Study procedures

This study utilizes longitudinally collected data from two double-blind, randomized controlled trials of intermittent preventive therapy during pregnancy in Uganda. The Dihydroartemisinin–Piperaquine for the Prevention of Malaria in Pregnancy study (NCT02163447) enrolled pregnant women from June to October 2014 and randomized them to preventive therapy with sulfadoxine-pyrimethamine given every 8 weeks, dihydroartemisinin-piperaquine given every 8 weeks, or dihydroartemisinin-piperaquine given every 4 weeks during pregnancy [21]. The primary endpoint of this study was the prevalence of histopathologically confirmed placenta malaria. The monthly sulfadoxine–pyrimethamine versus dihydroartemisinin–piperaquine for intermittent preventive treatment of malaria in pregnancy trial (NCT02793622) enrolled pregnant women between September 2016 to May 2017. They were randomly assigned to receive monthly sulfadoxine–pyrimethamine or monthly dihydroartemisinin–piperaquine for intermittent preventive treatment of malaria in pregnancy [22]. The primary endpoint of this study was to evaluate the risk of adverse birth outcomes in each cohort. All participants in both studies were HIV-negative women at least 16 years of age with a viable pregnancy between 12 and 20 weeks gestation.

Gestational age was determined by the first day of the last menstrual period (LMP) and an ultrasound performed on the day of enrollment. Per study protocol [21,22], if the ultrasound was consistent with a gestational age of 6 to 12 weeks and the ultrasound dating was within 7 days of the LMP, then the LMP was used to determine gestational age; if the ultrasound differed from the LMP by more than 7 days, then the ultrasound was used to determine gestational age. If the ultrasound was consistent with a gestational age of 13–24 weeks, and the ultrasound gestational age was within 14 days of the LMP, then the LMP was used to determine gestational age; if the ultrasound gestational age differed from the LMP by more than 14 days, then the ultrasound was used to determine gestational age.

In both studies, routine visits were conducted every 4 weeks. Routine laboratory testing was performed every 8 weeks for complete blood count and alanine aminotransferase. Women were encouraged to come to the clinic any time they felt ill. Those who presented with a documented fever (tympanic temperature ≥38·0°C) or history of fever in the previous 24 hours had blood collected for a thick blood smear for detection of malaria parasites. Blood smears were stained with 2% Giemsa and read by experienced laboratory technologists. A blood smear was considered to be negative when the examination of 100 high-power fields did not reveal asexual parasites. If the smear was positive, the patient was diagnosed with malaria and treated with artemether–lumefantrine. Collection of dried blood spots for molecular detection of asymptomatic parasitemia with quantitative PCR (qPCR) and loop-mediated isothermal amplification (LAMP) of P. falciparum DNA occurred every 4 weeks per study protocol. Details of how these molecular tests were operated and copy number of the parasite genome have been described for this dataset in previous studies and their protocols [21–23]. The results of these tests were not available until after completion of the study. Thus, asymptomatic parasitemia was not treated. Women were encouraged to deliver their babies 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 congenital anomalies, birth weight, and collection of biological specimens including placental tissue, placental blood, and maternal blood. Following delivery, women were followed for 6 weeks postpartum.

For assessment of placental malaria, two 1 cm-wide full thickness biopsies, obtained about 5 cm from the cord, were obtained within 1 hour of delivery and placed in 10% neutral buffered formalin. Biopsy specimens were embedded in paraffin wax, sectioned into 3 μm slices using a rotary microtome, fixed to glass slides, and dehydrated in sequential ethanol baths. Separate slides were stained in 0.1% hematoxylin/1% eosin (H&E) for 5 and 1 min, respectively, or in 2% Giemsa for 30 min. Placentas were graded into 5 categories using a standardized Rogerson criteria [24]. The presence of intervillous parasite-infected erythrocytes and of pigment in monocyte/macrophages or fibrin were noted. Quantitative assessments of placental malaria involved counting of 1000 intervillous blood cells under high power and determination of percentages of intervillous infected erythrocytes and monocyte/macrophages containing malarial pigment.

