Normative MRI Data Increase the Sensitivity to Brain Volume Abnormalities in the Classification of Fetal Alcohol Spectrum Disorder

Abstract

Objective

To evaluate the use of a large MRI normative dataset to quantify structural brain anomalies that may improve diagnostic sensitivity for atypical brain volume in youth with fetal alcohol spectrum disorder (FASD).

Study Design

Participants included 48 children with prenatal alcohol exposure (PAE) and 43 controls, ages 8–17 years, from the longitudinal Collaborative Initiative on Fetal Alcohol Spectrum Disorders. Recently published lifespan brain charts were used to quantify participants’ (per)centile for brain volumes (cortical and subcortical gray matter and cortical white matter), providing an index of (dis)similarity to typically developing individuals of the same age and sex.

Results:

Participants with PAE demonstrated lower mean centile scores compared with controls. Participants with PAE and scores ≤ 10^th centile on at least one brain volume metric demonstrated significantly lower performance on measures of intellectual function and aspects of executive functioning compared with participants with PAE and “typical” volumes (>10^th centile). Brain volume centiles explained a greater amount of variance in IQ and improved sensitivity to brain volume anomalies in FASD compared with the most commonly used diagnostic criterion of occipitofrontal circumference (OFC) ≤ 10^th.

Conclusion:

Age- and sex-adjusted brain volumes based on a large normative dataset may be useful predictors of functional outcomes and may identify a greater number of individuals with FASD than the currently used criterion of OFC.

Fetal alcohol spectrum disorders (FASD) are typically diagnosed on the basis of prenatal alcohol exposure (PAE), brain anomalies including microcephaly (small head size), physical growth restriction (height and weight), facial dysmorphology, and neurobehavioral impairment¹. Microcephaly (as determined by occipitofrontal circumference [OFC]) was a core finding in the earliest descriptions of fetal alcohol syndrome (FAS)² and is known to be associated with the degree of alcohol exposure³. Commonly-applied current diagnostic criteria for atypical brain volume include OFC ≤ 10^th percentile^4,5 or ≤ 3^rd percentile⁶ as determined by physical measurement of the head.

Although OFC is cost effective, quick to obtain, and widely used in pediatric clinical practice⁷, it has several limitations as a metric of brain volume. OFC can be difficult to interpret without reference to parental OFC⁸ (ie, due to potential impacts of shared genetic/environmental factors on OFC) which is often unavailable for youth with FASD (eg, those adopted or living in foster care). In addition, inter-rater reliability in OFC measurement varies by level of examiner expertise⁹. Although OFC is strongly correlated with total brain volume at younger ages in typically developing children, this association is attenuated after age 7^10,11. Emerging evidence suggests OFC may be suboptimally sensitive to PAE. In one study of youth (5 to 19 years) with and without PAE, researchers found that in 55 to 66 percent of youth with PAE who had total brain volumes ≤ 3^rd percentile (9.7% of PAE group) or ≤ 10^th percentile (15.3% of PAE group), OFC was in the average range (ie, ≥ 11^th percentile)³. In addition, studies comparing OFC and IQ in PAE samples have found inconsistent results^3,12. Together, these factors highlight the need for alternative, complementary approaches to characterizing anomalous brain volume associated with PAE.

Recent work with large MRI datasets provides a novel method for quantifying intra-individual brain volume in both typical and atypical development, providing important benefits over small, potentially unrepresentative comparison groups. In the current study, we used a recently published open-science online resource¹³ to compute (per)centile scores characterizing volumetric (dis)similarity to typically developing individuals of the same age and sex in our sample of youth with PAE and controls ages 8–17 years. Individual centile scores were generated for volumes of cortical and subcortical gray matter, and cortical white matter. We hypothesized that 1) brain centile scores would be lower for youth with PAE compared with controls, 2) youth with PAE who had atypical brain centile scores (i.e., ≤ 10^th percentile) would demonstrate greater neurocognitive and neurobehavioral impairment than those with typical brain centile scores, and 3) brain centile scores would identify a greater number of PAE participants with brain volume anomalies compared with the existing diagnostic criteria for anomalous brain volume (i.e., OFC ≤ 10^th percentile).

