Geophagy (Soil-eating) in Relation to Anemia and Helminth Infection among HIV–Infected Pregnant Women in Tanzania

Abstract

Geophagy, the regular and deliberate consumption of soil, is prevalent among pregnant women in sub-Saharan Africa. We examined the associations of geophagy with anemia and helminth infection among 971 human immune-deficiency virus (HIV)-positive pregnant women in Tanzania. About 29% of pregnant women regularly consumed soil. Occupation, marital status, and gestational age were associated with geophagy. Ascaris lumbricoides infection was associated with the prevalence of geophagy (adjusted-prevalence ratio 1.81; 95% confidence interval [CI] = 1.37–2.40); however, hookworm, Trichuris trichiura, and Strongyloides stercoralis showed no association. Anemia and red blood cell characteristics suggestive of iron deficiency were strongly correlated with geophagy at baseline. In longitudinal analyses, we found evidence suggesting that soil consumption may be associated with an increased risk of anemia (adjusted-relative risk 1.16; 95% CI = 0.98–1.36) and a lower hemoglobin concentration (adjusted-mean difference −3.8 g/L; 95% CI [−7.3, −0.4]). Pregnant women should be informed about the potential risks associated with soil consumption.

INTRODUCTION

Pica is the regular and deliberate eating of non-food substances seen worldwide. Geophagy is a type of pica, which involves the consumption of soil or clay. It is common among pregnant women and children in sub-Saharan Africa. Although geophagy is often regarded as a response to compulsive physiologic demand during pregnancy, it is a cultural and widely accepted practice in African societies.¹ The prevalence of geophagy among pregnant women ranged from 65% in Kenya, 46% in Ghana, 42% in Namibia, to 28% in Tanzania.^2–5 Women consume a variety of soil types, including hardened clay soil sold in shops and markets and soil from the walls of houses or termite mounds.

Anemia affects nearly half of the pregnant women in the world, and it is one of the most prevalent problems stemming from nutritional deficiency.⁶ Anemia and iron deficiency during pregnancy is an important risk factor for maternal mortality, poor pregnancy outcomes, including preterm delivery and low birthweight, and infant mortality in developing countries.^7–9 Furthermore, anemia is a common hematologic complication in human immunodeficiency virus (HIV)–infected women and associated with disease progression and an increased risk of mortality.^10,11 Numerous cross-sectional studies found that geophagy was correlated with anemia and iron deficiency among pregnant women in sub-Saharan Africa.^4,5,12,13 However, these cross-sectional studies do not explain whether geophagy causes anemia, or anemia induces a craving for soil. To our knowledge, no studies have prospectively examined the associations of geophagy with anemia among pregnant women. Geophagy is also speculated to be a risk factor for soil-transmitted helminth infection. Helminth infection affects over 1 billion people in tropical developing countries and contributes to severe morbidity. Women who practice geophagy may be particularly at risk of infection with Ascaris lumbricoides and Trichuris trichiura by ingesting eggs from contaminated soil. Hookworms and other intestinal helminths may cause anemia in pregnant women.^14–16 Hookworms and Strongyloides stercoralis are transmitted when the parasites burrow into the skin, and thus, these infections may not be a consequence of soil pica.

We examined the cross-sectional associations of geophagy with anemia, parasitic infection, and nutritional and sociodemographic factors among HIV-positive pregnant women in Dar es Salaam, Tanzania. We also researched the prospective relationships of geophagy with hemoglobin concentrations, the incidence of anemia, and pregnancy outcomes.

METHODS

Study design and population.

From April 1995 to July 1997, 1,078 HIV–infected pregnant women were enrolled in a trial to examine the effect of vitamin supplements on mother-to-child transmission (MTCT) of HIV, disease progression, and other health outcomes in Dar es Salaam, Tanzania. Details of the study design have been published.¹⁷ Eligible women were between 12 and 27 weeks gestation age at enrollment, according to the date of the last menstrual period. Women were randomly assigned in a two-by-two factorial design to receive a daily dose of vitamin A, multivitamins including vitamin A, multivitamins excluding vitamin A, or placebo. In accordance with the standard antenatal care services in Tanzania, all women received 400 mg of ferrous sulfate (equivalent to 120 mg of ferrous iron) and folate (5 mg) daily for anemia prophylaxis and 500-mg chloroquine phosphate (equivalent to 300 mg of chloroquine base) weekly for malaria prophyaxis during pregnancy. Antiretroviral therapy was not available in Tanzania at the time of the study.

