Prenatal and child malaria and child growth: evidence from a Ghanaian pregnancy cohort

Abstract

Introduction

Additional investigation is needed to better understand the associations between prenatal and early child malaria and early childhood growth outcomes.

Methods

We leveraged the Ghana Randomized Air Pollution and Health Study which enrolled 1414 pregnant women from Kintampo, Ghana, and followed the mother-child dyads through the child’s first year of life. We defined prenatal malaria exposure by histopathological examination of placenta tissue. We defined early child malaria exposure by active health surveillance over the child’s first year of life and malaria rapid diagnostic testing when clinically indicated. We employed latent class growth analyses to construct growth trajectories for each anthropometric measure (weight and height and corresponding WHO z-scores, mid-upper arm circumference and head circumference (HC)), which we measured at birth, 3, 6, 9 and 12 months of age. We then used unadjusted and adjusted ordinal and multinomial regression to identify associations between prenatal and early child malaria growth trajectories, considered separately.

Results

Of the 1306 live births, 1144 children had prenatal and early child malaria exposure and growth data and were included in the analysis; 250 (21.9%) had prenatal malaria exposure, and 525 (45.9%) had early child malaria infection. Children with prenatal malaria as compared to those without had higher risk for poorer length growth (OR=1.46, 95% CI 1.08 to 1.98; multivariable OR=1.46, 95% CI 1.05 to 2.02), poorer length-for-age z-scores (OR=1.43, 95% CI 1.09 to 1.88; multivariable OR=1.53 95% CI 1.14 to 2.05) and smaller HC (OR=1.56, 95% CI 1.16 to 2.09; multivariable OR=1.43, 95% CI 1.04 to 1.98). Children with early childhood malaria as compared to those without had higher risk for poorer length-for-age z-score (OR=1.13, 95% CI 1.03 to 1.24; multivariable OR=1.12, 95% CI 1.01 to 1.24). Children with prenatal malaria as compared to without showed a trend for persistent stunting (multivariable OR=1.63, 95% CI 0.98 to 2.71).

Conclusion

Malaria beginning in pregnancy was associated with poorer growth outcomes. Public health preventative strategies beginning in pregnancy may reduce malaria incidence and improve early childhood growth.

WHAT IS ALREADY KNOWN ON THE TOPIC

• Prior evidence finds that prenatal malaria is associated with worse birth outcomes but the impact on growth over early childhood remains unclear.

WHAT THIS STUDY ADDS

• These data add to the literature findings that malaria beginning prenatally is associated with early childhood growth, including stunting risk.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Malaria eradication programmes in sub-Saharan Africa that target pregnant women may have important benefits for early childhood growth.

Introduction

In sub-Saharan Africa, there are an estimated 144 million stunted and 47 million wasted children under 5,¹ defined as a height-for-age and weight-for-age, respectively, z-score of less than −2.² Poorer growth begins prenatally³ and has long-term health consequences, including impaired cognition, chronic disease development such as worse lung health⁴ and mortality.⁵ There is an urgent need to understand risk factors that are associated with poorer growth to inform public health policy.

Exposures prevalent in the low- and middle-income country context beginning prenatally can negatively impact child growth. For example, work by our group leveraging the Ghana Randomized Air Pollution and Health Study (GRAPHS), a pregnancy cohort in the Bono East Region of Ghana, identified that higher prenatal air pollution exposure was associated with poorer length/height growth trajectories over the first year of life.⁶ Another prevalent exposure is malaria, the leading cause of mortality for children under the age of 5⁷ in sub-Saharan Africa. In 2023, there were approximately 6 million pregnancies with malaria in West Africa alone.⁸

Malaria has contributed to 247 000 DALYs (disability-adjusted life years) per year in Ghana among children between the ages of 0 and 4.⁸ Malaria infection is transmitted by female mosquitos, and clinical symptoms include chills, vomiting, fever and sweating.⁹ Malaria infection during pregnancy can have health implications for both mother and developing fetus. Indeed, pregnant women and young children appear particularly susceptible to malaria and have the highest rates of morbidity and mortality.¹⁰ Although living in a malaria-endemic region may provide adults with some immunity to malaria, pregnancy causes this immunity to decrease, making pregnant people more vulnerable to the infection and severe disease.¹¹

