Risk factors for Plasmodium falciparum infection in pregnant women in Burkina Faso: a community-based cross-sectional survey

Background Malaria in pregnancy remains a public health problem in sub-Saharan Africa. Identifying risk factors for malaria in pregnancy could assist in developing interventions to reduce the risk of malaria in Burkina Faso and other countries in the region. Methods Two cross-sectional surveys were carried out to measure Plasmodium falciparum infection using microscopy in pregnant women in Saponé Health District, central Burkina Faso. Data were collected on individual, household and environmental variables and their association with P. falciparum infection assessed using multivariable analysis. Results A total of 356 pregnant women were enrolled in the surveys, 174 during the dry season and 182 during the wet season. The mean number of doses of sulfadoxine–pyrimethamine for Intermittent Preventive Treatment in pregnancy (IPTp-SP) was 0.4 doses during the first trimester, 1.1 doses at the second and 2.3 doses at the third. Overall prevalence of P. falciparum infection by microscopy was 15.7%; 17.8% in the dry season and 13.7% in the wet season. 88.2% of pregnant women reported sleeping under an insecticide-treated net (ITN) on the previous night. The odds of P. falciparum infection was 65% lower in women who reported using an ITN compared to those that did not use an ITN (Odds ratio, OR = 0.35, 95% CI 0.14–0.86, p = 0.02). IPTp-SP was also associated with reduced P. falciparum infection, with each additional dose of IPTp-SP reducing the odds of infection by 44% (OR = 0.56, 95% CI 0.39–0.79, p = 0.001). Literate women had a 2.54 times higher odds of P. falciparum infection compared to illiterate women (95% CI 1.31–4.91, p = 0.006). Conclusions The prevalence of P. falciparum infection among pregnant women remains high in Burkina Faso, although use of IPTp-SP and ITNs were found to reduce the odds of infection. Despite this, compliance with IPTp-SP remains far from that recommended by the National Malaria Control Programme and World Health Organization. Behaviour change communication should be strengthened to encourage compliance with protective malaria control tools during pregnancy. Supplementary Information The online version contains supplementary material available at 10.1186/s12936-021-03896-8.

are more susceptible to malaria and its adverse effects than non-pregnant women and infection with Plasmodium falciparum can lead to poor outcomes for the mother, the fetus and child [2]. Pregnant women are also more attractive to Anopheles gambiae, the most important African malaria vector, due to increased release of carbon dioxide and attractive volatiles, and hotter bodies [3,4]. Pregnant women infected with P. falciparum can develop placental malaria, with sequestration of the parasite in the placental vasculature [5]. This can lead to maternal anaemia and death, pre-term delivery, stillbirth, low birth weight, with long-term impacts on child growth and cognitive development. In areas of stable transmission, primigravid women are at greatest risk of malaria infection and have higher parasite densities [6]. Over successive pregnancies there appears to be a boosting of immunity to P. falciparum that reduces parasite density and prevents disease [6]. Prevention and management of malaria in pregnancy is three-pronged: the use of insecticide-treated nets [ITNs, distributed free-of-charge at Antenatal Clinic (ANC) visits], intermittent preventive treatment in pregnancy (IPTp) with sulfadoxinepyrimethamine (SP) and prompt access to diagnosis and effective case management [7].
Burkina Faso is a high burden country and despite high coverage of ITNs and prompt and effective treatment with anti-malarials is not experiencing declines in malaria [8]. Malaria transmission intensity is extremely high with children aged 5-15 years in south-west Burkina Faso receiving on average 3 infective bites per week during the 6 month transmission season [9]. Additional reasons for stagnating progress in malaria control may also include insecticide resistance [10,11], outdoor biting of malaria vectors [12,13], insufficient coverage and use of ITNs, or rapid loss of ITNs due to holing [14]. Burkina Faso has a high burden of malaria in pregnancy with one study conducted in a rural health district in southwest Burkina Faso in 2014 reporting an incidence of 39.2 per 1000 women-months, with the burden of infection in primigravids more than twice as high as that in multigravids at 88.6 per 1000 women-months [15]. In 2014, another study in Bobo-Dioulasso city in south-west Burkina Faso, identified 18.1% of pregnant women had P. falciparum infection [16]. According to national guidelines in Burkina Faso, pregnant women are advised to receive at least three doses of IPTp-SP starting from the second trimester, with a minimum interval of 1 month between doses [17]. IPTp-SP is administered at ANC visits by facility-based health workers and is provided free of charge, along with ITNs.
There have been many studies of risk factors for malaria in pregnancy in sub-Saharan Africa, where increased risk was reported to be associated with younger age in pregnancy, primigravidae, first trimester of pregnancy, non-use of ITNs, lack of education and HIV co-infection [15,16,18,19]. Few, however, have evaluated socioeconomic and environmental risk factors for malaria in pregnancy. For example, recently a number of studies have shown that malaria in children is associated with poor housing [9,20,21], but it is not known whether this is also true for pregnant women. The goal of the present study was to identify risk factors for P. falciparum infection in pregnancy in Saponé Health District, central Burkina Faso, including potential socioeconomic and environmental risk factors, during the wet season and dry season. Identifying risk factors for malaria in pregnancy could assist in developing interventions to reduce malaria burden in pregnancy in Burkina Faso and other countries in sub-Saharan Africa.

