Prevalence and risk factors of preconception anemia: A community based cross sectional study of rural women of reproductive age in northeastern Tanzania

Background Anemia is a major public health problem that adversely affects pregnancy outcomes. The prevalence of anemia among pregnant women before conception is not well known in Tanzania. The aim of this study was to determine the prevalence, types, and risk factors of preconception anemia in women of reproductive age from a rural Tanzanian setting. Methods Trained field workers visited households to identify all female residents aged 18–40 years and invited them to the nearby health facility for screening and enrolment into this study. Baseline samples were collected to measure hemoglobin levels, serum ferritin, vitamin B 12 , folate, C-reactive protein, alanine amino-transferase, the presence of malaria, HIV, and soil transmitted helminth infections. Anthropometric and socio-economic data were recorded alongside with clinical information of participants. Logistic regression analysis was used to determine the adjusted odds ratios (AOR) for the factors associated with preconception anemia. Findings Of 1248 women enrolled before conception, 36.7% (95% confidence interval (CI) 34.1–39.4) had anemia (hemoglobin <12 g/dL) and 37.6% (95% CI 34.9–40.4) had iron deficiency. For more than half of the anemic cases, iron deficiency was also diagnosed (58.8%, 95% CI 54.2–63.3). Anemia was independently associated with increased age (AOR 1.05, 95% CI 1.03–1.07), malaria infection at enrolment (AOR 2.21, 95% CI 1.37–3.58), inflammation (AOR 1.77, 95% CI 1.21–2.60) and iron deficiency (AOR 4.68, 95% CI 3.55–6.17). The odds of anemia were reduced among women with increased mid-upper arm circumference (AOR 0.90, 95% CI 0.84–0.96). Conclusion Anemia among women of reproductive age before conception was prevalent in this rural setting. Increased age, iron deficiency, malaria infection and inflammation were significant risk factors associated with preconception anemia, whereas increased mid-upper arm circumference was protective against anemia. Interventions to ensure adequate iron levels as well as malaria control before conception are needed to prevent anemia before and during pregnancy and improve birth outcomes in this setting. The study received ethical approval from the Medical Research Coordinating Committee of the National Institute for Medical Research (reference number NIMR/HQ/R.8a/Vol. IX/1717).Written informed consent or thumbprint (for illiterate women) was obtained prior to enrolment. All study procedures were performed according to good clinical and laboratory practices and the Declaration of Helsinki [19]. This cross sectional study was conducted as part of a community-based epidemiological study entitled “Foetal exposure and epidemiological transition: the role of anemia in early life for non-communicable diseases in later life” (FOETALforNCD) fromJuly 2014 to December 2016 in Korogwe and Handeni districts, Tanga region, Tanzania. The aim of theFOETALforNCD project was toevaluate fetal growth alterations, placental development, and newborn susceptibility to non-communicable diseases in later life, following exposure to maternal anemia before and during pregnancy. The study population composed of women of the reproductive age. The analyses presented here utilized baseline data from women enrolled before they became pregnant. Inclusion into this study was based on their likelihood to conceive during the study period. To be included, women had to be aged 18–40 years, not be using modern contraceptive methods (except condom), or not be sub-fertile (defined as failure to conceive for two or more consecutive years for women who were trying to become pregnant), or not be pregnant at the time of enrolment (negative urine pregnancy test, HCG Vista Care Company, Shandong China), or not have a baby less than nine months old and live in an accessible area, and be willing to receive antenatal care and deliver at Korogwe District Hospital. Different stakeholders including village leaders, health care providers, opinion makers as well as community members were sensitized about the study goals and aims through village and health facility meetings prior to the implementation of the FOETALforNCD study. The primary means of identifying and recruiting eligible women was through contact at the household level within each village. Trained field workers made door-to-door visits to each household to explain the study, enumerate all women of reproductive age, and issue invitation cards for them to visit the nearby health facility for screening and enrolment. Other awareness and recruitment strategies included regular home visits by trained field workers (to identify new women moving into established households) and screening women as they sought other health care services. Eligible women were informed that after conception, the intention was to follow them throughout pregnancy until delivery. Upon conception transabdominal