Choline enhances elicited imitation memory performance in preschool children with prenatal alcohol exposure: a cumulative report of 3 randomized controlled trials

Abstract

Background

Fetal alcohol spectrum disorders (FASDs) are associated with neurocognitive deficits for which there are no biological treatments. Choline supplementation may attenuate these deficits.

Objectives

This study was aimed to evaluate choline as a neurodevelopmental intervention for preschool-aged children with FASD.

Methods

We present combined data from 104 participants with FASD (aged 2.5–5.9 y) from 3 placebo randomized controlled trials (RCTs). Participants in RCT1 and RCT2 were randomly assigned to 9 mo choline (500 mg daily) or placebo. Participants in RCT3 were randomly assigned to 9 mo choline (19 mg/kg daily) or placebo. The primary outcome measure was an elicited imitation (EI) memory task.

Results

Adherence was high (78% doses received). Adverse effects were similar across groups except fishy body odor: choline group, 36%; placebo group, 8%. We observed a trend-level choline advantage; participants receiving choline performed 25% better on EI short-delay adjacent pairs (sequential memory) than those on placebo, with a steeper increase in scores between 6 and 9 mo (ŷ = −10.06; P = 0.03; 95% CI: −19.13, −0.99). No sex difference in response was seen, nor did we observe a dose–response relationship. Age-moderated response to choline between baseline and 9 mo (ŷ = 10.02; P = 0.01; 95% CI: 2.47, 17.57), with greater response in younger (≤4.2 y) than that in older (>4.2 y) participants. Overall, choline showed a beneficial effect on memory but no impact on executive functioning or intelligent quotient.

Conclusions

The results support choline as a neurodevelopmental intervention for improvement of memory in young children exposed to alcohol prenatally. Specifically, the use of choline bitartrate as a supplement in the range of 260–500 mg/d in children between 2.5 and 5.9 y of age is supported. Future studies are needed to further define appropriate dosage as well as optimal lengths and developmental windows for supplementation.

This trial was registered at clinicaltrials.gov as NCT01149538 and NCT02735473.

Introduction

Fetal alcohol spectrum disorders (FASDs) represent a serious public health burden, affecting 2.0%–5.0% of the European and North American populations [1,2]. FASD includes fetal alcohol syndrome (FAS), partial fetal alcohol syndrome (pFAS), and alcohol-related neurodevelopmental disorder (ARND) reflecting a range of dysmorphic facial features, growth deficits, neurologic abnormalities, and neurocognitive and neurobehavioral deficits [[3], [4], [5]]. Very few FASD treatments have been developed and rigorously tested [6,7].

Choline, an essential nutrient, has been studied as a neurodevelopmental intervention for FASD [8,9]. Preclinical models demonstrate effects of supplemental choline on brain development [10], hippocampal-dependent memory [[11], [12], [13], [14]], spatial learning [15], and motor development [16]. The handful of human prenatal and postnatal choline supplementation trials suggest beneficial effects on neurobehavioral functioning and brain volumes in infants [[17], [18], [19], [20]], with some variability in findings potentially related to methodologic differences [9]. The specificity of choline’s memory effects may be related to its propensity for altering hippocampal development [[21], [22], [23]]. A small number of postnatal trials have examined choline’s efficacy as a treatment in children with prenatal alcohol exposure (PAE). Our group initially established the feasibility, safety, and tolerability of postnatal choline in children with FASD [21]. We subsequently evaluated efficacy in a randomized, double-blind, placebo-controlled trial [randomized controlled trials (RCT)] of children with PAE ages 2.5–5.9 y, with 500 mg daily choline for 9 mo [22]. Growth curve analyses of the elicited imitation (EI) test (a hippocampal-dependent measure of delayed sequential memory) indicated greater improvement for the choline group than that for the placebo group, with a moderating effect of age (i.e., choline-treated children aged 2.5–3.9 y showed steeper learning than children aged 4–5.9 y). The findings of a greater memory benefit from earlier choline supplementation were consistent with another study showing no effect of choline on neurodevelopment when given in the 5- to 10-y-old range [23]. Two subsequent follow-up studies of our original cohort at 4 and 7 y after intervention showed sustained cognitive benefits and potential improvements in associated white matter microstructure for choline compared with those for placebo [24,25]. Because post hoc analyses suggested that higher daily choline doses per kilogram of body weight were associated with less improvement in memory at 9 mo, we tested a weight-adjusted dose in the subsequent third RCT. That RCT was interrupted by the COVID-19 pandemic, and the sample size was truncated.

In this study, we present combined analyses of data from all 3 RCTs in order to maximize statistical power and compare the dosing regimens in relation to cognitive outcomes. RCT2 and RCT3 were intentionally designed to allow for combined data analyses with RCT1, keeping critical factors consistent across all 3 studies including the following: age, diagnoses, inclusion/exclusion criteria, intervention, duration, outcome time points, outcome measures, monitoring procedures, and key personnel. We hypothesized children receiving choline would outperform children receiving placebo on a measure of hippocampal-dependent memory. We also hypothesized younger children treated with choline would show greater memory benefit than older children. We also evaluated choline’s potential influence on intelligent quotient (IQ) and executive functioning.