Study design
Risk factors for P. falciparum infection were measured during two cross-sectional surveys, one at the beginning of the dry season in December 2018 and the second one at the end of the rainy season from September to October 2019.

Study site
The study was conducted within the Saponé Health and Demographic Surveillance system (HDSS), situated in the central region of Burkina Faso, 45 km south-west of Ouagadougou city, the capital of Burkina Faso (Fig. 1). The Saponé HDSS covers 1600 km 2 , with a total estimated population of 102,000 living in 83 villages and 10,841 compounds [22,23]. Compounds are georeferenced and a census of the population is conducted annually. Malaria transmission is intense and highly seasonal [24], with the peak of malaria transmission occurring at the end of the rainy season (June to October) and markedly reduced transmission during the dry season (December to May) [25]. The main vectors are An. gambiae sensu stricto, Anopheles arabiensis and Anopheles funestus, and P. falciparum accounts for > 95% of all malaria infections [24][25][26]. This is a rural area of open Sudanian savannah, where farming is dominant and the major crops grown are sorghum and millet. Houses in the study area are typically constructed with mud walls and floors, with thatched or metal roofs [27]. The population is served by 23 local health facilities and a district hospital.

Screening and enrolment
The HDSS census listings were not up-to-date and so study personnel worked closely with Community Health Workers who identified women of child-bearing age in the study villages. Home visits were conducted in 20 Yaro et al. Malar J (2021) 20:362 Fig. 1 Map of the study site villages in the HDSS. The 20 villages were selected purposively because of their close proximity to Ouagadougou and their accessibility. Villages closest to Ouagadougou along the axis of the National Road 7 were selected until the sample size was reached. Enrolment at home was adopted, rather than screening at the ANC because ANC attendance is relatively low in the study area, with only 35% of women attending the ANC at least three times [28]. All study procedures were carried out at the home of the study participant. Women thought to be pregnant were invited to take a dipstick pregnancy test, which was performed by fieldworkers at the woman's home. Women identified as pregnant, but who had not yet visited their ANC, were reminded to attend their local health facility. The study protocol and procedures were explained by trained staff to potential participants in French or the main local language of Moore.

Inclusion and exclusion criteria
Pregnant women were enrolled if they were between 15 and 40 years, provided written informed consent and agreed with the study procedures, including taking of blood. Women found to have clinical malaria were eligible for enrolment in the study. Pregnant women with a known history of SP allergy or any other medical condition that in the opinion of a study physician may be a threat to her or the fetus were not recruited into the study. Women who were pregnant at both surveys were eligible for enrolment in the study, but no women were enrolled where this was the case.

Parasitological assessment
Each participant provided a finger prick blood sample (100 µL) for malaria infection detection and characterization. Plasmodium falciparum quantitative sexual and asexual parasite count and qualitative species identification was performed by microscopy. Two blood smears were prepared and read by two independent experienced microscopists based at the Centre National de Recherche et de Formation sur le Paludisme (CNRFP) according to established standard operating procedures. Discrepancies in positive and negative reads and parasite counts differing by more than ten-fold between the two reads were resolved by the supervisor. In case of fever (axillary temperature ≥ 37.5 °C or reported fever in last 24 h) or other symptoms/signs of clinical malaria, a rapid diagnostic test for malaria (SD BIOLINE Malaria Ag P.f/Pan, Abbott Laboratories, Illinois, USA) was performed. Subjects presenting with clinical malaria were referred to the nearest health facility and treated according to national guidelines [17].

