This cohort study was nested in a randomized partially placebo-controlled trial conducted from March 2018 to August 2019 involving 4680 pregnant women comparing monthly intermittent preventive treatment of malaria in pregnancy (IPTp) with sulfadoxine-pyrimethamine vs dihydroartemisinin-piperaquine, alone or combined with a single course of azithromycin at enrollment conducted in Kenya, Tanzania and Malawi [16]. Of these women, one-third were randomly selected into a nested cohort for fetal growth monitoring by serial ultrasound. In order to have a power of 80% to detect an expected proportion of women with STIs/RTIs was 40% in sulfadoxine-pyrimethamine arm compared to 30% in dihydroartemisinin-piperaquine/dihydroartemisinin-piperaquine + azithromycin, with alpha = 0.025, 432 women per treatment arm were needed. To allow for 13% loss to follow-up, 500 women were recruited per arm. Women attending antenatal care were enrolled if HIV-negative, had a viable singleton pregnancy between 16 and 28 weeks gestation, had no known heart disease, had not received sulfadoxine-pyrimethamine during the current pregnancy, and had no known allergy to the study drugs.

Details of data collection have been described elsewhere [16]. In brief, demographic data and medical history were collected at enrollment. Women were screened for urinary tract infection (using urine dipsticks) and hypertensive disorders (blood pressure >140/90 mmHg ± proteinuria), prior medication usage, and maternal anthropometrics were recorded at each antenatal visit. Hemoglobin level was assessed (Hemocue 301 or 201) at enrollment, in the third trimester, and at delivery.

Using ultrasound and standard methodology, gestational age was estimated based on crown-rump length until 13⁺⁶ weeks [17], and from 14⁺⁰ weeks by using an algorithm of head circumference and femur length [18], head circumference only [18] or femur length only [19], depending on availability of fetal biometrics. Serial ultrasound was performed at enrollment if gestational age was ≥22 weeks, at approximately 25–28 weeks gestation, and approximately 32–35 weeks, and fetal weights were estimated based on head circumference, abdominal circumference, and femur length [20].

Women were screened for malaria at enrollment using malaria rapid diagnostic tests (mRDTs) (CareStart^™ Malaria Pf/PAN (HRP2/pLDH) Ag Combo) as per national policy in Kenya and Tanzania. In all three countries, women with fever (≥ 37.5°C) or recent history of fever were also screened with mRDTs.

In Kenya and Malawi, regardless of treatment arm, women with positive mRDTs were treated with artemether-lumefantrine, and IPTp dosing was deferred for 4 weeks. In Tanzania, women with positive mRDTs in the sulfadoxine-pyrimethamine arm were treated with artemether-lumefantrine, and IPTp was deferred for 4 weeks. However, women in the dihydroartemisinin-piperaquine and dihydroartemisinin-piperaquine/azithromycin groups who had positive mRDTs at enrollment were given their first course of IPTp but at later visits artemether-lumefantrine was administrated if mRDTs were positive, and IPTp was deferred for 4 weeks.

Peripheral maternal venous blood was collected at all visits and at delivery, along with cord and placental blood. Thick and thin blood smears were prepared, Giemsa stained, and independently double-read by experienced microscopists; where results were discordant, a third reading was performed to determine the final result [16]. Dried blood spots were also prepared for quantitative real-time reverse transcription-polymerase chain reaction (qRT-PCR) [16]. Finally, placental biopsies were taken at delivery for malaria histology [16].

As part of standard care, pregnant women were pre-screened for HIV. Women living with HIV were provided treatment per national guidelines and excluded from the study. All women were subsequently screened for syphilis with SD-Bioline point-of-care tests and, if positive, they were treated with 2.4 million units intramuscular benzathine penicillin G. Additionally, clinic staff routinely asked women if they had experienced any symptoms associated with STIs/ RTIs. At any visit, if a woman responded in the affirmative, she was treated by the clinic staff according to national syndromic management guidelines recommended by the World Health Organization [8]. Apart from routine care, clinic staff collected vaginal swabs and stored them on site until the end of the trial, at which time the samples were shipped to a regional reference laboratory in East Africa for retrospective batch analysis. Serum and vaginal swab samples were collected at enrollment and between 32–36 weeks. Serum samples were tested for rapid plasma reagin and confirmatory syphilis testing with Treponema pallidum Hemagglutination assays. Vaginal samples were tested for chlamydia and gonorrhea DNA by RT-PCR (Artus^® CT/NG QS-RGQ Kit), trichomoniasis with SACASE^™ Real-TM Kit, and bacterial vaginosis using the Nugent scoring.