At the baseline visit, trained research assistants (nurses) interviewed women and obtained information about geophagy, which was defined as eating soil during their pregnancy, sociodemographic characteristics, and obstetric history. They also measured the women’s weight, height, and mid-upper arm circumference (MUAC). A study physician performed a complete medical examination. The stage of HIV disease was determined on the basis of the World Health Organization (WHO) Staging System.¹⁸ Blood, stool, and vaginal specimens were collected for laboratory analyses at baseline. Sexually transmitted diseases, intestinal helminths, and protozoan infections were treated at the time of diagnosis. Hemoglobin measurements were obtained at baseline and 6 weeks postpartum (delivery until 70 days postpartum). Immediately after birth, a research midwife measured the infant birthweight to the nearest 10 g on a standard beam balance. Low birthweight was defined as birthweight < 2,500 g, preterm birth as delivery before 37 weeks, and small for gestational age as birthweight below the tenth percentile of weight for gestational age by the standards of Brenner and others.¹⁹ Fetal death was defined as either miscarriage (delivery before 28 weeks of gestation) or stillbirth (delivery at or after 28 weeks).

Laboratory methods.

Hemoglobin was measured using either a CBC5 Coulter counter (Coulter Corp., Miami, FL) or the cyanmethemoglobin method with a colorimeter (Corning Inc., Corning, NY). To examine the extent of iron deficiency, trained laboratory technicians examined thin blood films with Leishman stain for red blood cell morphology. They coded the extent of the following cell characteristics: anisocytosis, poikilocytosis, hypochromasia, hyperchromasia, microcytosis, macrocytosis, normochromasia, and normocytosis. Cell characteristics were classified into five levels of severity, coded as either absent, < 25% (of the cells in the specified field are abnormal), 25–50%, 50–75%, or > 75%. Erythrocyte sedimentation rate (ESR) was determined using the Westergren method. Plasma levels of vitamin A and E were determined using reversed-phase high performance liquid chromatography (HPLC). Plasma selenium concentration was measured by neutron activation analysis at the University of Missouri Research Facility, Columbia, MO. The selenium concentration was adjusted for the sodium concentration of the sample.²⁰ Absolute CD4 cell counts were measured using the FACScount or the FACScan systems (Becton Dickinson, San Jose, CA). Infection with Plasmodium falciparum malaria parasites was identified using both thin and thick blood films with Giemsa staining. To identify intestinal helminths (hookworm, A. lumbricoides, T. trichiura, Strongyloides stercoralis, and Schistosoma mansoni) and pathogenic protozoan infections (Giardia lamblia, Entamoeba histolytica, and Cryptosporidium parvum), stool specimens were first examined macroscopically for general characteristics (pus, mucus, blood) and worms. Stools were then examined microscopically using saline wet mount for detection of eggs, larvae protozoan trophozoites, and cysts, followed by iodine wet mount to identify cysts. The formalin-ether concentration technique was used for further identification of eggs, larvae, and cysts. For diagnosis of HIV infection in infants, blood samples were collected at birth and 6 weeks. Details on the assessment of HIV status among children have been published.²¹ A child was determined to be HIV-1 infected if either a peripheral blood mononuclear cell specimen tested positive using a polymerase chain reaction (PCR). Infants who were HIV-positive at birth (from birth to 21 days) were presumably infected during the intrauterine period. Infants who were HIV-negative at birth, but positive at 6 weeks (from 22 to 49 days), were presumably infected during the intrapartum and early breast-feeding period.

Data analyses.

We first examined cross-sectional associations of geophagy with sociodemographic, nutritional, HIV related, and parasitic factors. Of the 1,078 HIV–infected women who were enrolled, we ascertained information about geophagy from 971 women at baseline, and they constituted for the cross-sectional analyses. Anemia was defined as hemoglobin < 110 g/L and severe anemia as hemoglobin < 85 g/L.²² As a proxy for iron deficiency anemia, findings from erythrocyte morphology testing were grouped into three levels. These categories were severe and moderate (hypochromasia with microcytosis), mild (hypochromasia without microcytosis), any other abnormality (anisocytosis and/or macrocytosis), and normal (normocytic and normochromic). The ESR was categorized as < 30, 30–60, and ≥ 60 mm/h. Plasma vitamin A was categorized as < 20 and ≥ 20 μg/dL, plasma vitamin E as < 9.7 and ≥ 9.7 μmol/dL (the median value), and plasma selenium into quartiles (< 108, 108–123, 123–140, > 140 μg/L). The WHO HIV disease stage was categorized as stage I versus stage II and III, and CD4 count as < 200, 200–499, and ≥ 500 cells/mm³. Prevalence ratios and 95% confidence intervals (CI) were estimated by binomial regression with a log link function.²³ Variables associated with geophagy at P < 0.20 were considered candidates for multivariate analyses. The adjusted prevalence ratio and 95% CI were estimated from the multivariate binomial regression models. The final multivariate models were built using backward selection procedure with P < 0.10 as the retention criterion. Because of collinearity, hemoglobin, erythrocyte morphology, or ESR was added separately to the multivariate model. We included participants with missing values by using the missing indicator method for all the multivariate analyses.

