[Skip to Content]
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
[Skip to Content Landing]
Figure.
Treatment Efficacy in Relation to Age at Diagnosis in Glucose Transporter Type 1 Deficiency
Treatment Efficacy in Relation to Age at Diagnosis in Glucose Transporter Type 1 Deficiency

Sample rates of full efficacy were 93.8% (size 16) for those 0 years of age, 52.6% (size 19) for those 1 through 3 years of age, and 54.5% (size 11) for those 4 years or older (P = .02, χ2 test).

Table 1.  
Distribution of G1D Manifestations and Other Findings in the G1D Registry
Distribution of G1D Manifestations and Other Findings in the G1D Registry
Table 2.  
Distribution of Overlaps Across 3 Types of Epilepsy
Distribution of Overlaps Across 3 Types of Epilepsy
Table 3.  
Categories of Dietary Treatment
Categories of Dietary Treatment
1.
Pascual  JM, Ronen  GM.  Glucose transporter type I deficiency (G1D) at 25 (1990-2015): presumptions, facts, and the lives of persons with this rare disease.  Pediatr Neurol. 2015;53(5):379-393.PubMedGoogle ScholarCrossref
2.
Marin-Valencia  I, Good  LB, Ma  Q,  et al.  Glut1 deficiency (G1D): epilepsy and metabolic dysfunction in a mouse model of the most common human phenotype.  Neurobiol Dis. 2012;48(1):92-101.PubMedGoogle ScholarCrossref
3.
Pascual  JM, Liu  P, Mao  D,  et al.  Triheptanoin for glucose transporter type I deficiency (G1D): modulation of human ictogenesis, cerebral metabolic rate, and cognitive indices by a food supplement.  JAMA Neurol. 2014;71(10):1255-1265.PubMedGoogle ScholarCrossref
4.
Arsov  T, Mullen  SA, Rogers  S,  et al.  Glucose transporter 1 deficiency in the idiopathic generalized epilepsies.  Ann Neurol. 2012;72(5):807-815.PubMedGoogle ScholarCrossref
5.
Arsov  T, Mullen  SA, Damiano  JA,  et al.  Early onset absence epilepsy: 1 in 10 cases is caused by GLUT1 deficiency.  Epilepsia. 2012;53(12):e204-e207.PubMedGoogle ScholarCrossref
6.
Hildebrand  MS, Damiano  JA, Mullen  SA,  et al.  Glucose metabolism transporters and epilepsy: only GLUT1 has an established role.  Epilepsia. 2014;55(2):e18-e21.PubMedGoogle ScholarCrossref
7.
Klepper  J.  Glucide metabolism disorders (excluding glycogen myopathies).  Handb Clin Neurol. 2013;113:1689-1694.PubMedGoogle Scholar
8.
Leen  WG, Klepper  J, Verbeek  MM,  et al.  Glucose transporter-1 deficiency syndrome: the expanding clinical and genetic spectrum of a treatable disorder.  Brain. 2010;133(pt 3):655-670.PubMedGoogle ScholarCrossref
9.
Tzadok  M, Nissenkorn  A, Porper  K,  et al.  The many faces of Glut1 deficiency syndrome.  J Child Neurol. 2014;29(3):349-359.PubMedGoogle ScholarCrossref
10.
Pearson  TS, Akman  C, Hinton  VJ, Engelstad  K, De Vivo  DC.  Phenotypic spectrum of glucose transporter type 1 deficiency syndrome (Glut1 DS).  Curr Neurol Neurosci Rep. 2013;13(4):342.PubMedGoogle ScholarCrossref
11.
Haspel  HC, Stephenson  KN, Davies-Hill  T,  et al.  Effects of barbiturates on facilitative glucose transporters are pharmacologically specific and isoform selective.  J Membr Biol. 1999;169(1):45-53.PubMedGoogle ScholarCrossref
12.
Kim  SK, Yang  H, Pascual  JM, De Vivo  DC.  Valproic acid enhances glucose transport in the cultured brain astrocytes of glucose transporter 1 heterozygous mice.  J Child Neurol. 2013;28(1):70-76.PubMedGoogle ScholarCrossref
13.
Klepper  J, Fischbarg  J, Vera  JC, Wang  D, De Vivo  DC.  GLUT1-deficiency: barbiturates potentiate haploinsufficiency in vitro.  Pediatr Res. 1999;46(6):677-683.PubMedGoogle ScholarCrossref
14.
Wong  HY, Chu  TS, Lai  JC,  et al.  Sodium valproate inhibits glucose transport and exacerbates Glut1-deficiency in vitro.  J Cell Biochem. 2005;96(4):775-785.PubMedGoogle ScholarCrossref
15.
clinicaltrials.gov. The Glucose Transporter Type I Deficiency (G1D) Registry. NCT02013583. https://clinicaltrials.gov/ct2/show/NCT02013583. Accessed March 10, 2017.
16.
Klepper  J, Flörcken  A, Fischbarg  J, Voit  T.  Effects of anticonvulsants on GLUT1-mediated glucose transport in GLUT1 deficiency syndrome in vitro.  Eur J Pediatr. 2003;162(2):84-89.PubMedGoogle Scholar
17.
Pascual  JM, Van Heertum  RL, Wang  D, Engelstad  K, De Vivo  DC.  Imaging the metabolic footprint of Glut1 deficiency on the brain.  Ann Neurol. 2002;52(4):458-464.PubMedGoogle ScholarCrossref
18.
Akman  CI, Yu  J, Alter  A, Engelstad  K, De Vivo  DC.  Diagnosing glucose transporter 1 deficiency at initial presentation facilitates early treatment.  J Pediatr. 2016;171:220-226.PubMedGoogle ScholarCrossref
19.
Pascual  JM, Wang  D, Yang  R, Shi  L, Yang  H, De Vivo  DC.  Structural signatures and membrane helix 4 in GLUT1: inferences from human blood-brain glucose transport mutants.  J Biol Chem. 2008;283(24):16732-16742.PubMedGoogle ScholarCrossref
20.
Kass  HR, Winesett  SP, Bessone  SK, Turner  Z, Kossoff  EH.  Use of dietary therapies amongst patients with GLUT1 deficiency syndrome.  Seizure. 2016;35:83-87.PubMedGoogle ScholarCrossref
Original Investigation
June 2017

