Menu
Menu
Menu
Menu
The reduction of food intake during pregnancy is part of many cultural and religious traditions around the world. The impact of such practices on fetal growth and development are poorly understood. Here, we examined the patterns of diet intake among Maasai pregnant women and assessed their effect on newborn morphometrics. We recruited 141 mother-infant pairs from Ngorongoro Conservation Area (NCA) in Northern Tanzania and quantified dietary intake and changes in maternal diet during pregnancy. We obtained measurements of body weight (BW) and head circumference (HC) at birth. We found that Maasai women significantly reduced their dietary intake during the third trimester, going from an average of 1601 kcal/day during the first two trimesters to 799 kcal/day in the final trimester. The greatest proportion of nutrient reduction was in carbohydrates. Overall, 40% of HC Z-scores of the NCA sample were more than 2 standard deviations below the WHO standard. Nearly a third of neonates classify as low birth weight (< 2500g). HC was smaller relative to BW in this cohort than predicted using the WHO standard. This contrasts markedly to a Tanzanian birth cohort obtained at the same time in an urban context in which only 12% of infants exhibited low weight, only two individuals had HC Z-scores < 2 and HC’s relative to birth weight were larger than predicted using the WHO standards. The surprising lack of head sparing in the NCA cohort suggests that the impact of third trimester malnutrition bears further investigation in both animal models and human populations, especially as low HC is negatively associated with long term health outcomes.
The present study focuses on Maasai pastoralists inhabiting the Ngorongoro Conservation Area (NCA) (S1 Fig). Maasai are among 60,000 full-time NCA residents who subsist on livestock husbandry and intermittent cultivation [38–40]. Access to natural resources is limited by climate, exclusion from water sources, human population growth, and disease [41]. In addition, federal policy limits opportunities for small-scale cultivation available to NCA Maasai communities [40]. In the NCA area the climate is bimodal, with seasonal precipitations in November and December, followed by longer periods of rainfall from early March to late May [42,43]. During the dry season, acute water shortages require women to travel up to 10 km daily to gather approximately 10–20 L of water for domestic use (personal communication and field observations, 2008–2010). During this period, it is common for significant numbers of cattle to succumb to starvation, leading to diminished milk production, a traditional staple of Maasai diet [43,44]. Although seasonal food insecurity is initially relieved during the rainy seasons, damp conditions and lower temperatures are associated with respiratory infections and zoonotic diseases that further reduce cattle populations [45,46]. While pastoralism remains the dominant mode of subsistence, NCA Maasai communities organize and transact through a closed market system of commodities that include bovine milk, traditional beadwork, clothing, and accessories such as spears, walking sticks, and indigenous medicine (field observations, 2008–2015). Rural to urban migration may also contribute to the economy. These activities allow some access to food markets although quantities are limited and prices are high within the NCA compared to elsewhere in the region (field observations, 2008–2015). Restrictions on cattle grazing, and periodic bans on cultivation also contribute to persistent food insecurity in this community [41]. Study participants came from Endulen Village and surrounding areas within a few hours walking distance. With the assistance of 12 Maasai traditional birth attendants (TBA) we used chain-referral sampling to recruit 141 mothers and their infants (71 males and 70 females). The TBAs were drawn from a radius of 10 km around Endulen (S1 Fig). All births occurred at participants’ homes during the dry season between June and September in 2010. Enlisting TBAs was critical to the success of this project. Access to newborn children is very limited by Maasai tradition and so measurement at birth by researchers would either not have been allowed or perceived as intrusive (field observations, 2008–2015). TBAs regularly provide care to expectant mothers and are present at birth in the majority of cases. Training TBAs to collect these data was the only method available to obtain the data necessary for this study. A further consideration is that Maasai households are dispersed over a large geographic area, and many households are accessible only by footpath. Maasai women rarely attend prenatal clinics, and instead seek the care of TBAs, who provide support during home deliveries. The TBAs were trained to collect data during multiple training sessions and focus groups [47]. They were equipped with data collection field kits that we developed according to community stakeholder input. For baseline comparisons, we used WHO growth standards [48]. In addition, we analyzed a sample of (n = 102) neonates born at Bugando Medical Centre (Mwanza, Tanzania). These