Zinc, Gravida, Infection, and Iron, but Not Vitamin B-12 or Folate Status, Predict Hemoglobin during Pregnancy in Southern Ethiopia^,

Abstract

The etiology of anemia during pregnancy in rural Southern Ethiopia is uncertain. Intakes of animal-source foods are low and infections and bacterial overgrowth probably coexist. We therefore measured the dietary intakes of a convenience sample of Sidama women in late pregnancy who consumed either maize (n = 68) or fermented enset (Enset ventricosum) (n = 31) as their major energy source. Blood samples were analyzed for a complete blood count, vitamin B-12 and folate status, plasma ferritin, retinol, zinc, albumin, and C-reactive protein (CRP). The role of infection and gravida was also examined. Dietary intakes were calculated from 1-d weighed records. No cellular animal products were consumed. Of the women, 29% had anemia, 13% had iron deficiency anemia, 33% had depleted iron stores, and 74 and 27% had low plasma zinc and retinol, respectively. Only 2% had low plasma folate (<6.8 nmol/L) and 23% had low plasma vitamin B-12 (<150 pmol/L), even though 62% had elevated plasma methylmalonic acid (MMA) (> 271 nmol/L). None had elevated plasma cystathionine or total homocysteine (tHcys). Women with enset-based diets had higher (P = 0.052) plasma vitamin B-12 concentration and lower (P < 0.05) cell volume, plasma cystathionine, and retinol than women consuming maize-based diets, but mean hemoglobin, plasma ferritin, MMA, tHcys, and folate did not differ. Plasma zinc, followed by CRP (≤5 mg/L), gravida (≤4), and plasma ferritin (≥12 μmg/L) status were major positive predictors of hemoglobin. Despite some early functional vitamin B-12 deficiency, there was no macrocytic anemia. Consumption of fermented enset may have increased vitamin B-12 levels in diet and plasma.

Introduction

Pregnant women from subsistence farming households in rural Africa usually consume predominantly plant-based diets; intakes of animal-source foods are low. Consequently, micronutrient intakes are often inadequate (1,2), a problem often exacerbated in these settings by malabsorption induced by infections and bacterial overgrowth (3,4). Hence, coexisting deficiencies of iron, zinc, vitamin A, vitamin B-12, and folate have often been reported (2,5,6) and associated with increased risk of maternal complications and poor fetal outcomes (7).

In the Sidama Zone of Southern Ethiopia, maize (Zea mays L.) and fermented enset (Enset ventricosum) products are the major staple foods, contributing up to 90% of energy. Animal products often provide <1% of the total energy (8). Nevertheless, dietary iron deficiency is not a major cause of anemia during pregnancy in this region (8). Such low intakes of animal products concomitant with infections and bacterial overgrowth may place pregnant women in Sidama at high risk for vitamin B-12 deficiency and possibly folate deficiency (3,6,9), especially in this malarial endemic area, because some antimalarial drugs interfere with folate metabolism (9). However, some vitamin B-12 and folate-producing microorganisms have been isolated from fermented enset (10). Hence, the consumption of fermented enset may reduce the risk for vitamin B-12 and folate deficiency. To our knowledge, however, there are no data on vitamin B-12 and folate status of pregnant women in Sidama.

Our earlier cross-sectional study of subsistence pregnant farmers from Sidama, Southern Ethiopia reported a high prevalence of protein and zinc deficiency but not iron deficiency (8). Here, we extend the research by investigating the vitamin B-12 and folate status of these same subjects using a combination of static (i.e. plasma vitamin B-12 and folate concentrations) and functional [i.e. plasma methylmalonic acid (MMA),⁸ total homocysteine (tHcys), and cystathionine] biochemical indices. We also examined their vitamin B-12 and folate status according to whether the women received more of their dietary energy from unrefined maize (i.e. maize diet group) or from fermented enset products (enset diet group). We also investigated the role of vitamin B-12 and folate status, as well as protein, other micronutrients (iron, vitamin A, zinc), infection, and gravida on hemoglobin concentrations.