Normal growth over early childhood is an important determinant of the life course and potentially of intergenerational health. Stunting that persists after early childhood has been shown to be associated with both cognitive impairments, delays in gross and fine motor skills and language development.¹² This may be especially important in low- or middle-income countries, where malaria infection is common and reduced educational attainment on a population level stifles economic growth.³ Lower length for age and stunting leads to increased risk for both non-communicable disease and communicable disease.¹³

Emerging evidence suggests that malaria during the critical window of pregnancy can impact growth outcomes. A cross-sectional study leveraging nationally representative household surveys in Uganda suggested that insecticide-treated bed nets were associated with reductions in population-level stunting.¹⁴ Another study identified associations between placental malaria, a histopathological method for detecting malaria infection during pregnancy, and birth outcomes such as premature birth and low birth weight; however, associations with growth over infancy were not explored.^{15 16} In a separate study, malaria during the first 2 years of life was associated with lower weight-for-age z-scores (WAZ).¹⁷ Additional studies are needed to better characterise how malaria beginning prenatally impacts longitudinal growth patterns.

To address these questions concerning how malaria beginning prenatally impacts longitudinal growth, we leveraged the GRAPHS cohort to examine the associations between prenatal and early child malaria and child anthropometry measured at birth and every 3 months from birth through the first birthday. We employed growth trajectories,⁶ and for each trajectory, we performed multinomial or ordinal regression to examine associations between prenatal malaria exposure, defined by placental malaria on placental histopathology, and early child malaria, determined through child health surveillance for illness followed by serologic testing, and growth trajectories.

Methods

Ghana Randomized Air Pollution and Health Study cohort

GRAPHS has been described elsewhere.¹⁸ Briefly, GRAPHS was a cluster-randomised controlled study investigating the impact of cleaner fuel stoves to reduce household air pollution exposure on birth outcomes and pneumonia risk in the first year of life. Participants included non-smoking pregnant women with a singleton fetus, in their first or second trimester of pregnancy at the time of recruitment, and the primary cook of their household. A screening ultrasound was used to determine gestational age (GA)^{19 20} and confirm a singleton pregnancy. Participants were recruited from 35 communities (clusters) in the Kintampo North Municipality or Kintampo South District, both in the then Brong Ahafo region in Ghana.^{6 18} All participants received insecticide-treated bed-nets (ITN) and health insurance. GRAPHS was approved by the Institutional Review Boards at the Kintampo Health Research Centre (Kintampo Health Research Centre Institutional Ethics Committee/2014-23), Columbia University (IRB-AAAF3520) and the Icahn School of Medicine at Mount Sinai (IRB 17-01625).

Prenatal malaria infection

We defined prenatal malaria as Plasmodium falciparum-infected red blood cells present in the placenta.²¹ Prenatal malaria was ascertained by placental histopathological analyses with both acute and chronic placenta malaria being a marker of prenatal malaria (yes/no), as previously described.²² Placental biopsies were collected within 4 hours of delivery and formalin-fixed then paraffin-embedded.²³ There were 1158 participants that had a placental biopsy; reasons for not having a placental biopsy included the placenta not being saved by the facility, pregnancy loss/participant lost to follow-up or the birth was not identified until 24 hours after delivery.²² To determine malaria infection, the biopsies were classified as no infection, acute infection and chronic infection, and past infection by independent histopathologists blinded to the study outcomes.²⁴

Child malaria infection

Child health surveillance in GRAPHS was performed via both active and passive methodologies.^{18 25} First, trained resident fieldworkers visited participant homes weekly to assess child health. Any child who was identified to be ill following the WHO Integrated Management of Childhood Illness evaluation was referred to the study clinic for evaluation by a trained study healthcare provider. In a passive referral approach, caregivers could self-refer their child to the resident fieldworkers if they believed their child was ill. If the healthcare provider had clinical concerns for malaria (eg, fever), a rapid diagnostic malaria test was performed on blood collected from a finger stick (SD Bioline). Thick and thin smear microscopy was also used for malaria infection ascertainment.²⁴ Early child malaria infection was defined by yes/no (having experienced one or more episodes of clinical malaria vs no episode of clinical malaria) and the number of malaria infections tallied during the health surveillance. Malaria episodes less than 14 days from the previous episodes were excluded.