ultrasound (5–2 MHz abdominal probe, Sonosite TITAN and Sonosite Turbo, US High resolution, Sonosite, Bothell, WA, USA) was used to estimate gestational age (GA). Gestational age estimation was based on measurement of crown rump length in the first trimester [20] and head circumference in the second trimester [21]. From July 2014 to December 2015, 2629 women were screened for eligibility for inclusion into the FOETALforNCD study and 1415 were enrolled. Of the 1214 exclusions, 313 (25.8%) were not eligible by age, 322 (26.5%) were still using modern family planning methods, 116 (9.6%) were sub-fertile, 208 (16.6%) were already pregnant, 34 (2.8%) refused, 51(4.2%) migrated out of study area and 93 (7.7%) had a child <9 months old, while 77 (6.3%) were excluded due to other reasons. Of the 1415women included, 72 were later excluded because venous blood was not collected, and 11 did not fulfil the inclusion criteria. Furthermore, 84 women who conceived during the follow up were excluded because; 34 were already pregnant at enrolment based on the ultrasound estimated GA and 50 had a miscarriage before the GA could be ascertained, leaving 1248 women for the present analysis (Fig 1). Socio-demographic data including age, educational level, marital status, and economic (household size, house ownership and type of roofing materials, main source of drinking water and its ownership (private or public),type of toilet facility) and lifestyle factors (smoking, alcohol and tea consumption)were collected using a structured questionnaire. Previous medical histories, including gynecological and obstetric details, were documented. In order to define SES, a principal component analysis was applied and the variables which showed relevant contribution (greater than 10%) to the combined SES score were regarded as the ones which sufficiently described the SES of a woman [22]. Variables included in the final principal component analysis were educational level, occupation, type of house ownership, roofing materials, source of domestic water and its ownership as well as the type of toilet facility. The respective SES scores were categorized in tertiles as low, medium and high. Weight (in kilograms) was measured while on barefoot and wearing light clothes (precision 0.1kg, digital weighing scales, SecaGmbh& Co. Kg, Hamburg, Germany). Height in centimeters (cm) was measured with a stadiometer (precision 1 cm) [23]. Mid-upper arm circumference (MUAC) was measured on the upper right arm at the midpoint of the acromion process and the tip of the olecranon (precision 1mm). For measurement of skinfold thickness, trained staff pinched the skin above triceps muscle group to raise a double layer of skin and the underlying adipose tissue without the muscle. The HARPENDEN skinfold caliper (BATY International, England) was then applied 1 cm above and at right angle to the pinch, and a reading in millimeters (mm) taken after a few second. Waist circumference was measured just above the iliac crest in the horizontal plane, and hip circumference was measured at the point yielding the maximum circumference over the buttocks, all using a standard measuring tape to the nearest 1mm[24]. At enrolment, 15ml of venous blood was collected in ethylenediamine tetra acetic acid coated and plain serum tubes, transported at 2° to 8°C to the NIMR Korogwe Research Laboratory and processed within two hours of collection. To avoid photo degradation during transportation, all plain tubes were wrapped in aluminium foil. Separated serum samples were stored at -80°C and later shipped in dry ice to University Hospital Sealand, Denmark for micronutrients analysis. Hemoglobin level was measured by using Sysmex KX-21N hematological analyzer (Sysmex Corporation Kobe, Japan).According to WHO’s definition, anemia was defined as Hb<12.0 g/dL, and further categorized as mild (10.1–11.9 g/dL), moderate (8.0–10.0 g/dL) and severe (<8.0 g/dL) [3]. Microcytosis was defined as mean corpuscular volume (MCV) value <80 fL and hypochromic as mean cell hemoglobin concentration (MCHC) value <32 g/dL. Anemia was further classified as normocytic-normochromic (Hb<12 g/dL, MCV 80–100 fL and MCHC>32 g/dL), microcytic hypochromic (Hb<12 g/dL, MCV< 80 fL and MCHC<32 g/dL), megaloblastic (Hb<12g/dL, MCV≥100) or as mixed types (normocytic-hypochromic, microcytic-normochromic macrocytic-normochromic and macrocytic-hypochromic) anemia. For clinical care, Hb levels were measured using HemoCue 301 Hb analyzer (HemoCue AB, Angelholm, Sweden). Anemic women received treatments as follows: mild anemic (Hb10.1–11.9 g/dL) women with no symptoms received dietary counseling whereas women with symptoms were offered one combination tablet of 200 mg ferrous sulfate (~ 43 mg elemental iron) and 400μg folate per day (Ferrolic–LF, Laboratory and Allied LTD, Mombasa, Kenya). Moderately anemic patients with Hb 9.1–10.0g/dL received 2–3 combination tablets of iron and folic acid (Ferrolic–LFLaboratory and Allied LTD, Mombasa, Kenya) per day and monitored