., Prankerd, T.A.J., Brant, H. A. & Menzies, a West Los Angeles population. Am. J. Perinatol. 15: 515–522. D. N. (1971) The value of folic acid supplements in pregnancy. J. Obstet. Singla, P. N., Tyagi, M., Kumar, A., Dash, D. & Shankar, R. (1997) Fetal growth Gynaecol. Br. Commonw. 78: 781–785. in maternal anemia. J. Trop. Pediatr. 43: 89 –92.Fukushima, M. & Wantabe, H. (1991) An observation on pregnancy outcomes Sood, S. K., Ramachandran, K., Mathur, M., Gupta, K., Ramalingaswamy, V., in relation to haemoglobin levels. Fukushima J. Med. Sci. 37: 23–27. Swarnabai, C., Ponniah, J., Mathan, V. I. & Baker, S. J. (1975) WHOGarn, S. M., Keating, M. T. & Falkner, F. (1981a) Hematologic status and sponsored collaborative studies on nutritional anaemia in India., 1. The effect pregnancy outcomes. Am. J. Clin. Nutr. 34: 115–117. of supplemental oral iron administration to pregnant women. Q. J. Med. 174:Garn, S. M., Ridella, S. A., Petzold, A. S. & Falkner, F. (1981b) Maternal 251–258. hematologic levels and pregnancy outcomes. Semin. Perinatol. 5: 155–162. Spinillo, A., Capuzzo, E., Piazzi, G., Nicola, S., Colonna, L. & Iasci, A. (1994)Gibson, H. M. (1973) Plasma volume and glomerular ﬁltration rate in preg- Maternal high-risk factors and severity of growth deﬁcit in small for gesta- nancy and their relation to differences in fetal growth. J. Obstet. Gynaecol. Br. tional age infants. Early Hum. Dev. 38: 35– 43. Commonw. 80: 1067–1074. Srinivasan, V., Radhakrishna, S., Sudha, R., Malathi, M. V., Jabbar, S., Ra-Goepel, E., Ulmer, H. U. & Neth, R. D. (1988) Premature labor contractions and makrishnan, R. & Venkata Rao, T. (1995) Randomized controlled ﬁeld trial
IS THERE A CAUSAL RELATIONSHIP? 601S of two antenatal care packages in rural south India. Indian J. Med. Res. 102: stronger than in white women. These data do not support the 86 –94.Steer, P., Alam, M. A., Wadsworth, J. & Welch, A. (1995) Relation between use of different hemoglobin cutoffs by ethnic group. maternal haemoglobin concentration and birth weight in different ethnic Finally, I disagree that the currently available evidence does groups. Br. Med. J. 310: 489 – 491. not support the practice of offering women supplemental ironSwain, S., Singh, S., Bhatia, B. D., Pandley, S. & Krishna, M. (1994) Maternal hemoglobin and serum albumin and fetal growth. Indian Pediatr. 31: 777–782. during pregnancy. Observational studies are biased by the lackTaylor, D. J., Mallen, C., McDougall, N. & Lind, T. (1982) Effect of iron of ability to control for unknown factors related to iron deﬁ- supplementation on serum ferritin levels during and after pregnancy. Br. J. ciency and birth outcomes. However, after controlling for Obstet. Gynæcol. 89: 1017. known factors that would inﬂuence this association, severalTchernia, G., Blot, I., Rey, A. & Papiernik, E. (1983) Carences maternelles en fer et folates: repercussions sur le nouveau-ne. Sem. Hop. Paris 59: 416 – 420. ´ ´ ˆ observational studies show a strong association between lowThomses, J. K., Prien-Larsen, J. C., Devandier, A. & Fogh-Andersen, N. (1993) hemoglobin, and in one study, iron-deﬁciency anemia, and Low dose iron supplementation does not cover the need for iron during adverse birth outcomes. As Rasmussen pointed out, the results pregnancy. Acta Obstet. Gynecol. Scand. 72: 93–98.Trigg, K. H., Rendall, E. J. C., Johnson, A., Fellingham, F. R. & Prankerd, T.A.J. of the intervention trials to date were biased toward false- (1976) Folate supplements during pregnancy. J. R. Coll. Gen. Pract. 26: negative ﬁndings. These biases include small sample sizes; the 228 –230. inability of the population to respond because of inadequateU.S. Preventive Services Task Force (1993) Routine iron supplementation during pregnancy: policy statement. J. Am. Med. Assoc. 270: 2846 –2854. duration, dose, or late start of iron supplementation; or a smallVerma, K. C. & Dhar, G. (1976) Relationship of maternal anaemia, birth weight proportion of women with iron deﬁciency. Results from poorly and perinatal mortality in low birth weight neonates: (a hospital study). Indian designed intervention trials do not outweigh the evidence Pediatr. 13: 439 – 441.Whittaker, P. G., Macphail, S. & Lind, T. (1996) Serial hematologic changes from well-designed observational studies. Until well-designed and pregnancy outcome. Obstet. Gynecol. 88: 33–39. intervention trials give evidence that it is not beneﬁcial, theYip, R. (2000) Signiﬁcance of an abnormally low or high hemoglobin concen- practice of iron supplementation during pregnancy is war- tration during pregnancy: special consideration of iron nutrition. Am. J. Clin. Nutr. 72 (suppl): 272S–279S. ranted by the strong association between anemia and adverse Downloaded from jn.nutrition.org by guest on February 21, 2012Zhou, L.-M., Yang, W.-W., Hua, J.-Z., Deng, C.