Risk factor data collection
All study participants completed a questionnaire (Additional file 1), where demographic data, medical and obstetric history, including previous ANC visits, IPTp-SP doses and use of anti-malarials within 14 days prior to study enrolment were recorded. ANC attendance and IPTp-SP doses were transcribed from the participants health card. Fieldworkers recorded information about the household, including whether the woman had access to an ITN and whether she slept under an ITN the previous night. Women sleeping under an ITN were asked about the bed net source and how many times they left their ITN during the previous night. ITN fabric integrity was also assessed by fieldworker observation and classified as entire/complete, with any hole, or torn. Women were asked to estimate the time they went to bed and the time they get out of bed in the morning. Social and economic risk factors for malaria were recorded, including ethnicity, education level and occupation, ownership of a radio or mobile phone, estimated distance to the nearest health facility, and use of other protective measures, including mosquito coils, insecticide sprays, traditional spatial repellent or commercial topical repellents.
House construction (metal or thatched roof, presence of open eaves, electricity supply to sleeping room), household size (number of persons) and the presence of clothes hanging in the sleeping room were recorded. The presence or absence of large domestic animals (donkey, horses, sheep, cows, goats, dogs) and rubbish within 5 m of each study participant's household was also recorded.

Sample size
Assuming a 20% prevalence of P. falciparum infection in pregnant women [16], a sample size of 346 women provided greater than 80% power to detect effect sizes of > 50% (odds ratio = 2) at the 5% level of significance, given a 50% prevalence of the risk factor of interest in the population. The sample size was calculated using the formula for calculation of sample size for un-matched crosssectional studies (www. opene pi. com/ Sampl eSize/ SSCoh ort. htm). In determining the effect size we considered housing type since improved housing has been found to reduce malaria prevalence by ~ 50% [29].

Data management and statistical analysis
Data were collected on Android personal digital assistants programmed using Open Data Kit (https:// getodk. org/) and included drop down boxes and consistency checks to reduce data entry errors. Following cleaning, the dataset was locked and saved in Microsoft Access and analysed with Stata 15 (Statacorp, Texas, USA).
Descriptive analysis was performed including calculation of percentages for categorial variables and mean and standard deviation for continuous variables. Geometric mean was calculated for asexual stage parasite density (geometric mean parasite density, GMPD). Categorical variables were compared using a Chi-squared test, and normally distributed continuous variables using a t-test. The primary outcome measure was the prevalence of microscopically confirmed P. falciparum infection in pregnant women during each cross-sectional survey. Logistic regression was used to investigate the association between independent variables (risk factors) and the primary outcome (dependent variable), adjusting for clustering by village. The wet and dry season surveys were analysed together with season evaluated as an independent risk factor. A multivariable model was constructed using a forwards stepwise process and models compared using a Wald test. Variables were tested in the multivariable model if p ≤ 0.05 in the univariable model. Correlations between variables were tested using a Pearson's coefficient (r) and variables that were correlated with r ≥ 0.5 or ≤ − 0.5 were omitted from the multivariable model. Interactions between independent variables were evaluated. Odds ratios and adjusted odd ratios with 95% confidence intervals were computed and Wald test p-values presented. P-values were not corrected for multiple comparisons.

Socio-demographic characteristics of pregnant women
A total of 356 pregnant women were enrolled in the surveys, 182 during the wet season and 174 during the dry season ( Table 1). The mean age of the study participants was 26.9 years, ranging from 15 to 40 years old, and was similar in both surveys. Of these women, 78 (21.9%) were in their first pregnancy, 74 (20.8%) in their second and 204 (57.3%) in their third pregnancy or more. Women aged under 20 years were primarily primigravidae ( Characteristics of the women surveyed in the dry and wet season tended to be similar, although in the wet season the distance to the nearest health centre was longer (40.2% > 3 km) than in the dry season (46.1% > 3 km), and in the dry season there were more likely to be clothes hanging in the sleeping room (80.5%) than in the wet season (49.5%).