Methods

Participants

All participants’ parents underwent a comprehensive informed consent process, and all procedures were approved by the University of Minnesota’s institutional review board. For all 3 trials, additional oversight was provided by the University’s clinical trial monitoring program as well as an independent Data Safety Monitoring Board. Choline was administered under the Federal Drug Administration Investigational New Drug application 107085. The trials were registered on clinicalttrials.gov as NCT01149538 (21 June, 2010; RCT1 and RCT2) and NCT02735473 (22 April, 2016; RCT3) before enrollment of the first participant in each respective study. Complete descriptions of the methods for the first 2 trials have been reported previously [21,22]. The sample sizes for RCT1 (target, n = 20; enrolled, n = 20) and RCT2 (target, n = 40; enrolled, n = 40) were initially chosen to achieve 80% power targeting a Cohen d effect size of 0.43—which would represent a meaningful group change in response to the study drug. The sample size for RCT3 (target, n = 60; enrolled, n = 44) was based on the actual effect size (d = 0.66) that was observed in the data from RCT1 and RCT2. Complete protocols will be provided upon request to the first author (JRW).

Participants, aged 2.5–5.9 y (younger than 6 y) at enrollment, were from 3 RCTs conducted at the University of Minnesota between June 2010 and September 2021. Participants were recruited from FASD diagnostic clinics, including the University of Minnesota’s clinic, as well as support groups for families raising children with FASD. Although not required for inclusion, all participants were living with adoptive families or in other permanent placements (none were in foster/transitional care). Participants were excluded for other neurodevelopmental disorders (e.g., autism and Down syndrome), neurologic conditions (e.g., epilepsy and traumatic brain injury), other medical conditions affecting the brain, or a known history of very low birthweight (<1500 g). The presence of psychiatric comorbidity, such as attention-deficit/hyperactivity disorder, was not exclusionary because comorbidity is common in FASD [26]. Polysubstance exposure is also common in FASD, and we did not exclude on the basis of comorbid drug use.

As shown in Figure 1, across studies, 149 potential participants were screened and 109 who were both eligible and interested were randomly assigned. Of these, 104 were enrolled and received the allocated intervention. Three participants in the choline group had invalid EI data (determined by the test administrator and the individual scoring the test, both before unblinding). Relevant characteristics for the remaining 101 participants who contributed some valid EI data are listed in Table 1. A total of 20 participants discontinued before the final study visit, resulting in a completion total of 84 participants, with 81 of those remaining having contributed valid and complete EI data sets. Ethnicity (Hispanic or non-Hispanic) and race (6 categories) in Table 1 were independently self-reported by participant families. Diagnoses in Table 1 were made using updated, modified Institute of Medicine (IOM) criteria [5], using all available information from the child’s medical records and from a dysmorphology evaluation conducted by the study physician (see procedures further). Of the 101 participants who contributed data to the current set of analyses, 43% met criteria for ARND, 45% met criteria for pFAS, and 12% met criteria for FAS.

FIGURE 1.

CONSORT diagram showing the flow of individuals through study phases including screen failures at enrollment, allocation failures following randomization, participant discontinuations at follow-up, and loss of primary outcome (EI) data due to poor child cooperation with the test and invalid scores. EI, elicited imitation.

TABLE 1.

Demographic characteristics of enrolled participants with valid elicited imitation data included in the analyses.

	Choline (n = 50)	Placebo (n = 51)	Statistical test
Age, mean (SD)	4.06 (0.85)	4.00 (0.87)	t(99) = 0.41
Intelligence quotient, mean (SD)	86.3 (14.4)	87.1 (18.5)	t(89) = −0.22
Female sex, n (%)	27 (54)	31 (61)	χ2 = 0.48
Ethnicity: Hispanic, n (%)	3 (6)	4 (8)	χ2 = 0.48
Race, n (%)1
American Indian/Alaska Native	9 (18)	7 (14)	χ2 = 6.40
Asian	2 (4)	2 (4)	—
Black or African American	7 (14)	17 (33)	—
White	20 (40)	18 (35)	—
Multiracial	11 (22)	7 (14)	—
Unknown	1 (2)	0
Diagnosis, n (%)2
Alcohol-related neurodevelopmental disorder	20 (40)	24 (47)	χ2 = 0.71
Partial fetal alcohol syndrome	23 (46)	22 (43)	—
Fetal alcohol syndrome	7 (14)	5 (10)	—
Global intellectual impairment	11 (22)	15 (29)	χ2 = 0.73
≥3 domains of impairment	49 (98)	49 (97)	χ2 = 0.32
Other drug exposure	37 (74)	40 (78)	χ2 = 0.27

¹The χ² tests reflect comparisons of each racial group to the proportion of participants who identified as White (eg, proportion of participants who identified as multiracial to proportion of those who identified as White).

²The diagnoses listed were made using updated, modified Institute of Medicine (IOM) criteria.