To examine the prospective relationships of geophagy with the incidence of anemia and hemoglobin concentrations, we excluded 254 women who were severely anemic at baseline from 956 women with known baseline hemoglobin. Among the remaining 702 women, 547 women who had a hemoglobin measurement at 6 weeks postpartum were included in the analyses. Binomial regression models were used to estimate relative risks of anemia and severe anemia. Linear regression models were used to estimate mean difference in hemoglobin concentration. The multivariate models were adjusted for MUAC, WHO HIV disease stage, CD4 count, gestational age, malaria infection, and treatment regimen. Inclusion of these variables in the multivariate models was based on prior knowledge and statistical considerations. We did not adjust for hemoglobin concentration and A. lumbricoides infection, because they were considered intermediate variables.

We also longitudinally examined the associations of geophagy with pregnancy outcomes and the risk of HIV transmission. Among 971 women, there were 24 sets of twin births and 1 pregnancy outcome was unknown. Of 946 singleton birth outcomes, there were 51 fetal deaths (11 miscarriages and 40 stillbirths). We analyzed 895 singleton liveborns for preterm birth and 804 for low birthweight and small for gestational age. For 91 infants, birthweights were not known because of delivery at home or at another medical facility. For pregnancy outcomes and HIV transmission at birth and 6 weeks, binomial models were used for univariate analyses. Poisson regression model with robust variance estimates was used in the multivariate analyses because the multivariate binomial model failed to converge.²³ All statistical analyses were conducted using SAS version 9.1 (SAS Institute, Cary, NC).

Ethical considerations.

The study protocol was approved by the Research and Publications Committee of Muhimbili University of Health and Allied Sciences, the Ethical Committee of the National AIDS Control Program of the Tanzanian Ministry of Health, and the Institutional Review Board of the Harvard School of Public Health.

RESULTS

Characteristics of the study population.

Of 1,078 women enrolled in the trial, 971 had baseline information about geophagy and were included in the analyses. The mean age of women was 24.7 ± 4.8 years (Table 1). The majority (87%) of the women had completed 5 to 8 years of primary education. Almost three quarters (73%) were not employed outside the home, 64% were married, and 85% shared meals with less than 5 people at home. Women were at an average of 20.3 ± 3.4 weeks gestation at enrollment. The mean weight was 57.3 ± 9.0 kg. Nearly 82.6% of women were anemic and one-third (33.9%) of the participants had plasma concentrations of vitamin A < 20 μg/dL. The majority (80.4%) of the women were in stage I of HIV disease and 12.5% had a CD4 cell count < 200 cells/mm³. One-third (34%) was primiparous. The prevalence of P. falciparum malaria infection was 18.4%. Intestinal helminths infections were present in 29.4% of the participants: hookworm 12.0%, A. lumbricoides 5.8%, T. trichiura 1.1%, S. stercoralis 1.6%, and S. mansoni 1.7%.

Table 1.

Baseline characteristics of HIV–infected pregnant women in Tanzania (N = 971)

Characteristics	N (%)* or Mean ± SD
Age (years)	24.7 ± 4.8
Marital status
Married, monogamous	560 (57.7%)
Married, polygamous	56 (5.8%)
Cohabiting	246 (25.3%)
Single, widowed, divorced	109 (11.2%)
Education (years)
None or adult	79 (8.1%)
Primary 1–4	50 (5.2%)
Primary 5–8	742 (76.4%)
Secondary or higher, ≥ 9	100 (10.3%)
Occupation
No outside employment	707 (72.8%)
Small business, professional or other	264 (27.2%)
Gestational age at enrollment (weeks)	20.3 ± 3.4
Number of prior pregnancy
0	330 (34.0%)
1–2	459 (47.3%)
≥ 3	182 (18.7%)
Hemoglobin (g/dL)
< 85: severe anemia	254 (26.6%)
85–109: moderate anemia	536 (56.0%)
≥ 110: normal	166 (17.4%)
WHO HIV disease stage
I	777 (80.4%)
II and III	189 (19.6%)
Parasitic infection
Malaria infection	177 (18.4%)
Hookworm infection	96 (12.0%)
Ascaris lumbricoides infection	46 (5.8%)
Trichuris trichiura infection	9 (1.1%)
Strongyloides stercoralis infection	13 (1.6%)
Schistosoma mansoni infection	14 (1.7%)
Any pathogenic protozoan infection	53 (6.6%)

^*Sums may not add up to 971 because of missing values.