METHODS

Participants

Participants in this study included children and adolescents ages 8 to 17 with PAE (n = 48) and non-exposed controls (n = 43) who were part of the longitudinal Collaborative Initiative on Fetal Alcohol Spectrum Disorders (CIFASD) study (see www.cifasd.org)¹⁴. We performed an analysis using data collected at the first study visit. Additional information regarding the larger parent study from which these data were drawn was previously reported¹⁵. Consent and assent procedures (IRB approved) were completed with participants and their guardian or parent at enrollment, and participants were provided monetary compensation. A total of 95 participants were enrolled in the study. Four participants (1 PAE, 3 control) were excluded due to either poor image quality, excessive head motion, or being an extreme outlier in multiple whole-brain metrics. A total of 91 participants were therefore included in the current study (PAE = 48; control = 43). Table I contains demographic information for the participants.

Table I.

Demographic characteristics of participants included in primary analyses.

	PAE (n = 48)	Control (n = 43)	Statistical Test
Age [mean(SD)]	12.34 (2.35)	12.79 (2.53)	t(86) = −0.88, p = 0.380
Intelligence Quotient [mean (SD)]	93.15 (14.87)	115.52 (12.05)	t(88) = −7.88, p < 0.001
Sex [n(%Female)]	25 (52%)	21 (49%)	χ2 = 0.10, p = 0.757
Ethnicity [n(%Hispanic)]	2 (4%)	2 (5%)	χ2 = 0.01, p = 0.910
Race *
[n(%American Indian/Alaska Native)]	4 (8%)	0	χ2 = 7.15, p = 0.008
[n(%Asian)]	2 (4%)	1 (2%)	χ2 = 1.40, p = 0.237
[n(%Black or African American)]	12 (25%)	0	χ2 = 18.18, p < 0.001
[n(%Native Hawaiian/Other Pacific Islander)]	1 (2%)	0	χ2 = 1.94, p = 0.164
[n(%White)]	26 (46%)	42 (98%)	χ2 = 29.21, p < 0.001
[n(%Multiracial)]	7 (15%)	0	χ2 = 11.67, p < 0.001
Handedness [n(%Right)] †	35 (73%)	36 (84%)	χ2 = 0.66, p = 0.719
Physical characteristics
aGrowth Deficiency	5 (10%)	3 (7%)	χ2 = 0.33, p = 0.563
bMicrocephaly	5 (10%)	0	χ2 = 4.15, p = 0.042
cDysmorphic Face	12 (25%)	2 (5%)	χ2 = 5.97, p = 0.014

NOTE: Demographics are from the 91 participants included in the primary analyses. Age range at baseline evaluation ranged from 8 to 17 years.

PAE: Prenatal alcohol exposure group

^*Chi squared tests reflect comparisons of each racial group with the proportion of participants who identified as White (eg, proportion of participants who identified as Multiracial to proportion of those who identified as White)

^†Handedness information was not available for 5 participants (4 PAE, 1 Control).

^aHeight or weight ≤ 10%ile.

^bHead circumference ≤ 10%ile.

^cAt least two of the following: palpebral Fissure Length ≤ 10%ile, thin vermillion border, smooth philtrum (4 or 5 on lipometer scale). The two Control participants who had “dysmorphic faces” had scores of 4 on the philtrum and 4 on the vermillion border; neither had any other facial features nor abnormal growth parameters.

Enrollment in the parent study (in both groups) required the absence of neurological and developmental disorders, severe psychiatric disabilities that would prevent participation, participant drug/alcohol use, birth weight <1500 grams, and contraindications to MRI scanning (eg, non-MR-safe medical devices, braces, claustrophobia). A diagnosis of attention-deficit/hyperactivity disorder (ADHD) was not exclusionary. Prenatal exposure to substances other than alcohol was not exclusionary for the PAE group because of the commonality of this pattern. Controls with prenatal alcohol and drug exposure (excluding tobacco and caffeine) were excluded from the parent study.

FASD Diagnosis

Research staff conducted phone screens and record reviews to determine PAE history. Inclusion criteria for PAE participants included evidence of heavy PAE (≥ 13 drinks per week or ≥ 4 drinks per occasion at least once per week during pregnancy). Individuals with “suspected” but unconfirmed PAE were included if they met diagnostic criteria for partial FAS (pFAS) or FAS based on dysmorphology and growth characteristics. For all participants, physical assessment by a trained investigator (JRW) included ratings of the upper lip and the philtrum; measurement of the palpebral fissure length (PFL), height, and weight. Physical examinations were completed within 6 months of the MRI scan and neurobehavioral assessment. Normative data for OFC measurements were taken from Nelhaus et al (1968)⁷. The Modified Institute of Medicine criteria were used for FASD diagnostic classification⁴. Neurobehavioral impairment was characterized as standard scores ≥ 1.5 standard deviations (SD) below the normative mean on neuropsychological assessments and parent-rated measures (i.e., either an overall Wechsler IQ score ≤ 78 or impairment in ≥ 2 specific neuropsychological domains).