Subjects and Methods

Study site and subjects

Demographic and clinical details were given earlier (8). Briefly, the women (n = 99) were subsistence farmers from 3 rural communities in the Sidama Zone of southern Ethiopia who volunteered to participate in the study. The women were all in the 3rd trimester of pregnancy (based on fundal height), afebrile, and aged 27.8 ± 4.6 y (mean ± SD). Ethical approval of the study protocol was obtained from the Human Ethics Committee of Hawassa University, Awassa, Ethiopia, and the Colorado Multiple Institutional Review Board of the University of Colorado, Denver, CO. Verbal informed consent was obtained from all the participants. After recruitment, trained research assistants interviewed the women using a pretested questionnaire and observations to obtain sociodemographic and health data. The sample size (n = 99) allowed estimation of the prevalence of anemia to within 10 percentage points of the true value with a confidence level of 95%.

Dietary and anthropometric assessment

Dietary intakes were calculated from 1-d weighed food records using an Ethiopian food composition table based on literature and analyzed values for calcium, iron, zinc, and phytate (Phy) (8,11). Observed intakes of energy and protein were adjusted for digestibility and protein intakes for amino acid score. The women were then classified into 2 groups according to whether maize (n = 68) or enset (n = 31) was the predominant energy source. Samples of each of the 2 fermented enset products, kocho and bulla, were purchased from local markets and analyzed for folate and vitamin B-12 content. We performed analysis for folate in duplicate on each sample by microbiological assay using Lactobacillus rhamnosus (formerly known as Lactobacillus casei, ATCC 7469) and the trienzyme extraction method (12). Vitamin B-12 was assayed using a newly developed liquid chromatography-MS method in the laboratory of S. Stabler (R. H. Allen and S. P. Stabler, unpublished data). Briefly, D2-labeled CN-cobalamin was added to samples of the kocho and bulla, which were partially purified prior to liquid chromatography. The peak areas for the internal standard were compared with that of the endogenous cobalamin to calculate the concentration in the sample. We measured weight and height and calculated BMI and height-for-age Z-scores (8).

Laboratory assessment

Anticoagulated whole blood and plasma were collected from nonfasting morning venipuncture samples using rigorous collection and separation procedures. Methodological details for the complete blood count, plasma albumin, ferritin, zinc, and C-reactive protein (CRP) were reported previously (8). Plasma vitamin B-12 was analyzed using the Solid Phase No Boil Dual count Radioassay B-12/Folic acid kit (Diagnostic Products). MMA, tHcys, and cystathionine were assayed in plasma by isotope dilution GC-MS (13,14). Plasma retinol was measured by HPLC (15). Plasma CRP concentrations >5 mg/L were used to indicate the presence of acute infection or inflammation.

The following interpretive criteria were used to define risk of biochemical micronutrient deficiencies in late pregnancy: plasma vitamin B-12 <150 pmol/L (16); MMA >271 nmol/L (17); plasma tHcys >13.9 μmol/L (18); plasma cystathionine >342 nmol/L (17); plasma folate <6.8 nmol/L (19); plasma zinc <7.6 μmol/L (20); and plasma retinol <0.70 μmol/L (21).

Anemia was defined as a hemoglobin concentration <115 g/L (cutoff adjusted upward by 5 g/L for an altitude of 1800 m) (22). Macrocytosis was defined as a mean cell volume (MCV) >100 fL (23) and microcytosis as MCV <81 fL (24). Anemia accompanied by a low plasma ferritin concentration (i.e. <12 μg/L) in the absence of infection or inflammation was considered to reflect iron deficiency anemia (IDA); a low plasma ferritin in the absence of anemia or infection was taken to indicate depleted iron stores (8).

