Rigor of non-dairy galactose restriction in early childhood, measured by retrospective survey, does not associate with severity of five long-term outcomes quantified in 231 children and adults with classic galactosemia

Abstract

One of many vexing decisions faced by parents of an infant with classic galactosemia (CG) is how carefully to restrict non-dairy galactose from their growing child’s diet. Until recently, many experts recommended vigorous lifelong dietary restriction of milk and all high-galactose dairy products as well as some non-dairy sources of galactose such as legumes and specific fruits and vegetables. Recently, experts have begun to relax their recommendations. The new recommendations, that restrict only high galactose dairy products, were made in the face of uncertainty, however, because no sufficiently powered study had been reported testing for possible association between rigor of non-dairy galactose restriction and severity of long-term outcomes in CG. Here we describe the largest study of diet and outcomes in CG reported to date, conducted using information gathered from 231 patients with CG and 71 unaffected sibling controls. We compared rigor of dietary galactose restriction, measured using a 4-point scale by a retrospective parent-response survey, with outcomes including growth, adaptive behaviors, receipt of speech therapy, receipt of special educational services, and for girls and women, a plasma marker of ovarian function (AMH). Our results confirmed the expected differences between patients and controls, but among patients showed no significant association between rigor of non-dairy galactose restriction in early childhood and any of the outcomes quantified. Indeed, some weak associations were seen suggesting that rigorous restriction of non-dairy galactose may be deleterious rather than beneficial. Despite limitations, these findings support the ongoing trend toward diet liberalization with regard to non-dairy sources of galactose for children and adults with classic galactosemia.

Introduction

Classic galactosemia (CG) (Fridovich-Keil & Walter, 2008) is one of the most common inborn errors of metabolism identified by newborn screening in the United States (CDC, 2012). CG results from profound deficiency of galactose-1-P uridylyltransferase (GALT, EC 2.7.7.12) (Isselbacher, Anderson, Kurahashi, & Kalckar, 1956), the middle enzyme in the highly conserved Leloir pathway of galactose metabolism. Affected infants may appear normal at birth, but following exposure to breast milk or a milk-based formula experience a rapid and devastating decline that can progress in days from vomiting, diarrhea, and jaundice to hepatomegaly, failure to thrive, E. coli sepsis, and neonatal death (G. Berry, 2014).

Early detection by newborn screening, which may or may not be pre-symptomatic, coupled with rapid dietary restriction of galactose can prevent or resolve the acute and potentially lethal symptoms of CG. However, despite early detection and intervention, by early to mid-childhood many treated patients experience one or more of a constellation of long-term complications that can include speech, cognitive, and behavioral disabilities in at least half of all patients, tremor and/or other movement problems in close to 40% of patients, growth delay and low bone mineral density in many patients, and primary ovarian insufficiency (POI) in >80% of girls and young women (G. Berry, 2014). The mechanisms and primary or secondary nature of these complications remain unclear, and the variability of long-term outcome severity among patients remains largely unexplained.

One factor proposed as a potential contributor to long-term outcome variability in CG is the degree to which dietary galactose exposure is restricted, especially in early childhood. Indeed, diet recommendations for children with CG have evolved over time largely due to a changing appreciation for the galactose content of some non-dairy foods (e.g. (Gleason, Rasberry, & Van Calcar, 2010)), and the extent of endogenous galactose production in children and adults (G. T. Berry et al., 1995; Schadewaldt, Kamalanathan, Hammen, Kotzka, & Wendel, 2014).

Essentially all infants diagnosed with CG drink low-galactose soy or elemental formula in the first months of life following diagnosis (G. Berry, 2014; Welling et al., 2017). Once solid foods are introduced, however, some children continue on a diet that rigorously excludes not only milk and all high galactose dairy products but also legumes, many fruits and vegetables, and other foods believed to contain concerning amounts of galactose. In contrast, other families of children with CG are advised to continue restricting milk and high galactose dairy foods, but to feed their child an otherwise unrestricted diet (Gleason et al., 2010; S. van Calcar & Bernstein, 2011; S. C. van Calcar et al., 2014; Welling et al., 2017).