Evaluations

Participants were evaluated at the University of Minnesota and administered the following measures: Wechsler Intelligence Scale for Children, 5th Edition (WISC-V)¹⁶ or Wechsler Adult Intelligence Scale 4th Edition (WAIS-IV)¹⁷ if they were 17 years old (measuring general intelligence [FSIQ, hereafter IQ]); Trail-Making, Verbal Fluency, and Color-Word Interference subtests from the Delis-Kaplan Executive Functioning System (D-KEFS; measuring sequencing, inhibition, and cognitive flexibility)¹⁸; Dimensional Change Card Sort (DCCS; measuring inhibition and cognitive flexibility), and the Flanker test from the NIH Toolbox¹⁹ (measuring inhibition); and the Vineland Adaptive Behavior Scales 3rd Edition (VABS-3; measuring independence in activities of daily living)²⁰. Performance on the WISC-V, WAIS-IV, and VABS-3 is measured with standard scores (M = 100, SD = 15), the D-KEFS is measured with scaled scores (M = 10, SD = 3), and the DCCS and Flanker is measured with T-scores (M = 50, SD = 10). Higher scores indicate better performance on all measures.

MRI Acquisition and Processing

Structural MRI data were acquired at the University of Minnesota’s Center for Magnetic Resonance Research on a Siemens 3T Prisma scanner (Siemens, Erlangen, Germany) equipped with a standard 32-channel head coil. T1-weighted and T2-weighted scans were acquired using custom pulse sequences including automatic real-time motion detection and k-space line rejection and replacement software. Pulse sequence parameters were chosen to match those used in the Lifespan Human Connectome Project Development (HCP-D) project²¹. The PreFreeSurfer and FreeSurfer stages of the HCP Minimal Preprocessing Pipeline (v4.0.1) were used to process the structural images²². Data were visually inspected by a trained operator (DJR) to ensure accuracy. In the case of significantly aberrant FreeSurfer processing (such as failed boundary identification), participant data were excluded from the analyses. The Euler value²³ averaged across hemispheres, was used to further assess for group differences in within-scanner motion. Measurements were obtained for total volumes of cortical gray matter, subcortical gray matter, and cortical white matter.

Normative Quantitative Measurements

Individual (per)centile scores were generated using BrainChart¹³, an interactive, open source online software tool that uses a large, aggregated neuroimaging dataset (95,536 control brains, with the total number of brains [cases and controls] across samples with age ranges relevant to the current study totaling over 9,000). Centile scores were calculated for cortical gray matter volume (GMV), subcortical gray matter volume (sGMV), and cortical white matter volume (WMV), allowing for quantification of each participant’s brain volumetric (dis)similarity to typically developing individuals of the same age and sex.

Statistical Analysis

Statistical analysis was performed with R version 4.1.1²⁴. Chi-square tests and independent samples t-tests were used to examine participant characteristics and group differences in centile scores. One-sample t-tests were used to compare the PAE group to a normative mean centile score of 0.50. Group differences in motion during the MRI scan (Euler values) were analyzed using independent-sample t-tests. Pearson correlation was used to examine relationships between OFC percentile and brain centile scores separately for each group (PAE and controls). Linear models were used to examine the variance in IQ explained by brain centile scores and OFC for PAE participants. Independent-sample t-tests were used to explore potential differences in neurocognitive and neurobehavioral function between PAE participants with atypical centile scores (ie, ≤ 10^th percentile) compared with PAE participants with more “typical” brain volumes (ie, scores at the 11^th to 99^th centile). Sensitivity and specificity analyses were conducted using the “caret” package²⁵ to examine sensitivity (true positive rate), specificity (true negative rate), positive predictive value (probability of condition being present with a positive test), negative predictive value (probability of condition not being present with a negative test), and accuracy (overall probability of correct classification). These analyses were performed individually for brain centile scores, height or weight anomalies, small OFC (≤ 10^th percentile), facial dysmorphology, low IQ (1.5 SD below the mean; standard score ≤ 78), and a combination of small OFC and atypicality in any brain volume centile score.