-Q., Tao, X. & Stoltzfus, R. J. birth outcomes. (1998) Relation of hemoglobin measured at different times in pregnancy to Dr. Haas: About the U-shape relationship that you are preterm birth and low birth weight in Shanghai, China. Am. J. Epidemiol. 148: ﬁnding with hemoglobin and either intrauterine growth retar- 998-1006.Zittoun, J., Blot, I., Hill, C., Papiernik, E. & Tchernia, G. (1983) Iron supple- dation or preterm, you have identiﬁed what appear to be two ments versus placebo during pregnancy: its effects on iron and folate status curves that were superimposed to create one curve. One curve on mothers and newborns. Ann. Nutr. Metab. 27: 320 –327. that might be related to iron deﬁciency or all the pathology associated with that—which includes anemia—shows the high DISCUSSION risk at low values; the other curve that is superimposed shows that as you decrease plasma volume expansion you have an Participants: Cogswell, Sazawal, Haas, Rasmussen, Beard, increase in hemoglobin and also get an increase in pathology.Habicht, Lynch, Stoltzfus, Schultink, Tielsch, Allen, Horton, Has anybody tried to look at the two curves separately and sayLozoff what is happening with the relationship between hemoglobin Dr. Cogswell: I have four points. First, I agree with Ras- and these outcomes when you have eliminated the plasmamussen that there is an association between hemoglobin levels volume problem or when you look at the plasma volume, whenand birth weight and preterm delivery, but the U-shaped you have eliminated hemoglobin problems? The nadir forrelationship between hemoglobin and low birth weight is due hemoglobin when there is just anemia may hit near the cutoffto two separate associations between low hemoglobin and that we have been using all along but may be obscured by thepreterm delivery and high hemoglobin and small-for-gesta- pathology associated with plasma volume at the higher end oftional age. In 173,000 pregnant women who attended publicly the hemoglobin distribution.funded health programs in 10 states, we found that the high Dr. Rasmussen: I agree with you, except that the numberhemoglobin during the ﬁrst and second trimester was not of plasma volume estimates we have in the literature is veryassociated with preterm birth but low hemoglobin was. On the small and most are from healthy Scottish women. So, we areother hand, we found that very high hemoglobin, that is, not going to be able to use those to answer the question thatϾ140 g/L, was associated with small-for-gestational age deliv- you had in mind. It certainly is something that is worth doing.ery. An elevated hemoglobin level is an indicator of possible Dr. Beard: Maybe another variation on that is to askpregnancy complications associated with poor plasma volume whether you can drive hemoglobin values up in the secondexpansion and should not be mistaken for good iron status. and third trimester with iron supplementation? Second, I disagree that the lowest proportion of low birth Dr. Rasmussen: Yes. Yes, clearly.weight occurs at maternal hemoglobin values below the cur- Dr. Beard: Can you give iron supplements to subjects whorent cutoffs for anemia. The use of lowest hemoglobin value in are following the normal dilution patterns of hemoglobin andseveral large studies biases the relationship between hemoglo- drive hemoglobin up into the pathological range?bin and birth outcomes. As shown in a study by Zhou and Dr. Rasmussen: You can drive the hemoglobin up. Thecolleagues, using the lowest value of hemoglobin artiﬁcially pathological range is open to question. You would have toshifts the relationship between hemoglobin and low birth deﬁne the pathological range quite a bit better.weight towards a lower distribution of hemoglobin. When Dr. Habicht: It is true that you can drive hemoglobin up byrandom hemoglobin values are used and stratiﬁed by trimester, iron overload—if you raise saturation levels. If you look at theas in a few recent studies, the lowest proportion of low birth saturation levels, you see that hemoglobin goes up veryweight is found among women with hemoglobin values above slightly, but that is in nonpregnant women. You can drive it upthe current cutoffs for anemia. somewhat, but nowhere near to these levels. If we extrapolate Third, few studies have contrasted the associations between from that ﬁnding to pregnant women, the answer is, no, youlow hemoglobin and preterm delivery in black and white cannot drive it up that high. It is almost certain, as far as I canwomen. In our data we found similar associations between low see, that those high levels are not due to iron overload. Thosehemoglobin and preterm delivery in black and white women. levels are due to inadequate plasma expansion. In which case,If anything, the odds for preterm birth in black women with it is irrelevant relative to recommendations about iron or formoderate to severe anemia during the second trimester was trying to estimate what we are after.