Healthcare providers and parents who advise continued vigorous dietary restriction of even non-dairy sources of galactose beyond infancy contend that any galactose exposure puts a growing child with CG at risk (P. Acosta & Yannicelli, 2001; P. B. Acosta & Gross, 1995). Those who advise a more liberal diet beyond infancy (S. C. van Calcar et al., 2014) counter that the trace levels of galactose in legumes, fruits, vegetables, other non-dairy foods, and even some hard cheeses, are negligible compared with endogenous galactose production (G. T. Berry et al., 1995), and children with CG, like other children, benefit from eating a diverse and well-balanced diet. The concern has even been raised that overly vigorous dietary restriction of galactose might worsen outcomes (Shaw, Mulle, Epstein, & Fridovich-Keil, 2016; S. C. van Calcar et al., 2014) potentially by negatively impacting glycosylation (Coman et al., 2010; Coss et al., 2013; Coss et al., 2014; Hughes et al., 2009; Knerr et al., 2015).

Diet recommendations for children and adults with CG were liberalized in recent years following a 2014 task force review of the literature, but the authors of that report clearly stated that their new recommendations were based not on data showing that a liberalized diet is safe, but rather on an absence of data showing that a liberalized diet is not safe (S. C. van Calcar et al., 2014). This reality left healthcare providers and many concerned parents, who had spent years scrutinizing food package labels, in a quandary. Here we present a retrospective study of diet and outcomes of 231 patients and 71 controls confirming, to the statistical power of our study, that restriction of non-dairy sources of galactose in early childhood does not associate significantly with any of the 5 outcomes we quantified. While limited in scope and depth, these results support the recommendation of a more liberalized diet with respect to low-galactose foods for children and adults with CG.

Materials and Methods

Study participants

Patients for this study were selected from among participants currently enrolled in our longitudinal protocol “Bases of Pathophysiology and Modifiers of Outcome in Galactosemia” (Emory IRB00024933; PI: JL Fridovich-Keil), which is approved by the Emory Institutional Review Board and has been enrolling both patients and controls continuously since 1992. Since that time we have recruited patients for this protocol via a combination of self-referral, predominantly from members of the Galactosemia Foundation (www.galactosemia.org), and referral from metabolic clinics, predominantly in North America. Selection criteria of patients from our larger cohort for the study reported here included availability of relevant diet and outcome information. Controls were unaffected siblings recruited from families already in the study.

Diet information

We collected current and retrospective diet information using a parent/guardian survey administered online via REDCap (https://projectredcap.org/), or over the phone, if requested by the family. Diet results used for this study included questions addressing the degree to which different categories of food were excluded from the diet in order to limit exposure to galactose (a) from birth to 12 months, (b) from age 1–5 years, and (c) from age 6 years and up. For each age range, we categorized dietary restrictions into 4 levels of increasing rigor: (1) no dietary restrictions of galactose, (2) excludes only dairy products, (3) excludes dairy and legumes, or (4) excludes dairy, legumes, and at least one other category of food, generally some fruits and/or vegetables.

Outcomes quantified

We included the following 5 outcome areas in this study: height, receipt of speech therapy, receipt of other special educational services, adaptive function as quantified by a parent/guardian or self-response survey, and for girls and women, a hormone marker of ovarian function (AMH).

Information about height was gathered for both patients and controls either by direct measurement, or from medical records, for the following age categories: 2–6 years old, 7–11 years old, 12–16 years old, and ≥17 years old. These age ranges were selected to divide childhood from ages 2–16 into 3 blocks, each of 5 years, roughly defined as early childhood (2–6 years), mid-childhood (7–11 years), and late childhood (12–16 years). We also collected parental heights and calculated age- and gender-adjusted Z-score differentials for all children based on mid-parental heights. If more than one height measurement in a given age range was available for a child, the available height Z-scores within that range were averaged for that child.

We collected information about receipt of speech therapy or other special educational services using a parent-response survey. For the purposes of this study, we reduced each of these complex and longitudinal outcomes to a binary parameter, reflecting whether the patient or control had, or had not, ever received speech therapy or other special educational services.

We collected information about adaptive behaviors using the Adaptive Behaviors Assessment System, 3^rd Edition (ABAS-3; http://www.pearsonclinical.com/); surveys were completed by parent/guardians for their children or by adults for themselves. Scored, the ABAS-3 survey yields a single normalized composite score for each patient or control as well as normalized sub-scores for each of the following areas of adaptive behavior: conceptual, social, and practical. All scores are internally adjusted for age and gender. We tested for association of diet with composite scores as well as each of the 3 sub-scores.