HIV = human immunodeficiency virus; WHO = World Health Organization.

Correlates of geophagy.

At baseline, the prevalence of reported geophagy was 28.5% (277/971). We examined the correlates of geophagy with sociodemographic, nutritional, HIV related, and parasitic factors in univariate and multivariate models (Table 2 and 3). Women’s age, education, and money spent on food were not related to the prevalence of geophagy. Women who were cohabiting with a partner were more likely to consume soil than those who were in monogamous marriage. Women who were employed outside the home were less likely to be geophageous than housewives. Other sociodemographic factors associated with geophagy included increasing gestational age at enrollment, the number of people sharing meals at home, and having had one or two prior pregnancies compared with primipara. At baseline, severely anemic women (hemoglobin < 85 g/L) were 75% more likely to be geophageous than those without anemia (hemoglobin ≥ 110 g/L) in the multivariate model (adjusted-prevalence ratio = 1.75; 95% confidence interval [CI] = 1.30–2.36; P < 0.0002). Women with hypochromasia with or without microcytosis, characteristics commonly caused by iron deficiency, were more likely to consume soil compared with those without hypochromasia. Higher ESR was also strongly associated with geophagy. High plasma selenium levels were associated with geophagy in the univariate model but did not remain significant in the multivariate model. Plasma vitamin A and E were not associated with the prevalence of geophagy. The HIV-related factors, including WHO HIV disease stage, CD4 count, and vaginal discharge, were not associated with the prevalence of geophagy. Women infected with malaria were less likely to consume soil compared with those not infected with malaria. Ascaris lumbricoides infection was strongly associated with geophagy (adjusted-prevalence ratio = 1.81; 95% CI = 1.37–2.40; P < 0.0001); however, other helminth infections showed no association with geophagy.

Table 2.

Univariate associations of factors with geophagy among HIV–infected pregnant women in Tanzania (N = 971)