Clinical Aspects of Glucose Transporter Type 1 Deficiency: Information From a Global Registry

Author Affiliations
  • 1Department of Mathematics, University of Texas at Arlington, Arlington
  • 2Rare Brain Disorders Program, Department of Neurology and Neurotherapeutics, University of Texas Southwestern Medical Center, Dallas
  • 3Department of Physiology, University of Texas Southwestern Medical Center, Dallas
  • 4Department of Pediatrics, University of Texas Southwestern Medical Center, Dallas
  • 5Eugene McDermott Center for Human Growth & Development/Center for Human Genetics, University of Texas Southwestern Medical Center, Dallas
JAMA Neurol. 2017;74(6):727-732. doi:10.1001/jamaneurol.2017.0298
Key Points

Question  What are the main diagnostic and therapeutic approaches to glucose transporter type 1 deficiency, and what factors determine outcome?

Findings  In this study of 181 patients with glucose transporter type 1 deficiency in an electronic registry, most patients underwent genetic analysis, with one-fourth of them exhibiting white matter abnormalities by magnetic resonance imaging. Anticonvulsant treatment was ineffective or detrimental, but modified diets were efficacious in two-thirds of cases, with the most important factor in terms of outcome being age at diagnosis.

Meaning  The results of this study indicate a departure from canonical ketogenic diets, emphasize the need for new pharmacologic agents for diet-unresponsive patients, and highlight the value of early diagnosis followed by initiation of dietary therapy.

Abstract

Importance  Case reports regularly document unique or unusual aspects of glucose transporter type 1 deficiency (G1D). In contrast, population studies from which to draw global inferences are lacking. Twenty-five years after the earliest case reports, this deficiency still particularly affects treatment and prognostic counseling.

Objective  To examine the most common features of G1D.

