Comparison of the FASD 4-Digit Code and Hoyme et al. 2016 FASD diagnostic guidelines

Background: As clinicians strive to achieve consensus worldwide on how best to diagnose fetal alcohol spectrum disorders (FASD), the most recent FASD diagnosstic systems exhibit convergence and divergence. Applying these systems to a single clinical population illustrates contrasts between them, but validation studies are ultimately required to identify the best system. Currently, only the 4-Digit Code has published comprehensive validation studies. Methods: The 4-Digit Code and Hoyme 2016 FASD systems were applied to the records of 1,392 patients evaluated for FASD at the University of Washington to: 1) Compare the diagnostic criteria and tools used by each system, 2) Compare the prevalence and concordance of diagnostic outcomes and assess measures of validity. Results: Only 38% of patients received concordant diagnoses. The Hoyme criteria rendered half as many diagnoses under the umbrella of FASD (n=558) as the 4-Digit Code (n=1,092) and diagnosed a much higher proportion (53%) as fetal alcohol syndrome/partial fetal alcohol syndrome (FAS/PFAS) than the 4-Digit Code (7%). Key Hoyme factors contributing to discordance included relaxation of facial criteria (40% had the Hoyme FAS face, including patients with confirmed absence of alcohol exposure); setting alcohol exposure thresholds prevented 1/3 with confirmed exposure from receiving FAS/FASD diagnoses; and setting minimum age limits for Alcohol-Related Neurodevelopmental Disorder prevented 79% of alcohol-exposed infants with neurodevelopmental impairment a FASD diagnosis. The Hoyme Lip/Philtrum Guides differ substantively from the 4-Digit Lip-Philtrum Guides and thus are not valid for use with the 4-Digit Code. Conclusions: All FASD diagnostic systems need to publish comprehensive validation studies to identify which is the most accurate, reproducible, and medically valid.


Introduction
As the field of fetal alcohol spectrum disorders (FASD) strives to achieve consensus worldwide on how best to diagnose FASD, the most recent versions of published guidelines (4-Digit Code, 2004 [1]) Canadian, 2015 [2], Hoyme et al., 2016 [3], and Australian, 2016 [4]) show both convergence and divergence. The new Canadian and Australian systems share many features in common, but diverge substantially from the 4-Digit Code and Hoyme et al. systems by dropping the growth deficiency criteria [5] and adopting a nomenclature (FASD with the face, and FASD without the face) that no longer reflects the spectrum of outcome. The 4-Digit Code [1] and Hoyme et al. [3] criteria continue to generate a spectrum of diagnoses under the umbrella of FASD (fetal alcohol syndrome (FAS), partial FAS (PFAS), Alcohol Related Neurodevelopmental Disorder (ARND), Static Encephalopathy/ Alcohol Exposed (SE/AE), Neurobehavioral Disorder/Alcohol Exposed (ND/AE), and Alcohol Related Birth Defects (ARBD)) and maintain the 3 original core diagnostic criteria (growth deficiency, facial anomalies, and CNS abnormalities). The 4-Digit Code and Hoyme et al. systems differ in their diagnostic nomenclature, diagnostic tools, and the specific criteria used to generate each diagnosis. Comparing the diagnostic outcomes generated by the different systems when applied to a single clinical population serves to illustrate the major contrasts and similarities between the systems, but empirical validation studies are ultimately needed to identify the best system.
The objectives of this study were to:

1.
Compare the tools and criteria used by the 4-Digit Code and Hoyme et al. 2016 FASD diagnostic systems.

2.
Administer each system to the records of 1,392 patients to:

a.
Compare the prevalence of FASD diagnoses produced by each system.

b.
Assess diagnostic concordance between the two systems.

c.
Compare measures of validity applied to each system.
The outcomes of Objective 1 helped guide the study design (methods and study population) for Objective 2. Thus, the methods and results for Objective 1 are presented first, followed by the methods and results for Objective 2.