Use of preventive measures against malaria
ANC attendance was higher in the wet season than in the dry season, with 20.3% (37/182) of women attending four or more times in the wet season, compared to 12.9%

Risk factors for P. falciparum infection
Univariable analysis indicated lower odds of P. falciparum infection among women who were in the 3rd trimester compared to the 1st trimester; aged 30-40 years compared to those aged less than 20 years; reported using ITNs; attended more ANC visits; and received more IPTp-SP doses. Univariable analysis indicated a higher prevalence of P. falciparum infection in literate women compared to illiterate women. Analysis of correlations between variables indicated that gestation was correlated with ANC attendance (r = 0.60) and number of IPTp-SP doses (r = 0.60), ANC attendance was correlated with number of IPTp-SP doses (r = 0.82) and gravidity was correlated with age (r = 0.59). Since women will receive additional IPTp-SP doses with increased ANC attendance and increased gestational age, we excluded gestation and ANC attendance from the variables considered. Based on this, reported ITN use, the number of IPTp-SP doses, education and age group were evaluated in the multivariable model. Multivariable analysis showed that the odds of P. falciparum infection in pregnancy was reduced among pregnant women who reported using ITNs (Odds ratio, OR = 0.35, 95% CI 0.14-0.86, p = 0.02) after adjusting for number of IPTp-SP doses, education and age group ( Table 3). The odds of P. falciparum infection was also reduced with use of IPTp-SP, with each additional dose reducing the odds by 44% (OR = 0.56, 95% CI 0. 39