Design, Setting, and Participants  In this study, data were collected electronically from 181 patients with G1D through a web-based, worldwide patient registry from December 1, 2013, through December 1, 2016. The study used several statistical methods tailored to address the age at onset of various forms of G1D, associated manifestations, natural history, treatment efficacy, and diagnostic procedures. These factors were correlated in a predictive mathematical model designed to guide prognosis on the basis of clinical features present at diagnosis.

Results  A total of 181 patients with G1D were included in the study (92 [50.8%] male and 89 female [49.2%]; median age, 9 years; age range, 0-65 years). As previously known, a relatively large variety of common phenotypes are characteristic of the G1D syndrome, including movement disorders, absence epilepsy (typical and atypical), and myoclonic and generalized epilepsies. The 3 main novel results are (1) the feasibility of effective dietary therapies (such as the modified Atkins diet) other than the ketogenic diet, (2) the relatively frequent occurrence (one-fourth of cases) of white matter magnetic resonance imaging abnormalities, and (3) the favorable effect of early diagnosis and treatment regardless of treatment modality and mutation type. In fact, the most important factor that determines outcome is age at diagnosis, as reflected in a predictive mathematical model.

Conclusions and Relevance  The results reveal several changing notions in the approach to G1D syndrome diagnosis and treatment, such as the perceived absolute requirement for a ketogenic diet, the assumed lack of structural brain defects, and the potential existence of genotype-phenotype correlations, all of which are contested by the registry data.

Introduction

The initial descriptions of glucose transporter type 1 deficiency (G1D; the result of mutations in the solute carrier family 2 [facilitated glucose transporter], member 1 gene [OMIM 138140]) emphasized its association with intractable infantile epilepsy responsive to a ketogenic diet. Partial deficiency of the glucose transporter 1 (Glut1) protein results in a restriction to the facilitated movement of glucose across brain capillary endothelial and glial cells.1 This restriction leads to a cerebral carbon (acetyl coenzyme A) deficit,2 which is at least partially amenable to therapeutic compensation by other sources of carbon.3

However, G1D affects all ages. In adults, G1D is increasingly recognized and manifests as a mild chronic encephalopathy with infrequent seizures and various degrees of spasticity and ataxia, which is heritable in an autosomal dominant fashion with variable penetrance, or as a residual severe epileptic encephalopathy with onset in early infancy that tends to ameliorate as postpuberal years progress. The former, milder cases frequently remain underdiagnosed and are often detected after the identification of more severely affected offspring. The latter, more severe adult cases are also at risk of missed diagnoses because of a lack of familiarity with the disorder or the limited application of genetic testing. Many patients diagnosed in the past 25 years with G1D are now reaching adulthood.

Although G1D is still considered to be a rare disorder, estimates suggest that it accounts for approximately 1% of idiopathic generalized epilepsies4 and approximately 10% of early-onset absence epilepsies.5 On the basis of these estimates and the fact that milder phenotypes that consist of less severe symptoms are often not subject to genotyping, G1D is believed to be more common than current prevalence rates would indicate.6,7 Cerebrospinal fluid glucose levels are not always determined and may even be normal, limiting their diagnostic reliability. Thus, G1D diagnosis is likely to be dominated by ascertainment biases that result from phenotypic identification.1 However, a variety of targeted, next-generation DNA sequencing approaches have resulted in the increased detection of G1D in individuals who have never received a lumbar puncture and in adults to whom diagnostic epilepsy DNA panels are applied.

Subtypes of the disorder are also best classified by phenotype. The most common phenotype currently known consists of infantile-onset seizures primarily of the absence type, acquired microcephaly, spasticity, ataxia, and abnormal neuropsychological development.8,9 Another common phenotype is characterized by motor and intellectual dysfunction without seizures.10 Seizures associated with G1D are medication refractory, although findings regarding which antiepileptic medications are beneficial, ineffective, or detrimental have been limited and are often contradictory.11-14

We set out to clarify challenges surrounding G1D diagnosis and treatment through the creation of a patient registry. The registry, the first of its kind to collect detailed medical and treatment history from a broad range of patients diagnosed with G1D, presents an opportunity to explore phenotypes, treatment history, and genotypes.

