Influence of the timing of malaria infection during pregnancy on birth weight and on maternal anemia in Benin

Although consequences of malaria in pregnancy are well known, the period of pregnancy in which infection has the highest impact is still unclear. In Benin, we followed up a cohort of 1,037 women through pregnancy until delivery. The objective was to evaluate the relationship between the timing of infection and birth weight, and maternal anemia at delivery. At the beginning of pregnancy, peripheral infections were associated with a decrease in mean birth weight (-98.5 g; P = 0.03) and an increase in the risk of anemia at delivery (adjusted odds ratio [aOR] = 1.6; P = 0.03). Infections in late pregnancy were related to a higher risk of maternal anemia at delivery (aOR = 1.7; P = 0.001). To fully protect the women during the whole pregnancy, already implemented measures (insecticide-treated nets and intermittent preventive treatment) should be reinforced. In the future, a vaccine against pregnancy-associated malaria parasites could protect the women in early pregnancy, which seems to be a high-risk period. Copyright © 2011 by The American Society of Tropical Medicine and Hygiene. The study took place in the district of Comé in the Mono province, located 70 km West from the economical capital of Benin, Cotonou (Figure 1). The climate is subtropical with two rainy seasons: from April to July, and from September to November. The annual rainfall is over 1,300 mm. The principal malaria vectors are Anopheles gambiae s.s., and Anopheles funestus. The setting is a high malaria transmission area with two peaks during the rainy seasons. The entomological inoculation rate ranges from 35 to 60 infective bites per person and per year.15 The predominant parasite species (97%) causing malaria in this region is Plasmodium falciparum. Study area (Charlotte Pierrat UMR216, IRD). In the study area, health care is provided through three health dispensaries, 11 private clinics, and a district hospital. Women were enrolled in the three dispensaries: Comé, Akodeha, and Ouedeme Pedah. Comé is a semi-rural site and the two other health centers are located in a rural setting. The principal occupations of the inhabitants are farming, fishing, and trading. The enrollment of pregnant women started in June 2008 and the last delivery occurred in September 2010. Local midwives provided clinical and gynecological examination. Five nurses were recruited and trained as “project assistants” to fill out questionnaires and to collect blood samples from the study participants. Midwives and project assistants worked in close collaboration. The inclusion criteria were gestational age under 24 weeks, living within 15 km from the dispensary for > 6 months, and having planned to deliver at the hospital. The study objectives were explained twice to the women: first by the midwife and then by the project assistant. Pregnant women were enrolled in the study after giving informed and signed consent. On the initial visit, assistants and midwives collected information regarding the reproductive history and the current pregnancy, medical history, socio-economic indicators, and the use of bed nets. Axillary temperature, blood pressure, weight (SECA scale, Hamburg, Germany), height, and the mid-upper arm circumference (MUAC) of mothers were measured. After clinical examination, rapid diagnostic tests (RDT) and thick and thin blood smears were systematically made and venous blood samples (3 tubes: ethylenediaminetetraacetic acid [EDTA] [2 mL], dry [2 mL], and citrate-phosphate-dextrose with adenine (CPDA) [4 mL]) were taken. Albumin and sugar in urine were measured with a dipstick (Uriscan 7strip, YD Diagnostics Corporation, Kyunggi-do, South Korea). According to the Beninese national recommendations, a kit including tablets of iron (200 mg to be taken daily during 1 month), folic acid (5 mg daily, 1-month treatment), mebendazole (500 mg during 3 days) for deworming, and an insecticide-treated net was given to the mother at enrollment. At each monthly ANV, symptoms experienced and illnesses having occurred since the last visit were noted. The same clinical and biological information were collected for inclusion. Following the national guidelines, two doses of sulfadoxine pyrimethamine (SP) (1,500 mg of sulfadoxine and 75 mg of pyrimethamine) given on the occasion of intermittent preventive treatment during pregnancy (IPTp) were administered at least 1 month apart in the second trimester of pregnancy under the supervision of midwives. Iron and folic acid tablets ensuring a daily intake for 1 month were given to the women at each ANV until delivery. Any participant with documented fever (axillary temperature ≥ 37.5°) and malaria infection assessed by RDT received a treatment dose of quinine, or SP if it was the scheduled visit for IPTp intake. In this case, a control based on a blood smear was performed 7 days later. In case of clinical symptoms between ANVs, women were asked to attend health facilities to get treatment. The same clinical and biological information were collected as during ANVs. These visits were identified as “unscheduled visits.” In case of an illness unmanageable by the medical team of the dispensary, the woman was sent to the district hospital to get appropriate care. Diagnostic and treatment received at the reference hospital were noted. Four ultrasound scans were planned and performed with a portable ultrasound system (Titan, Sonosite, Bothell, WA) by a midwife trained for ultrasound from August 2008 to March 2009 and by