Efficacy of maternal choline supplementation during pregnancy in mitigating adverse effects of prenatal alcohol exposure on growth and cognitive function: A randomized, double-blind, placebo-controlled clinical trial

Abstract

Background

We recently demonstrated the acceptability and feasibility of a randomized, double-blind choline supplementation intervention for heavy drinking women during pregnancy. In this paper, we report our results relating to the efficacy of this intervention in mitigating adverse effects of prenatal alcohol exposure on infant growth and cognitive function.

Methods

69 Cape Coloured (mixed ancestry) heavy drinkers in Cape Town, South Africa, recruited in mid-pregnancy, were randomly assigned to receive a daily oral dose of either 2 g of choline or placebo from time of enrollment until delivery. Each dose consisted of an individually wrapped packet of powder that, when mixed with water, produced a sweet tasting grape-flavored drink. The primary outcome, eyeblink conditioning (EBC), was assessed at 6.5 months. Somatic growth was measured at birth, 6.5, and 12 months; recognition memory and processing speed on the Fagan Test of Infant Intelligence, at 6.5 and 12 months.

Results

Infants born to choline-treated mothers were more likely to meet criterion for conditioning on EBC than the placebo group. Moreover, within the choline arm, degree of maternal adherence to the supplementation protocol strongly predicted EBC performance. Both groups were small at birth, but choline-treated infants showed considerable catch-up growth in weight and head circumference at 6.5 and 12 months. At 12 months, the infants in the choline treatment arm had higher novelty preference scores, indicating better visual recognition memory.

Conclusions

This exploratory study is the first to provide evidence that a high dose of choline administered early in pregnancy can mitigate adverse effects of heavy prenatal alcohol exposure on EBC, postnatal growth, and cognition in human infants. These findings are consistent with studies of alcohol-exposed animals that have demonstrated beneficial effects of choline supplementation on classical conditioning, learning, and memory.

Introduction

Prenatal alcohol exposure (PAE) has been linked to a broad range of deficits in growth and neurobehavioral function, collectively referred to as fetal alcohol spectrum disorders (FASD), which comprise the most common preventable cause of neurodevelopmental disabilities worldwide. Prevalence estimates range from 1.1–5.0% in the US (May et al., 2018) to 13.6–20.9% in South Africa (May et al., 2013). Fetal alcohol-related effects are seen on pre- and postnatal growth (Jacobson et al., 1994a,b; Carter et al., 2016a; Day et al., 2002), IQ (Jacobson et al., 2004; Mattson et al., 1997; Streissguth et al., 1990), and learning and memory (Mattson and Roebuck, 2002; Lewis et al., 2015). In infancy, PAE has been linked to poorer performance on the Bayley Scales of Infant Development (BSID; Streissguth et al., 1980; J. Jacobson et al., 1993) and recognition memory and slower information processing (S. Jacobson et al., 1993; Kable and Coles, 2004).

We have identified a strikingly consistent effect of PAE on eyeblink conditioning (EBC), a culturally-neutral, Pavlovian paradigm that involves contingent temporal pairing of a conditioned stimulus (e.g., auditory tone) with an unconditioned stimulus (e.g., air puff) (Jacobson et al., 2008). In our Cape Town, South Africa, cohort, not a single child with full FAS met criterion for conditioning at age 5 years, compared with 75% of non-exposed controls. Two-thirds of the other heavily exposed children also failed to meet criterion for conditioning. PAE-related EBC impairment was seen again in these children at 10 years (Cheng et al., 2017), in a second Cape Town cohort at 11 years (Jacobson et al. 2011a), and in an independent study of school-age children in the U.S. (Coffin et al., 2005).

The neural circuitry involved in EBC has been documented across a range of animal species and across the life-span (Woodruff-Pak and Steinmetz, 2000). By 5 months postpartum, typically-developing human infants reach the same terminal level in delay EBC as adults (Herbert et al., 2003). Heavy alcohol exposure during the equivalent of the 3^rd trimester of pregnancy in humans disrupts EBC in weanling rats (Stanton and Goodlett, 1998) and adults, a deficit that is mediated by dose-dependent cell loss and altered neural activity in the deep cerebellar nuclei (Green et al., 2002). Binge exposure during this period in rodents is also associated with loss of Purkinje and granule cells in the cerebellum (Hamre and West, 1993). Neuroimaging studies of humans, using functional MRI (Cheng et al., 2017), diffusion tensor imaging (Fan et al., 2015), and magnetic resonance spectroscopy (du Plessis et al., 2015), have documented adverse effects of PAE on structure and function of the neural network known to support EBC.

We have been conducting longitudinal research on children and mothers recruited during pregnancy in a Cape Coloured (mixed ancestry) community in Cape Town since 1999 (Jacobson et al., 2008). Despite widely disseminated public health advisories, including educational brochures and posters in antenatal clinics, information about risks of pregnancy drinking provided by nurse/midwives, and numerous psychosocial community-based interventions, heavy drinking during pregnancy continues to be highly prevalent in this community (May et al., 2013). Given the limited effectiveness of these interventions, there is a critical need for pharmacological and/or nutritional treatments that may mitigate the teratogenic effects of alcohol.

Laboratory studies have shown that choline supplementation in rats has remarkable potential to mitigate effects of PAE on a range of behavioral outcomes, including working memory, spatial learning and hyperactivity, and reversal learning (Thomas et al., 2000, 2004, 2007), and trace fear conditioning (Wagner and Hunt, 2006). The beneficial effects of choline supplementation early in development persisted even after treatment was completed, indicating long-lasting changes in central nervous system organization and development. Thomas and Tran (2012) have demonstrated that choline supplementation also mitigates alcohol’s effects on trace EBC. The deficit in acquisition of conditioned responses (CRs) seen in alcohol-exposed rats was not seen in the alcohol-exposed group treated with choline, which performed as well as non-exposed controls. Choline was administered during the equivalent of the 3^rd trimester of pregnancy in these studies. When administered earlier in pregnancy, treatment is even more effective in enhancing cognition in typically developing rats (Meck et al., 1989) and can mitigate effects on brain and body weight, reflexes, and motor coordination in alcohol-exposed rats (Thomas et al. 2009).