Methods

The G1D Registry (https://www.g1dregistry.org) is a self-assessment, voluntary, dynamic questionnaire available to patients worldwide. The registry is securely housed within the Health Insurance Portability and Accountability Act–compliant University of Texas Southwestern Medical Center medical records computer servers. The registry was approved by the University of Texas Southwestern Medical Center Institutional Review Board.15

Recruitment methods for the registry included websites (Rare Brain Disorders Program at University of Texas Southwestern Medical Center, clinicaltrials.gov), direct referral from the authors (J.M.P.) at a clinic visit, and links provided in the newsletters from the Glut1 Deficiency Foundation. Study participants provided written informed consent and were allowed to create a username and password for updates to their data. All data were deidentified. The registry questionnaire covered the following areas: (1) diagnosis, including results of any diagnostic tests that patients have completed (such as genotyping); (2) symptoms; (3) symptom history, including medications used; (4) family history; (5) test results (such as blood and spinal fluid tests and electroencephalography, and imaging examinations, including magnetic resonance imaging [MRI], computed tomography, and positron emission tomography); and (6) medical care, including age when symptoms began, when the diagnostic process began, and when a diagnosis was made. The questions were designed to be easily understood, and all technical terms were defined within the registry. We present findings derived from the first 3 years of the registry (December 1, 2013, through December 1, 2016).

R statistical software, version 3.1.0 (R Development Core Team) was used. Where appropriate, test statistics that used χ2 distributions assumed independent, normally distributed data. Whenever an association within or between data items could be expressed as a statistical model with variables to be estimated from a sample, the Wald test was used to test the true value of the variables based on the sample estimate. P values were determined with the χ2 test and z test (normal distribution). P < .05 (2-sided tests) was considered statistically significant.

Results

A total of 181 patients were included in the study (92 [50.8%] male and 89 [49.2%] female; median age, 9 years; age range, 0-65 years). The mean (SD) time of observation for any patient was 3.11 (3.61) years. The total observation time for the 181 patients was 563 patient-years. One hundred fifty-seven (86.7%) of all patients worldwide (predominantly from the United States and English-speaking European territories) were white, 11 (6.1%) were African American, 7 (3.9%) were Hispanic or Latino, and 6 (3.3%) were Asian American or of Asian origin. A total of 138 patients (76.2%) were 17 years or younger (eTable 1 in the Supplement).

Three distinct groups emerged from genotyping, accounting for 98 of 141 patients (69.5%) in our sample: 33 of 125 patients (26.4%) identified a single gene deletion, 45 of 125 (36.0%) identified a point mutation, and 12 of 125 (9.6%) identified no mutation. Only 7 of 98 patients (7.0%) had not undergone genotyping. The rest ignored whether genotyping had been performed. Generally, G1D was diagnosed at a young age, with 71 of 90 (78.9%) having their conditions diagnosed by the age of 3 years (eTable 2 in the Supplement). Among patients with G1D, the most frequent clinical diagnoses before the diagnosis of G1D were abnormal (delayed) neurologic development (108 of 155 [69.7%]), epilepsy or seizures (99 of 155 [63.9%]), and ataxia (87 of 155 [56.1%]). eTable 3 in the Supplement shows the frequency of these manifestations.

Patients reported 3 predominant types of epilepsy. Table 1 lists these types of epilepsy and the patient numbers for each type. Table 2 lists the overlap among these 3 types of epilepsy. The salient neurologic features of the disorder were investigated in terms of patient distribution relative to the persistence of each manifestation (permanent, reversible, or intermittent) under each category (eTable 3 in the Supplement), and manifestations reversible as the result of treatment initiation are noted.

We used the χ2 test to determine whether there was an association between age group at diagnosis and treatment efficacy. We found P = .02, which was less than the significance level of α = .05, indicating a statistically significant association (Figure). We divided the patients into 3 groups according to their ages at treatment initiation: 0 through 2 years, 3 through 5 years, and 6 years or older. Sample rates of full efficacy for the 3 age groups at diagnosis were 93.8% (size 16) for those 0 years old, 52.6% (size 19) for those 1 through 3 years old, and 54.5% (size 11) for those 4 years or older. The association between the age group and efficacy was moderately significant (P = .07, χ2 test) when a significance level of α = .10 was selected. Sample rates of full efficacy were 81.3% (size 16) for those 0 through 2 years old at treatment initiation and 55.6% (size 27) for those 3 years or older group.