At delivery, birthweight was measured using digital scales (Seca GmbH & Co. KG., precision 10 g or ADE M112600, precision 5 g) and head and abdominal circumferences using flexible tape. Birthweights recorded >1-hour post-delivery were adjusted for physiological weight loss [21].

Analyses were conducted using Stata software, v16 (Stata Corp, Texas, USA).

Malaria exposure was defined as testing positive at any time-point by any assay: mRDT, microscopy, qRT-PCR, and/or placental histology. STIs/RTIs exposure was defined for individual STIs/RTIs and as a composite variable with positive test for any STIs/RTIs at any time-point. For the longitudinal analyses, women were considered negative until their first malaria and/or STIs/RTIs episode and thereafter considered positive. Four unique exposure groups were generated to assess if malaria and STIs/RTIs co-infection affected growth trajectories; a control group with neither malaria nor STIs/RTIs; malaria-only; STIs/RTIs only; and malaria plus STIs/RTIs.

The primary outcome was Z-scores for fetal weights and birthweight using a sex-specific Tanzanian reference chart [22] based on previous evidence indicating that a local growth curve is more representative than the international growth curve [23]. Our approach aligns with recent recommendations by the International Federation of Gynecology and Obstetrics (FIGO) on the accuracy of growth curves [24]. Secondary outcomes were birthweight Z-scores alone, growth trajectories based only on fetal weights Z-score, SGA (birthweight <10^th percentile) [22], LBW (birthweight <2.5 Kg), preterm delivery (gestational age <37 weeks), and new-born abdominal circumference in millimeters and head circumference in millimeters or Z-scores based on INTERGROWTH-21^st reference [25].

Women with a non-viable pregnancy outcome (miscarriage, stillbirths), twin pregnancy, severe congenital malformations, or missing data on malaria and STIs/RTIs were excluded. Furthermore, observations with weights measured <14 days apart, gestational age <18 weeks or ≥45 weeks, birthweights <250 g or ≥6500 g, or fetal/birthweight Z-score >±5, were excluded.

Linear regression models and linear mixed-effects models were used to assess the effect of malaria and/or STIs/RTIs on birth size and growth trajectories respectively. All crude models were adjusted for study design factors (study arm, site, and gravidity [paucigravidae, i.e., primi- and secundigravidae, and multigravidae]). In mixed models, these same design factors were included as fixed effects, gestational age at visit was included as a time factor, and individual participant as a random effect to account for within-subject clustering. In addition, other potential confounders, selected based on the statistical analysis plan for the main trial, including rainfall patterns, malaria transmission intensity, patterns of parasite resistance to sulfadoxine-pyrimethamine, maternal age, gestational age at enrollment or delivery, socioeconomic status, maternal body mass index (BMI), bednet use, number of IPTp doses received, hemoglobin levels, and sex of the fetus/newborns, were considered if associated with the outcome variable with a P-value <0.2 in the univariate models and retained in final models if P-values were <0.1.

Malaria infection is more detrimental in paucigravidae and undernourished women than in multigravidae and well-nourished counterparts. Thus, we fitted models with interaction terms to investigate possible effect-modification between malaria and gravidity or malaria and maternal BMI. The interaction between malaria and STIs/RTIs was also assessed.

To assess if the effect on growth trajectories was due to poor growth close to delivery, models only including Z-scores for fetal weights but not birthweight, were also generated. Finally, as fetal weight gain is mainly in the third trimester, a linear regression model was generated with a single fetal weight Z-score in the third trimester as the outcome, and malaria infections or STIs/RTIs occurring before the fetal weight estimation as exposure.

Additionally, a dose-response relationship was assessed by comparing the impact of number of malaria episodes on birth weight Z-score using the group with one malaria episode as the reference group. Furthermore, the model on STIs/RTIs was repeated after categorizing STIs/RTIs exposure by (1) composite STIs/RTIs only at enrollment, between weeks 32 and 36, or both at enrollment and between weeks 32–36; (2) only one type of STIs/RTIs, or multiple STIs/RTIs. Finally, we assessed the effect of malaria and STIs/RTIs on SGA, LBW, and preterm delivery using Poisson regression with robust error variance.

