Association between maternal nutritional status in pregnancy and offspring cognitive function during childhood and adolescence; a systematic review

Our exposure of interest was maternal nutritional status during pregnancy: BMI, height and weight; status or intake of selected single micronutrients (vitamins D, B1, B6, B12 and folate) and iron and dietary intake of macronutrients (carbohydrate, protein and fat). Our outcome was any measure of cognitive function in children aged <18 years. We included observational studies and trials, published in English from January 1960 to October 2014, and excluded case reports and animal studies. We searched Medline/PubMed and the Cochrane Library using the medical subject headings (MeSH) terms and text word terms shown in Table 1. A lateral search (screening of reference lists of literature retrieved for review) was carried out.

We included trials if they used a single micronutrient or if it was a multiple micronutrient trial which included intervention groups that differed by a single micronutrient. We did not assess the effects of multiple micronutrient supplements or fatty acids which have been the subject of recent systematic reviews.

From the database search 16,143 articles were identified (Fig. 1). Their titles and abstracts were evaluated, and 57 were eligible. Another 8 articles were identified by lateral search, making a total of 65 for full review. 27 were excluded leaving 38 studies for final evaluation.

Data extraction and quality assessment of each article was carried out independently by SRV and SK. Quality assessment and risk of bias was assessed using a standardized form consisting of 22 criteria (Additional file 1) [7], including information about study setting, population and design, sample selection, baseline characteristics, losses to follow-up, reliability of exposure and outcome measurements, reporting, the appropriateness of data analyses, confounding factors adjusted for and the study results. Discrepancies between assessors were resolved by discussion. Scores indicated a low (>16), medium (12-16) or high (<12) risk of bias.

All were observational studies and from developed countries [8‐19]. Of the 12 studies, 10 examined pre-pregnancy BMI or weight, mostly self-reported [8‐17], and six examined gestational weight gain [8, 10, 13, 17‐19]. Sample size varied from 101 to over 30,000 mother-offspring pairs. Loss to follow-up was high (>30 %) in 7 studies [8, 10, 12, 14, 17‐19]. Some studies excluded children of underweight mothers (BMI < 16 or <18.5 kg/m²) [12, 13, 15, 16].

In the 10 studies of pre-pregnancy BMI and/or weight exposure data were collected during antenatal visits [8‐10, 13‐19] or up to nine months following the index pregnancy or delivery [11, 12]. The Bayley Scale of Infant Development (BSID) was used in three studies [11, 15, 16] but different instruments were used, and different cognitive domains were assessed, in all the others. Assessment was by trained examiners in all except one, where it was based on parental report [14]. None of the papers stated whether outcome assessors were blind to exposure status.

The associations of maternal pre-pregnancy BMI and/or weight, or gestational weight gain with children’s cognitive function were mostly consistent. Nine of the ten studies using BMI categories found that maternal pre-pregnancy obesity (BMI > 29 kg/m²) was associated with lower offspring cognitive function than normal maternal BMI (Table 2) [8‐13, 15‐17]. The effect size was small. For example, in one study, mental development scores at age ~2 years were 0.1 SD lower in the extremely obese maternal BMI category (BMI > 35 kg/m²) [11]. Children of obese mothers scored ~0.3 SD lower in general cognitive and non-verbal abilities but not in verbal or motor abilities at age 5.3 years [8], ~0.1-0.2 SD lower in reading and mathematics scores at age 5-7 years [10] and in mental but not psychomotor development scores at age 1-2 years [15]. In another study, children of obese mothers scored ~2-2.5 points lower in full-scale intelligence quotient (IQ) and verbal-scale IQ but not in performance-scale IQ at age 7 years [17]. Similarly, in a study of two datasets (both with a small sample, N = ~100), in one dataset children of obese mothers scored 0.6 SD lower in performance IQ but not in full-scale and verbal IQ at age 8 years. In the other dataset children of obese mothers scored 0.2-0.5 SD lower in motor, language and cognitive scores at age 2 years, but this was not significant [16]. In this study, the percentage of children who scored below the composite scores in BSID at age 2 years and Wechsler Intelligent Scale for children (WISC-III) at age 8 years was higher in children of obese mothers (BSID: 33 % v 13 %; WISC-III: 50 % v 17 %) [16]. In another study that included two birth-cohorts, risk of intellectual disability (IQ score < 70) was higher in children of obese mothers in one cohort (OR = 2.8; 95 % CI: 1.5, 5.3) at age 11.5 years, but there was no association between maternal obesity and offspring intellectual disability in the other cohort [9]. In the remaining studies effects were found in both overweight and obese categories [12, 13]. Children of obese and overweight mothers scored ~0.1-0.2 SD lower in general intelligence at age 5 and 7 years [12] and performed poorly in a test of executive function at age 7 years [13]. There was no significant association between maternal overweight/obesity and the child’s cognitive function in only one of the 10 studies that included two cohorts [14].