Statistical analysis

Values in the text are means ± SD or median [interquartile range (IQR)]. Plasma vitamin B-12, cystathionine, and tHcys showed significant departures from normality and, hence, we calculated medians and the Mann-Whitney U statistic to assess between-group differences. Spearman correlation coefficients were used to explore the relation between the variables. We assessed differences between the 2 diet groups using the Student’s t test for independent samples or Fisher’s exact test. Multiple linear regression analysis was used to examine the predictors of hemoglobin concentration in all the subjects. We used indicator variables in the regression analysis for infection, gravida, and ferritin because of the nonnormal distribution of these independent variables. Differences were considered significant at P < 0.05. Data were analyzed using SPSS version 12.0.

Results

Selected demographic and anthropometric characteristics of the pregnant women did not differ between the 2 diet groups (Table 1).

TABLE 1.

Selected demographic and anthropometric characteristics of pregnant women who are subsistence farmers from Sidamo¹

	Maize-based diet, n = 68	Enset-based diet, n = 31	Mean difference2 (95% CI)
Age, y	27.6 ± 4.3	28.4 ± 5.4	0.86 (-3.10, 1.38)
No schooling, %	77.9	74.2	0.793
Fundal height, cm	7.1 ± 1.0	7.0 ± 1.1	0.03 (-0.45, 0.50)
Gravida > 4, %	63.2	56.7	0.653
Weight, kg	52.0 ± 6.3	52.4 ± 5.7	-0.45 (-3.00, 2.11)
Height, cm	154.8 ± 6.4	154.9 ± 6.7	-0.15 (-3.03, 2.73)
Height-for-age Z-score	-1.48 ± 1.07	-1.46 ± 1.12	-0.03 (-0.51, 0.45)
BMI, kg/m2	21.7 ± 2.0	21.8 ± 2.0	-0.18 (-1.04, 0.69)
Rarely had enough food, %	39.7	32.3	0.513

¹Values are means ± SD or %.

²From 2-tailed t test. Equality of variances not assumed.

³Significance of difference in proportion from Fisher’s exact test (2-sided).

Dietary assessment

Both groups had a very monotonous diet, low in energy, with >85% provided by foods from only 2 food sources: unrefined maize and enset products. Less than 1% of energy in both diet groups was provided by dairy products and no cellular animal foods were eaten. The enset diet group had lower intakes (P < 0.001) of utilizable protein, fat, zinc, dietary fiber, Phy, and lower Phy:Zn and Phy:Fe molar ratios but higher intakes (P < 0.001) of calcium than the maize diet group (Table 2).

TABLE 2.

Intakes per day of energy, selected nutrients, dietary fiber, Phy, and Phy:zinc and Phy:iron molar ratios for 2 groups of women consuming unrefined maize- or enset-based diets¹

	Maize-based diet, n = 68	Enset-based diet, n = 31	Mean difference2 (95% CI)
Adjusted energy, kJ	4382 ± 1481	3884 ± 1273	497 (-83, 1078)
Utilizable protein, g	16.3 ± 5.9	11.3 ± 4.7	5.0 (2.8, 7.2)*
Fat, g	10.5 ± 4.8	4.4 ± 3.4	6.1 (4.5, 7.8)*
Carbohydrate, g	250.2 ± 85.6	228.1 ± 75.3	22.1 (-11.9, 56.2)
Calcium, mg	384 ± 9.3	773 ± 324.1	-388 (-520, -256)*
Iron, mg	28.2 ± 11.3	27.7 ± 9.9	0.5 (-4.0, 5.0)
Zinc, mg	6.3 ± 2.5	3.7 ± 1.6	2.59 (1.8, 3.4)*
Ascorbic acid,3 mg	2.2 (0.7, 5.0)	1.8 (1.0, 4.6)	0.87
Total dietary fiber, g	29.0 ± 11.8	17.0 ± 8.7	12.0 (7.7, 16.2)*
Phy, mg	1356 ± 603	446 ± 335	910 (722, 1098)*
Phy:zinc molar ratio	21.1 ± 4.4	10.8 ± 4.8	10.3 (8.2, 12.3)*
Phy:iron molar ratio	4.1 ± 1.2	1.3 ± 0.7	2.8 (2.4, 3.2)*
Energy from maize products, %	74.0 ± 19.1	20.8 ± 15.4
Energy from enset products, %	15.4 ± 16.2	66.1 ± 15.6

¹Values are means ± SD and mean differences (95% CI) unless noted otherwise.