Because IOM criteria do not specifically characterize cognitive functioning, we further applied Centers for Disease Control and Prevention (CDC) central nervous system criteria for FASDs [27]. Global intellectual impairment and domains of impairment in Table 1 were determined from the medical record and cognitive testing that was part of the study itself. Twenty-six participants (26%) met the central nervous system criteria for an FASD diagnosis on the basis of global cognitive impairment/low IQ (≥1.5 SDs below the mean), and 98 participants (97%) had deficits of >1 SD in ≥3 domains (eg, intellectual, language, motor, visual-perceptual, adaptive functioning, and behavioral domains).

Eighty-seven participants (86%) had confirmed prenatal alcohol exposure, including a self-report by the biological mother or social service records that indicated heavy maternal use during pregnancy. Confirmed prenatal alcohol exposure included maternal alcohol use at rank 3 or 4 in the University of Washington diagnostic system [3]. The remaining 14 participants (14%) had unconfirmed but suspected alcohol exposure, and all 14 of these participants had dysmorphic faces and/or cognitive deficits as previously defined. All 14 participants met the modified IOM criteria for FAS (n = 1), pFAS (n =10), or ARND (n = 3). In all cases, alcohol was the predominant substance of misuse, and alcohol use was extensive. However, in 77 cases (76%), other prenatal drug exposure was confirmed or suspected based on information in the medical record. There was not a significant difference in suspected drug use between the 2 treatment arms (Table 1).

Procedures

In all 3 RCTs, potential participants were initially contacted by telephone or letter or they responded to study advertisements. Interested individuals underwent a screening process to determine eligibility with a trained staff member. Eligible children were enrolled and came to the University of Minnesota for 3 in-person visits over the course of 9 mo (baseline, 6 mo, and 9 mo). At the baseline visit (before the allocated intervention), participants completed a set of cognitive tests administered by a trained psychometrist. Participants in the first [21] and second [22] RCTs were administered the Mullen Scales of Early Learning [28] as a measure of IQ, although participants in the third RCT were administered the Stanford–Binet Intelligence Scale, Fifth Edition (SB-5) [29]. The EI paradigm was administered to all participants at each of the 3 in-person visits in each RCT (baseline visit, 6-mo visit, and 9-mo visit). For each EI administration, new events (sets of toys) were used. In addition, parents completed the Child Behavior Checklist (CBCL) at each of the 3 in-person visits. For RCT1 and RCT2, the NIH Toolbox Dimensional Change Card Sort Task (DCCS) [30] was administered, and for RCT3, the Minnesota Executive Function Scale (MEFS) [31], a commercialized version of the DCCS, was administered.

At the 9-mo visit (study completion), children were readministered the IQ measure (Mullen or SB-5). Between baseline and study completion, 8 telephone visits were completed to check adherence and collect adverse event information at 2 wk following the baseline visit and monthly thereafter to monitor supplement adherence and adverse events.

Allocated intervention

All 3 RCTs incorporated a parent-administered supplement of choline or placebo to be given once daily over the course of 9 mo. The 9-mo treatment duration was selected to allow for sufficient developmental change that would be detectable by the outcome instruments. The University of Minnesota’s Investigational Drug Services Pharmacy randomly assigned the participants in each study to receive choline or placebo in a 1:1 allocation ratio based on preprepared computerized block-randomization schedules. A concealed allocation was implemented; the research team and the participants were blinded to group assignment. RCT1 and RCT2 participants received 500 mg daily of choline (from 1250 mg choline bitartrate) or a placebo for 9 mo. RCT3 participants received a weight-adjusted daily dose equal to 19 mg/kg of body weight or placebo; daily doses ranged from 260 to 500 mg choline or placebo. The study drug was supplied in coded light-blocking foil packets containing a powdered, fruit-flavored drink mix developed for the study. Choline and placebo were matched for color and flavor. Packet contents were concealed and identifiable with a code only known to the pharmacy. Parents were instructed to mix 1 packet with 4.0 fl oz (118.3 mL) of water per day and administer the child’s dose (custom premarked measuring cups were provided for each child). Families were sent home with their first 3-mo supply of the supplement and were mailed 2 more 3-mo supplies over the course of 9 mo.

Quality and stability of the supplement packets were evaluated by an outside laboratory using high-performance liquid chromatography. The dosage per packet was within 0.1% of the target, on average, and stable (within 5.2% of target dosage over the study duration).

Measures

Adherence

Adherence and adverse events were assessed through calendar log sheets, serum choline concentrations [21,22], packet counts, and phone visits. Caregivers used calendar log sheets to detail days the child did not take the supplement or consumed only a portion of it. They also completed 9 phone visits to monitor any adverse events or difficulty with supplementation.

Dietary intake

Detailed 24-h food and drink recalls were collected with the ASA24 Dietary Assessment Tool [32] administered by trained study staff with the caregivers at the baseline, 6-mo, and 9-mo visits to obtain data regarding daily dietary intake and to assess the potential confounding influence of changes in dietary choline intake. During phone visits, study staff asked caregivers about any supplements that were added to their child’s diet to ensure they did not contain choline. Families were instructed to refrain from adding dietary supplements during the study. Procedures for obtaining dietary intake data have been described previously [21,22].