an obstetrician from April 2009 until the end of the study. The first scan aimed to determine the exact term of the pregnancy and the following to evaluate the intrauterine growth and fetal morphology. At delivery, temperature, weight, and blood pressure of the mother were measured. Venous blood samples, thick and thin blood smears, and an RDT were obtained from the mother before delivery. The placenta was measured and weighed. The macroscopic aspect of the placenta was recorded. Blood placental samples and three placental biopsies were collected and two placental smears (thick and thin) were made. In case of positive RDT, a placenta perfusion with phosphate buffered saline plus CPDA solution was performed to collect placental Plasmodium parasites. Blood samples from the umbilical cord were drawn. The newborn was examined clinically (APGAR score, icterus, and malformation) and anthropological measures were performed by the midwife: weight (electronic scale Seca), height, MUAC, head circumference, abdominal circumference, and foot length. The gestational age was assessed clinically by using the new Ballard score.16 For women delivering outside the study frame, birth outcomes were collected from the antenatal care book. Ballard score was assessed if the women presented at the study center within 2 days after delivery. For the biological detection of plasmodial infections, we used Parascreen RDT (Zephyr Biomedical Systems, Goa, India) which detects P. falciparum histidine protein-2 (Pf HRP-2) and Plasmodium-species lactate dehydrogenase (Psp LDH). Daily, all RDTs, thick and thin smears, and blood samples from the three health centers were collected, kept at 20°C, and sent to the laboratory in Cotonou. Thick blood smears were stained with Giemsa and read by two experienced parasitology technicians. Smears were considered negative if no asexual-stage Plasmodium parasite was detected after 500 leukocytes had been counted. Malaria parasites were counted against 200 leukocytes and parasite densities were estimated using leukocyte count of the hemogram. If results were discrepant, the slides were read by a third microscopist. Blood samples were centrifuged and frozen for further immunological analyses and blood drops were deposited on filter paper for parasite genotyping. Analyses to determine hemogram and to measure alanine aminotransferase and creatinine concentrations were performed. Data were double-entered, validated, and cleaned using Access (Microsoft, version 2003, Redmond, WA). Stata version 11 for Windows (Stata Corp., College Station, TX) was used for all statistical analyses. Plasmodium infection status was based on the results of thick blood smear. Low birth weight was defined as a birth weight < 2,500 g. We considered three periods during pregnancy: early pregnancy (0–4 months of gestation); mid-pregnancy (5–6 months), and late pregnancy (> 6 months). Because of the low number of women enrolled in the first trimester, we chose a 4-month limit for the first period of pregnancy instead of the more usual categorization by trimesters. The women’s body mass index (BMI) was defined as the weight after delivery (kg)/height squared (m2) and then transformed into a binary variable: under and over the median (< 21; ≥ 21). Maternal anemia was defined as hemoglobin concentration under 11 g/dL. We excluded twin newborns from the analysis. Our primary outcomes were the newborn's weight (LBW and decrease in mean birth weight) and maternal anemia at delivery. We first studied the association between each of these outcomes and the occurrence of a Plasmodium infection during the three gestational periods. We then considered other variables susceptible to have an influence on birth weight or maternal anemia, such as the infection of the placenta or the number of Plasmodium infections during the course of pregnancy (no infection, 1–2 infections, and more than 2 infections). Differences in proportions and means were compared using the khi2 (or Fisher's exact test) and the Student t test, respectively. We used a logistic or a linear regression depending on the variable analyzed (LBW, birth weight, or anemia). Covariates were included in the initial models on the basis of the literature and on hypothesized underlying causal relationships directed acyclic graphs (DAG) method.17,18 We considered the following covariates: duration of pregnancy (transformed into a four class variable corresponding to the quartiles), sex of the newborn, parity (primigravidae or multigravidae), rainy season at delivery, human immunodeficiency virus (HIV) status, number of SP intakes, education, use of bed net, BMI, number of consultations (sum of the ANVs and unscheduled visits). We used a backward strategy to obtain the final multivariate model, a P value of < 0.05 was considered statistically significant. During the first trimester, African women usually consult late for their first ANV; therefore, malaria status was missing for some women. At delivery, mostly because of a health workers strike in the area, some mothers' weights and malaria status (placental and peripheral smears) were missing. The HIV status was unknown for some women. To deal with missing data, we used the method of multiple imputations by chain equations (MICE)19 and we carried out 20 imputed datasets with 15 iterations for each. This study was approved by the ethics committees of the Research Institute for Development (IRD) in France and of the Science and Health Faculty (University of Abomey Calavi) in Benin. Written informed consent was given by all participants.