Of the 1586 women randomly selected for fetal growth monitoring, 1435 were eligible for analyses. Of the 1435 participants, 573 (39.9%) were >22 weeks at enrollment and had fetal weight assessed, 1007 (70.2%) had fetal weight assessed between approximately 25–28 weeks, 1045 (72.8%) between approximately 32–35 weeks. Birthweights were available for 1325 (92.3%) participants. Thus, 3950 observations of fetal weight/birthweights were included in the longitudinal analysis (Figure S1).

The distribution of the 1435 women was similar across study arms and countries. The mean age was 24.9 (SD 5.8) years. Only 2.8% were underweight (BMI <18.5 Kg/m²) at enrollment, whereas 33.2% were overweight (25–29.9 kg/m²) or obese (≥30 Kg/m²). Among the newborns, 13.4% were SGA, 4.3% were preterm, and 8.1% were LBW (Table 1). Baseline maternal characteristics were similar between included and excluded mother-newborn dyads, except that; a higher proportion of excluded women were from Malawi and paucigravidae, the proportion of bed net use at enrollment also differed significantly between the two groups and this proportion was lower among excluded women (Table S1).

Malaria infection was common: 43.4% (623/1435) of women had at least one episode during pregnancy, and 46.3% (364/787) of paucigravidae had malaria (Table 2a). Malaria prevalence varied across study sites and arms, being highest in Malawi and the sulfadoxine-pyrimethamine arm (Table S2). Women with malaria had lower socioeconomic status, were younger, had lower BMI and hemoglobin levels at enrollment, and more often came from rural areas (Table S2).

Similarly, a high prevalence of STIs/RTIs was observed, with over half of the women having STIs/RTIs detected either at enrollment, in the third trimester, or at both timepoints (Table 2b). Bacterial vaginosis was the most common, with 34.6% (449/1297) of the women testing positive for bacterial vaginosis at least once during pregnancy. Only 1.9% (27/1407) and 4.2% (54/1298) of the women had syphilis and gonorrhea, respectively. Among women with STIs/RTIs a higher proportion were from Tanzania. Women with and without STIs/RTIs had similar demographic characteristics across study arms (Table S3). Fetal biometry in second and third trimesters by gestational age and gravidity is described in Table S4.

There was a trend toward lower mean birthweight Z-scores among women with malaria infection and STIs/RTIs compared to women without (adjusted mean difference [aMD] [95% confidence interval: CI] malaria:-0.10 [−0.22, 0.02], P = 0.09; STIs/RTIs: aMD = −0.09 [−0.21, 0.02], P = 0.12) (Table 3a and b). Malaria exposure was also associated with a higher proportion of newborns being SGA (adjusted risk ratio [aRR]: 1.50 [1.14–1.97], P = 0.004) (Table 3a). The effect was more evident among paucigravidae women with malaria or STIs/RTIs (Malaria: aMD for birthweight Z-score= −0.19 [−0.35,−0.03], P = 0.02 and SGA aRR = 1.84 [1.26–2.69], P = 0.002); STIs/RTIs: aMD for birthweight Z-score = −0.17 [−0.33,−0.01], P = 0.04) (Table 3a and b). There was a tendency toward a dose-response relationship between the number of malaria episodes and impact on birthweight Z-score, although this was not statistically significant (1 vs 2 malaria episodes aMD −0.12 [−0.41, 0.16], P = 0.39; 1 vs 3+ malaria episodes aMD −0.32 [−0.72, 0.09], P = 0.13). Infection with both malaria and STIs/RTIs in paucigravidae women had an even more pronounced effect on birthweight Z-scores (aMD = −0.34 [−0.57, −0.11], P = 0.003) (Table 4a) and SGA (aRR = 2.53 [1.37–4.67], P = 0.003) (Table 4b). The same effect on birthweight and risk of SGA was not observed among multigravidae (Tables 3 and 4).