Physical examination/adverse effects

Physical examinations, including a review of major organ systems to monitor any adverse events, were completed at each in-person visit by the study physician (JKE). The physician obtained participants’ height, weight, blood pressure, and heart rate. At the baseline visit, the physician measured head circumference and palpebral fissure length and rated the upper lip and philtrum using a 5-point Likert scale and lip and philtrum guide [33]. The review of major organ systems that was conducted by the study physician at baseline was readministered as a checklist by a trained research coordinator at each of the remaining study visits including the in-person visits and telephone visits in order to monitor for adverse effects of taking the study drink mix (choline or placebo). At each visit after the baseline, participants were asked to specifically report on new symptoms in each domain addressed by the review of systems.

Primary outcome: delayed sequential memory

Elicited Imitation (EI) is a play-structured paradigm measuring hippocampal-dependent explicit memory via the behavioral imitation of action sequences [[34], [35], [36]]. It is a nonverbal task designed to measure delayed recall memory in infants and young children. One caveat is that, like most childhood cognitive measures, attention may play a role in performance as well as memory. Nonetheless, preschool EI performance has been shown to predict later memory ability in school-aged children [37,38]. The measure involves 3 sets of toys, each of which has a theme (eg, playing in the park) and 9 actions (eg, catch the butterfly, throw the frisbee, and roll the ball). Item sets (events) were categorized according to salience (low, medium, or high-connectivity, with higher connectivity reflecting greater salience or natural connection between the steps). For example, catching a fish is a more ordinal connection with cooking the fish (higher connectivity) than it is with putting up a tent (lower connectivity). EI events were randomized and counterbalanced across treatment groups for each visit. RCT1 and RCT2 used a high-connectivity short-delay event and a medium-connectivity short-delay event, which were averaged together to maximize reliability of the measure. RCT3 used a medium-connectivity short-delay event and a low-connectivity short-delay event, which were averaged together. The EI paradigm used in RCT3 was, by design, more difficult and resulted in lower scores overall than that in RCT1 and RCT2. This was done to avoid potential ceiling effects on the EI in RCT3 as there were some participants who approached ceiling-level performance on the EI in RCT1 and RCT2. All 3 RCTs also used a medium-connectivity immediate-memory event, which was not examined in this study.

For each event, participants were given the toys to explore for 2–3 min. The experimenter demonstrated 9 actions in sequence and repeated the sequence twice. The 2 short-delay events were demonstrated followed by the immediate-memory event, where the child was then given the opportunity to play with the immediate-memory toy set. After a 15-min delay, the participant was given the toys from the first short-delay event and asked to repeat the previously demonstrated sequence, followed by the second short-delay event. For each trial, the number of components (actions) was recorded (maximum of 9 total per trial). A pair was recorded if the child performed 2 steps in the correct order (maximum of 8 per trial). An adjacent pair was recorded if the child performed 2 steps in consecutive order (1 right after the other; maximum of 8 per trial). Sessions were video recorded and later coded by trained raters. In our previous study, we found high interrater reliability (93%) using 20% of videos coded by multiple raters, suggesting a high interrater reliability with this method [22]. Outcome measures included the percentage of correct components, pairs, and adjacent pairs for each condition at the 6-mo and 9-mo visits. These 3 measures are correlated but evaluate semi-independent memory processes [39]. The components measure is thought to reflect the individual’s most basic stimulus-response association (seeing the object triggers an imitation response), whereas the pairs and adjacent pairs measures reflect deeper forms of memory in which ordinal information and linkage between items are also encoded via the hippocampus [36,40].

Secondary outcome: global cognitive functioning

The Mullen Scales of Early Learning is a measure evaluating global cognitive development using normative data from birth to 68 mo of age. The measure assesses visual reception, receptive language, expressive language, and fine motor abilities and provides T scores for each domain (mean: 50; SD: 10). Summed subtest scores are converted to an age-scaled intelligence Early Learning Composite (mean: 100; SD: 15). The SB-5 provides the following standard scores (mean: 100; SD: 15): full-scale IQ, verbal IQ, nonverbal IQ, abbreviated intelligent quotient (ABIQ), fluid reasoning, knowledge, quantitative reasoning, visual-spatial skills, and working memory. For 8 participants in RCT3, the ABIQ was used to estimate global cognitive function when the full measure was not able to be administered (eg, due to behavioral challenges and limited attentional capacity). The ABIQ is calculated using the object series/matrices and vocabulary subtest scores.

Tertiary outcomes: executive function and behavior

Executive function (cognitive flexibility) was assessed in RCT1 and RCT2 using the NIH Toolbox (version 2) DCCS Task [30] and, in RTC3, using the MEFS [31], a commercialized version of the DCCS. Conducted on iPads, the DCCS/MEFS require the child to place cards with colored animals on them into 1 of the 2 boxes depending on a specific rule for each level [31]. The highest level at which the child performed 80% or more of the trials correctly was the primary outcome for this test. For purposes of this analysis, the highest level attained on the DCCS (RCT1 and RCT2) and the MEFS (RCT3) were considered equivalent measures. These tasks were conducted at baseline and 9 mo.