We next investigated dietary treatment. Table 3 presents the distribution of patients under 3 categories of the dietary treatment: ketogenic diet (with a further subdivision by ketogenic diet ratio), modified Atkins diet, and other special diets, including the low glycemic index diet and its variants.

Benzodiazepines and barbiturates are often, but not universally, contraindicated on the basis of in vitro evidence.16 We thus asked about treatment-related adverse effects. eTable 4 in the Supplement lists the adverse effects (increased seizure frequency) of the relevant drugs as reported by patients. There was a relatively high incidence of adverse effects for some of these drugs in some of the patients.

We analyzed the possible interdependence of treatment efficacy (full, partial) on 3 predictor variables: ketogenic diet ratio, age at diagnosis, and type of (genetic) diagnosis. We used a logistic regression model to predict log odds of full efficacy at a given level of 3 dependent variables. In the first logistic model with 3 predictors, the coefficients for the 2 indicators for type of diagnosis were not significant (P = .97 [estimate = −0.0356, SD = 1.02, z value = −0.035] and P = .94 [estimate = −0.0639, SD = 0.924, z value = −0.069]), and the variable for type of diagnosis could be dropped from the model. We then considered a logistic regression of treatment efficacy depending on ketogenic diet ratio and age at diagnosis variables. This model was significant (P = .02, goodness-of-fit test, likelihood ratio χ2 test). We thus concluded that the association between age at diagnosis and treatment efficacy was significant; therefore, we reasonably believe that treatment initiation at a younger age likely yields higher efficacy. As further described below, this logistic model is an apt tool to predict treatment efficacy. Full treatment efficacy was found in 15 patients aged 0 years, 10 patients aged 1 through 3 years, and 6 patients aged 4 years or older. Partial treatment efficacy was found in 1 patient aged 0 years, 9 patients aged 0 through 3 years, and 5 patients aged 4 years or older (P = .02). We used a χ2 test to determine whether there was an association between age group at diagnosis and treatment efficacy. P = .02 was less than the significance level α = .05 (sample rates of full efficacy for the 3 age groups at diagnosis were 93.8% [size 16] for those 0 years old, 52.6% [size 19] for those 1-3 years old, and 54.5% [size 11] for those ≥4 years old). We thus conclude that the association between age at diagnosis and treatment efficacy is significant (Figure).

We next asked whether age at diagnosis affects established (later) disease manifestations. We divided the manifestations into 4 types: developmental, neurologic, psychiatric, and others. eTable 5 in the Supplement lists the frequencies of types of manifestations by age group according to their ages at diagnosis. We used a χ2 test to conduct hypothesis testing. The association between age group and manifestation type was not significant (93.8% vs 52.6% full efficacy rates for the 0-year age group vs the 1- to 3-year age group, P = .52, χ2 test).

We also investigated whether the type of genetic mutation affected treatment (diet) efficacy. We divided patients into 3 types according to their DNA genetic diagnoses: deletion or duplication, point mutation, and other types of mutation. eTable 6 in the Supplement gives the frequencies of efficacy under each type of genetic diagnosis by patient count. The association between type of genetic diagnosis and efficacy was not significant (28 with full efficacy and 15 with partial efficacy, P = .35, χ2 test).

After treatment with the ketogenic diet or the modified Atkins diet, the proportion of patients who still experienced incomplete efficacy was significant, reaching a mean (SD) of 31% (7%) (P < .001).

We next investigated potential associations between manifestations. For a patient with seizures, the most likely associated manifestations were abnormal neurologic development (65 [75.6%]), ataxia (47 [54.7%]), intellectual disability (44 [51.1%]), and infantile hypotonia (34 [39.5%]). For a patient presenting with intermittent involuntary gaze, the most likely associated manifestations were abnormal intellectual development (27 [87.1%]), ataxia (24 [77.4%]), seizures (21 [67.7%]), and infantile hypotonia (19 [61.3%]).