Cohort definition

Data was analyzed to create the following two cohorts from live, singleton births: malaria negative and any-malaria. The malaria negative cohort was defined as the absence of clinical malaria, asymptomatic parasitemia, and placental malaria. The any-malaria cohort included women with any of the following: clinical malaria, placental malaria, or asymptomatic parasitemia. Placental malaria was diagnosed by positive histopathology by Rogerson criteria [13], with or without symptomatic malarial infection. Asymptomatic parasitemia was defined as no clinical maternal malaria episodes but parasite DNA was detected in the blood by positive peripheral maternal blood qPCR or LAMP, with or without placental malaria.

The malaria negative cohort was used to create a sex-stratified birth weight standard. This birth weight standard was then used to determine the prevalence of SGA neonates, defined as <10% percentile, in the any-malaria cohort. The performance of this birth weight standard was then compared to six commonly used growth curves including: Schmiegelow [20] (based on a Tanzanian population), the WHO global reference (based on Mikolajczyk et. al algorithm) [19],Williams’ US-based curve [9], Verhoeff’s Malawi standard [8], INTERGROWTH-21^st (based on a multinational standard) [6], and WHO sex-stratified (based on an international standard detailed in Kiserud’s study [3]). We input our Ugandan specific data into the WHO global reference algorithm as detailed in Mikolajczyk’s paper [19]; the output was used in the above comparison and was labeled as WHO Uganda. The sex-stratified WHO curves based on Kiserud’s study were labeled WHO International.

The six comparison curves were chosen to compare our curve to regional standards (Tanzanian and Malawi), internationally inclusive standards (WHO and INTERGROWTH-21st), and internationally utilized curves from high-income countries (Williams). We aim to use these curves to provide commentary on the design of global birth weight standards and the generalizability of these standards.

Data analysis

Data were triple-checked and validated using Microsoft Excel 2019. The growth standard chart and curve were created using MathWorks MATLAB R2018b (The MathWorks, Inc., Natick, Massachusetts, United States).

The dataset was organized by completed gestational week in ascending order. For each completed gestational week, the birth weights were arranged in arrays of ascending order. A percentile function similar to the estimation-maximization model approach [25] was implemented due to its ability to estimate the distributions directly. Given our sample size, we opted for this model approach to limit the influence of outliers.

The percentile birth weights derived from our model were smoothed using the MATLAB curve fit tool to create the Ugandan standard growth curve, highlighting the 10th, 50th, and 90th percentiles. A 3^rd order polynomial regression was applied to create the growth curve using the least actual residual model.

A function was created to generate the proportion of SGA determined using our Ugandan standard for each cohort of interest incorporating sex-stratification. Using this SGA comparison function we were able to compare the proportion of SGA as determined by each of the six growth standards of interest.

Mann-Whitney U test was used to compare medians of non-parametric continuous variables and chi-squared analysis was used to compare proportions of non-parametric categorical variables. Wilcoxon signed rank test was conducted to compare the diagnosis of SGA by each curve and the percent agreement, relative risk and 95% confidence intervals (CI) were calculated. Statistical analysis was done in SPSS (IBM SPSS Version 23 Armonk, NY: IBM Corp). Forest plots and bar graphs were created in GraphPad Prism Version 8 (GraphPad Software, La Jolla California USA).

Ethical approval

The two randomized control trials were approved by the ethics committees of Makerere University School of Biomedical Sciences (Kampala, Uganda), the Uganda National Council for Science and Technology (Kampala, Uganda), and the University of California San Francisco (San Francisco, CA, USA). All study participants provided written informed consent.

Principal findings

We created a malaria negative birth weight standard based on longitudinally assessed maternal malaria status and histopathologic placental evaluation for assessing clinical malaria, placental malaria, and asymptomatic parasitemia. We found that our malaria negative Ugandan standard tended to diagnose more SGA neonates than WHO Uganda and Schmiegelow’s Tanzanian curve in women with malaria infection. The INTERGROWTH-21^st curve performed similarly to the Ugandan standard in diagnosing SGA while the WHO International curve statistically overestimated SGA in neonates with maternal malaria infection.

