INTRODUCTION
Each year, over 20 million infants worldwide are born with low birth weight (LBW), increasing their risk of neonatal mortality and childhood morbidity [Reference McCormick1]. LBW in resource-poor settings consists largely of small-for-gestational-age (SGA) births [Reference Kramer2–Reference Villar and Belizan4]. Although some SGA births are genetically determined, most result from intrauterine growth restriction (IUGR), which is characterized by insufficient transfer of nutrients and oxygen to the fetus and impaired growth of fetal organs and tissues. IUGR may result from limited availability of maternal micro- and macro-nutrients (maternal under-nutrition), or from medical conditions (e.g. hypertension or infection), that impede proper vascularization of the placenta and restrict the transfer of essential nutrients from mother to fetus [Reference Kramer2, Reference Resnik5, Reference Prada and Tsang6].
In resource-poor settings such as sub-Saharan Africa, pregnant women are frequently undernourished and at increased risk of malaria infection, making them particularly vulnerable to delivering an SGA infant. Malaria infection [Reference Sullivan7–Reference Steketee13] and maternal anthropometric indicators of under-nutrition, including short stature [Reference Kramer2, Reference Fedrick and Adelstein14], low pre-pregnancy weight [Reference Fedrick and Adelstein14, Reference Edwards15] or low body mass index (BMI) [Reference Kramer2, Reference Neggers and Goldenberg16], low upper arm fat mass [Reference Kramer2], and inadequate pregnancy weight gain [Reference Strauss and Dietz17, Reference Cedergren18] are independently associated with an increased risk of SGA at delivery.
To date, studies of fetal growth in sub-Saharan Africa have been limited to describing risk factors for occurrence of SGA births. Studies describing in utero fetal growth are limited, due largely to a lack of ultrasound resources necessary to diagnose IUGR. The objective of our prospective cohort study was to assess the effects of maternal malaria and under-nutrition on the risk of IUGR in an urban, resource-poor African population.
MATERIALS AND METHODS
Study population and recruitment
This longitudinal prospective cohort study was conducted between May 2005 and May 2006 among pregnant women seeking antenatal care at Binza Maternity Hospital in Kinshasa, Democratic Republic of Congo. Binza Maternity Hospital is one of the oldest maternities in Kinshasa and serves a predominately urban population. The study was designed to assess feasibility and refine ultrasound protocols in preparation for a larger subsequent trial; no formal sample size calculations were conducted as the sample was selected for convenience.
During routine antenatal registration, all women identified as aged ⩾18 years with a fundal height or last-menstrual-period-derived gestational age of ⩽23 weeks were invited to receive an ultrasound examination to confirm gestational age. All women with an ultrasound-confirmed gestational age ⩽22 weeks were invited to participate in the longitudinal study. Women with high blood pressure at baseline (systolic ⩾140 mmHg and/or diastolic ⩾90 mmHg), multiple gestations, or an ultrasound detected fetal abnormality were excluded. Participant HIV status was determined through Binza Maternity Hospital's routine HIV voluntary counselling and testing programme and reported to our study staff through a linked patient identification number. HIV-positive women were included in the longitudinal follow-up, but excluded from the analyses described here. All participants provided written informed consent; ethical and study protocol approval were obtained by the Institutional Review Boards at the University of North Carolina at Chapel Hill, USA and the University of Kinshasa School of Public Health, Democratic Republic of Congo.
Baseline and follow-up visits
During a baseline interview, sociodemographic characteristics, alcohol, tobacco and drug use, medical and obstetric history, malaria symptoms and current use of anti-malarial drugs were collected. Using standard techniques [19], maternal weight was measured to the nearest 0·1 kg (SECA digital scale Model 890), and maternal height (without footwear or head cover) and mid upper arm circumference (MUAC) to the nearest 0·1 cm. Blood pressure and body temperature were recorded and a malaria thick smear and haematocrit were prepared from a finger-prick blood sample. All women received an insecticide-treated bed net.