Choline, an essential nutrient that is a precursor to the neurotransmitter acetylcholine, plays an important role in cell membrane integrity, trans-membrane signaling, and lipid and cholesterol transport and metabolism (Zeisel and Niculescu, 2006). It also serves as a methyl-group donor needed for homocysteine metabolism and DNA methylation, a critical mechanism in epigenetic processes that have been implicated in alcohol teratogenesis (Zeisel, 2011). Choline is derived from dietary intake, principally eggs, liver, wheat germ, and milk, and from endogenous synthesis, catalyzed by the enzyme phosphatidylethanolamine-N-methyltransferase (PEMT; Resseguie et al., 2007). A common SNP variant (rs12325817) in the promoter region of the PEMT gene confers a markedly higher risk for choline deficiency (da Costa et al., 2006). Choline dietary intake in pregnant women is often much lower than the 450 mg/d recommended by the Institute of Medicine (IOM, 2006). U.S. National Health and Nutrition Examination Survey data show that only 7% of women achieve the adequate intake (AI) level for choline (Chester et al., 2011). In the U.S., the lowest quartile of choline intake in women of reproductive age is 25–50% of the AI (Wallace et al., 2014), and in developing countries, including South Africa, choline intake is even lower (Gossell-Williams et al., 2005; Carter et al., 2017).

In a previous paper (Jacobson et al., submitted), we have reported findings from a pilot, randomized placebo-controlled trial demonstrating the acceptability and feasibility of a maternal choline supplementation intervention for heavy drinking women during pregnancy. In this paper, we report our results relating to the efficacy of this intervention. The aims of this study were to assess (1) the efficacy of prenatal choline supplementation in mitigating adverse effects of PAE on our primary outcome, EBC, and (2) efficacy in mitigating deficits in three secondary outcomes—pre- and postnatal growth restriction, recognition memory, and information processing speed. We also examined the effects of choline on FASD diagnosis and the degree to which choline supplementation is particularly effective in women with choline deficient diets and in women who carry the rs12325817 variant of the enzyme PEMT, which limits endogenous choline synthesis.

Methods

Trial Design

Women were randomly assigned to either choline supplementation or placebo using a randomization list with variable blocks of 2 and 4 subjects and 1:1 allocation ratio generated by a biostatistician not otherwise involved in the trial. All investigators and research staff remained blind to the participants’ treatment groups throughout the trial.

Participants

Recruitment occurred between April 2012 and September 2014; the last infant was born in January 2015. 69 women from the Cape Coloured (mixed ancestry) community participated in the trial. Procedures for recruitment, ascertainment of maternal alcohol consumption during pregnancy, and randomization are detailed in Jacobson et al. (submitted). Briefly, maternal alcohol consumption was assessed using a timeline follow-back (TLFB) interview. Heavy drinking during pregnancy was defined as an average of at least 2 standard drinks (1.0 oz absolute alcohol (AA))/day or 2 or more incidents of binge drinking (4 or more drinks/occasion). Maternal exclusionary criteria were weeks gestation at recruitment >23; age <18 years; HIV or syphilis infection; multiple-gestation pregnancy; pharmacologic treatment for a serious pre-existing medical condition (e.g., diabetes, hypertension, epilepsy, or cardiac problems); use of methamphetamine, a popular drug at the time of recruitment; plans to leave Cape Town before study completion; and dietary choline intake ≥1.5g/day, assessed prior to randomization using a quantitative choline food frequency questionnaire (QFFQ) developed for this study (Carter et al., submitted). Infant exclusionary criteria were major chromosomal anomalies, neural tube defects, seizures, very low birth weight (<1500 g), and gestational age (GA) <32 weeks.

Human subjects approval was obtained from the Wayne State University, University of Cape Town (UCT) Faculty of Health Sciences, and Columbia University Medical Center Institutional Review Boards, and the South African Medicines Control Council. Informed consent was obtained from each mother and the father, if available. An independent data safety monitoring board (DSMB), comprised of a developmental psychologist, obstetrician, neonatologist, and statistician, met with SWJ, RCC, CDM, and JLJ via telephone/Skype prior to the initiation of the trial to review the research plan and every 9 months thereafter to review the progress of the trial and any study-related adverse events.

Treatment Protocol

Each participant was instructed to take 2 daily doses (1 in the morning, 1 in the evening) from time of enrollment until delivery. Each choline supplement dose consisted of 1.25g choline bitartrate, which contained 1g of bioavailable choline cation. The daily choline dose of 2g was chosen in consultation with SZ to maximize potential benefit while being well within the parameters for maternal and fetal safety during pregnancy based on the tolerable upper intake level (UL) for choline determined by the Institute of Medicine Food and Nutrition Board (IOM, 2006) (see Jacobson et al., submitted). Each dose consisted of an individually wrapped packet of powder that, when mixed with 8 oz of water, results in a sweet tasting, slightly fizzy, grape-flavored drink. The choline supplement and placebo were indistinguishable in terms of taste, smell, and appearance. When the participant was given the packets, CDM cautioned her that taking the drink mix would not make it safe to drink alcohol during pregnancy. At the initiation of the trial, the mother was given a box with a 1-month supply of drink packets. At the end of each month, the nurse collected the box (with used and unused packets), replacing it with a new 1-month supply, and number of packets used was tabulated in our laboratory. Adherence was good-to-excellent (median doses taken=74.0%; interquartile range=53.9–88.7) for most participants, and poor adherence (≤33.0%) was relatively rare (11.3% of participants). Plasma choline concentrations increased significantly more in the choline supplementation arm over the course of the trial than in the placebo arm (group by treatment phase interaction, t(105)=3.36, p=0.001; see Fig. 4 in Jacobson et al. (in press).