infants are of mixed socioeconomic background and from urban and peri-urban communities whose mothers attended antenatal clinics during the pregnancy period to monitor the progress of the pregnancy. During the antenatal visits, they were given health education regarding nutrition. The information provided by healthcare practitioners does not report any specific diet restrictions among these pregnant women. Although our previous work has documented significant growth faltering among Mwanza children [49], the lack of evidence of specific third-trimester food restriction makes this sample suitable for comparative purposes. The inclusion of this cohort in this paper is not intended as a control as there are obviously many factors that distinguish them from the Maasai cohort. Birth outcome data are very sparse in Tanzanian and elsewhere in African low-income countries. For this reason, there is value in providing a comparative dataset from Tanzania to aid interpretation of the results of this study in addition to the WHO standards which serve as the baseline comparison. Ethics approval was granted by the Conjoint Health Research Ethics Board (CHREB–Ethics ID: 23033), University of Calgary, and the National Institute of Medical Research (NIMR–Ethics ID: HQ/R.8A/Vol and HQ/R.8A/Vol I.107), Tanzania. The consent form was translated into KiSwahili and KiMaa. Because low literacy pervades the NCA, potential participants were verbally informed of the study details, and verbal consent was obtained by TBAs. Twelve TBAs were trained on the verbal consenting process over three training sessions. This verbal consenting process involving the TBAs was reviewed and approved by CHREB and NIMR. All participants (TBAs, mothers and infants) were assigned a non-linkable identifier. Interviews data were securely stored on two laptop computers that were exclusively dedicated to the study. Field notes were delivered to a secure storage site upon return from the field. All data were transferred and converted to electronic format, and then securely stored at the Department of Cell Biology & Anatomy, University of Calgary. No information or data were disclosed to members outside the approved roster of study personnel. To investigate maternal food intake throughout gestation, one of us (CP) first conducted a series of group interviews and generated observational field notes. Interviews were held with 12 TBAs who represented a cross-section of ages (30–50 years), were multiparous and had extensive knowledge of maternal practices. The initial group interview focused on the rationale for structuring and implementing a food frequency questionnaire. We recognized that the responses of the TBAs may have been influenced by social desirability. Maasai women may have been reluctant to disclose their actual dietary habits to avoid criticism if they did not conform the socially encourage practice of reducing food intake during the third trimester. To mitigate this potential bias, open-ended questions and informal dialogue to encourage active conversation were combined with participant-observation to confirm interview content. Using a refined version of the initial group interview, further interviews were done to confirm and elaborate on thematic content. Throughout these sessions, intersubjective meaning was established and conveyed as consensus by five or six participants on behalf of each group [50]. Validity and interpretation of data were verified through cross-case comparison, and continuous dialogue with TBAs about the topic of maternal health [51]. These data were transcribed and analyzed using NVIVO 9. On the basis of the group interviews, we developed a food frequency questionnaire (FFQ) to measure maternal dietary intake, and to describe variation in maternal diet by measuring frequency and serving size. Due to the remote and isolated location of Maasai households and pervasive low literacy among participants, logistical challenges were met by developing the FFQ according to previously validated methods that rely on images of traditional food containers [50,52–55]. Initial FFQ content was based on existing nutritional surveys of Maasai communities [20,56]. Food items were established through a series of group interviews with the 12 TBAs [51] based on open-ended questions about maternal diet. These discussions were interpreted in KiMaa, KiSwahili, and English by a community member who was experienced with implementing health research projects. TBAs identified food items commonly consumed during pregnancy. These items were recorded and compared to nutritional surveys reflecting typical maternal diets consisting of a narrow range of food items, and food consumption patterns of decline that mark the onset of third trimester food intake [20,56]. The initial FFQ was further reviewed by 36 women at various stages of pregnancy. These women independently confirmed the relevance of the initial list of items and suggested additional items for the second draft of the FFQ. The accuracy of the FFQ was improved by including habitual portion sizes [52,53]. Thumbnail photographs of FFQ items were presented to participants [55]. Items selected by the participants were pasted into the data collection booklet (S1 Appendix). Food items were measured using two 300 mL volumetric containers (one for dry food, and one for liquid food), which enabled participants to estimate serving sizes akin to those they regularly consumed. Maternal dietary intake data were collected individually by the TBAs 2–3 days postpartum. Women were asked about their diet during early-mid pregnancy and during the third trimester and the TBAs filled out two FFQs per woman, one for each period. Given that the maternal diet is not altered until five to six months gestation, recall for early to mid-pregnancy was no longer than 3–4 months, which is an acceptable window for recall when using the FFQ method [57]. Moreover, because of seasonal food insecurity and limited subsistence options, the NCA diet is monotonous, which lends to reproducibility and accurate recall because fewer types of food items are consumed on a regular basis [58,59]. Tanzania Food Composition Tables (TFCT) were used for estimating the composition of all reported food items [60]. Compiled by the Harvard and Tanzanian Food and Nutrition Centre, the TFCT lists 47 nutrients and more than 400 commonly consumed items for the purpose of assessing links between nutrition and health outcomes [60]. We described caloric intake as kcal/day, and macronutrients (i.e., protein, fat, carbohydrates) as g/day. Since dietary restriction did not commence until trimester three, dietary intake from early to mid pregnancy is defined as “T1-2”, and third trimester dietary intake is defined as “T3”. Linear models were used to compare T1-2 and T3 dietary intake, in which time of gestation was included as a fixed factor, while TBAs and mothers were included as random factors to account for differences among TBAs in data collection and base-line differences among mothers. The models had random intercepts and random slopes to account for differences in the baseline as well as in the responses to the fixed factor. The significance of the differences in calorie and macronutrient intake between T1-2 and T3 was estimated by comparing these models against the null models by a likelihood ratio test using the ANOVA function in R. The mixed models were performed using the function lmer from the package lme4 for R and the p-values were obtained using the lmerTest package [61]. Anthropometric data were gathered, by the TBAs, 48–72 hours postpartum. Using a portable medical hang-scale (Salter Breknell 235-S), BW was measured to 100 grams. Infants were wrapped in a cloth hammock and suspended from the anchored hang-scale. Hammock weight was subtracted from the measured BW. The HC (supraorbital ridge to occipital protuberance) was measured twice to 0.5 cm on supine infants [62] using medical-grade tapes. The TBAs were trained by one of us (CP). To facilitate the measurement procedure and reduce the observer error, we provided images of the scale-face and paper tape in the booklet so that numeric values could be circled instead of hand-printed (S1 Appendix). A total of 140 Maasai infants were measured. Body weight and HC were compared to WHO growth standards for two-day-old infants. The WHO data are based on a large international sample of infants who were born into ideal socioeconomic conditions [63]. An ANOVA test was performed first to assess the presence of systematic measurement error among TBAs. The linear model included the BW and HC as dependent variables, while TBA was set as the independent or fixed factor. The sample used for this analysis includes the 140 infants with anthropometric information. The results of these analyses showed a significant effect of TBA on both BW (F (11, 128) = 3.64, p<0.01) and HC (F (11, 128) = 5.41, p<0.01). Consequently, the variation associated with TBAs, was removed by standardizing all values to the grand TBA mean (mean of the TBA means) for each variable. This removes systematic error due to differences among TBAs in data collection. We checked the TBA standardized data for outliers and eliminated all values that fell four or more standard deviations (SD) from the mean. For our sample size, a cut-off point of 4 SD for outliers was recommended based on simulation studies [64]. Infants with no sex information were also excluded from the subsequent analyses due to the dependence of z-values estimations on this variable. After elimination of outliers and missing data, the NCA sample for anthropometric analysis was composed of 116 individuals. To assess whether the z-scores for HC and BW of the NCA sample differ from the WHO reference, we used a one sample T-test. In order to account for the allometric relationship between HC and BW, the expected HC given the BW was estimated for infants from the two Tanzanian samples on the basis of values derived from the WHO reference. A linear regression was used to describe the relation between both variables and the parameters (intercept and slope) obtained from this analysis were then applied to the estimation of HC from BW. For the NCA sample, the estimations were based on the values adjusted by TBA expressed in cm and grams, respectively.
