Reductions in malaria in pregnancy and adverse birth outcomes following indoor residual spraying of insecticide in Uganda

Tororo is a rural district in southeastern Uganda with an entomologic inoculation rate estimated at 310 infectious bites per person year in 2012 [21]. Following a universal LLIN distribution campaign in November 2013 [23], 95 % of household in the region reported owning at least one LLIN [24]. Between June and October 2014, 300 women were enrolled into a double-blinded, placebo-controlled trial of three-dose SP vs three-dose DP vs monthly DP for IPTp. Details of the parent study have been described elsewhere [22]. Briefly, participants were HIV-uninfected pregnant women at least 16 years of age of all gravidities with an estimated gestational age of 12–20 weeks confirmed by ultrasound. The sub-study described in this report includes all women followed through delivery (n = 289).

From December 2014 to February 2015, Abt, in cooperation with the Ugandan Ministry of Health, initiated IRS in Tororo district using bendiocarb wettable powder, a carbamate insecticide [23]. Homes were sprayed once and following spraying, houses were marked with the date of spraying. Home visitors obtained the exact date of spraying from each participant’s household. If a household was not sprayed, the reasons for a lack of spraying were obtained and if the surrounding households in the village were sprayed, the date of spraying was obtained from surrounding homes.

At enrolment, women underwent a standardized examination and received a LLIN. Study participants were randomized in a 1:1:1 ratio to three-dose SP vs three-dose DP vs monthly DP for IPTp, as previously described [22]. Women received all of their medical care at a study clinic open daily. Routine visits were conducted every 4 weeks, including collection of dried blood spots (DBS). Women were encouraged to come to the clinic any time they were ill. Those who presented with a documented fever (tympanic temperature >38.0 °C) or history of fever in the previous 24 h had blood collected for a thick blood smear. If the smear was positive, the patient was diagnosed with malaria and treated with artemether-lumefantrine.

Women were encouraged to deliver at the hospital adjacent to the study clinic. Women delivering at home were visited by study staff at the time of delivery or as soon as possible afterwards. At delivery a standardized assessment was completed, including evaluation for birth weight and collection of specimens, including placental tissue and DBS of placental blood.

DBS were tested for the presence of malaria parasites using loop-mediated isothermal amplification (LAMP) [25]. Placental tissues were processed for histological evidence of placental malaria as previously described [22]. Histopathology slides were read in duplicate using a standardized case record form by two independent readers and any discrepant results resolved by a third reader. Blood smears were collected in febrile women during pregnancy, stained with 2 % Giemsa and read by laboratory technologists. A blood smear was considered negative when the examination of 100 high power fields did not reveal asexual parasites. All slides were read by a second microscopist and a third reviewer settled any discrepant readings.

Entomological surveys were conducted monthly beginning in June 2011 from 100 households enrolled in a separate longitudinal cohort study in Nagongera sub-county, Tororo district, as part of the ICEMR programme in Uganda [21]. Each month, miniature CDC light traps (Model 512; John W Hock Company, Gainesville, FL, USA) were positioned with the light 1 m above the floor at the foot end of the bed where a cohort study participant slept. Traps were set at 19.00 and collected at 07.00 h the following morning by field workers.

Outcomes assessed during pregnancy included parasite prevalence by LAMP and the incidence of malaria, calculated as the number of episodes per person years (ppy) at risk. Outcomes assessed at birth included the prevalence of parasitaemia at delivery by LAMP and evidence of placental malaria (parasites or pigment) by histopathology. Placental histopathology was also classified according to whether moderate-high grade pigment deposition was present (defined as pigment detected in >5 % of high power 40× fields) [26]. Birth outcomes assessed included birth weight, LBW (<2500 g), pre-term delivery (<37 weeks), and fetal/neonatal deaths, including spontaneous abortion, stillbirth and neonatal death within 4 weeks of delivery. For women giving birth to twins, delivery outcomes were based on whether the outcome was present in either child/placenta.

Data were double entered into an Access database. Data analysis was done using Stata version 14 (Stata Corp, College Station, TX, USA). Baseline characteristics between groups were compared using the \(\chi^{2}\) test. Repeated prevalence measures and the daily risk of malaria during pregnancy were compared using generalized estimating equations with a log-binomial family, after adjustment for gravidity, age, IPTp arm, and gestational age when study drugs were started, and presented as adjusted risk ratios (aRR). For results stratified by IPTp arm, women randomized to either IPTp-DP arm were considered together. Dichotomous outcomes at delivery were compared using multivariate logistic regression. Continuous outcomes at delivery were compared using multivariate linear regression, with adjustment as above. As a secondary sensitivity analysis, propensity scores and inverse probability weighting were used to assess the average causal effect of IRS exposure with outcomes at delivery. All P-values were two-sided and values <0.05 considered statistically significant.

Between June and October 2014, 300 women were enrolled prior to IRS, and 289 women were followed through delivery from October 2014 to May 2015 (Fig. 1). From December 2014 to February 2015, 123,924 of 145,574 households (85.1 %) in Tororo district received IRS. Among women in the study cohort, 95 % of households were sprayed between 1 December, 2014 and 31 January, 2015, with the remaining sprayed in February 2015. Monthly measurements of female Anopheles mosquitoes collected per household per night in Tororo were lower from February to May 2015, after IRS, compared with February–May 2014 (5.4 vs 33.7 female Anopheles mosquitoes per house per night, P < 0.001) (Fig. 1).