Maternal Assessments

The TLFB interview was re-administered at 4- and 12-weeks following randomization, and data from the three interviews were averaged to provide continuous measures of alcohol consumption during pregnancy. Alcohol abuse and/or dependence was diagnosed based on DSM-IV criteria using the Structured Clinical Interview for DSM-IV. Extensive information was also collected regarding sociodemographic background and illicit drug use. Nutritional status was assessed using a multiple-pass 24-hr dietary recall interview. Adequate dietary caloric intake was defined as average daily nutrient intake (adjusted using the Nutrition Research Council [NRC] method, Dodd et al., 2006) below the estimated energy requirement (Prentice et al., 1996) for energy or, for nutrients, the estimated average requirement (EAR) or, where no EAR is available, the AI (Institute of Medicine, 2006). Dietary choline intake was assessed on the QFFQ prior to initiation of and twice during the trial. Adequate choline intake, defined as ≥450 mg/day (IOM, 2006), was determined from average intake reported on the three QFFQs. Maternal whole blood samples were genotyped for the PEMT SNP rs12325817, using real-time PCR performed on an Eppendorf Realplex 4.0 (Eppendorf North America, Westbury, NY, USA).

Sample Attrition

A flow diagram of the progression of participants through the trial is presented in Figure 1. 70 women were randomly assigned to condition, but 1 withdrew from the study prior to initiating treatment. Of the 69 in the trial, there were 4 non-study-related fetal deaths (1 spontaneous abortion, 2 stillbirths, 1 fetus whose mother was murdered during pregnancy); 2 women who met a priori exclusionary criteria were removed from the sample (1 twin pregnancy diagnosed after randomization, 1 very preterm delivery (<29 wk gestation)); and 1 woman withdrew after delivery but prior to the 6.5-month infant assessments. In this paper we present data on the 62 infants (31 choline, 31 placebo) who were assessed during the trial.

Figure 1.

Flow diagram of the progression of participants through the trial

Infant Assessments

EBC

EBC was assessed at 6.5 months (with correction for GA in cases of preterm birth (GA<37 weeks)), using the procedure developed by Ivkovich et al. (1999) and Herbert et al. (2003), in which the infant is entertained by a research assistant using a visual display of brightly colored moving objects and toys while being administered the EBC trials. We used the same commercially available human EBC system (San Diego Instruments, Model #2325-0145-W) from our two previous studies with older children (Jacobson et al., 2008, 2011a). The infant wore a headband which supported a flexible plastic tube that delivered an air puff to the right eye, at a distance of ~2.5 cm (Fig. 2). Eyelid closure was measured with a photodiode placed at the corner of the right eye. Above the head, ~45 cm to either side, two 7 Ohm speakers delivered a 1-kHz, 80-dB tone. The interface units generated the auditory conditioned stimulus (CS) and air puff unconditioned stimulus (US), processed the eyeblink signal, and integrated the peripheral devices with the personal computer.

Figure 2.

Headband supporting tube that (1) delivers air puff to right eye and (2) photodiode that measures right eyelid closure, at a distance of ~2.5 cm.

Each session consisted of 51 trials of the delay EBC procedure, in which the air puff was administered simultaneously during the last 100ms of a 750ms presentation of the tone (Fig. 3). The intertrial interval ranged from 8–16s, with an average of 1 trial every 12s. Every fifth trial in each block of 10 was an air puff-alone trial to test for somatosensory responsiveness. Every 10th trial was a tone-alone trial to test for conditioned responding in the absence of the subsequent air puff. In this paradigm the infant must learn to adjust the timing of the anticipatory blink, which occurred optimally between 300 and 650ms after the onset of the tone. Eyeblinks within 350ms prior to the air puff onset were considered CRs (Herbert et al., 2003). Responses that occurred in the first 200ms after the tone onset were coded as alpha or startle responses (see Jacobson et al., 2008). EBC sessions were administered on 3 days. Time between visits did not differ between groups: choline (M=4.0 days between visits) vs. placebo (M=4.5), t=1.22, p=0.226.

Figure 3.

CS and US onset and offset times (2000 milliseconds total) and data recording windows for CR and UR. CS, conditioned stimulus (tone); US, unconditioned stimulus (air puff); CR, conditioned response (blink to CS); UR, unconditioned response (blink to US); a, startle response.

Preprocessing of the data was similar to our previous reports with children (Jacobson et al., 2008; 2011a) with modifications to address movement and other signal-to-noise integrity artifacts that arise when studying EBC in infants (Ivkovich et al., 1999). Sessions in which at least 22 trials provided usable data were considered acceptable. Of the 62 infants assessed at 6.5 months, 11 were excluded (6 choline, 5 placebo) due to excessive movement (noisy data) and/or problems with sensor placement (e.g., sensor became misaligned during the session so that the air puff and photodiode sensor were no longer properly aligned with the eye). An infant met criterion for conditioning if s/he exhibited CRs in at least 50% of the trials administered during at least one of the three sessions. Data were excluded from an additional three infants who failed to meet this criterion in the first two sessions and did not provide usable data for the third session (i.e., did not have the opportunity to demonstrate whether they would have conditioned during the final session). There was no difference between the number of infants excluded from the EBC analyses in the two treatment arms, χ²(1) = 1.48, p = 0.224.

Infant growth

Birth weight, length, and head circumference were obtained from medical records. GA was based on early gestation ultrasound, which was available for 85.7% of the participants, or date of last menstrual period. Infant weight, length, and head circumference were measured at 6.5 and 12 months by research staff trained by RCC using standard WHO protocols; weight, length, and head circumference z-scores and percentiles were calculated using WHO norms, which adjust for age and sex (de Onis et al., 2004).

Fagan Test of Infant Intelligence (FTII)

Visual recognition memory was assessed at 6.5 and 12 months on the FTII (Fagan and Singer, 1983). The infant, seated on the mother’s lap, is first shown two identical photographs and then a novel photograph paired with the familiar one. The normative response, preference for the novel stimulus, indicates ability to recall the familiar stimulus and discriminate it from the novel one. Infant fixation was recorded on a computer, and novelty preference was computed by dividing duration of time looking at the novel stimulus by total time looking at the paired familiar and novel stimuli for each of the 10 problems. Mean length of look, a measure of information processing speed (Colombo et al., 1991), was computed for each problem by dividing the total duration looking time by the number of looks. A pattern of short looks is believed to reflect more rapid and efficient processing of information. Usable FTII data were available for 53 infants (85.5%) at 6.5 months and 52 (83.9%) at 12 months (e.g., fussiness, inattention).