Four of the 10 studies also reported an inverse association between maternal BMI (used as a continuous variable) and offspring cognitive function [8, 12, 13, 15]. For example, cognitive and psychomotor development scores fell by 0.01 SD per unit increase in maternal BMI [15].

In another study, for each kg increase in maternal pre-pregnancy weight there was a small reduction (0.004 SD) in offspring school entry assessment scores at age 4 years, IQ at age 8 years and odds of achieving adequate final exam results at age 16 years (OR = 0.99; 95 % CI: 0.98, 0.99) [19].

Six of the 10 studies also examined maternal underweight as a predictor of the children’s cognitive function. All were from developed countries and based on quite small numbers of underweight mothers. All six, however, found lower cognitive function among children of underweight (BMI <20 kg/m²) mothers compared to normal weight mothers [8‐12, 17]. In one, there was a very small non-significant difference in general cognitive, verbal and non-verbal abilities scores (0.06-0.1 SD) at age 5.3 years [8] and reading and mathematics scores (0.02-0.05 SD) at age 5-7 years [10]. In a study that included two birth-cohorts, risk of mild (IQ score 50-70), but not severe intellectual disability (IQ score < 50), was higher in children of underweight compared to normal mothers in one cohort (OR = 2.1; 95 % CI: 1.0, 4.1) at age 11.5 years, but there was no association in the other cohort [9]. In another study, with no adjustment for confounders, there was a significantly lower general intelligence score (0.2-0.3 SD) at age 5 and 7 years [12]. In the fifth, there was a significantly higher risk of delayed mental development at age ~2 years (risk ratio 1.36; 95 % CI: 1.04, 1.78), but no association when the scores were treated as a continuous variable [11]. In the sixth, children of underweight mothers had lower full-scale IQ scores (~0.6 points; 95 % CI: -1.5, -0.1) [17].

Three out of the six studies that used gestational weight gain as the exposure found fairly consistent effects on offspring cognitive function. Less than expected weight gain in normal weight mothers or more than expected weight gain in normal or already obese mothers had a negative impact on offspring cognition. One found poorer non-verbal intelligence scores in children of mothers who gained >30 pounds (N = 230) compared to those who gained 5-29 pounds (N = 1361) [18]. In another study, children of mothers who gained less weight than recommended had lower school entry assessment scores (0.08 SD) at age 4 years and were less likely to achieve ‘adequate’ final exam results (OR = 0.88) at age 16 years [19]. There was a ~0.07 SD increase in offspring school entry assessment scores and a 0.07 SD increase in 8- year IQ, per 400 g/week gain in maternal weight during early, mid and late pregnancy. In a third study, the association of gestational weight gain with the child’s cognitive function differed in normal and obese mothers [17]. Either inadequate or excessive weight gain in mothers of normal weight was associated with lower offspring IQ scores at age 7 years [17]. Among obese mothers, inadequate gestational weight gain was unrelated to the children’s IQ, but higher gestational weight gain >40 pounds compared with 21-25 pounds was associated with lower full-scale IQ (-6.5 points; 95 % CI: -0.2, -0.11) [17]. The other three studies (two with a small sample (N = 355 [8]; N = 174 [13])) found no association between gestational weight gain and offspring cognitive function [8, 10, 13].

Risk of bias was medium in all 12 studies (score: 13-16). Factors most often responsible for a high risk of bias score were inadequate control for socio-economic status (SES), unmeasured confounders such as maternal intelligence (MIQ) and home environment, low validity of the exposure, small sample and questionable selection, poor reporting about losses to follow-up and observer bias.

All three studies were conducted in developed populations and were observational [20‐22]. Sample size varied from 178 to over 1800 mother-offspring pairs. The proportion of participants lost to follow-up was very high (70 %) in one study [20], high in another (30-40 %) [21] and small (12 %) in the third study [22].

All the studies assessed serum vitamin D concentration using stored samples collected during the second or third trimester. The duration of sample storage was 5 years in one study [20], 20+ years in another [21] and unreported in the third [22]. In all, cognitive function was assessed by trained personnel, though the cognitive domains and the test battery used were different. None reported whether the outcome assessors were blind to exposure status.