RESULTS

Subject Characteristics

PAE and control groups did not differ significantly on age, sex, ethnicity, handedness or growth deficiency (Table I). More participants with PAE demonstrated microcephaly and dysmorphic facial features than controls. Participants with PAE had lower IQ scores than the control group (an approximate 22 point difference in mean score). Groups also differed significantly regarding racial composition (ie, more participants in the PAE group identified as American Indian/Alaska Native, Black/African American, and Multiracial compared with the control group). Finally, the control group had a higher proportion of participants that identified as White vs. Non-white compared with the PAE group. Within-scanner motion (estimated with Euler values) was not significantly different between groups, t(61) = −1.19, p = 0.240.

Group Comparisons in Brain Volume

Participants in the PAE group demonstrated lower mean centile scores than controls across GMV, WMV, and sGMV (Figure 2). One-sample t-tests indicated PAE centile scores were significantly below the 50th centile for GMV (t[47] = −3.19, p = 0.003), WMV (t[47] = −4.16, p < 0.001), and sGMV (t[47] = −5.09, p < 0.001).

Figure 2.

Significantly lower mean centile scores for participants with PAE than controls in cortical gray matter volume (GMV), cortical white matter volume (WMV), and subcortical gray matter volume (sGMV). Dashed line indicates the 50th centile. White circles represent group mean centile scores.

Relationship of Brain Volume and OFC

For both groups (PAE and control), brain volume centiles were significantly positively correlated with OFC centiles (Figure 3). However, a number of participants with PAE and OFC > 10th centile nonetheless had atypically small brain volume centiles for GMV (n = 3; 6% of PAE group), WMV (n = 8; 17%), and sGMV (n = 9; 19%).

Figure 3.

Significant correlations for PAE and control participants between OFC centile and brain volume centile across GMV (A-B), WMV (C-D), and sGMV (E-F). Black boxes are used to indicate participants with PAE and normal OFC (11^th to 99^th centile) who have small brain volumes (≤ 10th centile).

Relationship of Brain Volume and OFC to IQ

WMV centile scores were significantly correlated with IQ, and the model with all three brain centiles explained 15% of the variance in IQ (Table II). In contrast, OFC centile was not significantly correlated with IQ and explained only 3% of the variance in IQ (Table III).

Table II.

Variance in IQ explained by brain centile scores for the PAE group

Predictor	b	b 95% CI [LL, UL]	beta	beta 95% CI [LL, UL]	sr 2	sr2 95% CI [LL, UL]	r	Fit
(Intercept)	87.47**	[80.22, 94.71]
GMV centile	−2.17	[−27.29, 22.96]	−0.04	[−0.50, 0.42]	.00	[−.01, .01]	.21
WMV centile	29.89*	[4.70, 55.08]	0.53	[0.08, 0.99]	.11	[−.05, .27]	.36*
sGMV centile	−12.07	[−35.45, 11.31]	−0.22	[−0.63, 0.20]	.02	[−.05, .10]	.13
							R2 = .155
							95% CI[.00,.31]

Note. A significant b-weight indicates the beta-weight and semi-partial correlation are also significant. b represents unstandardized regression weights. beta indicates the standardized regression weights. sr² represents the semi-partial correlation squared. r represents the zero-order correlation. LL and UL indicate the lower and upper limits of a confidence interval, respectively.

^*indicates p < .05.

^**indicates p < .01.

GMV = gray matter volume, WMV = white matter volume, sGMV = subcortical gray matter volume

Table III.

Variance in IQ explained by OFC centile scores for the PAE group

Predictor	b	b 95% CI [LL, UL]	beta	beta 95% CI [LL, UL]	sr 2	sr2 95% CI [LL, UL]	r	Fit
(Intercept)	87.12**	[76.18, 98.07]
OFC centile	0.09	[−0.07, 0.26]	0.18	[−0.12, 0.48]	.03	[.00, .18]	.18
							R2 = .032
							95% CI[.00,.18]

Note. A significant b-weight indicates the beta-weight and semi-partial correlation are also significant. b represents unstandardized regression weights. beta indicates the standardized regression weights. sr² represents the semi-partial correlation squared. r represents the zero-order correlation. LL and UL indicate the lower and upper limits of a confidence interval, respectively.

^**indicates p < .01.

Relationship of Centile Scores to Neurocognitive and Neurobehavioral Function in PAE

Participants with PAE and atypical centile scores in GMV and WMV demonstrated significantly lower IQ compared with participants with PAE with no volumetric anomaly (Figure 4). Similarly, participants with PAE and atypical centile scores in WMV and sGMV demonstrated significantly lower working memory (Digit Span scaled score) than participants with PAE who had typical centile scores. No other comparisons were statistically significant.