As an accessible marker of ovarian function for girls and women in this study, ages 2–35 years old, we used plasma Anti-Müllerian Hormone (AMH), quantified in ng/mL, as described previously (Sanders et al., 2009; Spencer et al., 2013). For patients for whom we had multiple AMH measurements, we used the most recent value.

Statistical analyses

All statistical analyses were performed in R, version 3.3.1. Eleven outcome parameters (4 growth measures, AMH level, 4 behavioral scores, and 2 educational measures) were each tested individually for univariate association with each of the 5 following potential confounders: race/ethnicity, family income, parent highest education level, gender, and family location. We also tested for possible association with school setting, where relevant. The outcomes “Ever received speech therapy” and “Ever received Special Education” were dichotomous and analysis was performed via logistic regression. All other outcomes were continuous and tested via a linear regression. Whenever one of the potential confounders was significantly associated (p<0.05) with outcome in the univariate model, it was included in the final regression model (logistic regression for dichotomous outcomes, linear regression for continuous outcomes). Each of the 11 outcomes was then tested for association with 4 separate predictors (diet below age 1 year, diet at 1–5 years, diet at 6+ years, and patient versus control status) adjusting for any significant univariate covariates with a general linear model. Specifically, we used Gaussian family for all continuous outcome measures, and binominal family for dichotomous outcomes. Setting a conservative experiment-wide significance threshold of 0.05/(11*4) = 0.0011 suggested that patient/control status was always (or nearly always) experiment-wide significant, while no other association was close, i.e. rarely even nominally significant, and never significant corrected for even five tests.

Results

Gathering diet and outcome information from a relatively large cohort of patients and controls

Both patients and controls in this study were consented and enrolled in our ongoing research study “Bases of Pathophysiology and Modifiers of Outcome in Galactosemia” (Emory IRB00024933; PI: JL Fridovich-Keil). Patients were children and adults who have classic galactosemia; controls were the unaffected siblings of patients. Patients were recruited as described in Materials and Methods; controls were recruited from the families of enrolled patients.

Demographic characteristics of the patient and control populations included in this study are presented in Table 1. In brief, both populations were well balanced by gender and well matched by age. As expected, considering they derived from the same set of families, both groups were also well matched in terms of race/ethnicity, family socioeconomic indicators, and geography. More than 60% of both patients and controls for whom we had the relevant information attended exclusively public schools. Comparing patients to controls via a Chi-square test yielded no significant demographic differences (p>0.05 for all comparisons).

Table 1. Demographic characteristics of patient and control cohorts in this study.

Chi-squared analyses confirmed no significant difference (p>0.05) in any category between Patients and Controls.

Category	Patients n=322	Controls n=114	Patients with diet information n=231	Controls with diet information n=71
Gender
Female	59.9%	51.8%	56.3%	53.5%
Male	40.1%	48.2%	43.7%	46.5%
Age (years)
2–17	56.8%	59.6%	61.0%	80.3%
18–30	27.3%	27.2%	24.2%	15.5%
30+	15.9%	13.2%	14.8%	4.2%
mean ± SD	18.7 ± 12.4	18.0 ± 12.6	18.0±12.6	13.8±10.1
Race/Ethnicity1
White/not of Hispanic origin	81.1%	82.5%	88.7%	94.4%
White/of Hispanic origin	5.0%	4.4%	6.1%	5.6%
Asian/Pacific Islander	1.2%	0.9%	0.9%	0.0%
Mixed	3.7%	2.6%	4.3%	0.0%
Unknown/preferred not to answer	9.0%	9.6%	0.0%	0.0%
Family Income
<$25,000	1.9%	0.0%	2.2%	0.0%
$25,000–$49,999	6.2%	7.9%	7.4%	7.0%
$50,000–$74,999	9.6%	10.5%	11.6%	9.9%
$75,000–$99,999	14.3%	16.7%	18.2%	26.8%
$100,000–$149,999	21.1%	23.7%	26.4%	28.2%
≥$150,000	17.1%	19.3%	22.1%	19.6%
Unknown/preferred not to answer	29.8%	21.9%	12.1%	8.5%
Parent Highest Level of Education
High school graduate or GED	5.3%	0.0%	7.4%	0.0%
Tech/associate’s deg/some college	15.9%	11.4%	19.9%	12.7%
Bachelor’s degree	24.8%	32.5%	31.6%	36.6%
Graduate degree	30.1%	39.4%	38.1%	47.9%
Unknown/preferred not to answer	23.9%	16.7%	3.0%	2.8%
Location
Living in the USA	90.6%	96.5%	89.2%	95.8%
Living outside the USA	7.8%	3.5%	9.9%	4.2%
Unknown	1.6%	0.0%	0.9%	0.0%
School Setting
Public only	45.3%	40.4%	60.2%	64.8%
Private only	6.5%	7.9%	8.7%	12.6%
Combination	11.2%	7.0%	14.7%	11.3%
Unknown	37.0%	44.7%	16.5%*	11.3%*
Predicted residual GALT activity
<0.4%	48.8%	GALT activity not available for this cohort	48.1%	GALT activity not available for this cohort
≥0.4%	13.7%		13.4%
unknown	37.6%		38.5%