²From 2-tailed t test. Equality of variances not assumed.

^*Means differ, P < 0.001.

³Median (IQR). In the right-hand column, the test statistic is the 2-tailed asymptotic significance from the Mann-Whitney U test.

Median intakes of kocho were 174 (129, 254) g/d for the enset diet group compared with 54 (0, 132) g/d for the maize diet group; neither group consumed bulla. The corresponding median intakes of maize for the enset and maize diet groups were 65 (25, 96) g/d and 107 (82, 158) g/d, respectively. The mean analyzed content of folate for the samples of the 2 fermented enset products, kocho and bulla, were 33.4 ± 9.3 μg/100 gand 15.0 ± 4.3 μg/100 g (fresh weight), respectively, whereas the corresponding analyzed values for vitamin B-12 were 3.9 μg/100 g and 3.7 μg/100 g, respectively. Hence, kocho contributed 58.1 μg/d folate and 6.8 μg/d vitamin B-12 for the enset diet group compared with 18.0 μg/d folate and 2.1 μg/d vitamin B-12 for the maize diet group.

Biochemical indices

Women with an elevated plasma CRP indicative of acute infection or inflammation (>5 mg/L; n = 8) had a significantly lower mean hemoglobin, RBC count, and plasma retinol (P < 0.05) than did those with a plasma CRP concentration ≤5 mg/L, although plasma zinc, ferritin, albumin, and none of the biochemical indices of folate or vitamin B-12 status did not differ. However, because plasma zinc, ferritin, retinol, and albumin (but not folate and vitamin B-12 indices) are all known to be influenced by inflammation or infection (15), women with elevated plasma CRP were excluded from the data in Table 3 but not from the multiple linear regression analysis in Table 4, where an indicator variable for infection was included.

TABLE 3.

Selected hematologic and biochemical indices in the maize- and enset-based diet groups¹,²

	All subjects			Maize-based diet			Enset-based diet			Mean difference for maize vs. enset (95% CI)
	n	%		n	%		n	%
Hemoglobin, g/L	84	17	121 ± 11.1	56	14	122 ± 10.6	28	21	118 ± 11.9	3.7 (-1.6, 9.1)
RBC count, × 1012/L	84	37	3.9 ± 0.41	56	36	3.9 ± 0.40	28	39	3.9 ± 0.44	-0.01 (-0.21, 0.19)
MCV, fL	84	5	88.9 ± 4.2	56	2	89.6 ± 3.7	28	11	87.3 ± 4.8	2.30 (0.21, 4.39)*
Mean cell hemoglobin concentration, g/L	83	0	349 ± 10	55	0	350 ± 10	28	0	346 ± 10	4.36 (-0.32, 9.04)
Plasma zinc, μmol/L	84	74	6.99 ± 1.09	57	77	7.05 ± 1.03	27	67	6.86 ± 1.22	0.19 (-0.36, 0.73)
Plasma ferritin, μg/L	86	33	13.9 (10.7, 23.7)	58	38	13.6 (9.8, 24.4)	28	21	14.5 (12.5, 21.8)	0.396
Plasma retinol, μmol/L	85	27	25 (19, 31)	62	21	0.96 ± 0.35	23	39	0.76 ± 0.35	0.20 (0.04, 0.36)*
Plasma vitamin B-12, pmol/L	83	23	268 (152, 372)	56	25	238 (147, 345)	27	18	291 (220, 449)	0.888
Plasma MMA, nmol/L	94	62	407 ± 206	65	60	398 ± 211	29	66	425 ± 198	-27 (-117, 64)
Plasma cystathionine, nmol/L	83	0	126 (94, 165)	65	3	135 (109, 172)	29	0	111 (88, 132)	0.020*
Plasma folate, nmol/L	94	2	26.1 ± 12.6	56	0	25.4 ± 12.1	27	0	27.6 ± 13.7	-2.2 (-8.4, 4.0)
Plasma homocysteine, μmol/L	94	0	3.5 (3.1, 4.1)	65	0	3.5 (3.0, 4.2)	29	0	3.3 (3.1, 4.1)	0.703

¹Values are means ± SD or median (IQR). The right-hand column is the mean difference (95% CI) or the 2-tailed asymptotic significance for the differences in the medians from the Mann-Whitney U test.