Anthropometric measurements

Child weight, length/height, mid-upper arm circumference (MUAC) and head circumference (HC) were measured in the first 24 hours after birth and at 3, 6, 9 and 12 months of age by trained fieldworkers. The Tanita digital scale (Tanita digital scale model BD-590; Tanita Corp.) was used to measure weight to the nearest 0.1 kg, and the Ayrton Infantometer (Ayrton Infantometer model M-200; Ayrton Corp) was used to measure length to the nearest 0.1 cm. The children’s weight was measured with clothes on but shoes off; diapers were not commonly worn. The United Nations Children’s Fund MUAC tape was used to measure MUAC which was measured using the child’s right arm, and the Child Growth Foundation paper lasso tape was used to measure HC. For the HC measurements, field workers measured starting at the forehead, through the occiput and ending back at the forehead.

Beginning at 3 months, the measurements collected were used to calculate height/length-for-age z-score (LAZ), weight-for-length z-score (WLZ) and WAZ using the 2006 WHO child growth standards.² As previously described, we identified that the lowest LAZ trajectory was consistently on average at or below −2 z-score at each time point, consistent with the definition of stunting.⁶ Therefore, as a secondary endpoint, we collapsed the categories into stunted (lowest LAZ trajectory, consistently at or below −2 z-score) versus not stunted.²⁶

Covariates

Maternal height (continuous variable, in metres) and weight (continuous variable, in kilograms) were collected during the enrolment visit. We employed a questionnaire on enrolment to characterise maternal age (continuous variable, in years), education (categorical variable, completed primary school or above), ethnicity (categorical variable) and marital status (categorical variable, married). We employed a questionnaire that enumerates the household assets and then performed a principal component analysis following standard methodology to calculate an asset index (continuous variable, median (IQR)).^{6 27} GA at delivery (continuous variable, in weeks) was determined using the date of birth and the ultrasound-derived estimated date of delivery as described above.¹⁹ Season of birth was determined using the month of birth and dichotomising into rainy season (April through June and September through October) versus not. Biological sex of child was determined at birth. Field workers asked participants about their ITN use during their visits to participant homes (categorical variable, yes or no) and maternal IPTp-Sp (intermittent preventative treatment of malaria in pregnancy via sulfoxide pyrimethamine administration) dosage (categorical variables, none, one to two doses and three doses) administered during pregnancy.

Statistical analysis

We used the GRAPHS growth trajectories developed by Boamah-Kaali et al. Briefly, as previously described, we used repeated anthropometric measurements to construct latent class growth trajectories using latent class growth analyses.⁶ We determined the best fitting model by AIC (Akaike Information Criterion), BIC (Bayesian Information Criterion), entropy and the number of individuals in each class and then assigned each participant to the class for which they had the highest probability of correct fit. For each trajectory, we assigned a continuous, ordinal value from either 0–2 when three trajectories were identified or 0–3 when four trajectories were identified. Trajectories classified as ‘0’ represent the referent category with the highest anthropometry measures while trajectories classified as either ‘2’ or ‘3’ where applicable represented the smallest or lowest anthropometry measures.

To examine the associations between prenatal and early child malaria and each growth trajectory, we fit bivariate and multivariable multinomial and ordinal logistic regression models. Multivariable models included both prenatal and early child malaria variables in the same model to identify the associations independent of malaria at the other time period. We assessed the proportional odds assumption²⁸ to determine if an ordinal logistic regression model was appropriate, and all but two models met this assumption. For these models (weight and WAZ growth trajectories), bivariate and multivariable multinomial logistic regression models were employed to model each trajectory level compared with trajectories classified as ‘0’. Finally, we fit logistic regression models to examine associations between prenatal and child malaria and persistent stunting, as compared with not persistently stunted, over the first year of life. Since these were factors that could impact the associations between the relationships we were investigating, we adjusted multivariable models for maternal age, education, ethnicity and marital status, self-reported IPTp-Sp dose uptake and use of ITN during pregnancy, household asset index, GA at delivery, child sex, season of birth and prenatal and child carbon monoxide exposure. Covariate data completeness was high (see table 1), and we did not impute missing covariate data. We used the ‘MASS’ and ‘nnet’ packages in R V.4.3.3 to carry out these analyses.