FASD diagnosis

In 2016 and 2017 we organized diagnostic clinics, in which the infants were examined for growth and anomalies independently by HEH and one of four other experienced dysmorphologists, using the revised Institute of Medicine guidelines (Hoyme et al., 2005). All 69 mothers had confirmed heavy alcohol consumption during pregnancy. Infants with at least 2 of the 3 principal dysmorphic features (flat philtrum and thin vermilion [4 or 5 on the Astley and Clarren (2000) lip-philtrum guide] and short palpebral fissures (<10^th percentile); microcephaly (<10^th percentile); and weight or length restriction (<10^th percentile) were diagnosed with FAS. A diagnosis of PFAS required presence of 2 of the 3 facial anomalies and microcephaly or weight or length restriction. The other infants were classified as heavily exposed nonsyndromal (HE). The dysmorphologists, SWJ, RCC, CDM, and JLJ met daily in case conferences following the clinic to reach consensus regarding which infants met criteria for FAS or PFAS diagnoses.

Data Analysis

Statistical analyses were performed using SPSS software (v.22; IBM, Armonk, NY). Demographic background and exposure for the 62 infants whose data are reported in this paper were compared using t-tests and χ² analysis. χ² was used to compare the number of infants who met criterion for conditioning in the choline and placebo arms, and mixed model repeated measures regression was used to compare changes in percent CRs across the three EBC sessions in the two groups. Pearson correlation was used to examine the relation of protocol adherence to percent CRs in Session 3 for the choline and placebo arms separately. T-tests were used to compare anthropometric measures between treatment arms during the newborn period; analysis of covariance, to compare treatment arms on the 6.5- and 12-month measures after adjustment for their corresponding anthropometric measures at birth. Mixed model repeated measures regression was used to compare growth trajectory percentiles between groups from birth through 12 months. T-tests were used to compare treatment arms on the FTII measures. χ² was used to compare incidence of FAS or PFAS in the two groups. Two-way analysis of variance (ANOVA; choline vs. placebo X adequate vs. inadequate choline intake during pregnancy) was used to test the hypothesis that the effects of the choline treatment would be stronger in infants born to mothers with inadequate choline intake during pregnancy. Two-way ANOVA (choline vs. placebo X presence vs. absence of the maternal PEMT polymorphism rs12325817) was used to test the hypothesis that treatment effects would be stronger in infants born to mothers with the polymorphism.

Results

Sample Characteristics

The treatment and placebo arms did not differ in social class or maternal education, verbal and nonverbal intellectual competence, perceived stress, depression, or nutritional status (Table 1). About 30% of both groups were poorly nourished based on their dietary caloric intake, and more than 70% did not meet the AI for dietary choline intake. There were no group differences in alcohol consumption, which was very heavy for both groups at time of conception (≈10.0 standard drinks/occasion on an average of 2–3 days/week). Women in both arms continued to drink heavily across pregnancy (≈8–9 drinks/occasion) but reduced their frequency to about 1–2 days/week. Both groups smoked about ¼ pack/day on average, and cigarettes/day was only slightly higher among those in the choline group. Use of illicit drugs other than marijuana was rare; none of the women reported using cocaine, heroin, or methaqualone. Although women who used methamphetamine were excluded at time of recruitment, 4 reported using it later in pregnancy (2 choline, 2 placebo). There were no group differences in weeks gestation at time of booking, initial screening, or randomization or duration of choline supplementation. Weeks gestation at randomization ranged from 8.6–26.0 (Median=20.0); duration of choline supplementation ranged from 9–27 weeks (Median=18.6). The groups differed on sex of infant, with more girls in the choline and more boys in the placebo arm. There were no group differences in infant age at the 6.5- and 12-month assessments.

Table 1.

Comparison of maternal sociodemographic characteristics and pregnancy alcohol and drug use (N = 62)

	Treatment group		t or χ2
	Choline (n = 31)	Placebo (n = 31)
Maternal characteristics
Maternal age at delivery	26.6	27.0	0.33
	(5.6)	(5.3)
Socioeconomic statusa	20.7	20.8	0.06
	(8.4)	(7.5)
Education (years)	9.1	9.4	0.74
	(1.9)	(1.7)
Marital status (% married)	29.0	35.5	0.30
Parity	2.4	2.7	1.06
	(2.8)	(0.9)
Gravidity	2.9	2.7	0.46
	(1.9)	(1.4)
Peabody Picture Vocabulary Test IQ	46.5	46.9	0.15
	(7.7)	(9.6)
Raven Progressive Matrices	20.4	22.3	0.81
	(11.4)	(12.3)
Beck Depression Inventory	9.0	9.8	0.37
	(9.2)	(9.2)
Life Events (number)	7.5	8.3	0.73
	(3.4)	(4.5)
Perceived Life Stress	28.9	28.8	0.02
	(18.5)	(24.6)
Nutritional status
Dietary caloric intake (kJ/day)b	9319.9	10201.6	1.01
	(2916.8)	(3875.3)
Inadequate dietary caloric intake (%)	29.4	30.3	0.01
Rate of gestational weight gain (kg/wk)	0.4	0.4	0.58
	(0.3)	(0.3)
Inadequate gestational weight gain (%)	58.8	54.5	0.13
Dietary choline intake (mg/day)c
Baseline	368.5	370.9	0.04
	(221.5)	(261.0)
Post-treatment (average)	252.5	290.7	1.22
	(132.3)	(113.4)
Inadequate dietary choline intake (%)	73.5	71.4	0.04
Alcohol during pregnancy
Oz absolute alcohol (AA)/day at conception	1.8	1.6	0.55
	(1.8)	(0.9)
Oz AA/drinking occasion at conception	4.5	5.1	0.87
	(3.0)	(2.8)
Frequency of drinking at conception (days/week)	2.5	2.1	1.12
	(1.2)	(1.0)
Oz AA/day across pregnancy	0.8	0.8	0.17
	(0.9)	(0.7)
Oz AA/drinking occasion across pregnancy	4.1	4.3	0.27
	(2.3)	(2.4)
Frequency of drinking across pregnancy (days/week)	1.3	1.3	0.03
	(0.9)	(0.8)
History of alcohol abuse and/or dependence (%)	19.0	17.0	0.24
Cigarettes/day during pregnancy	6.4	4.2	2.19*
	(4.1)	(3.7)
Marijuana during pregnancy (days/month)	3.5	1.0	1.63
	(7.7)	(3.6)
Methamphetamine during pregnancy (days/month)	0.3	0.01	1.23
	(1.3)	(0.03)
Weeks gestation at 1st antenatal visit	15.7	14.9	0.67
	(4.2)	(4.8)
Weeks gestation at initial screening	17.3	17.4	0.11
	(4.5)	(4.3)
Weeks gestation at randomization	19.3	19.8	0.53
	(3.8)	(3.8)
Treatment duration (days)	131.9	128.6	0.58
	(28.3)	(17.0)
Breastfeeding duration (wk)	38.3	39.3	0.22
	(19.1)	(18.1)
Infant characteristics
Sex (% male)	32.3	61.3	5.25*
Age at testing (months)
6.5 months	6.6	6.7	1.01
	(0.4)	(0.5)
12 months	12.1	12.1	0.35
	(0.3)	(0.3)