Of the three studies, one found a two-fold increase in language impairment in children of mothers with vitamin D deficiency (<46 nmol/L) compared to children of mothers with vitamin D >70 nmol/L [21]. Another study found higher mental and psychomotor development scores (2-3 score points (0.1-0.2 SD)) in children of mothers with normal vitamin D status (>75 nmol/L) compared to those with deficiency (<50 nmol/L) [22]. It also found a positive association between maternal vitamin D concentrations and offspring mental and psychomotor development scores (0.8-0.9 score points (~0.06 SD) per 25 nmol/L increase). The third study, with a small sample (N = 178) found no associations [20].

Risk of bias was medium in all three studies (score: 13-16).

Of 14 studies (13 observational [23‐35] and one trial [36]), four (all observational) were conducted in developing countries [25‐28]. The trial was in three Europeans centres (Germany, Spain and Hungary)) [36]. The sample size varied from 32 to over 39,000 mother-offspring pairs. Loss to follow-up was high (30-50 %) in four studies [24, 28, 30, 36], ~10-20 % in five [23, 26, 29, 33, 34], 1 % in one [35], and unreported in the others [25, 27, 31, 32].

The exposure was measured in different ways, and at different gestational ages. Four studies [23‐26] used plasma/red cell folate and/or total homocysteine concentrations, assessed during the second and/or the third trimester. One African study used folic acid deficiency based on bone marrow or serum folate level [27]. Three studies used daily folate intake during the first and/or the second trimester (calculated from a food frequency questionnaire (FFQ) and/or supplement use) as the predictor [28‐30]. One of these [29] and the remaining five studies used folic acid supplement use (with/without other vitamins and/or minerals) [31‐35]. In two of them supplements were used 4 weeks-3 months prior to pregnancy and/or 2-3 months after conception [31, 32, 35]. In three others there were no details about initiation and duration of supplement use [29, 33, 34]. In the European trial pregnant mothers were supplemented daily with 400 μg of 5-Methyl-tetrahydrofolate alone, or fish oil with/without folate, or placebo from the 20th week of gestation until delivery [36].

Of the 14 studies, three used the Denver development scale [27, 31, 35], two used BSID [23, 28] and others used different batteries to assess cognitive function. The cognitive domains tested varied between studies but were assessed by trained investigators in all except two, which relied on parental report. [31, 32] Only two studies reported whether outcome assessors were blind to exposure status [24, 28].

The trial (N = 154, recruited in three European countries) found no difference in cognitive scores between the folic acid supplemented group and intervention group without folic acid at age 6.5 years [36].

Findings from the four observational studies that used plasma folate or homocysteine concentrations as the exposure were mostly consistently negative [23‐26]. Three found no association of maternal folate or homocysteine concentrations with offspring mental and psychomotor development at age 1-2 years [23], verbal, non-verbal and general IQ, and gross motor development at age 5 years [24] and non-verbal intelligence, attention and memory at age 9 years [25]. One of them had a small sample (N = 154) of well-educated and affluent mothers who had no folate deficiency [23]. In another, mothers were from a socially disadvantaged background, without much variation in folate status [24]. In the third study the sample was very small (N = 108) and information about folate status was inadequate [25]. The other observational study, from India, found a positive association between third trimester maternal folate concentrations and the children’s learning, visuo-spatial ability and attention score (0.1 SD/SD increase in folate concentrations), but none with reasoning, short-term memory or verbal fluency scores, at age 9-10 years. There were no differences between children of deficient (4 %) or non-deficient mothers [26]. In this study there were no associations between maternal homocysteine concentrations and offspring cognitive function. In the African case–control study folate deficiency was associated with abnormal or delayed motor and/or language development at age 6 weeks-4 years [27]. This study had a very small sample size (N = 32) and a high risk of bias (score: 6).

In general findings were fairly consistently positive in the studies in which dietary folate intake was the exposure. In one study, low maternal folate intake (<400 μg/day) was associated with a lower mental development index (MDI, -1.8 score points (~0.3 SD); 95 % CI: -3.6, -0.04), but not psychomotor development index (PDI) score in children of mothers who were carriers of the Methylenetetrahydrofolate reductase (MTHFR) 677 TT genotype, but not in others, at age 1-12 months [28]. In another study, for each 600 μg/day increase in maternal folate intake from food and supplements during the first trimester, children’s receptive language scores increased by 1.6 score points (0.1 SD); 95 % CI: 0.1, 3.1 at age 3 years [29]. There were no associations with folate intakes peri-conceptionally, or in the second trimester, and no associations with children’s visuo-motor scores [29]. In another study, there were no associations of maternal folate intake at any stage of pregnancy with children’s cognitive function at age 7 years [30].