In all 3 trials, parents completed the CBCLat baseline, 6 mo, and 9 mo to assess psychiatric symptoms [41]. Internalizing problems and externalizing problems were examined for treatment group effects.

Statistical analyses

Linear mixed-model analyses were used to test for treatment group differences (choline compared with placebo) in the trajectory of EI scores over visit (baseline, 6-mo, and 9-mo). Analyses, which modeled fixed effects and random child-specific intercepts and slopes, were conducted in SPSS, version 29.0.1.0, using the MIXED procedure with restricted maximum likelihood estimation (REML) [42]. REML was chosen because there were missing data points for some participants. Moreover, REML provides valid variable estimates without imputing missing values or using list-wise deletion. Models with child-specific intercepts only were compared with models incorporating both child-specific intercepts and slopes using Akaike Information Criterion (AIC), which balanced parsimony and fit (lower values indicate better model fit). The models incorporating child-specific intercepts alone demonstrated consistently lower AIC values, and therefore, child-specific slopes were not used. The covariance matrix structure was also determined using AIC [43]. Longitudinal analyses (across baseline, 6-mo, and 9-mo time points) were conducted as intention-to-treat analyses with all available valid data from participants who received the allocated intervention (choline: n = 50; placebo: n = 51). Participants were included regardless of the number of visits completed or compliance with the study (Figure 1). The linear slope terms in the models estimated change over the 3 time points (eg, in EI performance). Intercepts represented the estimated values at baseline before intervention (eg, baseline EI performance).

For each outcome measure (eg, EI adjacent pairs), 1 or 2 mixed-model analyses were conducted. First, a simple model tested for an effect of treatment over visit without demographic covariates or moderators. For outcome measures with an effect (trend-level, P < 0.10; or significant, P < 0.05), a second fully adjusted model tested for moderation effects by evaluating 2-factor interactions between treatment and sex, dose type (fixed 500 mg or 19 mg/kg), and age at enrollment as a continuous covariate, as well as 3-way interactions between treatment and visit number and these variables. Other demographic characteristics listed in Table 1 were not included as covariates in analyses.

Results

Study adherence was high for both choline and placebo groups

Adherence with the study drug was estimated by calendar logs and by counting empty packets that were physically returned at the conclusion of the study. On average, participants who completed the study returned 211 empty packets of the 270 that were provided (estimated 78% adherence). There was not a significant difference in empty packets returned between choline (mean: 203; SD: 64.5) and placebo (mean: 219; SD: 57.0). Some participating families reported that a few empty packets had been left behind or thrown away accidentally and, therefore, the estimate of 78% adherence is likely a slight underestimate of the actual adherence/consumption of the product by participants. In our previous reports on RCT1 and RCT2, we observed adherence in the 82%–88% range (days on which some study drug was consumed) [21,22]. We did not observe significant correlations between adherence and change in performance (baseline visit to 9-mo visit) on the primary outcome measure EI short-delay adjacent pairs (r = 0.097; P = 0.61), perhaps due to the fact that adherence was relatively high overall.

No significant differences in dietary intake or change in dietary intake of choline between the choline and placebo groups

There was no significant differential change (choline compared with placebo) in overall dietary intake of calories, protein, fat, carbohydrates, or choline. For context, the mean baseline dietary intake of choline for all participants was 208.8 mg/d (SD: 116.7 mg/d), and there was not a significant group difference. Compared with the dietary reference intake values for choline [44] for children aged 1–3 y (200 mg/d) and 4–8 y (250 mg/d), 61.5% of our participants were consuming below the recommended intake levels at baseline. We previously reported a high rate of inadequate dietary intake of choline (84%) in the participants comprising RCT1 and RCT2; we also noted that the general United States population has an equivalently high rate of inadequate choline intake [45].

Choline dose was associated with the prevalence of fishy body odor but not with other physical changes or adverse effects

Potential differential physical changes were examined for the choline and placebo groups. There was no significant differential change (choline compared with placebo) in height, weight, occipital–frontal circumference, heart rate, systolic blood pressure, or diastolic blood pressure.

For each of the major organ systems reviewed via parent report, adverse events were totaled across the 10 postbaseline visits for each participant (Table 2). Participants who received choline reported fishy body odor more frequently (a mean of 1.5 times of the 10 visits) than those who received placebo [a mean of 0.2 times; t(1,93) = 3.13; P < 0.01]. The incidence of fishy body odor was 36% in the choline group compared with 8% in the placebo group (Fisher exact test = 15.80; P < 0.01). For participants who received choline, there was a dose effect such that 57% of those who received 500 mg/d reported fishy body odor at least once, but only 11% of those who received the weight-adjusted dose of choline reported fishy body odor (Fisher exact test = 10.19; P < 0.01). Fishy odor is the result of trimethylamine being produced in the gut by bacteria and, as such, is an indication of the presence of excess choline. Parents generally reported that the odor was only noticeable to them at close range as it was in the child’s sweat or urine and that it was not noticeable by others. There were no significant differences in the incidence of any other potential adverse effect.

TABLE 2.