A total of 37 of 155 patients (23.9%) had an abnormal MRI result. These abnormalities were (in order of decreasing frequency) hyperintensity of subcortical U-fibers, prominence of perivascular Virchow spaces, and delayed myelination for age. The top 4 most likely associated manifestations among these 37 patients were abnormal intellectual development (33 [89.2%]), seizures (24 [64.9%]), intellectual disability (21 [56.8%]), and ataxia (21 [56.8%]).

A total of 28 surveyed patients (18.1%) had psychiatric abnormalities manifested as obsessive compulsive traits. The top 4 most likely associated manifestations were delayed intellectual development (22 [78.6%]), ataxia (22 [78.6%]), seizures (20 [71.4%]), and intellectual disability (18 [64.3%]).

The rate of adherence with the ketogenic diet treatment was 88%. The association between adherence and efficacy (full, partial, or none) was significant (sample rates for full efficacy were 73.8% [size 42] for the full adherence group and 33.3% [size 6] for the partial or none adherence group, P = .04, χ2 test). Full adherence results in higher efficacy.

We also asked whether the ketogenic diet effectively treats nonseizure manifestations. Patients receiving a ketogenic diet were divided into 2 groups: patients with and without seizures. The association between group and efficacy was not significant (sample rates of full efficacy were 64.5% [size 31] for patients with seizures and 68.4% [size 19] for patients without seizures, P = .78, χ2 test). The rates of full efficacy for the seizure and nonseizure groups were not significantly different. Thus, the ketogenic diet can effectively treat seizures and nonseizure manifestations.

Last, we analyzed treatment efficacy by considering the age at diagnosis and other potentially relevant variables. Specifically, we analyzed whether treatment efficacy (including any treatment) was a function of one or more of the following: age at onset, age at diagnosis, type of diagnosis, and/or treatment modality. We thus studied the dependency of treatment efficacy (full, partial) on 3 predictor variables: ketogenic diet ratio (2:1-5:1), age at diagnosis (0-15 years), and type of (genetic) diagnosis (deletion or duplication, point mutation, or other mutations). We used a logistic regression model to predict the log odds of full efficacy at a given level of the 3 dependent variables. In the model, the diet ratio takes a value from 2 to 5. Ages at diagnosis were divided into 3 groups (representing 0, 1-3, or ≥4 years). Age at diagnosis and type of diagnosis could be represented by 2 indicator variables in the model. We used R statistical software to fit the data for logistic models. In the first logistic model with 3 predictors, the coefficients for the 2 indicators for type of diagnosis were not significant (P = .97 [estimate = −0.0356, SD = 1.02, z value = −0.035] and P = .94 [estimate = −0.0639, SD = 0.924, z value = −0.069]), such that the type of diagnosis could be dropped from the model.

We then considered a logistic regression of treatment efficacy depending on diet ratio and age at diagnosis variables. The model was significant (P = .02, goodness-of-fit test, likelihood ratio χ2 test). Both ketogenic diet ratio (P = .06) and age at diagnosis (P = .06) were moderately significant for the logistic regression model:

log(πi/[1 − πi]) = 5.9967 − 1.1979 × x1 − 2.4365 × x21 − 1.2768 × x22

where πi is the probability of full efficacy for each patient, x1 is ketogenic diet ratio, x21 is the first indicator (group 2 if x21 = 1 and x22 = 0) for age at diagnosis, and x22 is the other indicator (group 3 if x21 = 0 and x22 = 1). We conclude that ketogenic diet ratio and age at diagnosis affect treatment efficacy. In the fitted logistic regression, the intercept is 5.9967 (95% CI, 1.3552-10.6382; P = .01, Wald test), the coefficient for ketogenic diet ratio is −1.1979 (95% CI, −2.4476 to 0.0518; P = .06), the coefficient for age group 2 is −2.4365 (95% CI, −4.441 to −0.4322; P = .02, Wald test), and the coefficient for age group 3 is −1.2768 (95% CI, −3.3027 to 0.7491; P = .22, Wald test). Age group 1 at diagnosis and age group 2 affected treatment efficacy significantly (P = .02), whereas age group 1 at diagnosis and age group 3 did not have any significantly different effect on efficacy (P = .21). All 3 coefficients were negative, indicating that a greater ketogenic diet ratio and a younger age at diagnosis increase efficacy (ie, the probability of full efficacy). Thus, the model can be used to predict the probability of full efficacy. For example, at a ketogenic diet ratio of 2.5:1 and an age at diagnosis of 3 years (group 2), the model predicts that the odds are 1.76, and therefore the probability of full efficacy is 0.6377.