Variable	Geophagy, n (%)	Unadjusted prevalence ratio (95% CI)*	P value
Sociodemographic factors
Age (years)
< 20	38 (32.2%)	1.00
20–24	109 (27.5%)	0.85 (0.63, 1.16)	0.32
25–29	87 (29.5%)	0.92 (0.67, 1.26)	0.59
≥ 30	43 (26.5%)	0.82 (0.57, 1.19)	0.30
Education (years)
None or adult	24 (30.4%)	1.00
Primary 1–4	18 (36.0%)	1.19 (0.72, 1.95)	0.50
Primary 5–8	213 (28.7%)	0.94 (0.66, 1.34)	0.75
Secondary or higher, ≥ 9	22 (22.0%)	0.72 (0.44, 1.19)	0.20
Marital status
Married, monogamous	145 (25.9%)	1.00
Married, polygamous	21 (37.5%)	1.45 (1.00, 2.09)	0.047
Cohabiting	78 (31.7%)	1.22 (0.97, 1.54)	0.085
Single, widowed, divorced	33 (30.3%)	1.16 (0.85, 1.61)	0.33
Occupation
No outside employment	216 (30.6%)	1.00
Small business, professional, or other	61 (23.1%)	0.76 (0.59, 0.97)	0.026
Number of people sharing meals
≤ 5 people	226 (27.3%)	1.00
> 5 people	51 (35.9%)	1.32 (1.03, 1.69)	0.028
Money spent on food per person a day
≤ 500 shillings	149 (28.9%)	1.00
> 500 shillings	99 (28.1%)	0.97 (0.78, 1.20)	0.78
Gestational age at enrollment (weeks)
≤ 16	29 (17.7%)	1.00
17–20	89 (27.6%)	1.56 (1.07, 2.27)	0.0194
21–24	131 (32.2%)	1.82 (1.27, 2.61)	0.0011
≥ 25	28 (35.9%)	2.03 (1.30, 3.16)	0.0018
Number of prior pregnancy
0	81 (24.6%)	1.00
1–2	146 (31.8%)	1.30 (1.03, 1.63)	0.0284
≥ 3	50 (27.5%)	1.12 (0.83, 1.51)	0.47
Previous preterm infant†
Yes	24 (30.8%)	1.01 (0.71, 1.44)	0.97
No	172 (30.6%)	1.00
Previous low birthweight infant†
Yes	27 (35.1%)	1.17 (0.84, 1.63)	0.35
No	169 (30.0%)	1.00
Nutritional factors
Weight (kg)‡ (quartile)
< 51	75 (31.5%)	1.00
51–55	71 (32.0%)	1.00 (0.76, 1.30)	0.97
56–61	64 (26.7%)	0.81 (0.61, 1.07)	0.14
≥ 62	58 (23.2%)	0.72 (0.54, 0.96)	0.0267
BMI (kg/cm2)‡
< 18.5	4 (13.3%)	0.48 (0.19, 1.19)	0.11
18.5–24.9	180 (31.1%)	1.00
≥ 25	84 (24.6%)	0.79 (0.63, 0.98)	0.0340
Mid-upper arm circumference (cm)
< 22	20 (37.0%)	1.31 (0.91, 1.89)	0.15
22–23.9	58 (27.5%)	0.97 (0.76, 1.24)	0.81
≥ 24	198 (28.3%)	1.00
Hemoglobin (g/L)
< 85: severe anemia	102 (40.2%)	2.02 (1.44, 2.84)	< 0.0001
85–109: moderate anemia	137 (25.6%)	1.29 (0.92, 1.80)	0.14
≥ 110: normal	33 (19.9%)	1.00
Erythrocyte morphology
Hypochromasia with microcytosis	47 (40.1%)	1.78 (1.35, 2.35)	< 0.0001
Hypochromasia without microcytosis	86 (33.1%)	1.47 (1.15, 1.86)	0.0017
Other abnormality	2 (28.6%)	1.27 (0.39, 4.13)	0.70
Normal	109 (22.6%)	1.00
Erythrocyte sedimentation rate (mm/h)
< 30	44 (21.3%)	1.00
30–60	83 (29.9%)	1.40 (1.02, 1.93)	0.036
≥ 60	126 (31.9%)	1.50 (1.11, 2.02)	0.0078
Plasma vitamin A (μg/dL)
< 20	61 (27.2%)	1.00 (0.90, 1.10)	0.94
≥ 20	120 (27.5%)	1.00
Plasma vitamin E (μmol/dL)
< 9.7	82 (25.9%)	0.96 (0.88, 1.05)	0.40
≥ 9.7	99 (28.8%)	1.00
Plasma selenium (μg/L) (quartile)
< 108	54 (24.8%)	1.00
108–123	51 (26.2%)	1.06 (0.76, 1.47)	0.75
123–140	62 (27.9%)	1.13 (0.82, 1.54)	0.45
> 140	75 (33.9%)	1.37 (1.02, 1.84)	0.037
HIV related factors
WHO HIV disease stage (%)
I	217 (27.9%)	1.00
II and III	59 (31.2%)	1.12 (0.88, 1.42)	0.36
CD4 count (cells/mm3)
< 200	34 (29.8%)	1.09 (0.78, 1.53)	0.61
200–499	149 (28.8%)	1.05 (0.83, 1.33)	0.66
≥ 500	77 (27.3%)	1.00
Reported vaginal discharge
Yes	31 (34.8%)	1.25 (0.92, 1.69)	0.15
No	246 (27.9%)	1.00
Parasitic factors
Malaria infection
Yes	33 (18.6%)	0.60 (0.43, 0.83)	0.0021
No	243 (31.0%)	1.00
Hookworm infection
Yes	24 (25.0%)	0.88 (0.61, 1.27)	0.50
No	199 (28.4%)	1.00
Ascaris lumbricoides infection
Yes	26 (56.5%)	2.17 (1.64, 2.87)	< 0.0001
No	197 (26.1%)	1.00
Trichuris trichiura infection
Yes	1 (11.1%)	0.39 (0.06, 2.51)	0.33
No	225 (28.2%)	1.00
Strongyloides stercoralis infection
Yes	4 (30.8%)	1.11 (0.49, 2.52)	0.81
No	219 (27.8%)	1.00
Schistosoma mansoni infection
Yes	4 (28.6%)	1.02 (0.44, 2.36)	0.96
No	222 (28.0%)	1.00
Any helminth infection§
Yes	74 (32.7%)	1.17 (0.93, 1.48)	0.19
No	164 (28.1%)	1.00
Any pathogenic protozoan infection¶
Yes	13 (24.5%)	0.87 (0.54, 1.42)
No	210 (28.1%)	1.00	0.58

^*From binomial regression models with a log link function.