Number of individuals reporting symptoms at baseline and number reporting new symptoms (adverse effects) at least once during the course of the study.

	Choline (n = 50), n (%)	Placebo (n = 51), n (%)	P (Fisher exact test)
General health	13 (26)	8 (16)	0.28
Skin	10 (20)	8 (16)	0.38
Ear, nose, throat	1 (2)	1 (2)	0.74
Cardiovascular	1 (2)	2 (4)	0.51
Respiratory	9 (18)	6 (12)	0.38
Gastrointestinal	21 (42)	17 (33)	0.53
Fishy body odor	18 (36)	4 (8)	0.01
Genitourinary	4 (8)	10 (20)	0.10
Musculoskeletal	1 (2)	1 (2)	1.00
Neurologic	10 (18)	9 (20)	0.94
Allergy	1 (2)	3 (6)	0.75
Other	22 (44)	21 (41)	0.98

Participants receiving choline and placebo did not differ in change in global cognitive functioning over the course of the study

The simple mixed-model analysis testing for a potential choline effect on IQ (Mullen Scales or SB-5), revealed a significant effect for visit [F(1, 71.13) = 8.21; p < 0.01; IQ increased with readministration of the test]. There was no significant treatment effect [F(1, 94.11) = 0.01; P = 0.92] or significant visit × treatment interaction [F(1, 71.13) = 0.02; P = 0.88], and no further analyses were conducted. IQs ranged from 47 to 128. Unadjusted cognitive scores are included in Table 3 for reference.

TABLE 3.

Unadjusted means, SDs, and numbers of participants for cognitive tests and behavioral measures by group across study visits.

	Mean (SD), n
	Baseline visit	6-mo visit	9-mo visit
IQ1
Choline group	86.3 (14.4), 45	not administered	89.8 (15.6), 35
Placebo group	87.1 (18.5), 46	not administered	90.7 (20.5), 37
Elicited imitation
Components
Choline group	79.3 (23.9), 49	84.5 (16.7), 33	87.1 (15.6), 39
Placebo group	77.5 (22.2), 51	85.3 (19.9), 33	86.2 (17.3), 39
Ordered pairs
Choline group	48.2 (21.2), 49	56.5 (18.1), 33	61.8 (21.6), 39
Placebo group	47.4 (24.3), 50	54.2 (19.6), 33	57.4 (17.9), 39
Adjacent pairs
Choline group	32.3 (20.0), 49	37.2 (19.6), 33	45.0 (24.1), 39
Placebo group	30.9 (20.5), 50	34.4 (18.2), 33	34.8 (17.9), 39
MEFS highest level
Choline group	2.2 (1.4), 37	3.0 (1.4), 36	3.6 (1.4), 28
Placebo group	1.8 (1.5), 36	3.0 (1.8), 39	2.9 (1.5), 31
CBCL
Internalizing
Choline group	60.7 (10.9), 50	59.8 (9.3), 42	59.6 (11.2), 39
Placebo group	63.0 (9.3), 51	61.4 (9.8), 42	62.7 (10.4), 40
Externalizing
Choline group	66.2 (10.9), 50	65.8 (13.0), 42	64.9 (11.5), 39
Placebo group	69.1 (10.5), 51	66.3 (12.0), 42	66.1 (12.5), 40

Abbreviations: CBCL, Child Behavior Checklist; IQ, intelligent quotient; MEFS, Minnesota Executive Functioning System.

¹IQ was from either the Mullen Scales of Early Learning (RCT1 and RCT2) or the Stanford–Binet Intelligence Scale, Fifth Edition (RCT3).

Participants receiving choline showed a trend toward greater improvement in memory on the primary outcome measure (EI) than participants receiving placebo

For EI short-delay adjacent pairs, the simple model revealed a significant main effect for visit—all participants improved over repeated administration [F(2, 141.64) = 4.49; P = 0.01]; a nonsignificant effect for treatment [F(1, 101.17) = 1.08; P = 0.30]; and a trend-level interaction [F(2, 141.64) = 2.46; P = 0.09]. Examining the fixed-effect estimates revealed a steeper increase in score over visits for choline compared with that for placebo, which was most apparent between the 6-mo and 9-mo visits (ŷ = −10.06; P = 0.03; 95% CI: −19.13, −0.99). Figure 2 illustrates the interaction with clear separation of the curves most evident at the 9-mo time point. Examining the predicted scores from the model, the individual (within-subject) improvement from baseline to 9 mo was 35.4% for the choline group and 10.9% for the placebo group.

FIGURE 2.

Elicited imitation short-delay adjacent pairs predicted scores by visit (simple mixed-model analysis).