Values are mean (standard deviation)

^aHollingshead (2011) Scale

^bBased on multiple pass 24-hr dietary recall interviews, quantified using FoodFinder3^®. 4.18 kJ = 1 kcal.

^cBased on choline-indicated food frequency questionnaire (QFFQ), adjusted for daily caloric intake

^*p < 0.05

Eyeblink Conditioning

Among the 48 infants with usable EBC data, a larger proportion of the choline group met criterion for conditioning compared with the placebo group, although the effect fell short of conventional levels of statistical significance (Table 2). When the four infants whose mothers’ adherence was very poor (<20%) were excluded from the analysis, the effect was significant despite the smaller sample size; two-thirds of choline-treated infants met criterion for conditioning, compared with only one-third of the placebo group. When percent CRs in the two arms was examined across the three sessions, there was a main effect for session, F(2,60)=32.66, p<0.001, but not for treatment, F(1,30)=0.19, p>0.20 (Fig. 4). The infants in the choline arm showed a significantly greater increase in percent CRs across the three sessions than the placebo group (group by session interaction, F(2,60)=4.85, p<0.01.

Table 2.

Effects of choline supplementation on eyeblink conditioning

	Whole sample			Adherencea > 20%
	Choline	Placebo	Total	Choline	Placebo	Total
Fail	9	17	26	6	17	23
	40.9%	65.4%	54.2%	33.3%	65.4%	52.3%
Pass	13	9	22	12	9	21
	59.1%	34.6%	45.8%	66.7%	34.6%	47.7%
Total	22	26	48	18	26	44
	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%
	χ2 = 2.88; p = 0.090			χ2 = 4.38; p = 0.036

^aAdherence to supplementation protocol

Figure 4.

Percent CRs across the three eyeblink conditioning sessions by treatment group.

The relation between protocol adherence and percent CRs in Session 3 for the choline treatment group is shown in Figure 5. An analysis of Mahalanobis distance identified one bivariate maternal outlier (p<0.007) with very poor adherence and high percent CRs. When that outlier was removed from the analysis, there was a clear linear relation between degree of adherence and conditioning by Session 3, r=0.63, p<0.01. By contrast, adherence was not related to percent CRs in the placebo group, r=−0.14, p>0.20.

Figure 5.

Relation of protocol adherence (% packets used) in the choline-treated group to percent CRs during eyeblink conditioning session 3.

Gestational Age and Somatic Growth

The majority of the women (93.7%) had full-term pregnancies, and GA did not differ between groups (Table 3). Although low birthweight (<2500 g) was common (26.1%), the incidence (choline 25.0%, placebo 32.3%) was similar between arms, χ²(1)=0.41, p=0.524. There were also no significant group differences in birth size. With the exception of length in the placebo group, birth percentile measures were markedly below the 50^th percentile, likely due to the high levels of alcohol to which this cohort was exposed.

Table 3.