Table 1. Characteristics of Ghana Randomised Air Pollution and Health Study (GRAPHS) cohort with PM (prenatal malaria) and IM (infant malaria) exposure.

Characteristic	All children, N=1144	PM+, N=250	IM+, N=525
Mother’s age; median (IQR)	27 (22, 34)	22 (19, 27)	27 (22, 34)
(Missing)	*	0	0
Child biological sex
Female	560 (49%)	127 (51%)	249 (47%)
Male	584 (51%)	123 (49%)	276 (53%)
Ethnicity
*	199 (17.4%)	33 (13.2%)	67 (12.8%)
†	149 (13.0%)	40 (16.0%)	68 (13.0%)
3	745 (65.1%)	170 (68.0%)	368 (70.1%)
4	51 (4.5%)	7 (2.8%)	22 (4.2%)
Asset index; median (IQR)	−0.36 (−1.25, 0.84)	−0.36 (−1.10, 0.83)	−0.39 (−1.29, 0.77)
Gestational age at delivery, in weeks	39.7 (38.9, 40.6)	39.4 (38.4, 40.3)	39.71 (38.86, 40.71)
Prenatal CO exposure, ppm	1.07 (0.65, 1.65)	1.05 (0.68, 1.59)	1.14 (0.67, 1.66)
Child CO exposure, ppm	0.53 (0.24, 1.02)	0.53 (0.25, 0.96)	0.55 (0.26, 1.04)
Presence of prenatal malaria, n(%)	250 (24%)	250 (100%)	119 (25%)
(Missing)	104	0	47
Infant malaria, # of episodes	0 (0, 1)	0 (0, 1)	1 (1, 2)
Birth during rainy season, yes	502 (44%)	127 (51%)	227 (43%)
Maternal ITN use
Yes	734 (64%)	159 (64%)	348 (66%)
No	410 (35%)	91 (36%)	177 (34%)
Maternal IPTp-Sp dose uptake
None	6 (0.5%)	2 (0.8%)	1 (0.2%)
1–2 doses	160 (14%)	32 (13%)	83 (16%)
3 doses	978 (85%)	216 (86%)	441 (84%)
Maternal marital status, married	647 (57%)	101 (40%)	295 (56%)
Maternal educational level, completed primary school or beyond	677 (58%)	179 (72%)	295 (56%)

Carbon monoxide (CO) measurements were averaged across four time points for prenatal CO exposure and three time points for child CO exposure.

^*Median (IQR); n (%).

^†Rainy season is characterised by seasonal rainfall from April through June and September through October.

CO, ppm, carbon monoxide in parts per million; IM+, at least one clinical malaria episode during first 12 months of infant’s life; IPTp, intermittent preventive treatment of malaria in pregnancy via administration of sulfadoxine-pyrimethamine (Sp); ITN, insecticide-treated net; PM+, prenatal malaria.

Patient and public involvement

Study results will be disseminated to study participants on publication of results.

Results

GRAPHS recruited 1414 pregnant persons resulting in 1306 live births (92.4%), and 1144 (87.5%) of these children had available data on prenatal or child malaria exposure and had anthropometric data collected at least once during their first year of life. Table 1 describes the characteristics of the GRAPHS children with at least one anthropometric measurement and prenatal or child malaria data included in these analyses. Online supplemental table 1 compares the baseline characteristics of the GRAPHS children included versus not included in the presented analyses. Of those included, n=250 (21.8%) had evidence of prenatal, and n=525 (45.9%) had evidence of child malaria. As described above, all participants were provided an ITN at enrolment, and 64% (n=734) of the mothers reported using the nets. Most of the mothers in the cohort received three doses of IPTp-Sp (n=978, 85%).