The fully adjusted model for EI short-delay adjacent pairs revealed a trend-level 3-way interaction between treatment group, visit, and age at enrollment [F(4, 123.57) = 2.34; P = 0.06]. Examining the fixed-effect estimates revealed that age at enrollment interacted with treatment group and visits, significantly between the baseline and 9-mo visits (ŷ = 10.02; P = 0.01; 95% CI: 2.47, 17.57). There were no other significant 3-way interactions, suggesting that the trend-level treatment effect for choline described earlier was moderated by the child’s age at enrollment but not by the child’s sex or by the dose type. Figure 3 illustrates the EI short-delay adjacent pair scores over visit with these moderating effects taken into account in the model. To most easily illustrate the moderating effect of age at enrollment on the relationship between choline and change in EI short-delay adjacent pairs over visits, participants were divided into 2 groups based on a median (4.2 y) split in age (younger compared with older participants). Figure 4 shows the relatively steeper slope for choline than for placebo in the younger participants and similar slopes in the older participants. Examining the predicted scores from the fully adjusted model, the individual (within-subject) improvement from baseline to 9 mo was 146.5% for the younger choline group, 21.1% for the older choline group, 100.5% for the younger placebo group, and −12.9% for the older placebo group.

FIGURE 3.

Elicited imitation adjacent pairs predicted scores by visit (fully adjusted mixed-model analysis).

FIGURE 4.

Elicited imitation adjacent pairs predicted scores by visit (fully adjusted mixed-model analysis) for 2 age-groups based on a median age split.

For EI short-delay components, the simple model revealed a significant main effect for visit—all participants improved over repeated administration [F(2, 139.68) = 8.01; P < 0.01]; a nonsignificant effect for treatment [F(1, 102.29) = 0.05; P = 0.82]; and a nonsignificant interaction [F(2, 139.68) = 1.68; P = 0.19]. Because there was no significant interaction, a fully adjusted model was not performed. Similarly, for EI short-delay pairs, the simple model revealed a significant main effect for visit—all participants improved over repeated administration [F(2, 141.24) = 10.45; P < 0.01]; a nonsignificant effect for treatment [F(1, 100.64) = 0.15; P = 0.70]; and a nonsignificant interaction [F(2, 141.24) = 0.57; P = 0.57]. Because there was no significant interaction, a fully adjusted model was not performed.

Choline was not associated with behavior change over the study relative to placebo

A simple model analysis of CBCL internalizing problems was conducted. No significant effects were seen for the visit [F(2, 161.24) = 1.40; P = 0.25], treatment [F(1, 102.41) = 2.40; P = 0.12], or interaction [F (2, 161.24) = 0.44; P = 0.65]. Similarly, a simple model analysis of CBCL externalizing problems found no significant effects for the visit [F(2, 169.57) = 1.99; P = 0.14], treatment [F(1, 110.47) = 1.05; P = 0.31], or interaction [F(2, 169.57) = 0.31; P = 0.73].

Choline was not associated with improvement in executive functioning relative to placebo

A simple model analysis examining MEFS highest level achieved was conducted. A significant effect was seen for visit [F(2, 129.84) = 21.41; P < 0.01; scores improved with repeated administration], but there were no significant effects for the treatment [F(1, 94.05) = 1.11; P = 0.30] or the interaction [F(2, 129.84) = 1.49; P = 0.23].

Discussion

In this study, we reported combined analyses of 3 double-blind RCTs evaluating choline as a neurodevelopmental intervention for children with FASD. This cumulative data set provides a unique opportunity to compare 2 choline dosing regimens and to examine the potential moderating effects of age and sex although increasing statistical power to evaluate choline’s efficacy. Across the 2 RCTs, treatment adherence was high and equivalent for choline and placebo as observed previously [21]. We continue to see no serious adverse effects from choline supplementation in young children with FASD. Fishy body odor, the sole adverse effect across studies, was clearly dose related, suggesting that it is manageable with dose adjustments and/or dose splitting.

Children receiving choline showed a trend-level benefit on a hippocampus-dependent delayed sequential memory task (EI short-delay adjacent pairs) at the 9-mo follow-up compared with placebo. Choline’s benefit appears to accrue with continued supplementation—the effect was evident primarily at 9 mo. The data show a 25% greater increase in memory development over 9 mo for choline (35.4%) compared with placebo (10.9%). This effect is large enough to be potentially clinically significant, as the long-term benefits of interventions that modify developmental trajectories early in childhood are well-established [46]. Developmental studies have shown the importance of early nonverbal memory (the type measured by EI) in setting the foundation for later memory and learning processes [[47], [48], [49]]. EI performance at 20 mo of age predicts verbal and visual memory abilities, attention, learning, and processing speed at 6 y of age [38]. For EI ordered pairs at 20 mo, correlations at 6 y of age were 0.49 for Woodcock–Johnson III pair cancellation and 0.54 for EI ordered pairs and Woodcock–Johnson III visual matching (both measures of processing speed associated with overall cognitive functioning). Our own longitudinal studies demonstrate that early choline supplementation and improvement in early sequential memory functioning were associated with significantly higher nonverbal IQ (8% higher than placebo), visual-spatial reasoning (29% higher), crossmodal learning (38% higher), and nonverbal working memory (27% higher) 4 y after treatment—at a mean of 8.6 y of age [24]. In our second longitudinal study, benefits of choline were observed in lower-order executive function skills (eg, information processing speed) in a subset of participants who returned 7 y after completing a choline trial—at a mean 11.0 y of age [25]. That study also demonstrated potentially beneficial white matter organization differences in the splenium of the corpus callosum for choline compared with that for placebo [25]. Together, findings suggest the benefits of early choline supplementation in FASD may persist and potentially generalize beyond memory functioning as development unfolds and compounds over years.