Effects of choline supplementation on gestational age and somatic growth

	Treatment group		t or Fa	d
	Choline	Placebo
	(n = 31)	(n = 31)
Birth
Gestational age (weeks)	38.8	38.9	0.28	0.05
	(1.6)	(2.5)
Weight (g)	2852.7	2844.2	0.06	0.02
	(450.9)	(658.0)
Weight-for-age z-scoreb	−1.0	−1.1	0.21	0.05
	(1.1)	(1.5)
Weight-for-age percentileb	25.9	28.3	0.32	0.08
	(25.2)	(31.6)
Length (cm)	47.2	48.9	1.82†	0.49
	(3.3)	(3.7)
Length-for-age z-scoreb	−1.2	−0.4	1.61	0.32
	(1.8)	(2.0)
Length percentileb	30.6	47.1	1.75†	0.45
	(34.2)	(38.7)
Head circumference (cm)	32.6	33.0	1.00	0.24
	(1.5)	(1.9)
Head circumference-for-age z-scoreb	−1.2	−1.0	0.66	0.14
	(1.2)	(1.5)
Head circumference percentileb	23.8	30.9	0.99	0.25
	(24.4)	(31.6)
6.5 months
Weight (kg)	7.5	7.3	0.80	0.19
	(1.0)	(1.1)
Weight-for-age z-scoreb	−0.4	−0.8	2.81†	0.26
	(1.2)	(1.3)
Weight percentile	43.8	27.1	5.76*	0.52
	(32.4)	(31.5)
Length (cm)	64.9	64.3	1.04	0.24
	(2.6)	(2.5)
Length-for-age z-scoreb	−1.1	−1.7	6.28*	0.56
	(1.1)	(1.2)
Length percentile	23.7	15.3	2.23	0.36
	(23.4)	(23.0)
Head circumference (cm)	42.6	42.2	1.45	0.29
	(1.5)	(1.3)
Head circumference-for- age z-scoreb	−0.3	−0.9	5.95*	0.61
	(1.2)	(1.0)
Head circumference percentileb	41.9	25.1	5.62*	0.59
	(30.9)	(26.1)
12 months
Weight (kg)	9.0	8.6	1.48	0.31
	(1.4)	(1.2)
Weight-for-age z-scoreb	−0.4	−0.8	2.76	0.25
	(1.4)	(1.2)
Weight percentileb	43.5	27.0	5.13*	0.53
	(33.7)	(28.3)
Length (cm)	71.9	71.0	1.66	0.30
	(3.2)	(2.8)
Length-for- age z-scoreb	−1.1	−1.7	5.28*	0.54
	(1.2)	(1.1)
Length percentileb	22.2	12.4	3.45†	0.45
	(23.2)	(20.0)
Head circumference (cm)	44.7	44.4	1.88	0.19
	(1.7)	(1.4)
Head circumference-for-age z-scoreb	−0.4	−0.9	3.54†	0.47
	(1.3)	(0.9)
Head circumference percentileb	39.2	25.1	4.00*	0.50
	(32.6)	(23.8)

Values are mean (standard deviation) for birth measures. For 6.5 and 12 months, values are mean (standard deviation), adjusted for birth size. Z-scores and percentiles are adjusted for age and sex.

^aT-tests were used to compare the birth measures; analysis of covariance, to compare the 6.5- and 12-month measures after adjustment for their corresponding birth size measures.

^bPercentiles based on WHO norms (de Onis et al., 2004)

^†p < 0.10

^*p < 0.05

Among infants in the placebo arm, z-scores and percentiles for weight, length, and head circumference decreased from birth to 6.5 months (Table 3), which is consistent with our previous studies in Detroit (Jacobson et al. 1994a; Carter et al., 2013) and Cape Town (Carter et al., 2012) that found worsening of fetal alcohol growth restriction in the 1^st year of life. By contrast, infants in the choline arm demonstrated considerable catch-up growth in weight and head circumference by 6.5 months, which continued to be evident at 12 months. In repeated measures analyses examining the percentiles, the choline group showed significantly greater increases across this age period than the placebo group in weight, F(2,116)=4.85, p=0.009, and head circumference, F(2,116)=5.40, p=0.006. Group differences were not significant for the raw weight (kg), length (cm), and head circumference (cm) measures, presumably because the choline arm contained a higher proportion of females, who tend to be smaller than males.

Fagan Test of Infant Intelligence

Infants in the choline treatment arm performed more optimally on the FTII than placebo infants at 12 months, with higher novelty preference scores, indicating better visual recognition memory function (Table 4). The Cohen’s d measure indicates a moderate-to-large effect size. The moderate effect size for the processing speed measure suggests that this effect might be significant in a larger sample.

Table 4.

Effects of choline supplementation on FTII

	Choline			Placebo			t	d
	n	M	SD	n	M	SD
Recognition memorya
6.5 months	26	61.2	6.5	27	61.4	6.7	0.13	0.04
12 months	25	63.5	6.4	27	59.1	7.5	2.23*	0.62
Information processing speedb
6.5 months	26	1.8	0.4	27	1.8	0.4	0.55	0.12
12 months	25	2.0	0.5	27	2.1	0.4	1.00	0.28

Values are mean (standard deviation),

^aPercent novelty preference = (duration looking at novel stimulus/time looking at familiar + novel stimuli) X 100

^bDuration looking time (s)/number of looks to stimuli during familiarization

^*p < 0.05

FASD Diagnosis

Within the choline arm, 8 infants met criteria for FAS and 2 for PFAS (32.3%), compared with 5 who met criteria for FAS and 2 for PFAS (22.6%) in the placebo arm, χ²(1)=0.73, p=0.393.

Impact of Maternal Dietary Choline Status

Only 9.7% (3 choline-treated, 3 placebo) of the 62 infants assessed in this study were born to mothers whose dietary choline intake during pregnancy was adequate (>450 mg/day), and none exceeded the UL of 3.5 g/day. 21.6% of the infants (6 choline-treated, 5 placebo) were born to mothers who carry the rs12325817 variant of the enzyme PEMT, which reduces endogenous choline synthesis. The six outcomes on which the choline and placebo groups differed were examined. None of the choline group by maternal PEMT allele interactions were significant (all ps>0.20).

DISCUSSION

The aim of this trial was to assess the efficacy of a high dose choline supplementation initiated early in pregnancy in mitigating effects of heavy PAE on EBC and other infant developmental outcomes known to be affected by alcohol exposure. Infants born to choline-treated mothers were more likely to meet criterion for EBC conditioning and showed greater increases in CRs across the three training sessions, compared with infants whose mothers received placebo. Moreover, within the choline arm, degree of maternal adherence to the supplementation protocol was a strong predictor of infant EBC performance. We selected EBC as our primary outcome based on evidence that it is a highly sensitive indicator of fetal alcohol impairment in humans (Jacobson et al., 2008, 2011a; Coffin et al., 2005), it is impaired in laboratory animals with PAE (Stanton and Goodlett, 1998; Green et al., 2002), and ethanol-induced EBC impairment can be mitigated by choline supplementation in the animal model (Thomas and Tran, 2012). Given that by 5 months of age, human infants reach the same terminal levels of conditioning as adults (Herbert et al., 2003), EBC provides an excellent early biobehavioral marker (Jacobson et al., 2011b) of a PAE-related deficit that can be used for evaluating the efficacy of a micronutrient intervention in pregnancy.