The latent class trajectories developed for the GRAPHS cohort have been previously described and are shown in online supplemental figure 1 and online supplemental figure 2. In general, the GRAPHS cohort demonstrates poor growth trajectories over the first year of life with a significant proportion of children assigned to trajectories that on average are persistently at-risk for or have growth abnormalities for the first year of life.

Associations between prenatal and child malaria and child length growth

Child length growth trajectories included length, LAZ and WLZ trajectories. In general, the GRAPHS cohort demonstrated poor length growth over the first year of life with; for example, over 50% of children belonging to trajectories that on average were stunted or at risk for stunting for the first year of life.

In both bivariate and multivariable models, we find that prenatal malaria was associated with higher odds of being assigned to a poorer length growth trajectory (bivariate OR 1.46, 95% CI 1.08 to 1.98; multivariable OR 1.46, 95% CI 1.05 to 2.02, table 2). We additionally observed that prenatal malaria was associated with higher odds of being assigned to a worse LAZ trajectory (bivariate OR 1.43, 95% CI 1.09 to 1.88; multivariable OR 1.53, 95% CI 1.14 to 2.05, table 2).

Table 2. Associations between prenatal malaria^*, infant malaria^† and growth trajectories^‡.

Trajectory measurement	N	Bivariate model						Multivariable model§
		Prenatal malaria			Infant malaria		N	Prenatal malaria		Infant malaria
		R (95% CI)	P		R (95% CI)	P		R (95% CI)	P	R (95% CI)	P
Length	1040	1.46 (1.08, 1.98)	<0.01	1144	1.11 (1.00, 1.29)	0.05	1035	1.46 (1.05, 2.02)	0.02	1.10 (0.98, 1.23)	0.10
Head circumference	1040	1.56 (1.16, 2.09)	<0.01	1144	1.03 (0.93, 1.13)	0.61	1035	1.43 (1.04, 1.98)	0.03	1.03 (0.93, 1.15)	0.55
MUAC	1040	1.01 (0.77, 1.33)	0.93	1144	1.02 (0.93, 1.12)	0.62	1035	1.05 (0.78, 1.40)	0.77	1.02 (0.92, 1.13)	0.70
Weight-for-length z-score	1025	1.09 (0.83, 1.43)	0.54	1127	0.96 (0.87, 1.05)	0.38	1020	1.06 (0.78, 1.42)	0.72	0.94 (0.85, 1.04)	0.21
Length-for-age z-score	1025	1.43 (1.09, 1.88)	0.01	1127	1.13 (1.03, 1.24)	0.01	1020	1.53 (1.14, 2.05)	<0.01	1.12 (1.01, 1.24)	0.03

^*Prenatal malaria is defined as presence of Plasmodium falciparum-infected RBCs (red blood cells) sequestered in placental intervillous space.

^†Infant malaria is defined as the number of malaria infection(s) during first 12 months of infant’s life.

^‡Ordinal regression where the ORs may be interpreted as the odds of being in a lower/smaller trajectory for infants born with evidence of prenatal (or child) malaria versus not.

^§Multivariable models adjust for child sex, maternal age, gestational age at delivery, asset index, ethnicity, maternal educational status, maternal marital status, intermittent preventive treatment of malaria in pregnancy via administration (IPTp) dose uptake, use of insecticide-treated net (ITN), birth by rainy season, prenatal carbon monoxide (CO) exposure and child CO exposure.

MUAC, mid-upper arm circumference.

In the bivariate model, we identified that child malaria was associated with higher odds of being assigned to a poorer length trajectory (OR 1.11, 95% CI 1.00 to 1.29, table 2); however, these findings did not persist in the multivariable model. We identified that child malaria was associated with higher odds of being assigned to a worse LAZ trajectory (bivariate OR 1.13, 95% CI 1.03 to 1.24; multivariable OR 1.12, 95% CI 1.01 to 1.24, table 2).