Discussion
This study aimed to identify risk factors for malaria infection in pregnant women living in an area of intense and stable seasonal malaria transmission in Burkina Faso with high levels of pyrethroid resistance in malaria vectors. The overall prevalence of P. falciparum infection during both surveys was 15.7% and is similar to that recorded in other studies in Burkina Faso (e.g. 18.1% [16]) and other high burden countries in sub-Saharan Africa, e.g. 20.1% in Kenya [30] and 21.6% in Ghana [31]. These results suggest that P. falciparum malaria infection is common in pregnant women in the community and that malaria burden in pregnancy remains high despite the use of standard malaria control interventions. The overall geometric mean of parasites density in the study area was 777.3/ µl (95% CI 496.0-1218.2). Fana and co-workers from Nigeria, another high burden country, recorded a similar mean parasite density of 800/µl [19].
No difference in parasite prevalence by gravidity was observed, which differs from other studies, which found higher prevalence in primi-and secundigravidae, compared to multigravidae [32,33]. This may be because in our study setting pregnant women are, irrespective of gravidity, exposed to same high level of transmission, and hence the probability of getting infected is quite similar. Parasite density was, however, higher in primigravidae and secundigravidae compared to multigravidae and in younger women compared to older women, since younger women are more likely to be primigravid. This may be because primigravid and secundigravid women are less immunologically competent to control parasite density compared to multigravidae. There was no  the wet season (25.8%), and ANC attendance was lower in the dry season. It may be that women perceive a lower risk of malaria transmission in the dry season due to less mosquito biting pressure. ITNs were associated with 65% reduction in the odds of P. falciparum infection, which is higher than other studies have found [34]. This indicates that ITNs are protective against malaria in pregnancy in Burkina Faso despite high levels of insecticide resistance present in the country [10,35]. This contrasts with findings from two studies conducted in south-west Burkina Faso; a cohort study in children aged 5-15 years in which showed no difference in malaria risk between ITN users and non-users [9], and a cross-sectional survey in all ages which found no difference in infection risk between ITN users and non-users (Yaro et al., unpublished). Overall, 91.3% of pregnant women owned an ITN, with 88.2% reporting using an ITN the night before the survey. This is similar to other surveys from Burkina Faso; in the Banfora Region, 80.6% of surveyed children reported sleeping under an ITN the previous night [9]. The high reported ITN use is encouraging, although accurately determining net use is challenging and reporting can be susceptible to response bias [36]. Women reported going to bed at 20.21 h during the dry season and 20.13 h during the rainy season. This finding contrasts with a study by Guglielmo and co-workers who reported that all females in south-west Burkina Faso (sample of 211 and 695 females observed in two villages) were outdoors until 22.00 h, after which point women started to move indoors to bed [12]. It may be that pregnant women tend to go to bed earlier and so those using ITNs are more likely to be protected from vector biting during the early evening which has been observed in Burkina Faso [12].
The finding that literate women had 2.54 times the odds of P. falciparum infection compared to illiterate women was unexpected. While literature on this is mixed, with some studies showing an association between literacy and reduced infection [32] and others showing no such association [33,37], to the best of our knowledge there are no other studies which found a higher odds of infection in literate women. There was no association between literacy and ITN use, ANC attendance or IPTp-SP doses, suggesting that literacy is not acting via reduced uptake of preventive measures. Again here, existing literature is mixed with some studies finding an association between increased literacy and increased use of ITNs and IPTp [38,39], others finding, as this study did, no such association, [40,41] or an inverse association, including a study of malaria indicator surveys from eight African countries [42]. Literate women were more likely to be primigravid than illiterate women, however, this cannot explain the finding since gravidity was not associated with P.
falciparum infection in the model. It may be that literate women have higher risk occupations or are spending more time outdoors, potentially subject to early evening vector biting [12]. This finding requires further assessment to understand the mechanism by which literacy can increase infection risk.
A strength of this study is that it adopted communitylevel recruitment of pregnant women. This was an unusual approach since most studies assess risk factors for P. falciparum infection among those women attending ANC. The study population was probably more representative of the true situation since it also included women that did not necessarily know they were pregnant or women who had not yet attended an ANC. The lack of sampling frame from which to select the women of childbearing age may have, however, introduced bias in the selection of the study participants. A limitation of the study is the sample size which may have been insufficient to identify minor risk factors such as house construction details which may have been less common in the study population than assumed or been protective to a smaller extent than the study was powered for. Study villages were selected purposively based on their accessibility to Ouagadougou which may have meant that the study participants were not representative of the Saponé HDSS, perhaps due to differences in socioeconomic status if living closer to the A7 road afforded better economic opportunities than the more isolated rural villages. While we did collect information on asset ownership, this did not allow us to determine a wealth index or socioeconomic position, which may have been an important determinant of P. falciparum infection. We did not collect information on knowledge, attitudes and practices in relation to malaria in pregnancy and use of protective measures. This is an area for future research and will be informative for improvement of delivery of existing interventions and for design of novel interventions. Multiple comparisons were made and, therefore, some of the associations noted may have occurred by chance alone.
What are the implications of this research for control of malaria in pregnancy in Burkina Faso? Behaviour change communication is necessary to ensure high ANC attendance and compliance with IPTp-SP and ITN use. Based on the study findings, messaging for ANC attendance in the study area should target primigravid women due to their higher GMPD, literate women due to apparent higher infection risk and encourage early attendance irrespective of season. As is common in sub-Saharan Africa, pregnant women are often unaware that they are pregnant and so do not attend or are unwilling to attend an ANC in the early stages of pregnancy. An association between early ANC attendance and a higher average number of IPTp-SP doses has been demonstrated in several studies [43][44][45]. One option to increase IPTp-SP coverage is community delivery by community health workers, rather than ANC. This delivery route has been shown in a clinical trial in Burkina Faso to increase IPTp-SP compliance from 2.1 to 2.8 doses in the community delivery study arm with no apparent decrease in ANC attendance [46]. Seven women reported receiving IPTp-SP despite being in the first trimester. IPTp-SP is not recommended in the first trimester due to concerns over congenital abnormalities [47]. If this is a true finding, then increased health worker sensitization is necessary. While ultrasound can determine gestational age with precision, it is often unaffordable or inaccessible for many women in Burkina Faso.

Conclusion
The prevalence of P. falciparum infection among pregnant women was 15.7% despite wide deployment of ITNs and access to IPTp-SP. Nonetheless, the odds of P. falciparum infection was reduced by 44% for each additional IPTp-SP dose taken and women who reported using an ITN had a 65% lower odds of P. falciparum infection. These findings suggest that IPTp-SP and ITNs use are effective at reducing malaria infection in pregnant women living in malaria high burden countries, even where there is high insecticide resistance, but that research is needed to increase uptake of IPTp-SP. The finding of literate women having 2.54 times the odds of P. falciparum infection compared to illiterate women requires further research to elucidate the underlying mechanism.