602S SUPPLEMENT Dr. Beard: That is what I was trying to get at—whether the difﬁcult. We have seen the long list of issues that inﬂuence lowright half of that pregnancy hemoglobin distribution is iron birth weight and birth outcome. There is no way that you canresponsive. If it is not an iron-responsive portion, then we are really expect to get a universal answer where you do a study inlooking at a different pathology from what we are looking at in Bangladesh or somewhere in Africa or some other place andthe left half. you give one group no iron and the other group iron. There is Dr. Lynch: Is there any evidence that inducing iron deﬁ- no way you could translate the outcome of one country to theciency is an effective way of treating inadequate plasma vol- other country because all the different factors inﬂuencing lowume expansion? birth weight vary enormously between regions. So, I am really Dr. Sazawal: I was not sure whether the negative results wondering— do we need to do this? I would not be able tothat we are seeing in pregnancy trials is related to the lack of justify this from a programmatic point of view.sample size. From the trials that were presented, three trials Dr. Tielsch: To turn the question around, there is serioushad a positive point estimate and three trials had a negative doubt about whether the programs are justiﬁed and it clearlypoint estimate, which suggests that if you were doing random makes a difference. You would think programmatically veryeffects analysis, you would end up with null: unrelated to the differently if 60% of the population of pregnant women needsample size issue. I thought that actually the data are incon- to be supplemented vs. 8% of the population. So, if there isclusive and there is a need for more studies. Whether those little evidence— or certainly uncompelling evidence—forcan be done is another question. women with mild-to-moderate anemia measured at some ap- Dr. Stoltzfus: Then how do we get the evidence we need? propriate time early in their pregnancy that supplementationDo we have to do smarter observational research or do we have does not affect reproductive outcomes, then why we are ship-to make a strong statement that randomized trials are needed ping containers full of iron supplements?and that it is ethical to do that in certain circumstances? Dr. Allen: From a public health point of view, pregnancy Downloaded from jn.nutrition.org by guest on February 21, 2012 I want to offer two ideas. One thing that has intrigued me is a window of opportunity when you have a woman comingis why more people are not looking at erythrocyte protopor- for care. If you can get her to take iron supplements, there isphyrin in pregnancy as an indicator of iron deﬁciency. It seems not much doubt that this improves iron stores postpartum.ideal because it is very physiologically deﬁned as opposed to There is also not much doubt in my mind that it improvessome other measures. It is also independent of plasma volume infant iron status postpartum.expansion. One way to do smarter observational studies would Folate is a big confounder in these studies. If one nutrientbe to do some of the same things that we have already been will reduce preterm delivery, I am quite convinced that it isdoing but not use hemoglobin as the sole risk factor. We folate, working through different mechanisms. You have tocannot interpret it very well. Make the primary potential risk remove the effects of folate if you are going to look at thefactor erythrocyte protoporphyrin. effects or iron supplements. Another idea is the timing, because if you go back to Dr. Habicht: Two points. The ﬁrst one comes from Allen’s.Allen’s paper, I was impressed by some of the outcomes being If you are going to give iron, you are always going to givelinked to things that are happening at 16 –20 wk gestation. Is folate. So, from a purely public health point of view, I actuallythis working through development of the placenta? We know would prefer to see an iron-folate study than an iron study. Itthat the placenta is changed in anemia. The placenta develops will not satisfy our intellectual curiosity relative to iron, but Iearlier than the fetus and if we want to change placental would prefer to see a package that makes some sense.development as the route to changing fetal development, we The other thing that bothered me is that Rasmussen ex-have to get in there faster because the placenta is growing cluded all studies that were not randomized intervention stud-rapidly in the ﬁrst half of pregnancy. The fetus grows rapidly in ies. It is so nice and neat to say there are the randomized trialsthe second half of pregnancy. So, the timing issue may be very here, and all the other goats are here. It seems to me thatimportant, and most of our data are coming from the second actually those goats all are not the same. We need to think ahalf of pregnancy and most of our supplementation trials get little bit more carefully how we think about looking at trialsstarted in the second half of pregnancy. where there is some greater plausibility and trials where there Dr. Lynch: There may be advantage in using transferrin is much less. I have actually made a claim that for programreceptor as well. Theoretically, the transferrin receptor might evaluation, you basically