Pre- and/or postnatal growth restriction was identified clinically as a critical feature of FAS when the syndrome was first described (Jones and Smith, 1973) and is a key element in the diagnosis of FAS and PFAS (Hoyme et al., 2005). The finding that choline treatment mitigated postnatal but not fetal growth restriction is consistent with evidence that the effects on growth restriction during these two developmental periods are mediated by different mechanisms (Jacobson et al., 1994a; Carter et al. 2016b; Middaugh and Boggan, 1991). We have previously shown that growth trajectory pattern predicts severity of fetal alcohol-related cognitive impairment (Carter et al., 2016a). Correlations between PAE and cognitive function in childhood were significantly stronger in infants with both pre- and postnatal growth restriction than in infants demonstrating postnatal catch-up growth. Our finding that choline treatment mitigated adverse effects of PAE on postnatal growth thus provides some evidence from infancy that choline supplementation may have potential to protect against a range of alcohol-related cognitive deficits that do not become evident until childhood, although beneficial effects in childhood will need to be confirmed in follow-up studies.

On the FTII, infants in the choline arm exhibited better visual recognition memory than those in the placebo group, as indicated by preferential looking at the novel stimulus. A larger sample would be needed to evaluate the effects on processing speed. By contrast to classical tests of infant development, such as the Bayley Scales, FTII performance has been repeatedly shown to predict intellectual function in childhood (Kavšek, 2004). Whereas Bayley assessments rely heavily on sensorimotor manipulation of objects through 12 months, the FTII assesses basic information-processing skills involving encoding and retrieval (Bornstein and Sigman, 1987). The predictive validity of the FTII is consistent with the assumption that the neural integrity and elementary cognitive processing ability necessary for later more complex cognitive function are already present in infancy and that the processes by which infants distribute their attention during visual processing of novel and familiar processes are essentially the same as those used by older children and adults.

Choline supplementation did not alter the incidence of FAS or PFAS. Mouse model studies have demonstrated that the dysmorphic craniofacial features required for these diagnoses are formed during the early embryonic period (Sulik, 2005). We did not predict that this treatment would protect against emergence of these anomalies, since the intervention was not initiated until the 2^nd trimester for all but three study participants, who were randomized at the end of the 1^st trimester.

Despite numerous studies showing positive effects in rodents, no clear evidence of beneficial effects of choline supplementation was seen in three previous human trials, two of which assessed postnatal supplementation in childhood. In a pediatric study, Wozniak et al. (2013) administered 500mg/day choline (2–2.5 times AI) or placebo to 60 children (2.5–5 yr) for 9 months. Protocol adherence was confirmed by the finding that higher serum choline and betaine concentrations were found in the choline supplemented group. Positive effects were not seen on the primary outcome, the Mullen Scales of Early Learning (Wozniak et al., 2015). Elicited imitation was assessed as a secondary outcome. There was no effect of choline on immediate imitation, which assesses initial memory encoding, and an effect on delayed imitation that was seen only in the younger group (2.5–4.0 yr) may have been due to a marginal baseline difference between the groups and/or a ceiling effect. In another pediatric study, Nguyen et al. (2016) administered 625 mg/day (1.7–2.5 times AI) to 55 children (ages 5–10 yr) with FASD for 6 weeks. No effects were seen on memory, executive function, attention, or hyperactivity in relation to mean dietary choline intake despite good adherence to the treatment protocol.

In a study of 367 infants in the Ukraine conducted by Chambers and colleagues, two groups (alcohol-consuming women; abstainers and women who drank minimally) were randomly assigned to receive a choline supplement (750 mg/day (1.7 times AI)) or placebo during pregnancy. No beneficial effect of the choline supplement was seen on newborn growth, after adjustment for smoking, or on BSID-II cognitive or psychomotor performance at 6 months (Coles et al., 2015). However, adherence estimates were uncertain since they were based on maternal report (Kable et al., 2015), and significant increases were not seen in the serum choline levels of the mothers in the choline-treated group during the course of the intervention (Coles et al., 2015). Data on cardiac orienting responses (ΔHR) were collected on a subset of these infants during auditory and visual habituation/dishabituation learning paradigms administered at 12 months (Kable et al., 2015). No choline effects were seen on the auditory task. On the visual task, which examines aspects of cognitive function similar to those assessed in the FTII, choline supplementation resulted in greater ΔHR (amplitude and latency) during habituation for all infants (i.e., when the alcohol-exposed and control infants were pooled) and on amplitude only for the infants with no PAE during dishabituation.

Several features of our intervention distinguish our protocol from those used in the previous human studies. First, the choline dose was higher (2g/day)—4.4 times AI—albeit well below the tolerable UL for choline specified by the Food and Nutrition Board (IOM, 2006). The efficacy of this high dose is consistent with the beneficial effects reported for the high doses used in the animal studies. For example, Thomas and Tran (2012) administered 100mg/kg bodyweight, which would be the equivalent of 6g/day in a 60-kg woman. Although it is difficult to compare between-species differences in dose, in one study the human effective dose in the brain was found to be underestimated by 36% using rat data (Tolvanen et al., 2010). If the 2g/day dose in our study were increased by 36%, the effective dose would be 2.7, still considerably below the dose used in the Thomas et al. rodent studies. Second, the treatment was initiated much earlier in development than in the pediatric studies and possibly earlier than in the Ukraine study. Our data are thus consistent with findings from laboratory animal studies suggesting that choline supplementation may be more effective when administered earlier in development (Meck et al., 1989; Thomas et al., 2009). Another strength of this trial was the use of culture-free assessments (EBC, Fagan, anthropometry) that permit generalizability to other populations.