^*Means differ, P < 0.05.

²For the hematologic variables, plasma zinc, plasma ferritin, and plasma retinol, only subjects with plasma CRP ≤5 mg/L are included. Cutoffs used: hemoglobin <115 g/L; RBC count <3.8 × 10¹²/L; MCV <81 fL; mean cell hemoglobin concentration <320 g/L; plasma zinc <7.6 μmol/L; plasma ferritin ≥12 μg/L and plasma retinol <0.70 μmol/L; plasma vitamin B-12 <150 pmol/L; plasma MMA >271 nmol/L; plasma cystathionine >342 nmol/L; plasma folate <6.8 nmol/L; and plasma homocysteine >13.9 μmol/L.

TABLE 4.

Multiple linear regression analysis with hemoglobin as the dependent variable¹,²

	Unstandardized coefficients		Standardized coefficients3	95% CI for B
Model	B	SE	β	Lower bound	Upper bound
(Constant)	6.93	0.74		5.36	8.50
Plasma zinc, μmol/L	0.42	0.10	0.37	0.23	0.62
CRP ≤ 5 mg/L	1.35	0.37	0.31	0.60	2.09
Gravida ≤ 4	0.73	0.22	0.28	0.28	1.17
Plasma ferritin ≥ 12 μg/L	0.60	0.23	0.23	0.14	1.05

¹n = 88, R² = 0.37, adjusted R² = 0.34.

²The pairwise correlation coefficients for all of the independent variables in the linear regression model are all nonsignificant (data not shown), suggesting that colinearity does not compromise the regression model.

³Standardized coefficients are estimates from the analysis performed on the standardized variables so that each variable has a variance of 1.0.

Mean plasma vitamin B-12 level tended to be higher (P = 0.052), whereas MCV and plasma cystathionine were lower in the enset diet group compared with the maize diet group (P < 0.05). There were no other significant differences between the 2 diet groups for hematological or biochemical indices or for the prevalence of abnormal values (Table 3).

Of the women, 29% (n = 27) had anemia from all causes; none had severe anemia (hemoglobin <70 g/L). A total of 13% (n = 12) of the women had IDA with a mean cell hemoglobin concentration of 343 ± 11.7 g/L and a MCV of 88.3 ± 5.6 fL. Interestingly, the mean values for MCV, plasma vitamin B-12, plasma folate, MMA, and tHcys concentrations for the anemic and nonanemic women were comparable (data not shown), whereas the mean plasma cystathionine was higher in the anemic group than the nonanemic group (173 vs. 128 nmol/L; mean difference: 43.8, 95% CI: 6.9, 80.7).

Interrelations among laboratory variables

Spearman rank correlation coefficients were significant between plasma folate and plasma cobalamin (r = 0.265; P = 0.015), plasma folate and MMA (r = 0.298; P = 0.006), and plasma zinc and hemoglobin (r = 0.328; P = 0.001). Negative correlations were noted between hemoglobin and plasma cystathionine (r = -0.281; P = 0.007) and plasma folate and plasma cystathionine (r = -0.357; P = 0.001). No significant correlations existed between plasma zinc and ferritin or between plasma tHcys and any of the other folate or vitamin B-12 indices.

Predictors of hemoglobin status

In the multiple regression model with the largest explanatory effect on hemoglobin (Table 4), there was no evidence of multiple colinearity or interactions among the independent variables. Plasma zinc had the largest standardized β coefficient, followed by CRP, gravida, and plasma ferritin (Table 4). None of the biochemical indices of folate or vitamin B-12 status, plasma retinol, or plasma albumin were significant predictors of hemoglobin concentrations. Removal of plasma zinc from the multiple regression model produced the largest reduction in the percentage of variance explained, from 37.3 to 24.0%.