Participants returned for monthly routine follow-up visits until delivery during which the ultrasound examination, medical, and laboratory examinations (including malaria thick smears) were repeated. In accordance with Congolese National Policy, intermittent presumptive therapy in pregnancy (IPTp) with sulfadoxine-pyrimethamine (SP) was provided to all women between 16–27 weeks' and 28–32 weeks' gestation, regardless of malaria status. In addition, all women with positive parasitemia were treated; SP was the first-line treatment, however, quinine, artesunate or camoquine was prescribed if the woman had received SP within the preceding month. Women were instructed to return to the maternity hospital between follow-up visits if they experienced any pregnancy complications or symptoms of malaria. At these interim visits, a malaria thick smear was prepared and medical care was provided at Binza Maternity Hospital's outpatient clinic.
Ultrasound measurements
Fetal biometric measurement of the biparietal diameter (BPD), head circumference (HC), abdominal circumference (AC), and femur length (FL) were taken to estimate gestational age and fetal weight. In the first trimester (gestational age <14 weeks), crown–rump length was used to estimate gestational age. All ultrasounds were performed by a single ultrasonographer (V.L.) with a GE Logiqbook System using standard fetal biometric techniques [Reference Stebbins, Jaffe, Jaffe and Bui20]. A 10% sample of ultrasound images was assessed for quality by a maternal-fetal medicine physician at UNC–Chapel Hill. Gestational age in weeks and days (first ultrasound scan only) and fetal weight in grams (all ultrasound scans) were calculated using Hadlock algorithms programmed into the Logiqbook software [Reference Hadlock21, Reference Hadlock22].
Laboratory methods
One microscopist read all malaria thick smears on site during study visits. However, for quality assurance, a 10% sample of all malaria thick smears was assessed independently by a second laboratory technician. Smears were stained with Giemsa and read counting the number of asexual parasites against 200 white blood cells and converted to numbers of parasites/μl under the assumption that there were 6000 white blood cells/μl [Reference Abrams23]. Anaemia was assessed from finger-prick blood samples and the percent haematocrit was recorded (Clay Adams Readacrit; Becton Dickinson & Co., Franklin Lakes, NJ, USA).
Definitions
IUGR was defined as a binary outcome of <10th percentile of fetal weight for attained gestational age using the Hadlock fetal weight nomogram [Reference Hadlock, Harrist and Martinez-Poyer24]. Socioeconomic status (SES) was defined as a binary composite variable, with those who were currently employed (participant or her partner) and living in a home with toilet facilities, a nearby water source and electricity characterized as high SES. Anaemia was defined as a haematocrit of <30% at each visit.
The ‘recent malaria parasitaemia’ variable is a time-dependent measure that represents whether a woman had a positive smear at any time (including an interim visit) in the interval from her previous routine visit up to and including the visit in which the IUGR measurement was obtained (roughly equivalent to the past month). The ‘cumulative malaria parasitaemia’ variable represents a time-dependent measure of the number of positive smears starting at enrolment up to and including the routine follow-up visit in which the IUGR measurement was obtained. Short stature was defined as height <150 cm and low MUAC as <23 cm at enrolment. BMI was calculated using maternal weight and height at enrolment with low BMI defined as <19·8 kg/m2. Change in maternal weight was calculated between each monthly visit, and categorized as low (<1·5 kg) or adequate (⩾1·5 kg) monthly weight gain. The change in MUAC between around 14 and 27 weeks (second trimester) and 27 and 40 weeks (third trimester) was calculated and dichotomized as a loss in MUAC (<0 cm difference over the trimester) vs. no change or gain (⩾0 cm difference over the trimester). These anthropometric cut-points have been used previously to indicate women at risk of poor birth outcome [25, Reference Krasovec and Anderson26].
Statistical methods
All analyses were performed in SAS, version 8.2 (SAS Institute, Cary, NC, USA). Log-binomial regression models were fitted to estimate risk ratios (RR) and 95% confidence intervals (CI) for IUGR before and after adjusting for potential confounders. To account for repeat outcome measures over the course of pregnancy, the regression models were estimated based on the method of generalized estimating equations with an exchangeable working correlation structure [Reference Liang and Zeger27]. Owing to reduced variability in fetal weight during the first trimester, IUGR is not typically seen until the second trimester; we therefore left-truncated all person-time observations at 22 weeks' gestation, resulting in 758 study visits for which data on both recent and cumulative malaria infections, and IUGR measurements, were available for analyses.