¹Only racial/ethnic groups represented in this study population are listed.

^*The majority of children with diet information for whom school setting was unknown were not yet school age. Considering only children ages 6 and older, school setting was unknown for only 3.5% of patients and 1.6% of controls.

Diet information

Diet information for both patients and controls was collected as described in Materials and Methods. Patients and controls included in this study distributed across the 4 diet categories as illustrated in Figure 1. Three patterns were clear. First, as expected, all controls consumed an unrestricted diet (category 1) while all patients consumed a diet that restricted at least high-galactose dairy (categories 2–4). Second, the proportion of patients following the most restrictive diet was largest (60%) in the youngest age group, and smallest (35%) in the oldest age group. Finally, the distribution of patients among diet categories in all 3 age groups was sufficiently balanced to enable a statistically powerful analysis testing whether rigor of dietary galactose restriction associated with severity of long-term outcomes, quantified as described, of the patients in our cohort.

Figure 1. Diet reported for patients and controls in this study.

Variation in rigor of dietary restriction of non-dairy galactose among patients and controls during three age ranges: birth to 12 months, 1–5 years, and 6 years and up. The darker the shading in a panel the more restrictive the diet (see key).

Outcome information

For this study we collected and quantified information from enrolled patients and controls about height, adaptive behaviors, receipt of speech therapy, and receipt of special educational services, as described in Materials and Methods. For girls and women we also collected information about plasma AMH level. The distribution of patients and controls with regard to outcomes quantified for this study is illustrated in Figure 2 or listed in Supplemental Table 1.

Figure 2. Severity of long-term outcomes among patients and controls in this study.

(A) Proportion of patients and controls who did (shaded areas) versus did not (unshaded areas) receive either speech therapy or other special educational services at some point in childhood. (B) Range of severity of representative long-term outcomes among patients and controls included in this study. The panel on the left illustrates child height when 12–16 years old, plotted as Z-score differential from a mid-parental height prediction. The panel on the right illustrates normalized General Adaptive Behavior Composite Scores derived from the ABAS-3. For box and whisker plots, the midline in each box indicates the median sample value while the top and bottom of each box indicate the 75% and 25%, respectively, of the sample values plotted.

For each outcome we observed two clear patterns. First, among patients, as among controls, we saw a spread of scores for each outcome parameter. Second, for each outcome parameter the median and range of scores for patients was shifted downward relative to controls (Figure 2 and Supplemental Table 1). This was expected, as we had selected outcomes to quantify for this study that represent known areas of difficulty for many children and adults with classic galactosemia (G. Berry, 2014). For the purposes of this study, however, it was the variation of outcome scores among patients, and not the difference between patients and controls, that was our focus of attention.

Rigor of non-dairy galactose restriction among patients with CG does not associate with most potential covariates included in this study

Before testing whether rigor of non-dairy galactose restriction associated with long-term outcomes in CG quantified for this study we first applied logistic regression to test whether diet category itself associated with potential covariates that might confound analysis. For each diagnostic category and age group we therefore tested whether any significant relationship was evident between diet category and each of the following potential covariates: gender, race/ethnicity, birth year, family income, parent/guardian’s highest level of education attained, presence or absence of at least 0.4% predicted residual GALT function determined from GALT genotype as described previously (Riehman, Crews, & Fridovich-Keil, 2001; Ryan et al., 2013; Spencer et al., 2013), and public versus private school attendance (if known).