In our first RCT [22], all participants received a fixed 500 mg daily choline dose (or placebo), but represented a range of body weights. We computed effective doses per kilogram of body weight and found an inverse relationship with memory performance. As doses were not assigned in that study, the conclusions drawn regarding the effect of dose on neurodevelopmental functioning were limited. In this study, combining data from the 3 RCTs allowed for a direct evaluation of a dose–response relationship. Moreover, we observed that—within the range of 260–500 mg/d— choline dose was not significantly related to improvement in memory performance. For context, adequate intake is 200 mg/d for children aged 1–3 y and 250 mg/d for children 4–8 y [50]. Practically, ensuring adequate dietary intake of choline and other nutrients should be the first priority—especially in children with neurodevelopmental disorders like FASD who have atypical food preferences as well as inadequate intake [45,51]. Furthermore, choline supplementation ≤500 mg/d may be beneficial for young children with FASD. An extensive discussion of dosing, formulations, and other practical considerations has been described previously [8].

Previously, we reported that age of choline supplementation moderated the effect of choline on EI performance [22]. A split sample analysis showed that, in the choline group, younger children (2.5 to ≤4.0 y) showed greater memory change than older children (4.0–5.9 y). The effect of age was significant but not large (P = 0.05). At present, with the combined data set and increased sample size, we included age at enrollment as a potential moderating variable and observed a trend-level 3-way interaction, suggesting that age does moderate the impact of choline on EI adjacent pairs. Younger children (≤4.2 y) showed a greater change in EI performance than older (>4.2 y) children. This was consistent with our hypothesis that age would moderate choline’s impact on memory change because hippocampal growth and synaptogenesis are rapid during the first 2 y of life but slow thereafter [52,53] and because hippocampal developmental windows of opportunity are thought to exist for choline [8,9]. For context, we note that the only other published postnatal choline RCT in FASD—evaluating 625 mg choline daily for 6 wk in individuals with PAE aged 5–19 y—failed to find significant treatment effects and did not observe moderation of effects by age [23]. For context, 625 mg choline would have resulted in lower milligrams per kilogram doses for these older (and heavier) individuals compared with our samples of 2.5- to 5.9-y-old children. Our current data, together with established developmental principles, continue to suggest that earlier choline supplementation (including prenatal supplementation) results in the most benefit. The field has yet to establish a definitive upper age limit for choline’s effectiveness, and additional research is therefore necessary.

Choline supplementation early in childhood during periods of rapid brain growth may confer benefits through 1 or more mechanisms: regulating DNA methylation and gene expression [9,[54], [55], [56]], acting as a precursor to acetylcholine (a neurotransmitter implicated in hippocampal memory function) [52,[57], [58], [59], [60], [61]], and supporting phospholipid formation and myelination [62,63]. Such effects of choline may optimize development during a foundational stage with significant later downstream benefits for the individual.

Presently, we are conducting a fourth RCT comparing multiple cumulative choline supplementation levels (3 mo compared with 6 mo; NCT05108974) in a study with the same design as RCTs 1–3. Ultimately, these data will yield additional insights about the minimum necessary duration of choline supplementation to achieve a neurodevelopmental impact in children with FASD.

In conclusion, after >100 preschool-aged children with FASD have been studied in a series of RCTs, we continue to conclude that choline supplementation is safe, tolerable, and likely effective in improving a specific form of early nonverbal sequential memory. Choline dose is related to 1 mild side effect (fishy odor) but not to the degree of memory improvement, suggesting that its use as a treatment will require balancing these factors. These results are most generalizable to supplementation with oral choline bitartrate for male and female children aged 2.5–5.9 y with PAE and neurobehavioral impairments across a wide range of IQ (from 47 to 128). Our studies included participants from the entire FASD diagnostic spectrum. As indicated by Table 1, the data were derived from a relatively ethnically and racially diverse sample and should generalize as such. Future work will benefit from evaluating potential long-term benefits of choline across various dosing regimens with regard to cognitive function across a variety of domains, which may elucidate important differences across clinical trial cohorts. Such an approach may also provide valuable information to inform optimal developmental timing and dosing of the intervention.

Author contributions

The authors’ responsibilities were as follows – JRW, JKE: designed and conducted research; JRW, BAG, KAT, EdW: analyzed the data; JRW: had primary responsibility for final content; AME, MAE, KAT: conducted the research; AME, BAG, MAE, KAT, EdW, SHZ, MKG: wrote the paper; EdW, SHZ, MKG: designed the research; SHZ, MKG: consulted on data collection; and all authors: read and approved the final manuscript.

Data availability

Data described in the manuscript, code book, and analytic code will be made available upon request pending approval by the corresponding author.

Funding

This work was supported by the National Institutes of Health (NIH) grants R21AA019580, R33AA019580, R01AA024123, and R56AA024123 (to JRW and MKG).

Conflict of interest

JRW reports financial support was provided by National Institutes of Health. All other authors report no conflicts of interest.