Four out of five observational studies of supplement use found positive associations between the use of folic acid supplements and offspring cognitive function [31‐34]. In a study in the USA, supplement use was associated with better gross motor, but not fine motor or language development in the children at age 3 years (OR = 0.51; 95 % CI: 0.28, 0.93) and a lower risk of poor psychomotor development (OR = 0.48; 95 % CI: 0.25, 0.94) only in the sub-group of African-American children [31]. In another study, maternal supplement use was associated with a reduced risk of severe (OR = 0.55; 95 % CI: 0.35, 0.86) and moderate (OR = 0.82; 95 % CI: 0.69, 0.97) language delay in the children at age 3 years, but there was no association with gross motor skills [32]. In the third, children of mothers who used supplements scored ~4-5 points (0.3 SD) higher in motor skills, verbal ability and verbal-executive function compared to children of non-users at age 4 years, but not in perceptive performance or memory [34], and had a lower incidence of omission (better attention) but not commission errors at age 11 years (incidence rate ratio: 0.80; 95 % CI: 0.64, 1.00) [33]. A small study (N = 96) in a high risk population, with no information about confounders, found no difference in cognitive scores between the supplemented group and the general population [35].

Risk of bias was high in two studies (score: 6 [27] and 11 [31]), low in two (score: 17 [32] and 20 [36]) and medium in the others (score: 13-16).

Of seven studies (all observational) [23, 25, 26, 28‐30, 37] three were conducted in developing countries [25, 26, 28]. The sample size varied from 108 to over 6,000 mother-offspring pairs. Loss to follow-up was 10-20 % in three studies [23, 26, 29] and high (~50 %) in two [30, 37].

The exposure was measured in different ways, and at different gestational ages. In three studies the exposure was plasma B12 concentrations [23, 25, 26], assessed during the third trimester in two [25, 26] and during the second and the third trimester in another (which also assessed holotranscobalamin concentrations) [23]. In four studies, the exposure was average daily dietary B12 intake (FFQ and/or supplement use, self-reported) during the peri-conceptional period or the first and/or the second or the third trimester [28‐30, 37]. Trained assessors assessed cognitive function in all the studies using a similar test battery (BSID) in two [23, 28] and by different instruments in all others. Only one study reported whether outcome assessors were blinded to the exposure [28].

The findings were inconsistent. In a rural Indian population with a very high prevalence (~70 %) of vitamin B12 deficiency, the study was conducted in a very small selected sample of children of mothers with extreme (low and high) vitamin B12 status [25]. Children of mothers in the lowest decile of B12 concentrations (N = 49) performed poorly in tests of sustained attention (182 vs. 159 seconds) and short-term memory (4.3 vs. 4.4 digits) compared to children of mothers in the highest decile (N = 59) at age 9 years. There were no associations between maternal B12 deficiency and scores in tests of intelligence and visual recognition. In an urban Indian study in which ~40 % of mothers were deficient, there were no associations between maternal B12 concentrations (either deficiency or across the range) and children’s cognitive function at age 9-10 years [26]. In a small Canadian study (n = 154), in which only ~ 8 % of mothers were B12 deficient, there were no associations with the children’s language, cognitive and motor skills at age 1.5 years [23]. In a Mexican cohort, low maternal B12 intake (<2 μg/day) was associated with lower MDI, but not PDI score (-1.6 score points (~0.3 SD); 95 % CI: -2.8, -0.3) in the children at age 1-12 months [28]. In an American study, maternal B12 intake from food and supplements during the second, but not during the first trimester and peri-conceptional period, was inversely related to offspring receptive language (-0.4 score points (0.03 SD)/2.6 μg/day; 95 % CI: -0.8, -0.1), but not visuo-motor abilities, at age 3 years [29]. But in the same cohort at seven years, with ~50 % attrition, B12 intake during the first and the second trimester was unrelated to offspring receptive language, verbal and non-verbal intelligence [30]. In a large well-nourished UK sample, with ~50 % attrition, there was no association of maternal B12 intake with offspring IQ at age 8 years. However, there was a significant association, with a very small effect size, between maternal genetic variants linked to plasma vitamin-B12 and offspring IQ [37].