We discovered a malfunction in the scale used to weigh mothers that affected measurements taken during a 3–4 week period. All suspect data-points were removed before calculation of the weight gain variable leaving 588 visits with complete data for all covariates. A second set of log-binomial models was fitted for these 588 visits; the distribution of maternal sociodemographics, under-nutrition, malaria status, and gestational age of IUGR for the excluded visits were similar to those of the entire study population (data not shown).
Models were fitted separately for the incident and cumulative malaria exposure variables and for several maternal anthropometric indicators of under-nutrition. As there is evidence that maternal weight gain and MUAC may have differential fetal effects depending on pre-pregnancy nutritional status and timing during pregnancy, models for these exposures were also stratified for baseline BMI status and trimester of pregnancy, respectively [Reference Hediger28, Reference Abrams and Laros29].
To evaluate effect measure modification of the malaria–IUGR relationship by under-nutrition, an interaction term for malaria and each anthropometric indicator was added to the models. A P value of <0·15 for the interaction term was selected a priori to indicate statistical significance. As previous studies have shown that the negative impact of malaria on birth outcome is more pronounced in primigravid women, we also investigated maternal gravidity as a possible modifier of the malaria–IUGR relationship. In this population however, gravidity was not a significant modifier and we therefore combined all gravidities together for the remainder of the analyses and considered gravidity only as a possible confounder. Other confounders included in adjusted models were identified using a backward elimination procedure with 10% change in estimation criterion or were determined to be a priori confounders of the effect as identified from previous studies of SGA or LBW. We considered maternal age, SES, gravidity and anthropometrics as potential confounders in the malaria exposure models and maternal age, SES, gravidity, and malaria exposure in the models utilizing maternal anthropometrics as the main exposure. In Table 3, risk ratios are presented with and without adjustment for maternal weight gain due to the loss of some data for the weight gain variable as described above.
RESULTS
Recruitment and follow-up
Of 1111 new antenatal care attendees, 33% (n=370) met all initial screening criteria and were scanned to determine gestational age. Of those, 182 were eligible and consented to enrol in the longitudinal study. Reasons for ineligibility included: gestational age >22 weeks (n=154), absent for dating ultrasound (n=24), twin pregnancy (n=6), and no viable fetus present (n=4). Five participants identified as HIV-positive (4%) were retained for longitudinal follow-up, but excluded from the analyses described here, yielding a final analysis sample size of 177 women. Women received on average five ultrasound scans (range 2–8). One maternal death and one loss to follow-up occurred before delivery. Mean gestational age at enrolment was 18 weeks [standard deviation (s.d.)=3].
Quality control
Of 140 malaria thick smears examined twice (10% of all smears), there was one discordant positive and one discordant negative between the two technicians (kappa=0·91, 95% CI 0·78–1·00). Of 750 ultrasound images assessed for quality by a maternal-fetal medicine physician at UNC–Chapel Hill (10% of all ultrasounds images), 92% were deemed adequate for clinical assessment; 7% of questionable quality and 1% poor quality (i.e. not all biometry landmarks clearly visible, shadowing in the image or poor tracing of the length or circumference).
Antenatal malaria
During the entire study period, a total of 157 malaria infections were identified (14 probable recrudescent episodes that occurred within 14 days of a previous positive, despite receiving treatment, were excluded from all analyses). Sixty percent of women had at least one positive smear during pregnancy; 38% had a single incident infection, 15% had two incident infections, and 8% were infected three or more times.
During the 758 study visits after 22 weeks' gestation included in this analysis, a total of 110 recent positive malaria parasitaemias occurred. Three quarters of these infections (78%) were identified at routine study visits and 22% during an interim visit. The majority of infections were P. falciparum only (96%); 1% were P. malariae only and 3% mixed infections (P. falciparum plus P. malariae or P. ovale). The parasite density ranged from 29 to 13380 parasites/μl (geometric mean 98 parasites/μl). The majority of infections were subclinical, with only one accompanied by fever.
Maternal under-nutrition
At baseline, mean BMI was 23·7 kg/m2 (s.d.=3·6). Eleven percent of women were underweight (BMI <19·8 kg/m2), 66% were normal weight (19·8–26 kg/m2), 14% overweight (26–29 kg/m2) and 8% obese (⩾30 kg/m2). Three percent of women had short stature (mean height 161·4 cm, s.d.=6·6) and 14% had a MUAC <23 cm. Mean monthly weight gain was 1·6 kg (s.d.=1·5).