Comparison with existing literature

Here we present the first birth weight standard that definitively excluded malaria infection (clinical malaria, placental malaria, and asymptomatic parasitemia) during pregnancy, a significant cause of low birth weight and fetal growth restriction. As supported by previous research, the malaria negative cohort was older, had more multigravidas, larger neonates, and used dihydroartemisinin-piperaquine for IPTp instead of sulfadoxine-pyrimethamine when compared to the any malaria cohort [11,21]. The creation of a birth weight standard based on a true malaria negative cohort has not previously been done as birth weight standards in malaria-endemic regions have not assessed for malarial infection (particularly by placental histopathology), purposefully included malaria positive patients, or only assessed the serum for parasitemia [3,6,8,19,20]. This is important and novel because placental malaria is the known linkage between maternal malarial infection and adverse obstetrical outcomes such as low birth weight (relative risk (RR), 3.45; 95% CI 1.44–8.23; p = 0.005), preterm birth (RR, 7.52; 95% CI 1.72–32.8; p = 0.007), small for gestational age (RR, 2.30; 95% CI 1.10–4.80; p = 0.03), and higher rates of maternal anemia (adjusted odds ratio, 2.22; 95% CI 1.02–4.84; p = 0.045) [11,12,14].

We found that Williams’s US-based curve overestimated SGA and fetal growth. This finding is congruent with some of the current literature. One study showed that birth weight standards based on high-income countries often overestimate growth in Sub-Saharan Africa leading to the predication of larger neonates than what is accurate [26]. However, this study did not exclude maternal malaria or HIV infection in the cohort used to create their population standard [26]. Their finding of smaller neonates with the inclusion of endemic diseases aligns with the underdiagnosis of SGA we found in the Malawi curve which screened for peripheral and placental malaria, but did not exclude women who were positive for malaria [8]. We created a malaria and HIV negative cohort to assess optimal growth in the Ugandan population. Our findings aligned with INTERGROWTH-21^st, even with lower percent agreement, which sought to globally evaluate fetal and neonatal growth in low-risk pregnancies. The discrepancies here highlight that international standards can be applicable if they include diverse patient populations, INTERGROWTH-21^st represented a multinational population while Williams’ completed curve is on a United States-only population [9]. Additionally, the decision to exclude or include endemic diseases impacts the population standard as it defines what “normal” is. We believe normal growth in Sub-Saharan Africa can be achieved in the absence of the disease processes currently disproportionately affecting the region and should be represented in the medical literature.

Our findings also add to the growing discussion on the types of birth weight standards and references (web-based algorithms) used to define SGA. Gardosi et al, have argued for the creation of a customized birth weight standard, based on individual maternal characteristics instead of population-based standards [27,28]. They created an algorithm where maternal height and weight, ethnicity, parity, fetal weight and sex, and gestational age are analyzed to produce a customized birth weight centile [2,28]. Many studies have investigated their approach to assess its validity. One meta-analysis found that the customized charts estimated higher odds ratios for mortality in SGA neonates in comparison to population-based charts; however, these differences had overlapping confidence intervals [29]. Another comparison between Gardosi’s customized model and INTERGROWTH-21^st population-based standard found that INTERGROWTH-21^st both overestimated and underestimated SGA in healthy cohorts from 10 countries, while the customized curve was closer to the tenth percentile for most of the cohorts [30]. These findings have raised doubts about the one-size fits all approach to population-based fetal growth standards [30].

However, these studies primarily evaluated the above birth weight standards in high-income countries and excluded African countries. Africa is home to the most ethnic diversity globally [7]; thus, validating birth weight standards on the continent is paramount. Care must be taken when including African cohorts to note the ethnic diversity that exists within countries given the history of continental migration and the creation of national borders primarily through colonialism. Through this lens, we recommend for the exclusion of endemic diseases in order to more accurately assess fetal growth, an approach demonstrated to have a higher sensitivity for identifying SGA [19,20].