We found no significant association (p≤0.05) between diet category and any of the potential covariates tested except birth year, which was strongly associated with diet in infancy (p=5.26E-08). Specifically, older patients were significantly less likely to have experienced dietary restriction of non-dairy galactose in infancy. The association was weaker between birth year and diet when 1–5 years old (p=0.0013), and there was no evidence of association between birth year and diet when 6 years or older. Whether the apparent association between birth year and diet in early childhood for patients reflects the reality of changing dietary recommendations over time, or other factors, is unclear. Of note, we did not find an association between diet and parent education (p=0.14, 0.29, 0.26, respectively, for the 3 age ranges) or family income (p=0.73, 0.67, 0.62, respectively, for the 3 age ranges). School category was nominally associated with diet at older ages only (p=0.21, 0.09, and 0.01) but was not significant for any age after multiple test correction. Taken together, these results are important because they suggest that diet practices with regard to non-dairy galactose restriction for children with CG, presumably informed by parent consultation with appropriate health care professionals, were not limited by family socioeconomic status.

Child and adult outcomes in CG, as quantified for this study, show no significant association with rigor of non-dairy galactose restriction in infancy or early childhood

Finally, we tested whether variation in rigor of non-dairy galactose restriction of patients in infancy or early childhood associated with variation in the outcomes quantified in our cohort, adjusting for covariates where needed. The result was clear: we saw no significant association between rigor of non-dairy galactose restriction and long-term outcome parameters of patients in the study (Table 2). This was true for all of the outcomes quantified including those relating to growth (height), receipt of speech therapy, adaptive behaviors (ABAS-3 scores), receipt of special educational services, and for girls and women, ovarian function (plasma AMH). To be clear, with the exception of height (Z-score) at age ≥17 years (explained below), none of the outcomes tested showed an association with diet category that was even nominally significant (p≤0.05), meaning the associations would not have been judged significant even if we had not applied any multiple test correction.

Table 2.

Testing for Possible Association of Rigor of Dietary Galactose Restriction With Severity of Long-term Outcomes in Classic Galactosemia.

Outcome	Effect size of control vs. patient status (p-value)	Effect size of strict diet score when <1 year old (p-value)	Effect size of strict diet score when 1–5 years old (p-value)	Effect size of strict diet score when ≥6 years old (p-value)
Height (Z-score)
2–6 years	0.692 (0.0024)*	−0.039 (0.8226)	0.095 (0.4968)	0.257 (0.0606)
7–11 years	0.816 (0.0002)*	0.078 (0.6751)	0.169 (0.3322)	0.082 (0.5692)
12–16 years	0.897 (0.0002)*	−0.327 (0.0699)	−0.155 (0.3285)	−0.173 (0.2473)
≥17 years	0.405 (0.0098)	−0.289 (0.0155)	−0.246 (0.0349)	−0.270 (0.0250)
Ovarian function (plasma AMH ng/mL)	AMH values not available for this control cohort	0.015 (0.8750)	−0.112 (0.1970)	−0.041 (0.6760)
Ever received speech therapy (odds ratio)	0.020 (0.0001)*	1.413 (0.1447)	1.430 (0.1023)	1.431 (0.0948)
Ever received special educational services other than for speech (odds ratio)	0.05 (5.74E-05)*	1.096 (0.632)	1.134 (0.5060)	1.215 (0.2900)
ABAS-3 score (normed to ave=100, SD=15)
Composite	12.00 (3.42E-05)*	−0.543 (0.8124)	0.081 (0.969)	−0.214 (0.922)
Conceptual	13.80 (3.22E-06)*	0.077 (0.974)	0.293 (0.889)	0.367 (0.869)
Social	11.60 (3.96E-05)*	0.942 (0.671)	0.186 (0.925)	0.801 (0.71)
Practical	8.89 (1.99E-03)*	1.57 (0.492)	−0.118 (0.955)	−1.020 (0.641)

^*Differences qualifying as significant after multiple test correction (threshold p=0.0011). Note that, other than height Z-score at age ≥17 years, none of the other associations were even nominally significant (p≤0.05).