Tools
Lip-philtrum guides: Both diagnostic systems provide 5-point, pictorial lip-philtrum guides for ranking the magnitude of philtrum smoothness and upper lip thinness. The 4-Digit Code provides two guides: Lip-Philtrum Guide 1 for Caucasians and all races with thinner upper lips like Caucasian, and Lip-Philtrum Guide 2 for African Americans and all races with thicker upper lips like African Americans (Figures 1 and 2). Hoyme et al. have also introduced two lip/philtrum guides: the North American Lip/Philtrum Guide [3] produced from a U.S. white population and the South African Mixed Race Lip/Philtrum Guide [6] produced from a Cape Coloured (mixed race) population in the Western Cape Province (Figures 1 and 2). Facial analysis software: The 4-Digit Code advises measuring the facial features from 2D digital photos using the FAS Facial Photographic Analysis Software [7]. The authors of the Hoyme et al. system "feel direct examinations of facial features are more practical in an office setting." Since empirical studies have already confirmed the superior accuracy of the photo versus direct method of facial measurement [8,9], a formal assessment of photo versus direct measurement of facial features was not repeated in this study.  [6], the South African Mixed Race Lip/Philtrum Guide is not appropriate for use on an African American population and thus was not used to address Study Objective 2. The study population for Objective 2 was adjusted accordingly (as described below) to accommodate this finding.

Contrasts in diagnostic criteria
Growth deficiency: The Hoyme et al criteria use the same cut-off (prenatal or postnatal height and/or weight ≤ 10th percentile) to define growth deficiency as the 4-Digit Code, but the Hoyme et al. criteria classify growth deficiency on a dichotomous scale (present/absent), whereas the 4-Digit Code ranks growth deficiency on a 4-point ordinal scale with emphasis on short stature; a method that is confirmed to be highly predictive of CNS dysfunction [5].
Facial phenotype: When compared to the 4-Digit Code Rank 4 FAS facial phenotype, the Hoyme et al. FAS facial phenotype is substantially relaxed. This is best illustrated using the 4-Digit Code Facial ABC-Score printed on the backside of the 4-Digit Code "Caucasian" Lip-Philtrum Guide 1 ( Figure 5). The 4-Digit Code FAS facial phenotype is defined by a single ABC-Score (Facial ABC-Score CCC, Face Rank 4) ( Figure 5A). The three letters "CCC" reflect the magnitude of expression of the short palpebral fissure length (PFL), smooth philtrum, and thin upper lip in that order. C reflects severe expression in the FAS range, B reflects moderate expression, and A reflects normal expression. The Hoyme et al. FAS facial criteria are relaxed relative to the 4-Digit Code in three ways:

1.
Only 2 of 3 cardinal features are required.

2.
The PFL is relaxed to the 10th percentile.

3.
As shown in our analysis above, the Rank 4 lip on the Hoyme et al. North American Lip/Philtrum Guide has a circularity equivalent to the Rank 2 lip on the 4-Digit Lip-Philtrum Guide 1.
This results in almost every 4-Digit Code Facial ABC-Score meeting the relaxed Hoyme et al. facial criteria ( Figure 5B) including 13 of the 15 ABC-Scores that depict the 4-Digit Code Rank 2 (mild) facial phenotype and 3 of the 8 ABC-Scores that depict the complete absence of all three FAS facial features (Rank 1). Clinically, the 4-Digit Code classifies Rank 1 and 2 facial phenotypes as being within the normal range. The practical clinical impact of this relaxation is illustrated in Figure 6 in which an adolescent with high function (e.g., FSIQ 123) and confirmed absence of prenatal alcohol exposure met the Hoyme et al. criteria for the full FAS facial phenotype.
In addition to the contrasts in facial criteria, the scales of measurement used to clinically classify the facial phenotype also differ. The 4-Digit Code documents the full continuum of expression of the FAS facial phenotype (Face Ranks 1 through 4), a continuum confirmed to be highly predictive of CNS dysfunction [5,10]. In contrast, the Hoyme et al system documents the facial phenotype as simply present or absent.
weeks during pregnancy or ≥3 drinks per occasion on ≥2 occasions during pregnancy). The 4-Digit Code requires a confirmed exposure, but does not set thresholds because recall and reporting of quantity, frequency, and timing of exposure have been confirmed highly unreliable in a clinical setting and exposure below a designated threshold has not been confirmed safe for all fetuses [11]. The [7], and new more accurate growth norms have been adopted (CDC growth charts [17] and Stromland Scandinavian PFL charts [18].
For the purposes of research, all patients' clinical 4-Digit Codes are updated to "research" 4-Digit Codes to reflect the most current tools and norms available at the time of the research study. For this study, all 4-Digit Codes were updated to reflect the most current third edition of the 4-Digit Code [1].
FASD diagnoses. The Reader is encouraged to familiarize themselves with the diagnostic criteria specific to each diagnostic system [1,3] as space does not permit replication of the criteria here.