IUGR
After 22 weeks' gestation, a total of 52 fetuses (29%) experienced 76 episodes of IUGR. Of these, 17% were IUGR at only one scan, 8% were IUGR at two scans and 5% at three or more ultrasound scans. Eighty-two percent of the IUGR episodes occurred in the third trimester, with peak prevalence between 28 and 33 weeks' gestation. Receiving antimalarial treatment in the previous month (for IPTp or treatment) was significantly protective against IUGR (RR 0·5, 95% CI 0·3–0·7) (Table 1). Male fetuses were less likely to have IUGR than female fetuses (RR 0·6, 95% CI 0·3–1·0). Maternal anaemia, younger (18–24 years) and older (⩾30 years) maternal age, and gravidity <3 were not significantly associated with IUGR on univariate analyses (Table 1).
Table 1. Baseline and visit-specific characteristics of pregnant women and risk of intrauterine growth restriction (IUGR), Democratic Republic of Congo, 2005–2006
RR, Risk ratio; CI, confidence interval.
* Maternal age, socioeconomic status and gravidity recorded at enrolment only (n=177); treatment recorded at enrolment and each follow-up visit (n=758); haematocrit recorded at enrolment and every other follow-up visit (n=388).
Fetuses that ever experienced an episode of IUGR were significantly more likely to be born LBW (<2500 g) or SGA (Table 2). They were also more likely to be born pre-term, however this finding was not statistically significant. Nearly 50% of fetuses who had IUGR at one or two ultrasound scans, and 86% of fetuses who had IUGR three or more times, went on to be born SGA. IUGR fetuses also had significantly smaller length, head circumference, abdominal circumference and ponderal index at birth. There was no difference in gestational length by IUGR status.
Table 2. Association between intrauterine growth restriction (IUGR), birth outcome, length of gestation and fetal anthropometrics at delivery, Democratic Republic of Congo, 2005–2006
CI, Confidence interval.
* Small for gestational age defined as birth weight for gestational age less than the 10th percentile of the Wilcox standard curve [Reference Wilcox40].
Malaria and IUGR
We observed no significant effect of a recent malaria infection on IUGR risk in either the unadjusted analysis (RR 1·1, 95% CI 0·7–1·9) or after adjustment for maternal age, gravidity and weight gain in the past month (RR 1·5, 95% CI 0·9–2·5) (Table 3). The cumulative malaria parasitaemia data suggest a trend of increasing risk of IUGR with increasing number of malaria episodes; however this trend is not statistically significant. In the fully adjusted cumulative malaria model, a three-fold increase in the risk of IUGR was observed among women infected three or more times (RR 3·3, 95% CI 1·3–8·2) compared to women with no previous infections.
Table 3. Risk ratios and 95% confidence intervals for the association between intrauterine growth restriction (IUGR) and incident and cumulative malaria infection, Democratic Republic of Congo, 2005–2006
RR, Risk ratio; aRR, adjusted risk ratio; CI, confidence interval.
* Malaria parasitaemia recorded at all visits after 22 weeks' gestation (n=758).
† Adjusted for age and gravidity.
‡ Adjusted for age, gravidity and weight gain. Note that this fully adjusted model was fitted using only the 588 observations that had complete weight gain data (see Methods section for further explanation).
Under-nutrition and IUGR
None of the baseline anthropometric indicators of under-nutrition was significantly associated with IUGR (Table 4). Inadequate maternal weight gain (defined as <1·5 kg/month) was more strongly associated with an increased risk of IUGR in women with low baseline BMI (RR 2·7, 95% CI 0·9–8·5) compared to women with adequate baseline BMI (RR 1·4, 95% CI 0·9–2·2).
Table 4. Risk ratios and 95% confidence intervals for the association between intrauterine growth restriction (IUGR) and maternal anthropometric indicators, Democratic Republic of Congo, 2005–2006
RR, Risk ratio; aRR, adjusted risk ratio; CI, confidence interval; BMI, body mass index; MUAC, mid upper arm circumference.
* BMI, height and MUAC as recorded at enrolment (n=177); weight gain data reflects change per month (n=588 for all women, n=75 for low baseline BMI, n=513 for normal BMI).
† Adjusted for age.