^†Among multiparous women only.

^‡Adjusted for gestational age at enrollment.

^§Includes hookworm, A. lumbricoides, T. trichiura, S. stercoralis, and S. mansoni.

^¶Includes G. lamblia, E. histolytica, and C. parvum.

HIV = human immunodeficiency virus; CI = confidence interval, BMI = body mass index; WHO = World Health Organization.

Table 3.

Multivariate associations of factors with geophagy among HIV–infected pregnant women

Variable	Adjusted prevalence ratio (95% CI)*	P value
Marital status
Married, monogamous	1.00
Married, polygamous	1.24 (0.88, 1.75)	0.22
Cohabiting	1.29 (1.04, 1.59)	0.018
Single, widowed, divorced	1.09 (0.82, 1.47)	0.54
Occupation
No outside employment	1.00
Professional, small business or other	0.77 (0.62, 0.96)	0.020
Number of people sharing meals
≤ 5 people	1.00
> 5 people	1.25 (0.99, 1.59)	0.061
Gestational age at enrollment (weeks)
≤ 16	1.00
17–20	1.37 (1.00, 1.88)	0.053
21–24	1.54 (1.14, 2.09)	0.005
≥ 25	1.66 (1.11, 2.47)	0.013
Number of prior pregnancy
0	1.00
1–2	1.26 (1.02, 1.56)	0.032
≥ 3	1.06 (0.81, 1.39)	0.65
Hemoglobin (g/L)
< 85: severe anemia	1.75 (1.30, 2.36)	0.0002
85–109: moderate anemia	1.18 (0.88, 1.59)	0.26
≥ 110: normal	1.00
Erythrocyte morphology
Hypochromasia with microcytosis	1.59 (1.23, 2.06)	0.0004
Hypochromasia without microcytosis	1.34 (1.07, 1.67)	0.010
Other abnormality	1.26 (0.40, 3.95)	0.69
Normal	1.00
Erythrocyte sedimentation rate (mm/h)
< 30	1.00
30–60	1.27 (0.95, 1.68)	0.10
≥ 60	1.36 (1.04, 1.78)	0.0238
Malaria infection
Yes	0.61 (0.46, 0.80)	0.0004
No	1.00
Ascaris lumbricoides infection
Yes	1.81 (1.37, 2.40)	< 0.0001
No	1.00

^*From a multivariate binomial regression model. The basic multivariate model was adjusted for marital status, occupation, number of people sharing meals, gestational age at enrollment, number of prior pregnancies, hemoglobin, malaria infection, and A. lumbricoides infection. In separate models, hemoglobin was replaced by erythrocyte morphology or ESR, and entered into the multivariate model.

HIV = human immunodeficiency virus; CI = confidence interval.

Associations of geophagy with the incidence of anemia and pregnancy outcomes.

Geophagy was associated with a marginally significant increased risk of anemia after adjusting for potential confounding factors (adjusted-relative risk [RR] = 1.16; 95% CI = 0.98–1.36; P = 0.08; Table 4). Geophagy was not associated with the risk of developing severe anemia. Women who consumed soil had significantly lower hemoglobin concentration at 6 weeks postpartum compared with women who did not consume soil (adjusted mean difference = −3.8 g/L; 95% CI [−7.3, −0.4]; P = 0.03). These analyses were not adjusted for the baseline hemoglobin level given that it may be an intermediate variable on the causal pathway. Given that the baseline hemoglobin concentration may also be a confounder, we further present these associations including it in the regression models. The adjusted associations were attenuated; for anemia, adjusted-RR = 1.08; 95% CI = 0.93–1.26; P = 0.32, and for hemoglobin concentration, adjusted-mean difference = −2.3 g/L; 95% CI (–5.6, 1.0); P = 0.17.

Table 4.

Prospective relationships of geophagy and hemoglobin concentrations or the incidence of anemia^*

Endpoint	Geophagy		Unadjusted mean difference or RR (95% CI)	P value	Adjusted mean difference or RR (95% CI)†	P value
	Yes	No
Hemoglobin
Mean g/L	102.7 ± 17.8	106.7 ± 17.4	−4.0 (−7.5, −0.5)	0.02	−3.8 (−7.3, −0.4)	0.03
< 85 g/L. n/N (%) severe anemia	17/128 (13.5%)	41/419 (9.8%)	1.36 (0.80, 2.31)	0.26	1.30 (0.77, 2.22)	0.33
< 110 g/L. n/N (%) anemia	79/128 (61.7%)	226/419 (53.9%)	1.14 (0.97, 1.35)	0.10	1.16 (0.98, 1.36)	0.08

^*Among pregnant women (N = 547) who were not severely anemic at baseline.

