This study was conducted from September 2016 to June 2018, in the relatively high-malaria-burden districts Kamonyi and Huye, in Southern Province, Rwanda. The incidence of confirmed malaria cases during 2018 was 499 and 414 cases per 1000 population in Kamonyi and Huye districts, respectively, ranking them 7th and 11th highest among the 30 districts in Rwanda (MOPDD, unpublished data). IRS with a carbamate insecticide was conducted in April 2017 in Huye, but not Kamonyi District, with coverage of 96.6%; residual efficacy is expected for at least 4–5 months after spraying. All public health centers in the districts (8 in Huye, 6 in Kamonyi) were included. Within each district, facilities were pair-matched based; 1 facility in each pair was randomly allocated to ISTp and the other to the control arm.

Sample size was calculated using Stata version 13 software (StataCorp, College Station, Texas) for 2-sample comparison of proportions to detect a 50% reduction in the prevalence of placental malaria between control (estimated 10% prevalence) and intervention groups (5% prevalence), accounting for clustering by facility. Assuming a statistical significance of 0.05, power ≥0.80, and intracluster correlation coefficient of 0.005, an estimated 714 women were needed per arm. Allowing for 20% loss to follow-up, 893 women per arm, a total of 1786 participants, were targeted for enrollment.

Pregnant women presenting for their first ANC visit were consecutively enrolled. Eligibility criteria included age ≥18 years, residence in the study area, willingness to have a supervised delivery, and providing signed informed consent. In control health centers, an ITN was provided and baseline hemoglobin (Hb) concentration was determined at first ANC, while clinical screening for malaria (ie, asking about history of fever in the past 48 hours) was performed at each visit. Only women reporting fever were tested (by microscopy); treatment was provided according to national guidelines if microscopy positive. At ISTp health centers, the same care was provided, except all women were tested for malaria with an RDT (SD BIOLINE Malaria Ag P.f./PAN [05FK67] RDT, Standard Diagnostics, Yongin-si, Gyeonggi-do, Korea) [16] at each regularly scheduled visit, regardless of symptoms, and treated per national guidelines if the RDT was positive (Supplementary Methods). At the time of delivery, maternal peripheral and placental blood samples were collected on filter paper for malaria testing in all facilities. Birth outcome, birthweight, and prematurity were recorded (Supplementary Methods). Peripheral blood was tested for Hb concentration (HemoCue Hb201+; HemoCue AB, Ängelholm, Sweden). Peripheral and placental dried blood spots were tested for malaria and human actin using TaqMan polymerase chain reaction (PCR) [17].

Maternal anemia was classified as any (Hb <11 g/dL) or moderate to severe (Hb <10 g/dL) [18]. Prematurity was defined as delivery <37 weeks’ gestation, and LBW as <2500 g. Analyses were done using SAS version 9.4 software (SAS Institute, Cary, North Carolina) (proc genmod) and R4.1.2 (R Foundation for Statistical Computing, Vienna, Austria) (R package “gee”) [19] accounting for facility-level clustering; P values <.05 were considered statistically significant. Relative risks (RRs) were derived from a log-binomial regression model, mean differences were calculated with linear regression. Study arm (based on cluster allocation), gestational age at enrollment, self-reported treatment for malaria prior to enrollment, fever during pregnancy (≥1 episode; self-reported), IRS in the 6 months prior to or during the study, age, education, marital status, human immunodeficiency virus, total ANC visits, ITN use, and district were assessed as potential covariates and retained if significant. Gravidity was included in all models given the known association with maternal malaria infection and infant outcomes [10, 12]. Models for maternal malaria were adjusted for study arm, gravidity, total ANC visits, fever during pregnancy, and IRS. Maternal anemia outcome models were adjusted for study arm, gravidity, baseline Hb level, and IRS. Infant models were adjusted for study arm, gravidity, baseline Hb level, treatment for MIP prior to enrollment, and IRS; the model for LBW was additionally adjusted for number of ANC visits. Attempted inclusion of number of ANC visits to the preterm model rendered the model unstable. The number needed to test to identify 1 positive case among afebrile women was calculated by dividing the number of cases by number of women tested.