Discussion

The most notable finding related to this group of women from subsistence farming households in Sidama, Ethiopia was the low prevalence of IDA and the absence of any hematopoietic defects associated with folate or vitamin B-12 deficiency during pregnancy. This is not withstanding the negligible consumption of animal products. Consumption of fermented enset products may have increased vitamin B-12 levels in the diet and plasma and thus reduced the risk of vitamin B-12 deficiency anemia, although there was some evidence of an early stage of functional vitamin B-12 deficiency, based on elevated MMA values.

Interrelationship between diet and vitamin B-12 and folate status

Several vitamin B-12-producing microorganisms have been isolated during the fermentation of enset, including Lactobacillus planatarum (10). Hence, the tendency for women in the enset diet group to have higher plasma vitamin B-12 concentrations than those in the maize diet group might be associated with the microbial contribution of vitamin B-12 from fermented kocho. Certainly, based on our intake data and chemical analysis of kocho, vitamin B-12 intakes in the ensetdiet group were more than twice the level set by WHO/FAO (25) for the estimated average requirement during pregnancy (i.e. 2.2 μg/d). Even in the maize diet group, the contribution of vitamin B-12 from fermented enset was considerable (i.e. 2.1 μg vitamin B-12/d).

Notwithstanding the apparently large contribution to the diet of vitamin B-12 by fermented enset, mean MMA levels did not differ between the 2 diet groups. Indeed, the prevalence of elevated MMA concentrations, considered to be the most sensitive indicator of vitamin B-12 deficiency (26), was very high in both groups (Table 4). This apparent discrepancy suggests that the fermented enset might have contained some nonfunctional analogs of vitamin B-12 as well as the cobalamins that were measured in the radioassay. This may have led to the apparently normal plasma vitamin B-12 concentrations in some of the women whose plasma MMA values were unexpectedly high.

L. planatarum is also a folate-producing microorganism (27) so it is not surprising that fermented enset also contributed some folate, albeit a much smaller amount in relation to the WHO/FAO (25) estimated average requirement for folate during pregnancy (i.e. 520 μg/d). There were several additional sources of folate in the diets of both groups, most notably kale (Brassica carinata), kidney beans (Phaseolus vulgaris L.), and haricot beans (Phaseolus adenguarre) (8), which probably accounted for the absence of folate deficiency, based on low plasma folate and elevated tHcys concentrations, in both diet groups (Table 3).

Predictors of hemoglobin

Plasma zinc concentrations were the strongest predictor of hemoglobin, followed by the indicator variables for infection (CRP ≤5 mg/L), gravida (≤4), and plasma ferritin (≥12 μg/L) (Table 4). Moreover, there were no significant pairwise correlations linking the independent variables, including plasma ferritin and zinc (data not shown). The predictors of hemoglobin in Table 4 are discussed in turn below, along with protein and vitamin A, also required for normal hematopoiesis.

The strong positive association between plasma zinc and hemoglobin was unexpected, although it has been reported in some earlier studies (28-31). There have also been a few intervention studies in which the addition of zinc alone (32), or zinc and iron relative to iron alone (33,34), has improved the hematological response of young children who were thought to be zinc deficient. Several mechanisms may be involved whereby zinc affects hemoglobin concentrations. Zinc is implicated in hemoglobin synthesis through the activity of several zinc-dependent enzyme systems, including aminolevulinic acid dehydrase that mediates a step in the synthesis of heme (35) and thymidine kinase and DNA polymerase, which are involved in DNA synthesis. More recently, the zinc-finger transcription factor, GATA-1, has also been confirmed as essential for normal erythropoiesis (36). Other potential mechanisms may involve the stimulation of hematopoiesis by zinc-induced increases in plasma insulin-like growth factor-1 levels (37) and the role of zinc in stabilizing cell membranes (38).