1.
Height and weight normal growth charts: Height and weight percentiles were generated from the Hall [19] birth weight and length growth charts by gestational age; the World Health Organization (WHO) [20] height and weight growth charts for children 0-2 years of age, and the Centers for Disease Control (CDC) 2000 [17] height and weight growth charts for patients 2 years of age and older. The height percentile was adjusted for mid-parental height [21] when both parents' heights were reported.

Facial features:
At the time of each patient's FASD diagnostic evaluation, three standardized, digital facial photographs ( Figure 7) were taken and measured using the FAS Facial Photographic Analysis Software [7]. As a result, each patient's research record included the following facial measures: PFLs in millimeters, philtrum smoothness (Rank 1 to 5 on the 4-Digit Code Lip-Philtrum Guide 1) and upper lip circularity (perimeter 2 /area) and corresponding Lip Rank (Rank 1 to 5 on the 4-Digit Code Lip-Philtrum Guide 1).

1.
PFL: For the purposes of this research study, all PFL z-scores were updated to reflect the Stromland Scandinavian PFL growth charts. The Stromland charts are confirmed valid for use on a North American population [8] and address the full age span (birth through adult) represented in our study population. In addition, the Stromland PFL growth charts were generated from digital images, thus meeting the recommendation by Hoyme et al. [3] that PFLs measured from photos should be compared to PFL normal growth charts generated from photos.   [12,23]. For this reason, this diagnostic classification was not included in this study.
Statistical analyses-Descriptive statistics (valid percentages) were used to profile the study population. Chi-square tests were used to compare groups and linear trends across groups for outcomes measured on nominal or ordinal scales. One-way analysis of variance (ANOVA) was used to compare means and detect linear trends across three or more groups when outcomes were measured on a continuous scale. T-tests were used to compare means between two independent groups.

Results
Objective 2: Compare the diagnostic outcomes between the two systems.

Study population:
The socio-demographic profile of the study population (n=1,392) is presented in Table 1. The population spanned the entire age range from newborn to adult with 57% Caucasian and 44% female. Thirty-five percent (379/1,092) of the patients who received a diagnosis of FASD using the 4-Digit Code did not receive a diagnosis of FASD using the Hoyme et al. system ( Figure  8B). They all had confirmed prenatal alcohol exposures (e.g., birth mother reported drinking throughout the pregnancy), but their record of exposure did not meet the more stringent criteria (e.g., intoxication confirmed by BAC; positive biomarker test like analysis of FAEE; positive outcome on a validated screening tool like the T-ACE or AUDIT; or number of drinks per week or occasion reported) or the level of exposure (e.g., ≥ 6 drinks/week or ≥ 2 weeks or ≥2 drinks/occasion on ≥ 2 occasions) required by the Hoyme et al. system. In some cases, as illustrated in Figure 9, a patient diagnosed with severe FAS (4-Digit Code 4443) did not receive a diagnosis under the umbrella of FASD using the Hoyme et al. system because the exposure level reported directly by the birth mother (1 drink/week throughout pregnancy) was not high enough to meet the Hoyme et al. alcohol-exposure criteria (≥6 drinks/week for ≥2 weeks during pregnancy).
Among the subset of 141 alcohol-exposed patients under 3 years of age, the 4-Digit Code classified 70% (98/141) under the umbrella of FASD ( Figure 8C). The prevalence of SE/AE and ND/AE was 21% (n=29) and 41% (n=58) respectively. In contrast, the Hoyme et al. Objective 2b: Assess diagnostic concordance between the two systems-Diagnostic concordance was observed in 38% (n=528) of the 1,392 patients ( Figure 10). The two diagnostic systems ruled-out FASD in 239 patients and both rendered the same diagnosis under the umbrella of FASD for 289 patients. Diagnostic discordance was observed in 62% (n=864) of the 1,392 patients. The discordance ranged from subtle differences (e.g., the patient received a diagnosis of FAS by one system and PFAS by the other system) to marked contrasts (e.g., the patient received a diagnosis of FAS by one system and no diagnosis under the umbrella of FASD by the other system).
To illustrate some of the more striking contrasts, of the 21 patients that received a diagnosis of FAS/Alcohol Exposed using the 4-Digit Code, 10 had FASD ruled-out altogether using the Hoyme et al. system (see the 4-Digit Code FAS/AE column in Figure 10). All 10 patients were less than 5 years of age. They presented with CNS structural abnormalities (e.g., microcephaly: OFC ≤ 3 rd percentile), but early development was broadly within the normal range. All ten were too young to engage in the necessary level of testing to accurately ruleout moderate or severe CNS dysfunction. The Hoyme et al. system require both CNS structural abnormalities (e.g., OFC ≤ 10 th percentile) and evidence of CNS dysfunction for a diagnosis of FAS.
Among the 208 patients that were classified "Not FASD" by the 4-Digit Code, 39 received a FAS (n=16) or PFAS (n=23) diagnosis using the Hoyme et al. system ( Figure 10). The  The prevalence of each of the four core features that define FASD (growth deficiency, FAS facial phenotype, CNS abnormalities, and alcohol exposure) differed between the two diagnostic systems ( Figure 11). Both systems identified 32% of patients with growth deficiency (height and/or weight ≤10 th percentile The prevalence of the individual FAS facial features also differed between the two diagnostic systems ( Figure 12). More patients were classified with short PFLs using the Hoyme et al. system (77% ≤10 th percentile) than the 4-Digit Code (59% ≤3 rd percentile).  Figure 12B).
Cross tabulation of the CNS structural abnormalities and alcohol exposure classification document further contrasts between the two systems ( Figure 13).  Figure 14).