Figure 4.

Lower IQ standard scores in participants with PAE and atypical centiles (i.e., ≤ 10^th percentile) compared with participants with PAE and typical centiles (i.e., 11^th percentile to 99^th percentile) for GMV, WMV, and sGMV (A). Lower Digit Span scaled scores in participants with PAE and atypical centiles compared with participants with PAE and non-atypical centiles for GMV, WMV, and sGMV (B).

Sensitivity and Specificity Analyses

With acknowledgement that diagnostic criteria are never used in isolation, we present them here individually to illustrate the incremental utility of MRI-derived brain volumes for diagnosis. In these pre-identified groups, the commonly used metrics of anomalous growth, small OFC, and dysmorphic facial features, as well as low IQ scores, had low sensitivity but were highly specific in discriminating participants with PAE from controls, with low overall accuracy (Table III). Of these commonly used metrics, low IQ had the highest overall accuracy (56.67%). In contrast, atypicality in any brain volume centile had substantially higher sensitivity (35.42%), and only marginally lower specificity (95.35%) as well as improved accuracy (63.74%) compared with growth metrics, OFC, and facial features. A combination of atypical OFC or atypical brain volume centiles slightly improved sensitivity (36.96%) but slightly reduced specificity (94.44%) and accuracy (62.20%). Atypical OFC accurately classified only 5 cases with PAE, while atypical brain volume centiles accurately classified 17 cases (340% increase in cases correctly identified).

DISCUSSION

We demonstrate the value of leveraging a large normative MRI dataset to improve diagnostic sensitivity for the physical brain anomalies that occur in FASD. Using an open-source online tool¹³ we generated individual (per)centile scores for cortical and subcortical GMV and cortical WMV—placing each participant’s brain volumetrics on a “brain growth chart” to compare against typically-developing individuals of the same age and sex. We demonstrate that these individual brain volume centile scores have several advantages over OFC, the commonly used proxy measure for brain volume in FASD diagnostic systems, with regard to explaining neurobehavioral impairment and classification sensitivity and accuracy. We also find that within the PAE group, atypical brain volume centile scores were associated with poorer neurobehavioral function, suggesting these metrics may be predictors of important functional outcomes in this population.

As expected, participants in the PAE group demonstrated lower mean centile scores compared with controls and the normative mean for GMV, WMV, and sGMV, consistent with a large body of evidence indicating broadly reduced brain size in individuals with PAE^{26, 27}. Consistent with previous work with PAE³ and typically-developing samples^{10, 11} we found that brain volume centiles were significantly positively correlated with OFC for both PAE and controls. Importantly, the relationship between brain volume and OFC is known to be attenuated after age 7 due to the fact that skull thickness and non-neuronal tissues continue to grow during adolescence, while brain volume expansion peaks in early childhood³. Consistent with this, we demonstrated important limitations in the use of OFC as a reliable proxy of brain volume in a sample of youth with PAE ages 8–17 years. A meaningful proportion of participants with PAE and a “typical” OFC had atypically low brain volume centiles for GMV (6% of PAE group), WMV (17%), and sGMV (19%). In addition, within-group analyses indicated that for the PAE group, brain volumes explained 15% of the variance in IQ, while OFC explained only 3% of the variance in IQ. We also found that those with PAE and atypical brain centile scores demonstrated significantly lower IQ and working memory performance compared with PAE participants with typical centile scores. Together, these findings provide converging evidence that OFC measurements do not fully capture the range of central nervous system damage resulting from PAE and suggest that norm-adjusted MRI brain volumes may provide a complementary measure that also has the advantage of better predicting important functional outcomes.

We also explored the use of brain volume centiles in improving the classification of anomalous brain volume used in the diagnosis of FASD. Small OFC had low sensitivity but high specificity in discriminating participants with PAE from controls, with low overall accuracy. In contrast, atypicality in any brain volume centile had substantially higher sensitivity and only marginally lower specificity as well as improved accuracy, classifying 17 cases with PAE compared with only 5 cases classified by small OFC. While preliminary at this stage, these results suggest that MRI-derived gray and white matter volumes adjusted for age and sex meaningfully increase the sensitivity of the diagnostic criterion for brain volume anomalies in FASD compared with the traditionally used measure of OFC.