‡ Adjusted for age. Models that included weight gain as the exposure were fitted using only the 588 observations that had complete data for this variable (see Methods section for further explanation).
Low monthly weight gain during the second trimester (RR 5·7, 95% CI 1·3–25·0), but not the third trimester (RR 1·1, 95% CI 0·7–1·7) was significantly associated with an increased risk of IUGR. A similar pattern was observed for MUAC change per trimester. Women with a loss in MUAC in the second trimester were nearly three times as likely to have an IUGR fetus (RR 2·7, 95% CI 1·0–7·7) as women who had no change or gain in MUAC, whereas a third trimester loss in MUAC was not associated with IUGR (RR 1·1, 95% CI 0·6–1·9).
Combined effects of malaria and under-nutrition on IUGR
A detrimental effect of both recent and cumulative malaria infection on IUGR risk was stronger among undernourished women (Table 5), regardless of which anthropometric indicator was examined. For example, among women with low baseline BMI, the risk of IUGR associated with a recent malaria infection was nearly three times that (RR 2·8, 95% CI 1·0–8·1) of women unexposed to malaria. However, at normal baseline BMI levels, there was no observed association between malaria and IUGR (RR 0·9, 95% CI 0·5–1·7). A similar pattern was seen among shorter women, women with low MUAC, and women with monthly weight gain <1·5 kg. Analyses of the cumulative malaria parasitaemia variable resulted in a similar pattern, with the joint effect of low baseline BMI and cumulative malaria associated with the largest risk (RR 8·2, 95% CI 3·6–18·5).
Table 5. Risk ratios and 95% confidence intervals for the association between intrauterine growth restriction (IUGR) and malaria, stratified by maternal anthropometrics, Democratic Republic of Congo, 2005–2006
RR, Risk ratio; aRR, adjusted risk ratio; CI, confidence interval; BMI, Body mass index; MUAC, mid upper arm circumference.
† Adjusted for age and gravidity.
‡ Adjusted for age and gravidity. Models that included weight gain were fitted using only the 588 observations that had complete data for this variable (see Methods section for further explanation).
Risk ratio pairs highlighted in bold indicate a significant P value for the interaction term between malaria and the anthropometric indicator (P value <0·15).
DISCUSSION
A longitudinal study of IUGR has never previously been carried out in a malaria-endemic area. We measured fetal growth in utero and identified IUGR in nearly one third of fetuses in this urban, sub-Saharan Africa population. This analysis focused on two component causes of IUGR that have heightened relevance in resource-poor settings: malaria infection and maternal under-nutrition. In this Congolese population, we found that a significant independent effect of malaria was seen only among women with three or more incident infections during gestation. We also found that the effect of maternal malaria varied significantly by maternal nutritional status, and that the highest risks of IUGR were evident among the most undernourished women.
Antenatal malaria may lead to IUGR through accumulation of P. falciparum-infected erythrocytes, and immunity-related monocytes and pro-inflammatory cytokines, in the placental intervillious space [Reference Rogerson30]. Haemozoin, a product of parasite haemoglobin digestion, can also be found in phagocytic leucocytes and within fibrin deposits in the intervillous space. This build-up can lead to thickening of the trophoblast basement membrane and effect uteroplacental arterial development, thus decreasing maternal–fetal nutrient exchange [Reference Brabin31].
Previous studies conducted in areas of high P. falciparum transmission have consistently reported associations between SGA and both antenatal [Reference Sullivan7, Reference Verhoeff12, Reference Steketee13] and placental malaria infection [Reference Sullivan7, Reference Menendez8, Reference Okoko10, Reference Verhoeff12, Reference Steketee13]. Our findings are at variance with these earlier studies, with differences probably stemming from the fact that we screened for malaria at monthly intervals and treated all positive antenatal parasitaemia. Further, virtually all women received two presumptive doses of SP. Routine screening and treatment may have eliminated parasites before they had adequate time to sequester in the placenta and cause damage to the placental vasculature, potentially minimizing the effect of malaria infection. In this study, receiving an antimalarial drug (for treatment or IPTp) in the previous month was independently protective against both incident malaria infection and IUGR, and led to higher attained fetal weight (data not shown), further supporting this hypothesis. Our findings are consistent with two studies of low malaria transmission areas (the Thai–Burmese border and highlands of Ethiopia) that also had frequent monitoring and treatment of antenatal parasitaemia [Reference McGready32, Reference Newman33]. Collectively, these data lend support to the WHO recommendations for malaria prevention and control during pregnancy by showing that even in areas of high malaria transmission, the combination of IPTp, prompt identification and treatment of subclinical malaria infections (case management), and use of insecticide-treated bed nets may prevent fetal growth restriction [34].