Risk of bias was medium in all seven studies (score: 13-16).

Of eight studies (6 observational [31, 38‐42] and 2 trials, one in Australia [43] and another in China [44], three were conducted in developing countries [40, 42, 44]. The sample size varied from 63 to over 10,000 mother-offspring pairs. Loss to follow-up was ~30 % in two studies, including the Australian trial [38, 43], ~10-23 % in four studies, including the Chinese trial [39, 40, 42, 44], and unavailable in the remainder [31, 41].

The exposure varied in all six observational studies [31, 38‐42]. One used intake of iron supplements, with/without other vitamins/minerals (dose not available), three months prior to pregnancy and/or three months after conception [31]. In this study women also used folate supplements (details presented in folate section). Three studies used serum ferritin and/or haemoglobin concentrations and/or anaemia assessed during the first or the second and/or the third trimester [38, 39, 42]. In one of them, 90 % of mothers consumed a diet rich in iron and took iron supplements (27 mg) daily in the third trimester [38]. The fifth study (in mothers exposed to prenatal methyl mercury) used total body iron stores assessed (based on the ratio of the serum transferrin receptor to serum ferritin) at enrolment [40]. The sixth used haemoglobin level assessed before 18 weeks and after 28 weeks in pregnancy and maternal genes linked to iron or haemoglobin concentrations [41].

In the Australian trial, pregnant mothers received iron supplements (20 mg/day) or placebo from 20 weeks gestation until delivery [43]. In the Chinese trial, pregnant mothers were supplemented daily with 400 μg of folic acid alone (control group), or 400 μg folic acid with 60 mg iron, or 400 μg folic acid with 30 mg iron and multiple micronutrients, from enrolment (<28 weeks of gestation) until delivery [44].

Cognitive instruments differed between studies. Cognitive function was assessed by trained researchers in all except one, in which school scores were self-reported at age 14 years and teacher-rated at age 16 years [39]. In three studies, including the Chinese trial, outcome assessors were unaware about the exposure [39, 40, 44]; in the remaining studies this information was unavailable.

Findings from these studies were fairly consistent. Both trials, and four of the six observational studies found no associations of maternal iron status with offspring cognitive function [31, 38, 40, 41, 43, 44]. This was despite a good sample size and/or follow-up rates in three studies [31, 41, 44]. The remaining two observational studies found positive associations between maternal iron status and offspring cognitive function [39, 42]. One large study in Finland (N = ~10,000), found a small increase (0.03-0.06 SD) in children’s school performance score at age 14 and 16 years for each 10 g/L increase in maternal haemoglobin concentrations during the ninth (but not third or seventh) month of gestation [39]. Children of non-anaemic mothers had 0.04 to 0.07 SD higher school performance scores than children of anaemic mothers. In the other study in Vietnam, infants of anaemic mothers scored lower (-11.6 points (-0.7 SD); 95 % CI: -23.0, -0.2) in BSID composite score at age 6 months than the infants of non-anaemic mothers [42]. Both studies adjusted for multiple confounders (Table 6) but did not adjust for maternal IQ, home environment or the child’s own iron status.

Risk of bias was high in one study (score: 11) [31], medium in three (score: 15-16) [38, 41, 42] and low in the others, including the trials (score: 18-19).

The only study that we found was a trial conducted in rural Gambia [45]. The sample size was 1459 mother-offspring pairs. The children’s age at assessment was 16-22 years; 285 children were <18 years of age. Loss to follow-up was ~20 %.

Mothers in the intervention group received 2 biscuits/day fortified with protein-energy (providing 1015 kcal energy and 22 g protein/day) from 20 weeks gestation until delivery.

Mothers in the control group received the same supplements for 20 weeks postpartum, but not during pregnancy. Cognitive domains were assessed using standard tests by trained examiners who were unaware of the exposure. Additionally, information about school achievement was obtained by questionnaire. Information about compliance was not reported. This trial showed no difference in cognitive test scores and school achievement between the prenatally supplemented group and controls, either unadjusted or after adjustment for confounders. Risk of bias was low (score: 19).

Since the exposures and cognitive tests varied between studies, we were unable to perform a meta-analysis for any of the nutrients.

The review was conducted following CRD recommendations [5] and PRISMA guidelines [6].