at each scheduled visit. Those with Hb 8.0–9.0 g/dL received a daily dose of 20 mL Hemovit multivitamin syrup (200 mg Ferrous sulfate, 0.5mg B6, 50 μg B12, 1500 μg Folic acid and 2.33mg Zinc per 5mL, Shelys Pharmaceuticals, Dar es Salaam, Tanzania) and monitored at each scheduled visit. Severely anemic (Hb<8g/dL) women were referred to Korogwe District Hospital for further evaluation. Serum ferritin, vitamin B12, folate, alanine aminotransferase (ALT) and bilirubin levels were measured using Dimension Vista 1500 biochemical analyzer (Siemens Healthcare Diagnostics, Inc, New York, USA) [25]. C-reactive protein (CRP) was measured by using Afinion AS 100 analyzer (Axis Shield PoC, AS, Oslo Norway) and inflammation defined as serum CRP >5 mg/L and/or ALT >45 U/L [10]. To account for elevated serum ferritin due to sub-clinical infection and other inflammatory conditions, three approaches were applied to define ID and results compared. The first approach utilized arithmetic correction factor (CF)as proposed by Thurnhamet al.[26] to adjust for the increased serum ferritin levels due to inflammation. In this approach CF of 0.67 was applied only for samples that had evidence of inflammation (CRP>5 mg/L), and a cut-off of <15 μg/L was then applied to the adjusted ferritin levels to define ID. If CRP was not available, serum ferritin was coded as missing. In the second approach, a higher ferritin-cutoff (<30μg/L) was applied to the subset of individuals with inflammation (CRP >5 mg/L) to define ID, as proposed by the WHO [27]. In this approach ID was defined as serum ferritin <15μg/L (no inflammation) or 30μg/L (inflammation present). The third approach utilized the higher serum ferritin cutoff (<30μg/L) if CRP>5mg/L and/or ALT>45 U/L which is considered as a sign of liver disease [10]. In the core analyses presented here, ID anemia was defined as Hb<12g/dL in the presence of ID based on Thurnham approach. Vitamin B12 deficiency was defined as serum cobalamin <150 pmol/L, and folate deficiency as serum folate<10 nmol/L without adjusting for inflammation [28]. Malaria was diagnosed using malaria rapid diagnostic test (mRDT) kit, ParaHIT (span diagnostics, Gujarat, India) or CareStart Malaria Pf (HPR2), ACCESS BIO, New Jersey, USA) according to manufacturer instructions. In addition, thick and thin blood films were prepared for the detection and quantification of parasitemia. Malaria patients received oral artemether-lumefantrine, (Lumartem 20mg/120mg (Cipla Ltd, Patalganga, India), quinine or artesunate injections according to Tanzanian standard treatment guideline. Human immunodeficiency virus infection was tested by using DetermineHIV-1/2 test kit (Alere ltd, Stockport, UK) and seropositive cases were confirmed using Unigold test kit (Trinity Biotech Plc, Wicklow, Ireland) according to the manufacturers’ instructions. Newly diagnosed HIV patients were referred to the nearby care and treatment clinics for long-term care. Considering low prevalence of STH infestations in north eastern Tanzania [29], stool samples were collected from a subgroup of 434 women at the time of enrolment and preserved in 10% neutral buffered formalin solution. Formol-ether concentration technique was used to detect presence of STH infestations [30]. All confirmed (on stool samples)or clinical suspected cases of STH infestation received a single dose of albendazole (400mg) or mebendazole (500mg) tablets according to the existing Tanzanian standard treatment guideline. Microsoft Access software 2007 (Microsoft corporation, Redmond’s, USA) was used for data entry and validation. Stata version 13 (StataCorp, Lake Way drive, College station, USA) software was used for statistical analyses. Continuous variables were visually inspected for normality using histograms and described using mean and standard deviation if normally distributed or median (interquartile range—IQR) for skewed data. Univariate analysis was done using Student’s t-test or Mann-Whitney test for continuous parametric and non-parametric variables, and Chi-square (χ2) or Fisher's exact test for categorical variables. Factors associated with preconception anemia were determined using logistic regression analysis and expressed as unadjusted odds ratio (OR) and adjusted odds ratios (AOR). All variables with P-value <0.20 in the univariate analysis were entered into the multivariate models [31]. Using a stepwise backward elimination approach final models including variables with a P-value <0.10 were obtained. A P-value of <0.05 was considered statistically significant. Due to missing data on HIV infections in 350 (28.8%) women and considering HIV infection being an important confounding factor, two different models with and without adjusting for HIV infection were generated and compared. Finally, in order to illustrate the association between a risk factor and anemia, trendline figures were generated for each continuous risk factor found to be statistically significant or borderline significant in the multivariate model.