Discussion

The data from this global registry strongly suggest that there should be a reassessment of G1D syndrome despite the limitation that participants were not examined in a clinic setting. In this 181-patient registry, most patients underwent genetic analysis, with one-fourth of them exhibiting white matter abnormalities by MRI. Anticonvulsant treatment was ineffective or detrimental, but modified diets, such as the modified Atkins diet or the low glycemic index diet, were efficacious in two-thirds of cases. This rate of efficacy resembles that of the canonical ketogenic diet. Adherence with these diets may lead to a better outcome. The most important prognostic determinant in terms of outcome was age at diagnosis. In summary, these results indicate a departure from canonical therapeutic ketogenic diets, emphasize the need for new pharmacologic agents for diet-unresponsive patients, and highlight the value of early diagnosis followed by initiation of a dietary therapy. Of note, it is possible to predict long-term clinical outcomes using our regression model and age of diagnosis as the dominant variable. This finding suggests that the age at treatment initiation is more important than the type of dietary treatment followed thereafter.

Manifestations of G1D

The registry data allow for a relatively sharp definition of G1D syndrome; G1D has no obvious sex or racial predilection. Most cases are diagnosed in early infancy, a phenomenon that may have prognostic implications. Most patients (69.7%) experience abnormal intellectual development. Seizures follow in frequency. These seizures are most often convulsive and generalized, followed by absence seizures coexisting with generalized convulsive seizures. Ataxia is also common (56.1% of cases). A variety of nonspecific MRI abnormalities are found in as many as one-fourth of cases. Episodes of full-body weakness and intermittent involuntary gaze affect one-fourth of cases. In approximately half of patients, seizures prove to be a transitory phenomenon that subsides with therapy. The other half continues to experience appreciable seizures (however, less frequently) despite treatment. This finding is unexpected within the paradigm that considers the ketogenic diet an effective substitutive intervention. In addition to seizures, recurrent features that may be refractory to treatment include ataxia, alternating hemiplegia, dyskinesia, and episodic generalized weakness. Psychiatric manifestations (predominantly obsessive-compulsive trait) appear in childhood and persist.

Diagnostic Methods

The diagnosis of G1D is primarily based on genotyping. Other alternative methods, such as the determination of cerebrospinal fluid glucose concentrations, are prone to uncertainty and have become less frequently used. Approximately 8% of all patients have been diagnosed on the basis of positron emission tomography findings,17 although the sensitivity and specificity of this technique remain unknown. Red cell assays for glucose transport are subject to variability and are thus not generally used.

Effect of Early Diagnosis

The most important factor for predictive outcome is age at diagnosis, which corroborates sequential observations made in a series of 133 patients.18 Neither the type of genetic mutation nor the ketogenic diet ratio (when treated with a ketogenic diet) exert any strong effect on phenotype or response to therapy, with ketogenic diet ratio having only a modest effect on treatment outcome. This finding militates against any genotype-phenotype correlations.19 It also argues against recommendations for higher-ratio ketogenic diets, especially when these high ratios are not well tolerated.

Treatments

Most patients receive a ketogenic diet at a 3:1 ratio. However, a significant fraction receive a modified Atkins diet or a low glycemic index diet and its variants. This finding is consistent with a survey of 92 patients receiving dietary therapies.20 However, diet effectiveness decreases when initiated throughout late childhood.

Several anticonvulsants have been discouraged on the basis of in vitro experimental concerns related to direct inhibition of the Glut1 transporter. Among these, benzodiazepines and barbiturates are often used in infantile epilepsies. Consistent with the recommendations to refrain from using benzodiazepines, these agents were, as a group, associated with aggravation of seizures in a relatively disproportionate number of patients, although our data are preliminary in this regard.