Clinical implications

Our findings highlight the importance of a globally inclusive patient population when creating a birthweight standard. We found that the WHO International, based on 10 countries, overestimated fetal growth in our cohort and had the lowest percent agreement of 20%. The cohort in their study was 57.8% Caucasian [3]. However, the WHO Uganda curve, based on the WHO’s global reference, was similar to our Ugandan standard but had a percent agreement of just 51.8%, which may reflect the additional considerations of sex and malaria in the Ugandan standard. The WHO’s global reference algorithm is a combination of data collected and validated from 24 countries, covering Africa, Latin America and Asia, and the utilization of Hadlock’s fetal growth equation [19]. These findings illustrate that diversity in the creation of global reference algorithms (used to insert local data) and standards is important. Based on this, we find, similar to Gardosi [27], that a growth reference algorithm may be more applicable than a population standard. However, we found that a growth reference can be global, if the populations included are representative of the global community; thus, customized curves to each individual may not be needed. Additionally, customized curves requiring multiple data points on maternal characteristics may not be practical in low- and middle-income countries where that information may not be readily available.

With an applicable global reference, low- and middle-income countries can utilize appropriate data to categorize which neonates are SGA and therefore at high risk of morbidity and mortality. Interventions designed to alleviate these risks will have better population data on the magnitude of the issue, highlighting the larger population of neonates that require assistance. Furthermore, allocation of resources can adequately be restructured when the scope of the epidemiological impact of SGA is better defined.

Our study also addresses the exclusion of placental malaria infection, in malaria-endemic regions, for the purpose of obtaining a healthy growth standard. Schmiegelow’s curve, based on a population in a malaria-endemic country, performed similarly to our Ugandan standard, although they only excluded clinical malaria and asymptomatic parasitemia by blood smear and rapid diagnostic testing (RDT) [20]. However, there was a trend towards under diagnosis of SGA which may be a limitation of the small sample size. We found that 74% of cases with asymptomatic parasitemia also had placental malaria. Based on these findings, assessing for asymptomatic parasitemia may be an effective proxy for placental malaria in relation to creating healthy population standards for neonatal growth. This result is consistent with prior studies that demonstrated an association between peripheral parasitemia during pregnancy and placental malaria [12,31]. Positive testing with RDT, used to assess parasitemia in the Schmiegelow study, has been found to be strongly associated with placental malaria [12,31]. Based on these results, it may be reasonable to utilize RDT to assess for parasitemia as a proxy for placental malaria, but future studies with larger sample sizes are needed.

Future research directions

Our findings highlight the need for further research into a global birth weight reference or standard that is based on a diverse global population and is validated in cohorts from various high-, middle-, and low-income countries. Additionally, a larger cohort of placenta malaria negative neonates is needed to assess for differences that our small sample size may be underpowered to evaluate. Also, robust data from prospective fetal ultrasound data in well-characterized populations is needed to generate ideal fetal growth curves, which can potentially identify at risk pregnancies. Lastly, learning more about what factors contribute to normal fetal growth trajectory, with data on fetal ultrasound measurements during pregnancy, is needed to better assess deviations from the norm and subsequently identify potential identification markers.

Strengths and limitations

Our study was limited by a small population size and we may have been underpowered to find differences between similarly performing birth weight standards and references. All of the patients in our database were taking intermittent preventative therapy in pregnancy (IPTp) for malaria and thus, we may have underestimated the true prevalence of placental malaria in the general population where IPTp uptake is not 100%. Additionally, because women in this study volunteered to be involved in these trials, there may an element of selection bias that may limit the generalizability of our outcomes. Because the study was based upon birthweights, it is unclear how well these values will translate to estimated fetal weights and characterize fetal growth restriction. Future research is needed in this area.

To our knowledge, this is the first study to strategically create a Sub-Saharan birth weight standard based on obstetrical ultrasound dating that excluded maternal malarial infection over the course of pregnancy in HIV-negative women. The women were closely followed from 12–20 weeks through delivery. Because of the prospective collection of data, we were able to create a Ugandan standard based on a true healthy reference population.

Conclusions

Our findings, along with the current literature, highlight the importance of utilizing healthy populations to establish reference standards. Excluding women with malaria in malaria-endemic areas, creates birth weight norms at higher values and increases the detection of SGA. We found that excluding malaria does not have to be as extensive as doing placental pathology, but should include assessment of clinical malaria and asymptomatic parasitemia. Ensuring a true healthy population reference will better identify neonates who are pathologically small and require intervention.