Of 289 women followed through delivery, 155 (53.6 %) were exposed to IRS before the birth of their child; 137 of these women’s households were directly sprayed, and 18 households (11.6 %) were indirectly protected through spraying of the surrounding village. Of the 134 women not exposed to IRS during pregnancy, six lived outside of the IRS spraying area in neighbouring Busia district, and 128 were exposed to IRS after the birth of their child. The median duration of pregnancy under protection of IRS was 46 days (IQR 20–68 days), and women were stratified into three groups based on the proportion of pregnancy under the protection of IRS: 0 (n = 134), >0–20 % (n = 90), and >20–43 % (n = 65). At enrolment, the prevalence of malaria parasites was >55 % and similar between the three groups (Table 1). Women with no IRS protection during pregnancy were slightly younger, more likely primigravid, and more likely to have study drugs initiated at 20 weeks gestation vs 16 weeks gestation, than women with IRS protection during pregnancy. IPTp treatment assignments, maternal weight gain, and household wealth were similar between groups (Table 1).

Parasite prevalence at enrolment and before study drug initiation was stable and high prior to IRS (see Additional file 1). After study drug initiation, the incidence of malaria during pregnancy was significantly higher before IRS (0.49 episodes ppy) compared to after IRS (0.10 episodes ppy, aRR 0.20, P = 0.02) after adjustment for gravidity, age, IPTp arm, and gestational age when study drugs were started. Similarly, the prevalence of malaria parasites by LAMP during pregnancy was significantly higher before IRS (20.0 %) compared to after IRS (8.9 %, aRR 0.40, P < 0.001). Declines were observed in women randomized to both IPTp-SP and IPTp-DP, although there was evidence of interaction between IPTp and IRS (P = 0.07, Fig. 2). Among women randomized to IPTp-SP, the prevalence of parasitaemia was 44.0 % before IRS, before declining to 22.0 % after IRS (aRR 0.49, P = 0.003). Among women randomized to IPTp-DP, parasite prevalence was 8.7 % before IRS vs 1.1 % after IRS (aRR 0.13, P = 0.003). Women living in households directly sprayed vs those indirectly sprayed were both protected by IRS. Among women living in households directly sprayed, parasite prevalence was 20.2 % before IRS vs 8.5 % after IRS (aRR 0.42, 95 % CI 0.26–0.70, P = 0.001). Among women living in households indirectly sprayed, parasite prevalence was 34.2 % before IRS vs 5.0 % after IRS (aRR 0.12, 95 % CI 0.03–0.59, P = 0.009), consistent with community-wide benefits of IRS on parasitaemia [11].

At delivery, the prevalence of malaria parasites in placental blood by LAMP was significantly higher in women with no IRS protection (16.8 %) compared to women with >0–20 % (1.1 %, P = 0.001) or >20–43 % IRS protection during pregnancy (1.6 %, P = 0.006) (Table 2). The prevalence of any placental malaria by histopathology, as well as the prevalence of moderate to high-grade pigment deposition, was higher in women with no IRS protection compared with women with IRS protection (Table 2), but the differences were not statistically significant. All 105 placentas that were positive for placental malaria by histopathology had pigment in fibrin that was indicative of past infection, and seven (11 %) had parasites indicative of concomitant active infection. All seven with active infection were not exposed to IRS prior to delivery.

The prevalence of LBW was significantly higher in infants born to mothers with no IRS protection (20.9 %) compared to those with >0–20 % (10.0 %, P = 0.01) or >20–43 % IRS protection (3.1 %, P = 0.002) (Table 2). Compared to infants born to women with no IRS protection, mean birth weight was 196 g higher among infants born to women with 0–20 % IRS protection (95 % CI 51–340 g, P = 0.008) and 257 g higher among infants born to women with >20–43 % IRS protection (95 % CI 105–409 g, P = 0.001) (Fig. 3). When restricting the analysis to women without histologic evidence of placental malaria, the prevalence of LBW was similarly higher in infants born to mothers with no IRS protection (18.8 %) compared to those with >0–20 % (7.9 %, P = 0.07) or >20–43 % IRS protection (2.2 %, P = 0.03).

The prevalence of pre-term birth was significantly higher in infants born to women with no IRS protection (17.2 %) compared to those with >0–20 % (3.3 %, P = 0.01) or >20–43 % IRS protection (1.5 %, P = 0.006) (Table 2). Of 11 fetal/neonatal deaths (three spontaneous abortions, three still births, and five neonatal deaths), ten (7.5 % of pregnancies) occurred among those with no IRS protection, and none occurred among those with 20–43 % IRS protection (P = 0.03). Given the imbalances noted between IRS protection strata, a sensitivity analysis was performed to evaluate associations between IRS and outcomes at delivery using propensity scores with inverse probability weighting, with similar results (see Additional file 2). In all analyses, there was no significant interaction observed between IRS and assigned IPTp group.