Objective 2c: Assess measures of validation
Correlation between the FAS facial phenotype and prenatal alcohol exposure: If the FAS facial phenotype is specific to (caused only by) prenatal alcohol exposure, the FAS facial phenotype should be more prevalent among those with higher exposure and should not occur in individuals with confirmed absence of prenatal alcohol exposure. One would also expect that the majority of (if not all) individuals presenting with the FAS facial phenotype would meet criteria for a diagnosis under the umbrella of FASD.
The 4-Digit Code Rank 4 FAS facial phenotype was significantly more prevalent among patients with higher prenatal alcohol exposure ( Figure 15C, D). In contrast, the prevalence of the Hoyme et al. FAS facial phenotype was not more prevalent among patients with higher alcohol exposures. (Figure 15A, B). The mean number of days/week of drinking during pregnancy increased significantly with increasing magnitude of expression of the 4-Digit Code FAS facial phenotype ( Figure 16A). The mean number of days/week of drinking during pregnancy was only marginally higher among those with the Hoyme et al. FAS facial phenotype, but this was driven largely by the inclusion of 65 patients who also met the more stringent 4-Digit Code FAS facial criteria ( Figure 16B). When these 65 patients were removed, there was no longer a significant contrast in alcohol exposure between those with and without the relaxed Hoyme FAS facial phenotype ( Figure 16C).
When the Hoyme et al. and 4-Digit Code FAS facial criteria were applied to an adolescent with high function (FSIQ 123) and confirmed absence of prenatal alcohol exposure (4-Digit Four factors accounted for the greatest contrasts in diagnostic outcomes between the two systems.

1.
The more stringent Hoyme et al. alcohol exposure criteria prevented many with confirmed exposures from receiving a diagnosis of FASD. These more stringent criteria prevented almost one third (339; 29%) of the 1,177 patients with confirmed exposure from being able to receive a diagnosis under the umbrella of FASD ( Figure 14). As we illustrated in Figure 9, individuals with reported prenatal alcohol exposures below the Hoyme et al. threshold can and do present with full FAS when using the 4-Digit Code. Either this patient was particularly vulnerable to the teratogenic impact of alcohol, or the reported exposure was not accurate. In a clinical setting, one is never in a position to know how accurate the exposure was recalled and reported. Setting a threshold implies the details of all reported exposures are accurate and no fetus can be harmed by exposures below the threshold. Neither statement is true and the latter sends a dangerous public health message that lower levels are safe. Recognizing this, the 4-Digit Code requires a confirmed exposure, but does not set a threshold.
It is interesting to note that Petryk et al., [26]  criteria for FAS because the patients were microcephalic, but too young (<5 years old) to engage in the types of testing needed to identify moderate or severe CNS dysfunction. The 4-Digit Code has confirmed that over 90% of alcoholexposed infants and toddlers who present with one or more of the sentinel physical features of FAS as defined by the 4-Digit Code (microcephaly ≤3 rd percentile, a Rank 4 FAS facial phenotype, or Rank 4 growth deficiency) will present with severe CNS Rank 3 dysfunction later in childhood [5]. FAS is a birth defect syndrome, thus, by definition, it is present at birth. Failure to identify and diagnose FAS in newborns and infants will prevent these highest-risk children from receiving the benefits of early intervention. The 4-Digit Code allows evidence of CNS structural or functional abnormality to meet the CNS criteria. This allows FAS to be diagnosed in the newborn/infant who presents with the physical features (growth deficiency, the FAS facial features, and microcephaly), knowing these sentinel features are highly predictive of underlying CNS dysfunction that will manifest later in childhood.