Quality assessment was done by two independent reviewers. Most studies had a medium risk of bias and only 2 had a high risk. Exclusion of non-English language literature may have resulted in some important studies being missed. Although some studies with null findings were published (probably because of increased recent interest in this topic), publication bias is another potential limitation. We could not perform meta-analyses due to methodological differences in the published research. We did not include multiple micronutrient trials as there were already two recently published systematic reviews on this topic [3, 4], or trials of fatty acid supplementation which have also been systematically reviewed.

Nine of the ten studies, all from developed countries showed an association of high maternal BMI with poorer cognitive function in the children [8‐13, 15‐17]. The findings are consistent with earlier systematic reviews [2, 46]. Since then three more studies have been published [13, 15, 17]. Although the threshold BMI at which the effect was observed varied between studies, there was evidence of a dose response effect in the majority. For example, compared to the normal-weight category the effect was significant in the extremely obese [11] or obese categories, but not in the overweight category [8‐10, 15‐17] or overweight and obese categories combined [12, 13]. The effect size was generally small, ~0.1-0.2 SD lower IQ/cognitive test scores in children of obese rather than normal mothers, and similar in most of the studies [10‐12, 17]. Most of the studies found the effect for one or more of the mental development domains [10, 12, 13, 17]. In some the effect was found for mental development but not for motor development [8, 11, 15, 16]. Only one study found no association between maternal overweight and offspring cognitive function at age 2-3 years [14], possibly due to the young age at assessment (<3 years); most psychologists think 4 years is probably the earliest age for reliable estimates of cognitive function [47].

The association of maternal adiposity with reduced offspring cognitive function could be due to trans-placental transfer of inflammatory factors from maternal adipose tissue to the fetus [48, 49]. These inflammatory factors, which cross the blood brain barrier, could lead to inflammation of the brain, a reduction in fetal neurotrophic factors, and adversely affect neuronal differentiation, plasticity and function. Rodent studies in which obesity has been induced during pregnancy using high fat diets, have demonstrated increased inflammatory cytokines, lower levels of brain-derived neurotrophic factor (BDNF) in the offspring brain and poorer learning [48, 49]. Such experiments in humans are clearly impossible.

Confounding is another possible explanation for the findings. This is especially important in studies of obesity and cognitive function, both of which are strongly influenced by SES. Despite adjustment for SES, residual confounding could not be ruled out, since SES variables were often limited to fairly crude measures like income or occupation. Furthermore, in developed countries cohort studies have shown that lower IQ during childhood is linked with higher BMI/obesity in adulthood [50, 51]. Lack of adjustment for maternal intelligence (MIQ) could mean that any link between higher maternal BMI/obesity and offspring cognitive function was due to confounding. Only three of the nine studies that showed an association adjusted for MIQ [8, 10, 13].

Five studies from developed countries showed an association of low BMI with lower cognitive scores in the children, with a difference of 0.01-0.3 SD between children of underweight and normal-weight mothers [8, 10‐12, 17]. But these differences were mainly non-significant, possibly due to lack of power, as the underweight category tended to be small. However, in two studies and in one of the two cohorts in another study there was a significantly higher risk of delayed mental development (risk ratio = 1.36) or lower IQ scores or mild intellectual disability (OR = 2.1) in children of underweight mothers [9, 11, 17]. Fetal exposure to nutrient deficiencies might lead to alterations in the neurotransmitter and neuroendocrine systems, and structural brain development [1, 52‐56] and subsequent reduced cognitive function. Since the home environment, parental care and stimulation, in addition to socio-economic factors, influence cognitive function, inadequate adjustment for these factors could leave residual confounding.

Of the six studies that examined gestational weight gain as the exposure, all from developed populations, three found associations of reduced cognitive function which differed according to maternal weight status. Less than expected weight gain in normal weight mothers and more than expected weight gain in normal or already obese mothers was associated with lower offspring cognitive function [17‐19]. This could indicate a causal association of maternal undernutrition or overnutrition with poorer offspring cognitive function for the reasons explained above. However the findings could also be due to confounding for the above explained reasons.