Covariates that showed evidence of association with specific outcomes are listed with effect sizes and p-values in Supplemental Tables 2–5; these were accounted for as described in Materials and Methods. For growth, associated covariates included gender (males were taller than females at all ages), family income, and geography; the latter two associated with height in early but not later childhood. For adaptive behaviors quantified using ABAS-3 scores the covariates that associated with outcomes included family income, which associated with higher scores in all sub-sections; parent education, which associated with higher scores in the conceptual and social sub-sections in univariate testing but was not significant when included in the final multivariate model; and geography (US versus non-US), which associated with higher scores in the social sub-section. For receipt of speech therapy, school setting showed a nominally significant association with attendance at private school (p=0.04; ANOVA with 2 degrees of freedom). For receipt of other special educational services, family income associated weakly with increased likelihood of a child having received special educational services at some time in their life, as did attendance at private school (p=0.01; ANOVA with 2 degrees of freedom). Presence of ≥0.4% predicted residual GALT activity, quantified as described previously (Riehman et al., 2001; Ryan et al., 2013; Spencer et al., 2013) associated strongly with higher AMH values, and weakly with never having received speech therapy.

Considering the statistical power of our study, which varied among parameters depending on the number of patients for whom we had relevant data, that we did not find any significant associations between diet and outcomes quantified here means that if rigor of non-dairy galactose restriction influenced these outcomes, the impact accounted for considerably less than 10% of the observed variation for all of the analyses, and less than 5% of the variation for the best powered parts of the study (measures for which we had data for the most patients).

One outcome – height when 17 years or older – actually showed a slight negative association with dietary restriction of non-dairy sources of galactose. In brief, patients who had maintained stricter diets as children appeared to be shorter as adults compared with prediction from age, gender, and mid-parental height than did their counterparts who had consumed more liberal diets in childhood. However intriguing, this apparent association did not reach the level of significance after multiple test correction was applied (Table 2).

Discussion

Our goal in conducting this study was to test whether patients with CG who consumed diets more restricted for non-dairy sources of galactose also experienced milder long-term outcomes, quantified as described in Materials and Methods. The answer for the outcomes we quantified and tested here was no. These outcomes included receipt of speech therapy, growth, ABAS-3 (adaptive behavior) scores, receipt of special educational services, and for girls and women, a hormone marker of ovarian function. While not unexpected given the results of earlier, smaller studies (Jumbo-Lucioni et al., 2012; S. C. van Calcar et al., 2014), the current study provides the most statistically powerful result reported to date and supports the ongoing movement toward greater diet liberalization with regard to non-dairy galactose for patients with classic galactosemia.

Testing for possible association between diet and long-term outcome severity was the principal goal of our study; however, testing possible associations between diet and potential covariates was also revealing. Specifically, that neither family income nor parent education level associated with rigor of child dietary galactose restriction (p≥0.14 for all comparisons) confirmed that, at least in our cohort, differences in child diet were not a reflection of healthcare disparities driven by family socioeconomic factors. Presumably, the range of early childhood diets reported for patients in our study reflected the diversity of recommendations offered over time and in different clinics by healthcare providers to families of children with CG (Gleason et al., 2010; S. van Calcar & Bernstein, 2011; S. C. van Calcar et al., 2014; Welling et al., 2017).

Of note, some healthcare providers and families living with CG have historically considered more rigorous dietary restriction of galactose the “safer” option, at least for young children. However, a closer look at the data presented here, and elsewhere (Shaw et al., 2016), may challenge that assumption. Specifically, we saw a suggestive negative trend between diet and growth such that patients on a more restrictive diet in all 3 age ranges of early childhood actually tended to show more significant growth limitation later in life (≥17 years) compared with patients who ate a less restrictive diet as young children (nominal 0.0155≤p≤0.0349). This result was reminiscent of the suggestive trend we observed previously between rigor of dietary galactose restriction in early childhood and prevalence of gastrointestinal problems later in life for patients with CG (Shaw et al., 2016). While not statistically significant following multiple test correction, both of these trends suggested that a more lenient diet in early childhood might not be simply well tolerated by a child with CG – it might be beneficial. Larger studies are needed to test the significance of these apparent trends.