The Hoyme et al. criteria prevent children under 3 years of age from receiving a diagnosis of ARND.
As a result, 84% of the 87 alcohol-exposed infants/toddlers under 3 years of age that presented with moderate to severe CNS dysfunction and received a 4-Digit diagnosis of ND/AE or SE/AE ( Figure 8C) did not receive a diagnosis anywhere under the umbrella of FASD using the Hoyme et al. system. Since ARND, by definition, is Neurodevelopmental Disorder caused by prenatal alcohol exposure, individuals with ARND are born with ARND. Failure to diagnose ARND in alcohol-exposed infants less than 3 years of age may prevent them from receiving the benefits of early intervention. This is particularly concerning because 68 (61%) of these patients had 4-Digit Code Rank 1 or Rank 2 facial phenotypes that are, by our definition, clinically "normal". The Rank 1 and 2 phenotypes have no specificity to prenatal alcohol exposure [27]. The only reason FASD diagnostic systems allow a diagnosis of FAS to be made when prenatal alcohol exposure is unknown is because the facial phenotype is so highly specific to (caused only by) prenatal alcohol exposure, the face serves to confirm the exposure. If the facial phenotype defined by the diagnostic system is not confirmed to be highly specific to alcohol, then: 1) the diagnosis cannot be validly labeled FAS or PFAS because a causal link cannot be confirmed between the patient's alcohol exposure and their adverse outcomes, and 2) the facial phenotype cannot be validly used to confirm prenatal alcohol exposure when the history of exposure is unknown.

-
The 4-Digit Code allows a diagnosis of FAS to be made when prenatal alcohol exposure is unknown because the 4-Digit Code Rank 4 FAS facial phenotype is confirmed to be >95% specific to prenatal alcohol exposure [11,28]. The   Code advises measuring the facial features from 2D digital photos using the FAS Facial Photographic Analysis Software [7]. Empirical studies have confirmed the superior accuracy of the photo versus direct method of facial measurement [8,9]. normal phenotypes with no specificity to prenatal alcohol exposure. This was clearly illustrated in our FASD MRI study [27]. Sixteen high-functioning adolescents with confirmed absence of prenatal alcohol exposure were enrolled as controls in that study. Ten presented with Rank 1 facial phenotypes and 6 presented with Rank 2 facial phenotypes (one of which is illustrated in Figure 6).  Figure 17C). As stated in the Elements of Morphology [31], "A thin upper lip vermilion may be associated with a smooth philtrum and an absence of the Cupid's bow, but these should be assessed separately." In the absence of published validation studies supporting this proposed change in one of the cardinal facial features of FAS, clinical teams should adhere to the thin upper lip vermilion feature that is thoroughly validated [11,28].
One anticipated critique of our use of lip circularity in this analysis is that Hoyme et al. may intend for their lip-philtrum guide to be used for in-person visual comparison, not for photographic analysis using an objective measurement of lip thinness. We considered using retrospective visual comparison with clinic photographs using the Hoyme lip-philtrum guide, but determined that since the lips on the Hoyme guide do not become progressively thinner as Rank increases, and there is some confusion as to whether lip thinness or flat Cupid's bow is being assessed with this guide, it would be too difficult to achieve adequate inter-rater reliability without relying upon the more objective measure of lip circularity.
The quintessential role of the FAS facial phenotype. Why are the criteria used to define the FAS facial phenotype so important to the medical validity of all diagnoses under the umbrella of FASD, not just the diagnosis of FAS? When one makes a diagnosis of FAS, one is stating explicitly that the individual has a syndrome caused by prenatal alcohol exposure [11]. One is also stating explicitly that the biological mother drank alcohol during pregnancy and, as a result, harmed her child. These are bold conclusions to draw and are not without medical, ethical, and even legal consequences. When the FAS face is not specific to FAS and prenatal alcohol exposure, the validity of the entire FASD diagnostic system collapses. Here is why.