An elevated CRP (>5 mg/L) was negatively associated with hemoglobin, notwithstanding the relatively small number of subjects (n = 8) with elevated values indicative of underlying acute infection or inflammation. The latter is often accompanied by anemia, commonly termed anemia of chronic disease (ACD), which results from the effects of cytokines. Apparently normal or increased iron stores are present in ACD (4).

Gravida also had an independent and significant inverse association with hemoglobin (Table 4), a relationship that has also been reported in pregnant women in rural Malawi (2). This negative relationship has been attributed to the cumulative demands on iron stores of successive pregnancies and short interpregnancy intervals (39).

Plasma ferritin was also a significant predictor of hemoglobin, although its importance, based on the absolute value of its standardized coefficient, was less than that for plasma zinc (Table 4). Dietary factors probably account in part for the positive association between plasma ferritin and hemoglobin noted here. Certainly, dietary iron intake in Ethiopia is characteristically high (8,40,41), some of which is absorbed. In Sidama, absorption of nonheme iron may be facilitated by colonization of L. planatarum from fermented enset (42). Several other reports have confirmed that IDA is not a serious problem in Ethiopia, even during pregnancy (40,41), compared with other countries in Sub-Saharan Africa (e.g. Malawi) (2). Of the nondietary factors, helminth parasitic infections simultaneously reduce hemoglobin and ferritin concentrations and thus could account for the positive β coefficient observed here. In contrast in ACD and malaria there is a negative relationship between hemoglobin and ferritin (2,4). Hence, our observations imply that neither ACD nor malaria were significant in these pregnant women, a suggestion supported by the relatively small number of subjects (n = 8) with elevated CRP values.

Unlike ferritin, neither plasma vitamin B-12 nor MMA was a significant predictor of hemoglobin, despite the high prevalence of elevated concentrations of MMA (i.e. 62%). This discrepancy is attributed to nonfunctional vitamin B-12 analogs present in the plasma of women consuming fermented enset, as discussed earlier, as well as the absence of any hematopoietic defects associated with vitamin B-12-deficiency anemia. Plasma folate was also not a significant predictor of hemoglobin, although levels were negatively correlated with plasma cystathionine (r = -0.357; P = 0.001) (but not tHcys). This finding may indicate some folate (or pyridoxine) deprivation, although plasma folate may not reflect true folate status in malarial endemic areas such as Sidama, where levels may be elevated through erythrocyte hemolysis induced by malaria.

Plasma retinol concentrations were not positively associated with hemoglobin (Table 3) in contrast to some earlier findings in Ethiopia (43) and elsewhere (44). Likewise, plasma albumin levels were also unrelated to hemoglobin concentrations despite low levels, attributed to low protein intakes (8), hemodilution of pregnancy, infection, and zinc deficiency.

Our findings are based on an observational study conducted on a self-selected sample of women. Hence, causal inferences cannot be made from the associations reported in this study and the results cannot be extrapolated to all women in the 3rd trimester of pregnancy living in the Sidama Zone of Ethiopia (45). Also, our sample size was small and this may have contributed to the null findings of differences in micronutrient status between the 2 groups. However, fermented enset products are reported to be the staple food for more than more than 10 million Ethiopians (46) and further investigation into the presence of vitamin B-12 analogs in the plasma of women consuming fermented enset and in the food itself is warranted.

In summary, there was some evidence of early functional deficiency of vitamin B-12 but not folate deficiency among these pregnant women and no vitamin B-12 deficiency anemia, despite very low intakes of animal products. The prevalence of IDA was also low compared with elsewhere in Africa. Consumption of enset fermented by vitamin B-12-producing microorganisms may have increased vitamin B-12 levels in the diets and in plasma while at the same time enhancing nonheme iron absorption. Two micronutrients (zinc and iron status) and 2 nondietary factors, infection and gravida, were the major predictors of hemoglobin. A well-designed, randomized, controlled trial of multi-micronutrient supplements with and without zinc, conducted in a population group at high risk for zinc deficiency such as the Sidama pregnant women studied here, is needed to establish the role of zinc deficiency in the etiology of anemia.