Overall, 1786 pregnant women were enrolled, 975 in the intervention arm (ISTp) and 811 in the control arm. Of these, 1688 (94.5%) completed the study: 94.4% in the intervention arm and 94.7% in the control arm (Figure 1). Baseline characteristics did not differ significantly between the 2 study arms, except for self-reported treatment for malaria during the current pregnancy (9.6% in the ISTp group vs 7.0% in the control group, P < .001) and bed net utilization the night before the first ANC visit (93.7% and 98.4% for ISTp and control, respectively, P < .001; Table 1). While mean age did not differ between groups, there was a significant difference in age categories, with more women aged 20–34 in the control arm (76.8% vs 71.7%, P = .044).

Among women with delivery data, 72.0% and 66.8% of those in the ISTp and control groups, respectively, attended ≥3 ANC visits, with a mean of 2.98 visits for women in the ISTp arm versus 2.87 visits in the control arm (P = .011).

Among women in the ISTp arm, RDT positivity decreased from 10.6% to 3.2% from the first to fourth ANC visit (Table 2). Approximately half of all women with a positive RDT reported fever in the preceding 48 hours. RDT positivity among those without reported fever was 5.6% at first ANC, 2.8% at second ANC, and 1.1% at fourth ANC. The prevalence of self-reported fever in the preceding 48 hours also decreased with each successive ANC visit from 7.0% at first visit to 2.5% by the fourth visit. The number of afebrile persons needed to be tested to diagnose 1 person with malaria increased from 17.8 persons initially to 91.3 persons by fourth ANC.

Among all women with delivery data, 12.8% of women enrolled in the intervention group received an artemisinin combination therapy (ACT) during the study compared to only 2.2% in the control group. Only 14 women in the intervention arm received >1 course of ACT, and the maximum any woman received was 2 courses. None of the women in the control group received >1 course of ACT. The risk of LBW was not associated with receipt of ACT treatment.

ISTp was not associated with detection of either placental (15.0% and 15.3% in the ISTp and control arms, respectively; aRR, 0.94 [95% confidence interval {CI}, .59–1.50]; P = .799) or peripheral (10.5% vs 10.3%; aRR, 0.96 [95% CI, .69–1.34]; P = .825) malaria at delivery by PCR or peripheral malaria by RDT compared to control (4.5% vs 4.4%; aRR, 0.91 [95% CI, .34–2.43]; P = .853) (Table 3). Compared to PCR, the sensitivity and specificity of RDT at delivery were 21.6% and 98.2%, respectively (Supplementary Results). ISTp was also not associated with any significant difference in any anemia (aRR, 1.08 [95% CI, .57–2.04]; P = .821) or moderate to severe anemia (RR, 0.87 [95% CI, .37–2.08]; P = .760) at delivery.

Fever during pregnancy was associated with an increased risk of placental malaria at delivery (aRR, 1.90 [95% CI, 1.32–2.74]; P = .001), whereas IRS was associated with a decreased risk (aRR, 0.33 [95% CI, .23–.48]; P < .001) (Table 4). Having had 1 previous pregnancy reduced the risk of moderate to severe anemia (aRR, 0.37 [95% CI, .21–.68]), as did higher baseline Hb (aRR, 0.71 [95% CI, .62–.81]). IRS was marginally associated with a reduced risk of moderate to severe anemia (aRR, 0.43 [95% CI, .18–1.02]) (Table 4).

There were 930 newborns (including 20 twins) delivered in the intervention arm and 776 newborns (including 16 twins) delivered in the control arm. Excluding twins and stillborn infants, there were 897 live singleton newborns in the intervention arm and 747 in the control arm. The mean (standard deviation) birthweight of live singletons did not vary significantly between study arms (3054 g [448] vs 3096 g [443] for the ISTp and control arms, respectively; P = .395; Table 3). Compared to routine care, ISTp was associated with a higher proportion of LBW (7.3% vs 4.4%, P = .035) and preterm births (8.1% vs 5.1%, P = .207). In adjusted analyses, women in the ISTp group remained at a statistically significantly higher risk of LBW (aRR, 1.59 [95% CI, 1.02–2.49]; P = .042); while the point estimate for preterm birth was similar, this remained nonsignificant (aRR, 1.48 [95% CI, .79–2.76]; P = .219). Treatment for malaria before enrollment was associated with a statistically significantly increased risk of LBW (aRR, 1.54 [95% CI, 1.03–2.30]; P = .035; Table 5).