Our findings raise intriguing questions about the potential use of individualized age- and sex-adjusted brain volumetric data in clinical decision-making regarding FASD diagnosis. Practically speaking, OFC measurements present a number of limitations, including sub-optimal inter-rater reliability in clinicians without expert training⁹. Measurements of small OFC can also be difficult to interpret without reference to parental OFC⁸ –information that is often unavailable for youth who present for FASD diagnostic evaluation. The use of brain volumetric centile scores also have several limitations for use in clinical practice. Brain MRI scans are not routinely obtained for youth presenting for medical and/or neuropsychological evaluation for FASD, and the average clinical scan does not include quantitative image processing. However, emerging tools that combine automation with deep learning (e.g., SynthSeg²⁸) may provide opportunities for MRI processing required to calculate brain volume centile scores for youth for whom clinical MRI data are available. At the very least, the methods described here may be useful in research for identifying a larger percentage of individuals with brain anomalies resulting from prenatal alcohol exposure.

This study has several limitations. Given that the sample in this study was recruited on the basis of the presence or absence of PAE, there is a clear need for replication in more representative mixed clinical samples. Further research would benefit from comparing brain volumetric atypicality in youth with PAE to potential atypicality in youth with other neurodevelopmental conditions (e.g., attention-deficit/hyperactivity disorder). In addition, in contrast to high-quality normative data for OFC in young children²⁹, currently available normative data for calculating OFC in school-age and adolescent populations (as used in this study) are based on decades-old research with unrepresentative samples with regard to race and ethnicity⁷, which may have contributed to our finding of generally higher than expected OFC centiles in both groups (PAE and controls). Replication of our findings with updated normative data for OFC will be important in the future. We observed that control participants in our sample had brain volumes above the 50th percentile, which may reflect recruitment and screening procedures (i.e., exclusion of individuals with PAE). In addition, studies used in BrainChart may include participants with PAE (whether known or unknown). Lastly, while some diagnostic systems use a cutoff of the 3^rd percentile to classify atypical head circumference³⁰, our sample size did not permit us to examine this alternative criterion. Future research may benefit from a comparison of these two approaches with regard to diagnostic sensitivity and specificity.

In conclusion, we demonstrate the potential value of leveraging a large normative MRI dataset to improve diagnostic sensitivity for the physical brain anomalies that occur in PAE. Youth with PAE who have atypical brain volume also demonstrate greater impairment across several neurobehavioral domains, suggesting these metrics may be useful predictors of functional outcomes. Additionally, many individuals with PAE who would not otherwise meet existing diagnostic criteria for atypical brain volume are identified with this approach. MRI-derived brain volume metrics may be considered as an additional criterion for FASD diagnostic systems, warranting further study.

Figure 1.

Coronal and axial MRI illustrating parcellation used to generate total volumes for cortical gray matter volume (GMV; purple), cortical white matter volume (WMV; blue), and subcortical gray matter volume (sGMV; yellow). Ventricular volume (shown in orange) was not analyzed in the current study.

Table IV.

Sensitivity and specificity results

	Sensitivity	Specificity	PPV	NPV	Accuracy
Height or weight ≤ 10%ile	10.42%	93.02%	62.50%	48.19%	49.45%
OFC ≤ 10%ile	11.11%	100.0%	100.0%	46.67%	50.0%
≥ 2 facial features	26.67%	94.29%	85.71%	50.0%	56.25%
3 facial features	6.67%	100.0%	100.0%	45.45%	47.50%
Low IQ (≤ 1.5 SD below mean)	18.75%	100.0%	100.0%	51.85%	56.67%
GMV or WMV or sGMV ≤ 10%ile	35.42%	95.35%	89.47%	56.94%	63.74%
OFC ≤ 10%ile or brain atypicality	36.96%	94.44%	89.47%	53.97%	62.20%

Note: PPV positive predictive value; NPV negative predictive value; OFC occipitofrontal circumference; GMV gray matter volume; WMV white matter volume; sGMV subcortical gray matter volume

Abbreviations:

FASD

fetal alcohol spectrum disorder

PAE

prenatal alcohol exposure

OFC

occipitofrontal circumference

GMV

gray matter volume

WMV

white matter volume

sGMV

subcortical gray matter volume

WISC-V

Wechsler Intelligence Scale for Children, 5th Edition

D-KEFS

Delis-Kaplan Executive Functioning System

DATA AVAILABILITY STATEMENT

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. Additional information can be found at cifasd.org.