^†Multivariate linear regression or binomial models were adjusted for mid-upper arm circumference (MUAC), World Health Organization human immunodeficiency virus (WHO HIV) disease stage, CD4 count, gestational age at enrollment, malaria infection, and treatment regimen.

RR = relative risk; CI = confidence interval.

Geophagy was not associated with the risk of low birthweight, preterm birth, being small for gestational age, or fetal death after adjusting for potential confounding factors (Table 5). Our findings did not suggest an association between geophagy and the risk of in utero or intrapartum transmission of HIV.

Table 5.

Association of geophagy with pregnancy outcomes and the risk of HIV transmission

	Geophagy
Endpoint	Yes, n/N (%)	No, n/N (%)	Unadjusted RR (95% CI)	P value	Adjusted RR (95% CI)*	P value
Pregnancy outcomes
Birthweight < 2500 g	28/228 (12.3%)	60/576 (10.4%)	1.18 (0.77, 1.80)	0.44	1.23 (0.77, 1.94)	0.38
Preterm birth (< 37 weeks)	64/248 (25.8%)	156/647 (24.1%)	1.07 (0.83, 1.37)	0.62	1.14 (0.85, 1.54)	0.37
Small for gestational age†	26/228 (11.4%)	64/576 (11.1%)	1.03 (0.67, 1.58)	0.91	1.05 (0.66, 1.68)	0.83
Fetal death‡	16/264 (6.1%)	35/682 (5.1%)	1.18 (0.67, 2.10)	0.57	1.09 (0.60, 1.98)	0.78
HIV transmission§
HIV-positive at birth	22/200 (11.0%)	38/508 (7.5%)	1.47 (0.89, 2.42)	0.13	1.55 (0.90, 2.65)	0.11
HIV-positive at 6 weeks¶	15/106 (14.2%)	48/307 (15.6%)	0.91 (0.53, 1.55)	0.72	0.92 (0.51, 1.67)	0.79

^*Multivariate Poisson regression models were adjusted for mid-upper arm circumference (MUAC), World Health Organization human immunodeficiency virus (WHO HIV) disease stage, CD4 count, gestational age at enrollment, malaria infection, and treatment regimen.

^†Small for gestational age is defined as birthweight below the tenth percentile of weight for gestational age.

^‡Fetal deaths include miscarriages and stillbirths.

^§HIV infection at birth (from birth to 21 days) or at 6 weeks (from 22 to 49 days) was assessed among liveborn babies.

^¶Among HIV uninfected at birth.

RR = relative risk; CI = confidence interval.

DISCUSSION

About 29% of HIV–infected pregnant women reported regularly eating soil in Dar es Salaam, Tanzania. Cross-sectional analyses showed strong correlations of geophagy with iron deficiency anemia and A. lumbricoides infection at baseline. We also found that soil consumption was associated with an increased risk of anemia from longitudinal analyses. Geophagy was not related to adverse pregnancy outcomes or the risk of HIV transmission.

Pregnant women living in Dar es Salaam, Tanzania commonly ate hardened clay soil sold in the local market. Previous studies in sub-Saharan Africa reported that women consumed an average of 30 to 50 g daily during pregnancy.^3,12,24 We characterized sociodemographic correlates of geophagy. Women’s age, education, and money spent on food were not related to the prevalence of geophagy. Most previous studies also found no association of age or socio-economic status with the prevalence of geophagy.^24,25 These findings confirm the notion that soil eating is a widely accepted practice among all women in sub-Saharan Africa.¹ Geophagy is a female practice that symbolizes reproduction and fertility in African societies. In Kenya, a majority of pregnant women reported that they ate soil simply because they enjoyed its smell and taste and the remaining said that because they were pregnant.¹ Nonetheless, we also found that marital status and occupation may have influenced the soil eating practice. Differences in marital status could be partly because husbands disapproved the practice, which lead to women avoiding the consumption of soil.

We observed higher prevalence of geophagy with increasing gestational age and this finding was consistent with previous studies.^1,12 Women increasingly started practicing geophagy around the middle of the second trimester. This may indicate women’s response to an increased physiologic need for iron because of an increase in red blood cell mass and growth of the fetus. However, this finding could also be a result of a culturally common practice during pregnancy. It is difficult to conclude from our study whether women had started eating soil because of a physiologic response, a cultural practice, or both.