In districts in Rwanda with moderate malaria transmission, ISTp, compared to testing only symptomatic women for malaria, did not provide protection against malaria infection or anemia at delivery. Newborn outcomes were not improved by ISTp compared with routine ANC; on the contrary, women who received ISTp were more likely to have LBW infants than those who received routine care. That ISTp had no impact on placental malaria or birthweight was surprising, as one would expect that routine testing would have a greater impact than testing only symptomatic women. Prior studies suggested that ISTp performs similarly to IPTp-SP regarding the prevention of maternal malaria and adverse newborn outcomes [11, 13, 14], although not against febrile episodes during pregnancy [4, 10, 12]; as IPTp is better than placebo, one would expect that ISTp would provide benefit over passive screening for malaria.

The paradoxical failure to detect a reduction in malaria at delivery and the lower birthweight in the ISTp arm is likely related to greater exposure to malaria, indicated by the higher proportion of women reporting malaria prior to enrollment as well as the higher number of fever cases reported in the ISTp arm. Although facilities were randomized, the relatively small number of facilities may not have adequately eliminated confounding with respect to discrete hotspots of malaria transmission.

Alternatively, that half of all RDT-positive women had fever might indicate that in this population, asymptomatic malaria is rare, limiting the utility of universal screening. Although the areas chosen had higher burden of malaria than other areas in Rwanda, they have a much lower burden of malaria than other areas where ISTp has been studied, where malaria at delivery ranged from 16.9% to 38.2%, versus 15.3% [15]. Due to the lower prevalence of malaria in Rwanda, women have less pre-existing immunity; thus, a high proportion of cases were symptomatic. It is likely that ISTp would be more beneficial if the proportion of asymptomatic, RDT-positive women were higher. Furthermore, the overall proportion of women with fever was only 7% at the first visit and subsequently declined.

It is possible that more sensitive diagnostic tests would improve the efficacy of an ISTp strategy. In this study, placental and peripheral PCR detected approximately 2 and 3 times as many malaria infections at delivery as peripheral RDTs. Had a more sensitive test detected these subpatent malaria infections during ANC screenings, treatment of these presumably lower-level infections might have improved outcomes. Submicroscopic infections, which may be asymptomatic and remain undetected by routine RDTs, have been shown to be associated with poor birth outcomes [15]. Similar findings have been observed in prior studies from Malawi and Kenya. RDTs detected about 45% of the PCR-positive infections in paucigravidae and about 30% in multi-gravidae, allowing the majority of infections to persist in the placenta, potentially making ISTp less effective owing to a failure to detect low-level parasitemias. Highly sensitive RDTs could improve the sensitivity, and potentially the impact, of ISTp and deserve further study.

While ISTp cannot be recommended based on results from this study, consideration could be given to testing asymptomatic women at the first ANC visit only. RDT positivity is consistently higher at first visit [10, 12, 13, 20]; in this study, RDT positivity among asymptomatic women declined sharply from 5% at the first ANC visit to 1% by the fourth ANC visit, meaning that the number of afebrile women tested for each 10 malaria infections detected by RDT increased from 177 initially to 913 by the fourth ANC visit. In addition to potentially improving outcomes of women and their infants, screening at first ANC visit could provide highly granular data for surveillance and monitoring, as several studies have demonstrated that parasitemia prevalence in pregnant women corresponds well to that among children aged <5 years [21–24].

Though not the goal of the study, as 1 district conducted IRS while the other did not, there was an opportunity to assess the impact of IRS on malaria and anemia among pregnant women. Similar to findings from Uganda [25], IRS was associated with significant reduction in the risk of malaria at delivery and a marginal reduction in moderate to severe anemia after controlling for other factors.

ISTp did not reduce placental malaria or maternal anemia at delivery compared to routine care. Paradoxically, ISTp was associated with a statistically significantly higher risk of LBW (though not mean birthweight) compared to passive case detection for malaria only. This was most likely related to higher malaria exposure in the intervention group, as suggested by the significantly higher proportion of women who reported having had malaria in pregnancy prior to the study enrollment as well as the higher proportion of fever during the study. This study failed to show any benefit of implementing ISTp for malaria control in pregnancy. It may still be worthwhile to screen asymptomatic women at first ANC visits to provide surveillance data for monitoring parasitemia prevalence. This may also provide some benefit to pregnant women by detecting low-density infections early in pregnancy. RDTs with greater sensitivity could potentially identify more asymptomatic malaria cases among pregnant women and improve the efficacy of ISTp; further investigation into the utility of highly sensitive RDTs for ISTp is warranted.