Prenatal and child malaria and child head circumference and mid-upper arm circumference growth

We identified in bivariate and multivariable models that prenatal malaria was associated with higher odds of being in a smaller HC trajectory over infancy (bivariate OR 1.56, 95% CI 1.16 to 2.09; multivariable OR 1.43, 95% CI 1.04 to 1.98, table 2). We did not identify an association between child malaria and HC growth trajectories. We did not identify an association between prenatal or child malaria and MUAC trajectories.

Associations between prenatal and child malaria and child weight and weight-for-age z-score

Using multinomial regression, we did not observe an association between prenatal or child malaria and being in the lowest weight category in both the bivariate and multivariate models. Child malaria was associated with a lower WAZ only in the bivariate model (OR: 1.31, 95% CI: 1.05 to 1.64, table 3).

Table 3. Associations between prenatal malaria^*, infant malaria^† and weight trajectories^‡.

Trajectory measuremente		N	Bivariate model					Multivariable model§
			Prenatal malaria			Infant malaria		N	Prenatal malaria		Infant malaria
			OR (95% CI)	P	N	OR (95% CI)	P		OR (95% CI)	P	OR (95% CI)	P
Weight	0 versus 1	1040	1.60 (0.72, 3.56)	0.25	1144	1.17 (0.90, 1.51)	0.25	1035	1.67 (0.70, 3.96)	0.25	1.14 (0.86, 1.51)	0.37
	0 versus 2	1040	1.75 (0.80, 3.82)	0.16	1144	1.04 (0.81, 1.35)	0.75	1035	1.72 (0.73, 4.02)	0.21	1.01 (0.77, 1.34)	0.92
	0 versus 3	1040	1.97 (0.87, 4.48)	0.10	1144	1.28 (0.98, 1.67)	0.07	1035	2.00 (0.81, 4.92)	0.13	1.22 (0.91, 1.64)	0.19
Weight-for-age z-score	0 versus 1	1028	1.24 (0.70, 2.21)	0.46	1130	1.11 (0.91, 1.35)	0.31	1023	1.10 (0.59, 2.03)	0.77	1.08 (0.88, 1.34)	0.45
	0 versus 2	1028	1.43 (0.81, 2.55)	0.22	1130	1.03 (0.84, 1.27)	0.76	1023	1.31 (0.71, 2.42)	0.39	1.00 (0.81, 1.23)	0.97
	0 versus 3	1028	1.56 (0.80, 3.05)	0.19	1130	1.31 (1.05, 1.64)	0.02	1023	1.47 (0.71, 3.07)	0.30	1.20 (0.94, 1.53)	0.14

^*Prenatal malaria, defined as presence of Plasmodium falciparum-infected RBCs sequestered in placental intervillous space.

^†Infant malaria is defined as the number of malaria infection(s) during first 12 months of infant’s life.

^‡Multinomial regression models where the ORs may be interpreted as the odds of being in a smaller/lower trajectory for infants born with evidence of prenatal (or infant) malaria versus not, compared with the highest trajectory. Weight trajectories are the following: 0: heaviest, 1: heavy-moderate, 2: light-moderate, 3: lightest. Weight-for-age trajectories are the following: 0: normal-high, 1: normal, 2: at risk for underweight, 3: underweight.

^§Multivariable models adjust for child sex, maternal age, gestational age at delivery, asset index, ethnicity, maternal educational status, maternal marital status, intermittent preventive treatment of malaria in pregnancy via administration (IPTp) dose uptake, use of insecticide-treated net (ITN), birth by rainy season, prenatal carbon monoxide (CO) exposure, and child CO exposure.

Associations between prenatal and child malaria and stunting

There was no association identified between prenatal malaria and stunting in the bivariate model, but we did identify a trend between prenatal malaria and odds for stunting in the multivaiable model (OR 1.63, 95% CI 0.98 to 2.71, table 4). Additionally, we identified an association between child malaria and stunting in the bivariate model (OR 1.19, 95% CI 1.03 to 1.37, table 4), but this did not persist in the multivariable model.

Table 4. Associations between prenatal malaria^*, infant malaria^† and stunting.