Neither head circumference nor abdominal circumference differed significantly among malaria or STIs/RTIs exposed compared to non-exposed newborns (Tables 3 and 4). No statistically significant effect of the individual STIs/RTIs on birthweight was observed, albeit there was a trend toward lower birthweight Z-score among newborns whose mothers had bacterial vaginosis (crude MD = −0.13 [−0.21, 0.08], P = 0.06) (Table S5).

The effects of malaria and STIs/RTIs on growth trajectories were investigated using mixed-effect regression models on fetal weights and birthweight Z-scores (Table 5). Malaria infection was associated with a lower weight Z-score over time (aMD=−0.12 [−0.22, −0.03], P = 0.01) (Table 5a). The effects differed significantly by gravidity strata (P_interaction = 0.01) and were more pronounced among paucigravidae (weight Z-score [95% CI] over time aMD = −0.17 (−0.31, −0.04), P = 0.01) than multigravidae (aMD = −0.07 [−0.21, 0.07], P = 0.34) (Table 5b and c). There was no significant interaction between BMI and malaria (P_interaction = 0.48). STIs/RTIs also reduced weight Z-score over time (aMD = −0.11 [−0.20, −0.01], P = 0.03), again with paucigravidae being most affected (Table 5d and e).

The magnitude of the effect on growth trajectories was similar after exposure to malaria-alone (aMD = −0.18 [−0.31, −0.04], P = 0.01), STIs/RTIs alone (aMD = −0.14 [−0.26, −0.03], P = 0.01) or to malaria plus STIs/RTIs (aMD = −0.20 [−0.33, −0.07], P = 0.003) (Tables 5g), and there was a non-significant interaction between malaria and STIs/RTIs (P_interaction = 0.18). Again, infection with both malaria and STIs/RTIs impacted growth trajectories more in paucigravidae than multigravidae (aMD = −0.30 [−0.48, −0.11], P = 0.001 vs −0.11 [−0.30, 0.09], P = 0.28) (Table 5h and i).

Models containing only fetal weight Z-scores but not birthweight yielded similar results (Tables 3 and 4a).

Fetal weight in the third trimester, assessed by a single measure, was also lower among paucigravidae after malaria (aMD = −0.25 [−0.47, −0.03], P = 0.02), but not after STIs/RTIs (Table S6). Fetal weight gain over time was lower among women with STIs/RTIs at enrollment than women with STIs/RTIs both at enrollment and in the third trimester (Table S7). The individual STIs/RTIs were not significantly associated with impaired fetal growth, although there was a trend toward lower fetal/birthweight Z-score for trichomoniasis (aMD = −0.11 [−0.23, −0.02], P = 0.09) (Table S7). Finally, having multiple STIs/RTIs did not further reduce fetal weight gain compared to having a single STI/RTI (Table S7).

This is the largest study to date utilizing ultrasound for fetal weight estimation concurrently with in-depth testing for malaria and STIs/RTIs. High-quality obstetric ultrasound was ensured by thorough training of sonographers, review of all ultrasound images at the beginning of the study, and thereafter 10% randomly selected scans – all performed by a medical doctor with extensive experience in obstetric ultrasound (CS). All anthropometric measurements were performed twice, with a third reading for discrepancies and the average of the two closest readings was considered definitive. Birth weight measured >1 hour after delivery were also adjusted for physiological weight loss [21].

However, this study also has some limitations. First, fetal weight and birthweight were converted into Z-score using the Strategies To Prevent Pregnancy Associated Malaria (STOPPAM) reference chart, as we have previously demonstrated this reference chart to be more appropriate for the setting [23]. However, a similar reference for head circumference and abdominal circumference is not available, and the INTERGROWTH-21^st was therefore used for head circumference [25]. Second, previous studies indicated that malaria in either the first or second trimester might be the most detrimental [4,5]. However, women were enrolled from the second trimester onward. Thus, malaria infections occurring in the first trimester were not accounted for, and some women may wrongly have been classified as malaria-negative, resulting in an underestimation of the true burden. Third, miscarriage and stillbirth may be due to malaria and/or STIs/RTIs but were excluded from the analyses. Fourth, the prevalence of STIs/RTIs at enrollment was lower among the excluded women and may represent some selection bias. Finally, some residual confounders could not be ruled out, including genetic factors. However, these are unlikely to have influenced the results as they would be expected to be relatively infrequent and balanced between study exposure groups.