While our current study is compelling, it has notable limitations. For example, our sample size of 231 patients with diet information, though large for a rare disorder, nonetheless limited statistical power. Our patients and controls also tended to be predominantly white, better-educated, and higher income than the general US population (Heinrichs, Bertram, Kuschel, & Hahlweg, 2005). We cannot rule out that the results might have looked different in a more diverse cohort. Further, we only quantified a small number of outcomes and for some the measurement was crude and/or indirect. Specifically, we gathered information about relevant outcomes in ways that allowed easy quantitation for a large number of patients and controls at low cost and long distance. For example, we quantified adaptive behaviors using the ABAS-3 parent/guardian-response or self-response (for adults) survey rather than by direct assessment by a trained professional. While the ABAS-3 is a normed and validated instrument (http://www.wpspublish.com) that effectively stratified our patient cohort, it did not provide a comprehensive assessment of neurocognitive outcomes. Rigor of non-dairy galactose restriction may therefore have associated with specific neurocognitive or other outcomes that we did not test.

Further, our assignment of individual patients or controls to diet categories for each age range was based on parent/guardian responses to a diet survey; many of these responses were retrospective. Anecdotally, we have found that parents of children with classic galactosemia tend to recall what their child did and did not eat at different stages of life with amazing detail, perhaps because they worried about it so much. Nonetheless, we have no way to validate the responses.

As mentioned above, some of the outcomes included in this study, such as adaptive behaviors and receipt of speech therapy or other forms of special educational service, were also quantified based on parent/guardian or adult self-response; prior studies document that while generally useful, parent reports about their children can be biased (Kroes, Veerman, & De Bruyn, 2003; Pless & Pless, 1995). Future studies involving direct assessment of specific outcomes by trained professionals will be required to address the possible relationship between diet and outcomes more rigorously than we were able to accomplish here. For the present study, however, it is reassuring to see that diet also did not associate with an outcome that was quantified by a fully objective means – plasma AMH level.

Perhaps the most significant limitation of our study is that it was retrospective observational rather than prospective, randomized, and interventional. Specifically, families reported their ongoing or prior diet practices to us; patients were not randomized prospectively into diet categories and then followed longitudinally over time. Therefore, our study tested for the presence or absence of association, not causation, between diet categories and outcome parameters. Had we detected a significant association we would not have been able to say whether the association was causal. Of course, since we did not detect a significant association this limitation is less concerning.

Despite the limitations explained above, the results of our study should help to empower patients, families, and healthcare providers to make more informed decisions about diet for children and adults with classic galactosemia. However, it is important to stress that all of the patients in our study were on diets that carefully restricted high-galactose dairy; the diet categories compared among patients in our study differed only in the degree to which they did or did not restrict non-dairy sources of galactose. Our study therefore did not test whether consumption of high galactose dairy products by children or adults with CG associates with any negative outcomes. Some case observational and small interventional studies have begun to address that question (e.g. (Lee, Lilburn, Wendel, & Schadewaldt, 2003);(Bosch, Bakker, Wenniger-Prick, Wanders, & Wijburg, 2004; Knerr et al., 2015; Panis et al., 2006)) but our study did not.

Finally, what, if anything, do our results say about modifiers of long-term outcome in CG? The absence of a significant association between rigor of non-dairy galactose restriction in early childhood and long-term outcome measures later in childhood, or in adulthood, suggests that chronic exposure to non-dairy galactose in childhood is not a modifier of these later outcomes – at least to the degree we were able to test here. This result is consistent with, but also distinct from the result reported previously by Hughes and colleagues (Hughes et al., 2009), who demonstrated that exposure to high levels of galactose shortly after birth does not associate with severity of long-term outcomes in patients. Here we addressed the impact of chronic low-level dietary galactose exposure across months or years in early childhood. Nonetheless, both results suggest that factors other than dietary galactose play a key role in determining long-term outcome severity in CG. These might include genetic and other intrinsic or environmental factors, including differences in the level of endogenous galactose production or bypass galactose metabolism, or differences in the downstream pathways that are perturbed by a deficiency of GALT. Identifying those modifiers, and leveraging them as targets for improved intervention, will be a focus of future research.

1 sentence take-home message.

Dietary restrictions to exclude non-dairy sources of galactose in early childhood do not associate with severity of five long-term outcomes quantified for 231 children and adults with classic galactosemia.