depend upon plausibility most of theactually be more attractive. The problem with the ratio of time. You cannot do it through probability trials.erythrocyte protoporphyrin to heme is that it depends on the Dr. Brabin: We have been looking at a prospective cohorttime when the cell was made. So, it does persist beyond the study of pregnant women and, ﬁrst, half of the babies borntime of the iron deﬁciency, whereas the transferrin receptor is have hemoglobins Ͻ 125 g/L—and normal is ϳ165 g/L. So,going to be more sensitive to rapid changes. It is too early to one third of their hemoglobin mass is missing. Second, thereknow whether that is true, but it might be. Certainly, trans- was a highly signiﬁcant association between the seasonal pat-ferrin receptor does increase in pregnancy, but it does seem to tern of iron deﬁciency and the pattern of future anemia inbe sensitive to iron deﬁciency. infants, which is fairly suggestive that this is iron deﬁciency. Dr. Sazawal: Even if you ended up using some of these Third, the pattern of infant anemia is associated with the birthmeasures to do “smart observational studies,” you would be weight. Perhaps more importantly in terms of outcomes, aftersitting at this table 4 years down the line again advocating the 1st mo of life, infants were more at risk of dying if they hadneed for a clinical trial. Maybe we need smarter designs of low birth weight plus fetal anemia than low birth weightclinical trials. For example, you could look at different doses. alone. We have to think beyond birth weight. Despite theirYou do not have to have a placebo control but you can have limitations, observational studies can be very important.other control groups that are meaningful. Ultimately the issue Dr. Lynch: I was going to make the same point. It is awfullyis going to be resolved by good, well-done clinical trials, which important to look at the whole picture. One of the ﬁguresthis area does not have. Rasmussen showed includes the study by Preziosi et al., with Dr. Schultink: If we want to argue that we need to do very little effect on birth weight. Now, that study showed veryplacebo-controlled trials, meaning you give one group no iron clearly that children at 3 and 6 mo whose mothers receivedand you give the other group iron, this is going to be really placebo were much more iron deﬁcient. In fact, although not
IS THERE A CAUSAL RELATIONSHIP? 603Scommented on by the authors, the neonatal death rate was we want to have is some indicator of the infant’s status atmuch higher than in the supplemented individuals. If you put birth? There are not many studies that have that. Studies areit altogether, as you are pointing out, this is a major effect. focusing on birth weight, using a proxy that is not really very Dr. Sazawal: We do not realize when we discuss these good.issues as research priorities how they affect what happens in Dr. Lozoff: I do not think so. Some investigators havethe ﬁeld. Saying, well, this is a good time to get the woman shown cognitive differences across the entire birth weightand why not give her iron assumes unlimited resources. I was continuum up into the normal range. Now, people did not asksitting in the Ministry of Health with the UNICEF ofﬁcer and whether that is an iron effect or birth weight effect, but studiesdiscussed what can be done—what interventions you can do in are considering birth weight in relation to child developmentpregnancy. The Secretary of Health said that we do not have across the birth weight range, not just in this low end.enough money for iron. Give me some iron and forget about Dr. Horton: Birth weight is of interest in its own right?the rest. So, iron may be good, but it is an issue of what it Dr. Tielsch: Birth weight is the compelling reproductivedisplaces and what effect would be lost. outcome of interest because it has such strong association with Dr. Tielsch: This is why understanding the magnitude of both development and early mortality.the effect in solid, qualitative terms is absolutely critical. You Dr. Sazawal: It is the single strongest predictor in its ownhave got to provide program planners with some information right for survival, for anything you see. In fact, it is the greatestthat they can use to make rational decisions. Now, do they single predictor in any study we have done, including themake rational decisions all the time? Of course not. We all do effect of intensive feeding in the 1st y of life or the growth atnot make rational prioritization decisions all the time. At least 1 y.we have to give them some tools they can use to rationalize Dr. Horton: What if in addition to having birth weight,their resource allocations. you also have some information about iron status? Downloaded from jn.nutrition.org by guest on February 21, 2012 Dr. Lynch: That is particularly why you must look at the Dr. Tielsch: You are absolutely right. Not every interven-whole effect. tion that affects early infant mortality operates through birth Dr. Tielsch: Absolutely. You are absolutely correct. weight. There are lots of interventions that do not operate Dr. Horton: Are we using birth weights because we know through birth weight. Neonatal tetanus immunization, forthey are related to other things in infancy, when really what example, operates independently.