We observed a negative association between malaria and geophagy. The negative association could be because women with malaria were sick and avoided eating soil at that time. Iron supplementation may increase severity of malaria among pregnant women, and one may speculate that those who were sick with malaria may have avoided iron-containing foods, in particular soil.²⁶ Nevertheless, it is uncertain that women regard soil as an iron-containing substance in this region.

We found that severe anemia and red blood cell characteristics suggestive of iron deficiency were strongly associated with the prevalence of geophagy. More importantly, geophagy was associated with decreased hemoglobin concentration and a marginally increased risk of anemia. The cross-sectional association of geophagy with anemia among pregnant women has been reported from Kenya, Sudan, and Namibia.^4,12,13 One of these studies also showed an association with lower serum ferritin concentrations.¹² These findings were also demonstrated among school children in cross-sectional studies.^27–29 Furthermore, a recent trial examined the effect of iron supplements on geophageous behavior among school children in Zambia.²⁹ The authors found that children did not stop eating soil despite iron supplements. They speculated that geophagy is not a consequence but most likely a cause of iron deficiency.

There is an ongoing controversy regarding whether soil eating is nutritionally beneficial or harmful, however, our study showed that it may be harmful and could result in anemia.³⁰ The practice of geophagy has been long thought to help supplement mineral nutrients and have detoxifying effects.³¹ The nutritional benefit has been speculated, because soil contains large quantities of macronutrients and micronutrients.^32,33 However, recent studies showed that soil can impair the absorption of micronutrients and cause micronutrient deficiency.^34,35 The authors replicated the gastrointestinal condition in vitro and demonstrated that soil can effectively bind and remove nutrients that were already present particularly iron and zinc.^34,36 Alternatively, the soil contains high levels of aluminium and a study showed that aluminium could reduce serum ferritin and deplete iron stores.³⁷ Eating soil can also physically damage the intestinal mucosa and reduce the absorption of nutrients. These possible mechanisms may explain the association between geophagy and the increased risk of anemia in our study.

Ascaris lumbricoides infection was strongly associated with geophagy, but no significant associations were noted between geophagy and other helminth infections, including hookworm, T. trichiura, and S. stercoralis. Because A. lumbricoides infection is less likely to cause anemia or any other physiologic stimulus that may lead to craving, our cross-sectional association suggests that A. lumbricoides infection may be a consequence of soil pica. One prospective cohort study examined intestinal helminth infections among pregnant women in Kenya.²⁴ Consistent with our results, they found that geophagy was associated with A. lumbricoides reinfection, but it was not associated with reinfection with hookworm and T. trichiura. However, one cross-sectional study in Zanzibar, Tanzania found no association with infection with A. lumbricoides, T. trichiura, or hookworm.²⁵ Association of geophagy with A. lumbricoides infection was also found among school children in Kenya, South Africa, and Guinea.^38–40 People become infected with A. lumbricoides and T. trichiura by ingesting the fully developed eggs, whereas people become infected with hookworm and S. stercoralis when larvae of the parasites penetrate through the skin. Unlike T. trichiura eggs, the albumin coating of A. lumbricoides eggs permits them to stick to soil longer and thus facilitates their ingestion in the process of soil consumption.²⁴

We found no adverse effects of geophagy on pregnancy outcomes. No studies have examined the effect of soil consumption on pregnancy outcomes in sub-Saharan Africa. Two small studies in Argentina and the United States, where women commonly consumed ice, found that these practices were not associated with pregnancy outcomes but associated with lower hemoglobin levels.^41,42 Because anemia continues to be a significant cause of adverse pregnancy outcomes in sub-Saharan Africa, more studies will be needed to examine the effects on pregnancy outcomes.⁴³

Our cohort consisted of HIV-positive women and the results may not be generalizable to HIV-negative women. Further confirmatory studies are warranted to prospectively examine the association between geophagy and anemia among pregnant women and also children. About one-half of school children are affected by anemia in Africa and 40–70% of children were reported to be regularly consuming soils.^27,29,39,40

In conclusion, we found that soil consumption is a common practice associated with an increased risk of anemia and helminth infection among HIV–infected pregnant women in Tanzania. Further efforts are needed to treat and prevent anemia among pregnant women through micronutrient supplementation, treatment of malaria and intestinal helminth infections, and recommending diverse diets rich in iron. In addition to those efforts, our results suggested that pregnant women should be advised for a potential risk of anemia and helminth infection from soil consumption during antenatal care.