•
The term (FAS) is rendered invalid. If the face is NOT specific to (caused only by) alcohol, one can no longer label the condition fetal alcohol syndrome. One can no longer confirm alcohol is causally linked to any of the outcomes (growth, brain, or face) in an individual patient.

•
The diagnosis (FAS with unknown alcohol exposure) is also rendered invalid.
The FAS face can no longer serve as the confirmation of alcohol exposure when the exposure history is unknown.
• FAS is no longer distinct from ARND. ARND is essentially "FAS without the face." But if there is no FAS face, there is no distinction between FAS and ARND. Thus, one can no longer justify classifying FAS and ARND separately.

•
The term "ARND" remains problematic. Since ARND has no feature specific to prenatal alcohol, one is in no position to declare the Neurodevelopmental Disorder is "Alcohol-Related" (ARND) in an individual patient.
There are ethical consequences to the FASD diagnostic nomenclature. With a term like ARND, one feels compelled to require a significant exposure to alcohol to increase the odds that the individual's impairments may be caused, at least in part, by their alcohol exposure. This is a dangerous road to go down.
• Setting a threshold of significant exposure for ARND does not confirm the patient's alcohol exposure is related to their neurodevelopmental disorder.

•
Alcohol is never the only risk factor contributing to the neurodevelopmental disorder. In this study population, 92% were exposed to other prenatal risks including poor prenatal care, pregnancy complications, and exposure to illicit drugs and tobacco. One percent presented with other syndromes (Down, Williams, Sticklers, etc.). Ninety-six percent experienced postnatal risks including trauma, neglect, multiple home placements, and physical/sexual abuse. Seventy-seven percent were in foster/adoptive care at the time of their FASD diagnosis. These other risk factors are so important in the differential diagnostic process, the 4-Digit Code Ranks the severity of these prenatal and postnatal factors on 4-point scales (just like it does for growth, face, CNS, and alcohol exposure) and reports them in full in the patient's medical record.
• One is sending a dangerous message that lower levels of alcohol exposure are safe. As we illustrated in Figure 9, individuals with reported prenatal alcohol exposures below the Hoyme et al. threshold do present with full FAS. Either this patient was particularly vulnerable to the teratogenic insult of alcohol, or the reported exposure was not accurate. In a clinical setting, one is never in a position to know how accurate the exposure is recalled and reported. Setting a threshold implies the details of all reported exposures are accurate and no fetus can be harmed by exposures below the threshold.

•
And one is blaming a woman for harming her child, when they have limited ability to make/defend such a claim.
The 4-Digit Code introduced the terms ND/AE and SE/AE back in 1997 [14]. These terms state the verifiable facts; the individual presents with a disorder and the individual was exposed to prenatal alcohol exposure. The terminology does not explicitly state their disorder is related to their alcohol exposure. In fact, the 4-Digit Code formally Ranks all other prenatal and postnatal risks factors to make clear that alcohol is never the only risk factor contributing to an individual's neurobehavioral disorder of static encephalopathy. In 2013, the DSM5 [32] took a similar nosological approach when it introduced the new term "Neurobehavioral disorder associated with prenatal alcohol exposure" (ND/PAE) as a condition for further study.

When is it a FASD?
Fetal Alcohol Spectrum Disorders are, by definition, adverse outcomes caused by prenatal alcohol exposure. In the absence of an outcome that is specific to (caused only by) prenatal alcohol exposure (like the Rank 4 FAS facial phenotype), one cannot confirm or rule-out the role prenatal alcohol exposure played in an individual's CNS dysfunction. So…