Two of the three available studies, all from developed populations, showed an association of higher maternal vitamin D concentrations with better cognitive function in the children [21, 22]. Both showed evidence of a dose–response relationship. The effect size was modest. For example children of mothers with normal vitamin D status scored 0.1-0.2 SD higher in mental and psychomotor development tests compared to children of deficient mothers [22]. However, the percentage with language impairment was double in children of vitamin D deficient mothers compared to normal mothers in one study [21]. The effect was specific to language impairment in one study [21] and in the other effects were found for both mental and psychomotor development [22]. These findings are consistent with animal studies which have demonstrated poor learning and memory, and alterations in attention, in response to vitamin D deficiency before conception and/or during gestation [57, 58]. The findings are plausible due to a variety of biological actions of Vitamin D fundamental to neurodevelopment, including a signalling role in cell differentiation and synaptic formation [59], gene expression [59], regulation of the metabolism of neurotrophic and neurotoxic factors [60] and a protective role during brain inflammation [61]. Although both studies adjusted for confounders, socio-economic variables were limited to income or occupation. Another limitation was that maternal vitamin D status was available only during the second trimester; since vitamin D is known to fluctuate with sunlight exposure, the timing of deficiency may be crucial in determining cognitive function. A lack of trial data and a lack of studies from developing countries were notable omissions in the literature. Thus, based on a limited number of observational studies evidence linking maternal vitamin D deficiency with reduced offspring cognitive function is not conclusive.

Among the 14 studies reviewed, the findings were mixed. Of the 13 observational studies, mainly from developed populations, seven cohort studies and a case–control study in an African population showed positive associations of maternal folate (plasma concentrations or dietary intake or supplement use) with offspring cognitive function [26‐29, 31‐34]. Specificity varied between the studies. For example, in one there were associations with both mental and psychomotor development [34], while in another it was specific to mental development especially in children of mothers who were carriers of MTHFR677 TT genotype [28] and in others it was found for one or more of motor or mental development domains [26, 29, 31‐33]. One study showed evidence of a dose–response relationship [26]; in others it was impossible to evaluate dose–response effects. Among the positive studies, the effect was quite large. For example, children of mothers who used folic acid supplements scored ~0.3 SD higher in mental and psychomotor development scores than non-users [34]. Children of mothers with lower dietary folate intake (<400 μg/day) scored 0.3 SD lower in MDI compared to children of mothers with adequate folate intake [28].

Folate plays a role in a number of biological actions that could influence neurodevelopment, such as myelination, and maintenance of tissue levels of neurotrophic and neurotoxic cytokines [62, 63]. However, confounding is a concern. Higher dietary intake and/or use of supplements may be an indicator of higher SES or higher MIQ and/or education. Although, the majority of the studies adjusted for confounders, adjustment was usually limited to income, occupation or education. Using self-reported exposure as a measure of nutrient status is another concern. As the majority of studies used exposures measured by self-reported questionnaires, there may be measurement error due to either under- or over reporting. This could introduce bias and limit interpretation. Other limitations of these studies included, parent-reported outcomes, potential observer bias and selective reporting. Furthermore, information about MIQ, home environment, level of adherence among supplement users and the child’s current folate status were generally unavailable. Most of the null studies had low power due to a small sample [23, 25, 35] and sample selection was of concern. For example, mothers had little variation in folate status in one study [24] and no folate deficiency in another [23]. A double blind randomised controlled trial overcomes many of these methodological issues, and the European trial was negative [36]. The trial, though adequately powered, was of reasonable quality, but it had high attrition rates, and compliance was not reported. Maternal folate status was not reported, and a trial in Europe, where women are likely to be relatively folate replete, does not rule out an effect in populations with high rates of folate deficiency.

The seven studies had inconsistent findings. Two of them, both from developing countries (India and Mexico) with high rates of B12 deficiency, found reduced cognitive function in children of deficient mothers or those with low dietary intakes [25, 28]. It was impossible to evaluate dose–response effects. Where present, there was quite a large effect. For example, compared to children of mothers with adequate dietary B12 intake children of mothers with lower B12 intake (<2 μg/day) scored 0.3 SD lower in MDI [28]. The effect was found for one or more mental development domains. The findings may indicate a biological relationship (mechanisms are similar to those of folate reported above). However, residual confounding remains a concern.

Of the remaining five studies, mostly in developed countries, two, conducted in the same cohort at different ages showed an association between maternal vitamin B12 dietary intake and offspring cognition at age 3 years [29] but not at age 7 years [30]. In the remaining three there was no evidence of an association [23, 25, 37]. This could be due to a young age at assessment (<2 years), a small sample, or insufficient variation in B12 status [23]; and the use of self-reported FFQ-based dietary B12 intake [37].