Outcome	N	Bivariate model					Multivariable model‡
		Prenatal malaria			Infant malaria		N	Prenatal malaria		Infant malaria
		OR (95% CI)	P	N	OR (95% CI)	P		OR (95% CI)	P	OR (95% CI)	P
Stunting	1025	1.37 (0.87, 2.13)	0.17	1127	1.19 (1.03, 1.37)	0.02	1020	1.63 (0.98, 2.71)	0.059	1.11 (0.94, 1.30)	0.23

^*Prenatal malaria, defined as presence of Plasmodium falciparum-infected RBCs sequestered in placental intervillous space.

^†Infant malaria is defined as the number of malaria infection(s) during first 12 months of infant’s life.

^‡Multivariable models adjust for child sex, maternal age, gestational age at delivery, asset index, ethnicity, maternal educational status, maternal marital status, intermittent preventive treatment of malaria in pregnancy via administration (IPTp) dose uptake, use of insecticide-treated net (ITN), birth by rainy season, prenatal carbon monoxide (CO) exposure and child CO exposure.

Discussion

Understanding risk factors for poorer growth trajectories in sub-Saharan Africa is urgently needed to inform public health policy. By leveraging the GRAPHS cohort, we identified associations between prenatal malaria and poorer length, HC, LAZ growth trajectories and stunting over the first year of life. We also identified associations between early child malaria and poorer LAZ growth trajectories over the first year of life, but they were smaller in effect size than prenatal malaria. Our data add malaria exposure in both the prenatal, which is represented by infection of the placenta after fetal exposure to malaria antigens and early childhood periods as an additionally important factor to consider and supports prior literature suggesting the importance of early childhood malaria on growth.

Our data highlight the associations between prenatal malaria and poorer length growth over childhood, suggesting that the impact of malaria on early childhood DALYs may be even more significant than previously understood.^{8 29} If replicated, this work suggests that targeted malaria vaccination and eradication programmes could positively impact the burden of stunting and wasting in endemic areas. In Ghana, the National Malaria Elimination Strategic Plan (NMESP) 2024–2028 was created to focus on stopping the transmission of malaria; the NMESP plans to reduce malaria mortality by 90%, malaria incidence by 50% and rid areas that have low malaria burden of all of the disease by 2028.³⁰ Given our study results, it would be prudent for public health officials to examine the secondary health benefits of malaria disease risk reduction, for example, whether early childhood growth abnormalities improve with successful implementation of malaria eradication programmes.

Our study adds to the prior literature supporting associations between early childhood malaria and child growth and additionally implicates prenatal malaria as a determinant of early child growth. The mechanisms by which prenatal malaria impacts fetal development and child growth are poorly understood. Two of the previously researched mechanism are malaria-infected red blood cells in the intervillous space of the placenta which may cause inflammation,³¹ or malaria-infected red blood cells that may impact placental angiogenesis leading to placenta growth restriction and placental dysfunction.32,34 Of note, prenatal malaria has been associated with malaria risk over early childhood, hypothesised through potential immune priming pathways.³⁵

There were several strengths in our study. By leveraging the GRAPHS cohort, we were able to leverage a longitudinal cohort with data on placental malaria and child malaria among a group living in a malarial endemic area. Longitudinal follow-up over early childhood with repeated anthropometric measures allowed us to generate growth trajectories and then associate these with prenatal and child malaria exposures. The GRAPHS data repository includes important covariates to minimise confounding, including IPTp-Sp dosing.

We also acknowledge limitations. Inadequate nutrition impacts stunting and growth trajectories, but we do not have detailed nutrition data over pregnancy and early childhood. We also note the possibility of residual confounding by unmeasured variables and reverse causalities between infant malaria and anthropometric measurements. Additional studies are needed to better characterise how malaria infections beginning prenatally impact longitudinal growth patterns with implications for life course health. For example, previous studies have focused on the additional role of malnutrition on malaria and growth, which we are unable to do in our GRAPHS cohort.^{14 36} Children impacted by both malaria and stunting are particularly vulnerable to other detrimental child health outcomes.³⁷ Therefore, replication and expansion of our findings is important.

Conclusion

These data find that malaria infection beginning in pregnancy is associated with worse growth over early childhood. Public health interventions to reduce malaria beginning in pregnancy are critical for life course health.

Data availability statement

Data are available upon reasonable request.