• Do all individuals with SE/AE, ND/AE (or ARND) have FASD?
Not necessarily. Only the subset of individuals whose CNS dysfunction was caused (in whole or in part) by their alcohol exposure has FASD.
• Which subset is that? We currently have no way of knowing. This is why the 4-Digit Code refers to SE/AE and ND/AE as 'broadly" under the umbrella of FASD. Those with SE and ND caused by their alcohol exposure have FASD. Those with SE and ND that was not caused by their alcohol exposure do not have FASD.
• But if they are exposed to high alcohol levels, can't we just assume alcohol caused their disability? No. Not everyone exposed to high levels of alcohol presents with adverse outcomes. Among 2,576 alcohol-exposed patients evaluated in the University of Washington FASDPN Clinic to date, 26 with high exposures presented with full FAS (4-Digit Codes 4444) while 41 with high exposures presented with normal growth, face, and brain development (4-Digit Codes 1114). We also see discordant outcomes among fraternal twins. Among 20 twin pairs with identical high exposures, 5 had normal CNS function while their twin had moderate to severe CNS dysfunction.
When an individual presents with high alcohol exposure and severe CNS dysfunction, but no FAS facial phenotype, as depicted in the diagnosis SE/AE (4-Digit Code 2134): -If their CNS dysfunction is caused (at least in part) by their alcohol exposure, then their SE/AE is an FASD.
-If their CNS dysfunction was caused by other risk factors, not their alcohol exposure, then their SE/AE is NOT an FASD.

-
The only way we can link alcohol to an individual's CNS dysfunction is if they present with a highly specific Rank 4 FAS face (FAS 2434).
• If we cannot confirm alcohol caused their disabilities, does this impact our ability to provide them appropriate intervention? No. Intervention recommendations and a patient's access to services and supports should be based on their disabilities, not on what caused their disabilities. Twenty years of patient surveys [33] confirmed patients with a diagnosis of ND/AE and SE/AE were as likely to access and benefit from interventions as patients with FAS/PFAS. We did not have to label their disorder FAS or PFAS to qualify them for intervention and support services in Washington State.
• Does this impact our ability to prevent FASDs? No. To prevent FASD one must prevent prenatal alcohol exposure. To confirm efforts to prevent prenatal alcohol exposure are working, one needs to document prenatal alcohol exposure in a patient's medical record (regardless of outcome) and track the prevalence of prenatal alcohol exposure by birth cohort annually [34]. If one is reducing the prevalence of prenatal alcohol exposure, one is reducing the prevalence of FASD.

Sensitivity versus specificity
Hoyme et al. (3) reported "Sensitivity and specificity are 2 sides of a diagnostic coin. Theoretically, the guidelines presented here (the Hoyme et al. 2016 guidelines [3]) demonstrating increased sensitivity could lead to over-diagnosis; thus, our advocacy for a structured expert multidisciplinary approach. On the other hand, strict diagnostic cutoffs associated with increased specificity could lead to under-diagnosis of affected children. Children with FASD are subject to a host of societal, educational, health, and judicial problems, all of which are affected by the time of diagnosis. Because early diagnosis and initiation of intervention should be of paramount importance, the authors assert that improved, sensitive, and inclusive diagnostic criteria for FASD should continue to be imperatives in the diagnostic process." As demonstrated in the current study, strict diagnostic cutoffs (e.g., 3 facial features rather than 2, PFL 3 rd percentile rather than 10 th , lip Rank 4 rather than Rank 2; OFC 3 rd percentile rather than 10 th ) associated with increased specificity did not lead to under-diagnosis of affected children when using the 4-Digit Code. The 4-Digit Code uses stringent cutoffs for the FAS facial phenotype to achieve diagnostic accuracy/validity. If the face is not specific to (caused only by) alcohol, one cannot validly label the condition FAS (or PFAS) because one cannot link the patient's adverse outcomes to their alcohol exposure. High specificity does not prevent individuals at risk for FASD from being identified and diagnosed. The 4-Digit Code is able to document the full continuum of outcomes and exposures (from 1113 to 4444) across the entire age span because it is not constrained by the implication of causation that comes with the term ARND. Aase and colleagues [35] urged "simple recording of the verifiable conclusions. If prenatal alcohol exposure has taken place, but FAS cannot be substantiated, the exposure still should be indicated, and any nonspecific abnormalities or problems noted." This is the approach the FASD 4-Digit Code adopted when it was first introduced in 1997 [14]. This approach ensures no one is missed and no one is misdiagnosed.
The two diagnostic systems produce different outcomes, but which one, if either, is correct? Validation studies are required to confirm the accuracy, reproducibility, and medical validity of a diagnostic system. Validity is the degree to which a tool (or diagnostic system) is measuring what it purports to measure [36]. When the 4-Digit Code was introduced in 1997 [14,16], it was published as an empirical study confirming its superior performance to the gestalt [37] approach it was designed to replace. Since then, two decades of more extensive laboratory, clinical, and public health empirical studies have comprehensively affirmed the validity of the FASD 4-Digit Code [11]. A clinician's guide for how to fully assess the performance of FASD diagnostic systems was introduced in 2013 [11] and replicated below ( Table 2). The guide proposes 12 questions clinicians should ask to assess the performance of FASD diagnostic systems. The  The single most important form of validation required of a FASD diagnostic system is confirmation that the FAS facial phenotype is highly specific to (caused only by) prenatal alcohol exposure. As discussed above, in the absence of a highly specific facial phenotype, the validity of the entire FASD diagnostic system collapses. The labels FAS and PFAS are rendered invalid. Diagnoses of FAS and PFAS can no longer be validly rendered when prenatal alcohol exposure is unknown. And FAS and PFAS are no longer distinguishable from ARND. Several published studies have confirmed that the Rank 4 FAS facial phenotype is highly specific (>95% specificity) [11] to prenatal alcohol exposure while the Hoyme et al. FAS Facial phenotype has a reported level of specificity (71% to 75%) [24,25] that is insufficient to confirm it is caused only by prenatal alcohol exposure. This absence of association between the Hoyme et al. FAS facial phenotype and prenatal alcohol exposure was further confirmed in the current study. To date, the FASD 4-Digit Code is the only FASD diagnostic system that has a published record of validation.