Maternal under-nutrition was both an independent risk factor for IUGR and a significant modifier of the association between malaria and IUGR. Chronic pre-pregnancy under-nutrition, low weight gain and inadequate accumulation of fat stores during pregnancy can render a woman incapable of meeting the substantial metabolic demands of pregnancy [Reference Rosso and Rosso35]. The mean monthly weight gain of 1·6 kg for these Congolese women was similar to weight gain reported in other resource-poor settings [Reference Krasovec and Anderson26]. As suggested in previous studies, we found that insufficient maternal weight gain was more strongly associated with IUGR in women with low baseline BMI [Reference Cedergren18, Reference Abrams and Laros29], and that low weight gain in the second trimester increased IUGR risk [Reference Hickey36, Reference Abrams and Selvin37]. Our data also corroborate previous findings that failure to accrue arm fat during the second trimester, but not the third trimester, is associated with lower fetal weight [Reference Hediger28].
The association between malaria and IUGR was consistently two- to eight-fold higher among women with evidence of under-nutrition. In resource-poor settings, it has long been recognized that childhood malnutrition and attendant sequelae influence susceptibility to and severity of malaria infection [Reference Caulfield38]. Repercussions of childhood under-nutrition and malaria infection, such as stunting and low BMI, place pregnant women at increased risk of poor birth outcomes. Further, the joint effects of adult under-nutrition and malaria infection may act on similar physiological pathways to reduce uteroplacental blood flow [Reference Rosso and Rosso35] and decrease maternal–fetal oxygen transfer [Reference Menendez, Fleming and Alonso39].
Limitations and strengths
Although malaria and under-nutrition are common causes of IUGR, other risk factors, such as chromosomal abnormalities, pre-eclampsia or substance use may have played a role. We attempted to minimize the effects of other medical factors through our exclusion criteria and found that reported tobacco, alcohol and drug use were minimal. The extent to which fetuses in our cohort were constitutionally SGA vs. truly pathological IUGR cases remains unknown. Moreover, our IUGR definition utilized a fetal weight-for-age nomogram created from an industrialized country, which may have overestimated the proportion of IUGR fetuses in this resource-poor population. However, sensitivity analyses utilizing other industrialized nomograms, as well as a Congolese-specific nomogram we developed from these ultrasound data, showed that the results were highly robust to different definitions of IUGR (data not shown). We may have also overestimated true pre-pregnancy weight by using maternal weight at enrolment as a proxy; however, any resultant bias is probably minimal because participants were enrolled early in pregnancy, before women in resource-poor countries tend to gain significant pregnancy weight [Reference Krasovec and Anderson26]. Last, this study was designed as a pilot to prepare laboratory, ultrasound and clinical operating procedures for a larger subsequent trial, and thus the sample size was selected for convenience, rather than to maximize power. A larger longitudinal study to replicate these findings is warranted.
CONCLUSIONS
The heightened risk of IUGR seen among women who were both undernourished and malaria-infected underscores the importance of incorporating maternal anthropometric screening and nutritional supplementation into existing antenatal care and IPTp programmes in malaria endemic areas.
ACKNOWLEDGEMENTS
This study formed part of the doctoral dissertation of S. H. Landis, submitted to the Graduate School of the University of North Carolina at Chapel Hill, NC, USA. We acknowledge Odile Muniaka, Crispin Fela, and Hélène Matondo for their hard work and dedication in performing the clinical aspects of the study, David Nanlele for administrative assistance, Paluku Kitsa for laboratory assistance, Dr Luyeye Godefroid and the staff of Binza Maternity Hospital in Kinshasa for assistance with recruitment and delivery procedures, and Dr Amanda Horton for assisting with quality control of the ultrasound images. S. H. Landis is indebted to the NEWAID Foundation, the PEO Foundation, and the Graduate School of the University of North Carolina at Chapel Hill for pre-doctoral funding related to this project.
DECLARATION OF INTEREST
None.