Several studies of effects of ethanol with laboratory animals have demonstrated beneficial effects of choline supplementation on working memory and learning (Thomas et al., 2000, 2004, 2007), and conditioning (Wagner and Hunt, 2006; Thomas and Tran, 2012). Although these effects were initially attributed to increased release of acetylcholine in the hippocampus, the amount of choline that accumulates in the fetal brain following maternal supplementation does not appear to be sufficient to enhance acetylcholine release (Garner et al., 1995). Zeisel and colleagues have suggested two alternative mechanisms: (1) Choline is a potent source of biologically labile methyl-donor groups, which play an important role in DNA methylation, a critical mechanism in epigenetic processes leading to altered gene expression (Zeisel and Niculescu, 2006). Through changes in gene expression, choline may enhance a range of embryological and fetal developmental processes, including neuronal precursor cell proliferation, differentiation, and apoptosis (Wang et al., 2016). Gestational choline supplementation has been shown to prevent PAE-modulation of histone and DNA methylation in hypothalamic proopiomelanocortin neurons (Bekdash et al., 2013) and to reduce hyper-methylation associated with PAE in the hippocampus and prefrontal cortex (Otero et al., 2012). (2) Choline’s role as a precursor to two cell membrane constituents (phosphatidylcholine, sphingomyelin) and to cell signaling factors (platelet activating factor, sphingosylphosphocholine) has also been suggested as a mechanism contributing to its neuroprotective properties (Zeisel and Niculescu, 2006).

In their laboratory animal study on effects of postnatal choline supplementation on EBC, Thomas and Tran (2012) examined two groups of rats during adulthood—one group trained on delay EBC, the procedure used with the infants in our study; the other, on trace EBC. In delay, the CS precedes, overlaps, and co-terminates with the onset of the US, whereas in trace, there is a brief stimulus-free “trace interval” between the offset of the CS and the onset of the US. In our study, we administered delay conditioning because, although learning rates are slower, typically-developing infants demonstrate patterns of acquisition for delay that are indistinguishable from adults in terms of asymptote and timing of response, whereas trace conditioning during infancy is at best rudimentary (Herbert et al., 2003). Thomas and Tran found that, although performance of adult rats exposed to ethanol early in development was significantly impaired on both tasks, choline supplementation mitigated the effect on trace but not delay conditioning. They suggested that the selective deficits in trace conditioning that they observed were likely related to hippocampal dysfunction and that choline did not attenuate effects on delay because it is more dependent on the functional integrity of the cerebellum and less on the hippocampus. However, although animals can perform delay EBC even after the hippocampus has been removed, an intact hippocampus has been found to modulate delay conditioning in both animals and humans (e.g., Solomon et al., 1993; Cheng et al., 2017). The hippocampus may actually be more involved in delay conditioning during early development, as hippocampal injury has been shown to impair delay conditioning in weanling and juvenile rats (Ivkovich and Stanton, 2001). Thomas and Tran’s failure to find the beneficial effect of choline on delay conditioning seen in our study may be related to differences in the period of exposure to alcohol or choline, in the age when EBC assessments were performed, or a difference between rodent models and humans.

In the U.S., choline supplementation is not part of standard prenatal care and dietary choline intake in pregnancy is often inadequate at rates similar to those seen in Cape Town, which can increase the risk of delivering an infant with a birth defect (see Shaw et al., 2004). PAE can further disturb the metabolism of choline and other methyl donors, thereby potentially exacerbating dietary inadequacy (Zeisel, 2011). The American Medical Association has recently issued a policy recommendation for inclusion of choline in prenatal supplements (https://wire.ama-assn.org/ama-news/ama-backs-global-health-experts-calling-infertility-disease).

Zeisel (2011) has suggested that choline supplementation is likely to be most effective in pregnant women with choline deficiency, based on findings in the animal model (Thomas et al., 2000; Idrus et al., 2017). Choline deficiency can be caused by either inadequate dietary choline intake and/or the presence of the rs12325817 variant of the enzyme PEMT, which reduces endogenous choline synthesis. We found a lower incidence of this polymorphism in the Cape Coloured population compared with the prevalence reported in a Chapel Hill, NC, sample (da Costa et al., 2006). Given the small number of women with adequate choline intake and/or the PEMT variant, our sample lacked power to determine whether choline supplementation is differentially effective in these women. It should be noted that the high prevalence of inadequate choline intake in this population (>88%; Carter et al., 2017) and in this sample (>70%) may have enhanced the efficacy of the choline supplementation, although, as noted, dietary choline deficiency is also found in 93% of U.S. women (NHANES survey, Chester et al., 2011; also see Wallace et al., 2014).

Limitations

One limitation of this study was small sample size; the findings need to be confirmed in a larger trial. In addition to the large effect sizes on EBC, postnatal growth, and 12-month recognition memory, the moderate effect size on processing speed suggests another effect that warrants examination in a larger sample. Although there is a greater likelihood of subject loss in infant studies, due to movement artifact, fussiness, etc., compared to older children, our subject loss was somewhat lower than in other comparable studies (e.g., Herbert et al., 2003; Taylor and Herbert, 2013). A multi-site trial will be needed to assess generalizability to other populations, including those from differing ethnic backgrounds and populations that do not have the high incidence of poor nutrition found in this sample. Whether the choline supplementation prevents or repairs fetal alcohol-related impairment to brain structure and function is not clear and needs to be examined using animal models and in human neuroimaging studies. It is also not clear whether choline supplementation specifically targets alcohol-related impairment or improves development in all infants. A human study in which both alcohol-consuming and abstaining mothers are randomly assigned to receive choline at high doses or placebo early in pregnancy is needed to address this question.

Conclusions

This exploratory study is the first to provide evidence that a high dose of choline administered early in pregnancy can mitigate adverse effects of heavy PAE on EBC, postnatal growth, and cognition in human infants. These findings are consistent with studies of ethanol-exposed animals that have demonstrated beneficial effects of choline supplementation on classical conditioning, learning, and memory. Given that psychosocial interventions are often ineffective with heavy drinking pregnant women, the evidence that early choline supplementation can mitigate alcohol effects on the developing fetus provides an important breakthrough in treatment of maternal drinking during pregnancy, which is often intractable to intervention. This supplementation program was effective even among highly disadvantaged, poorly educated heavy drinking women. These findings have important public health implications since FASD continue to be a leading cause of intellectual disabilities, and infant outcomes showing positive effects in this trial are predictive of cognitive function in childhood. Follow-up of this cohort will be important to determine whether the positive effects of choline supplementation persist into childhood.