Conclusions
The FASD

Supplementary Material
Refer to Web version on PubMed Central for supplementary material.      [1] are as follows: FAS/Alcohol Exposed (Category A); FAS/Unknown Alcohol (B); PFAS/Alcohol Exposed (C); SE/AE (E,F); and ND/AE (G,H). ARND: Alcohol Related Neurodevelopmental Disorder; ARBD: Alcohol Related Birth Defects; FAS: fetal alcohol syndrome; FASD: fetal alcohol spectrum disorders; ND/AE: neurobehavioral disorder/alcohol exposed; pFAS: partial fetal alcohol syndrome; SE/AE: static encephalopathy/alcohol exposed.         Cross-tabulation of CNS structural abnormalities and alcohol exposure classification between the 4-Digit Code and Hoyme systems. To aid in interpretation; A) 315 patients met the Hoyme criteria for CNS structural/ neurological abnormalities. Seventy-nine of the 315 did not meet the 4-Digit Code criteria for CNS Rank 4 because the head circumference was between the 4 th and 10 th percentiles. The 4-Digit Code requires a head circumference ≦3 rd percentile. B) 552 patients were classified as having moderate prenatal alcohol exposure using the 4-Digit Code (Alcohol Rank 3). Of the 551, only 167 met the Hoyme criteria of alcohol exposure. The remaining 384 had confirmed exposures, but details on quantity, frequency, timing, blood alcohol levels, etc. were not available to meet the more stringent Hoyme criteria. OFC: occipital frontal circumference.  Table 2 published in the Hoyme et al. diagnostic guidelines [3]. Patients in this study were classified into only one of these categories starting with the top category (e.g., if a patient was exposed to ≥6 drinks/week ≥2 times and had a DUI, they were classified only in the ≥6 drinks/week ≥2 times category). Overall, 55% of the patients met the Hoyme alcohol criteria, whereas 85% met the 4-Digit Code alcohol criteria.  The 4-Digit FAS face was significantly associated with days/week of prenatal alcohol exposure. A) A strong, significant, linear association was observed between the mean number of days/ week of drinking during pregnancy and increasing magnitude of expression of the 4-Digit Code FAS facial phenotype (Face Rank 1 to 4) (One-way ANOVA Linear term: F=12.7, p=.000; n=615). Patients with the full Rank 4 FAS facial phenotype were exposed, on average, 1.3 more days per week than the patients with no FAS facial features (Face Rank 1). B) A much weaker, but significant, association was observed between alcohol exposure and the Hoyme et al.  The absence of the Cupid's bow is not a more precise method for documenting a thin upper lip The Cupid's bow (black line) is the contour of the line formed by the vermilion border of the upper lip, resembling an archer's bow in the frontal view. These images demonstrate that 1) the lower end of the philtrum groove and ridges form the Cupid's bow [30], and 2) the absence of the Cupid's bow is not a more precise method for documenting a thin upper lip [29]. The presence of a Cupid's bow is dependent on the depth of the philtrum, not the thinness of the upper lip. A deep philtrum will form a Cupid's bow even when the upper lip is thin. A and B) Examples of a deep philtrum creating a Cupid's bow in the contour of a