Of the eight studies, two observational studies, one in a developed [39] and another in a developing country [42], found an association between maternal haemoglobin concentrations/anaemia and offspring school performance score [39] or infant cognitive development [42]. The effect size varied, with a difference of 0.04-0.7 SD between the children of non-anaemic and anaemic mothers. The remaining studies, including two trials, found no associations of maternal iron (supplements/haemoglobin/ferritin levels) status with offspring cognitive function [31, 38, 40, 41, 43, 44]. Iron is required for cell differentiation, myelination and neurotransmitter synthesis, and could thus influence neurodevelopment [64]. Iron uptake by the brain is high during the third trimester of gestation, corresponding to the peak of myelinogenesis. However, our review of the available data provides little support for an effect of maternal iron status at this time. This might be due to methodological limitations. For example, in one study maternal supplement use was based on self prescription, the dose of the supplement was unavailable, maternal reported outcome data, and cognitive tests were designed to screen for developmental delay rather than to capture variations within the normal range [31]. In two studies, the sample was small and the children were young (6 months-to- < 3 years) [38, 40]. In one of them and in another large well conducted study there was little variation in maternal iron status [38, 41]. Neither of the trials showed an effect of iron supplementation on the children’s cognitive function, providing even stronger evidence for a lack of effect. The low dose of iron in the Australian trial [41], and high losses to follow-up in the Chinese trial [42] mean that there is still a need for more high-quality trial-based evidence, especially in iron-deficient populations.

In the only study, a trial, generally well conducted but with no information about compliance, there was no evidence of benefit of maternal carbohydrate/protein supplementation on offspring cognition [45]. In a rural Gambian population where women are likely to be malnourished, the dose of the supplement might not be adequate enough to have a long-term benefit; however this trial showed a large benefit for birthweight and infant survival. There is a need for more trials to evaluate the benefit of maternal carbohydrate/protein intake on offspring cognitive function.

The primary and most important limitation was the small number of studies from developing countries. This is very important for several reasons. Nutritional deficiencies tend to be more common and more extreme in developing than in developed countries due to poverty and poor diets. The confounding structure in developing countries is often different from developed countries. In developing countries the burden of poor SES, low literacy and unemployment are more severe than in developed countries; the socio-cultural environment or dietary practices vary between developed and developing populations. Thus the impact of these factors in predicting maternal nutritional status and cognitive function are likely to differ from developed populations. Studies in developing countries may reveal associations between maternal nutritional status and children’s cognitive function that are not detectable in developed populations. A lack of experimental evidence, measurement error due to self-reported exposures and a young age at outcome assessment were major limitations. Only four out of the 34 observational studies and three out of the four trials reported a power calculation, and there is a need for better reporting in the literature on the adequacy of sample sizes. Although the quality score was medium in the majority of studies, some reported sample selection, attrition, power and observer bias inadequately, thus limiting the conclusions that can be drawn.

Experimental studies would provide better evidence. However, such studies are expensive and ethical issues are an important barrier. Iron and folic acid supplementation in pregnancy is now standard in most countries. As already described, a recent systematic review evaluated six multiple micronutrient trials, all conducted in low-income settings, to assess evidence linking maternal multiple micronutrient supplementation and offspring cognitive function. Three of them showed that maternal multiple micronutrient supplementation during pregnancy was beneficial for offspring cognitive function. The authors concluded that the evidence was inconclusive due to transient findings, methodological limitations and inadequate reporting and suggested further research [4]. Three more multiple micronutrient trials, all conducted in developing populations, have been published following this review [65‐67]. Again, the findings were inconsistent and inconclusive. A study in Nepal [65] showed no associations between maternal multiple micronutrient supplementation during pregnancy and intellectual functioning in the offspring at age 7-9 years. Another in Indonesia showed a beneficial effect of maternal multiple micronutrient supplementation during pregnancy on children’s motor and attention/spatial ability at age 3.5 years only in the children of undernourished or anaemic mothers [66]. The third study in Viet Nam showed no associations between twice weekly maternal multiple micronutrient supplementation during pregnancy and offspring cognitive development at age 6 months [67].

Other than vitamins and minerals, fatty acids are another important and essential micronutrient required for optimal neurodevelopment and function. Findings from some observational and experimental studies suggest a beneficial association between maternal fatty acids status during pregnancy and offspring cognitive function. However, recommending routine supplementation of fatty acids and the amount required still remains a topic of debate [68].

There is debate about recommending multiple micronutrient supplementation as a routine instead of iron and folic acid, based on birthweight effects. It could be argued that if all the micronutrients are being supplemented there is no need to study effects of single micronutrient. However, a better understanding of which specific nutrients are important for neurodevelopment, and specific requirements in different settings and populations is required. There has also been concern that there are adverse interactions between micronutrients when supplied in a single preparation [69].