Outcome

Our regression model allows for a general prognostic estimation using the age at diagnosis as the principal relevant variable, which holds true especially when diagnosis is followed by any form of tolerable dietary treatment. Mutation type and ketogenic diet ratio are subordinate to age at diagnosis, highlighting the relevance of early diagnosis and therapy.

Limitations

Our analysis is limited by the fact that patients were not available for interview and that the data were not independently corroborated.

Conclusions

The results reveal several changing practices in the approach to disease diagnosis and treatment and underscore the relatively large variety of common phenotypes that comprise the G1D syndrome. The 3 principal findings of our study are as follows: (1) the feasibility of effective dietary therapies other than the ketogenic diet; (2) the relatively frequent occurrence (one-fourth of cases) of white matter MRI abnormalities, including subcortical U-fiber hyperintensity, prominent perivascular Virchow spaces, and delayed myelination for age; and (3) the favorable effect of early diagnosis and treatment regardless of treatment modality and mutation type. In fact, the most important factor that determines outcome is age at diagnosis, as reflected in our predictive mathematical model.

Back to top
Article Information

Corresponding Author: Juan M. Pascual, MD, PhD, University of Texas Southwestern Medical Center, 5323 Harry Hines Blvd, Mail Code 8813, Dallas, TX 75390-8813 (juan.pascual@utsouthwestern.edu).

Accepted for Publication: February 23, 2017.

Published Online: April 24, 2017. doi:10.1001/jamaneurol.2017.0298

Author Contributions: Dr Pascual had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Kelly, Pascual.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: All authors.

Critical revision of the manuscript for important intellectual content: Pascual.

Statistical analysis: Hao, Su.

Obtained funding: Pascual.

Administrative, technical, or material support: Kelly.

Study supervision: Pascual.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was funded by the Glut1 Deficiency Foundation. Dr Pascual is supported by grant NS077015 from the National Institutes of Health and by the Office of Rare Diseases Research G1D Collaboration, Education, and Test Translation program for rare genetic diseases.

Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and the decision to submit the manuscript for publication.

Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Additional Contributions: Nicolai DeGuzman, BS, and Abner Figueroa, BS, Department of Neurology and Neurotherapeutics, University of Texas Southwestern Medical Center, provided technical registry programming support. They were compensated for their work. The generous support of the Glut1 Deficiency Foundation is gratefully acknowledged.

References
1.
Pascual  JM, Ronen  GM.  Glucose transporter type I deficiency (G1D) at 25 (1990-2015): presumptions, facts, and the lives of persons with this rare disease.  Pediatr Neurol. 2015;53(5):379-393.PubMedGoogle ScholarCrossref
2.
Marin-Valencia  I, Good  LB, Ma  Q,  et al.  Glut1 deficiency (G1D): epilepsy and metabolic dysfunction in a mouse model of the most common human phenotype.  Neurobiol Dis. 2012;48(1):92-101.PubMedGoogle ScholarCrossref
3.
Pascual  JM, Liu  P, Mao  D,  et al.  Triheptanoin for glucose transporter type I deficiency (G1D): modulation of human ictogenesis, cerebral metabolic rate, and cognitive indices by a food supplement.  JAMA Neurol. 2014;71(10):1255-1265.PubMedGoogle ScholarCrossref
4.
Arsov  T, Mullen  SA, Rogers  S,  et al.  Glucose transporter 1 deficiency in the idiopathic generalized epilepsies.  Ann Neurol. 2012;72(5):807-815.PubMedGoogle ScholarCrossref
5.
Arsov  T, Mullen  SA, Damiano  JA,  et al.  Early onset absence epilepsy: 1 in 10 cases is caused by GLUT1 deficiency.  Epilepsia. 2012;53(12):e204-e207.PubMedGoogle ScholarCrossref
6.
Hildebrand  MS, Damiano  JA, Mullen  SA,  et al.  Glucose metabolism transporters and epilepsy: only GLUT1 has an established role.  Epilepsia. 2014;55(2):e18-e21.PubMedGoogle ScholarCrossref
7.
Klepper  J.  Glucide metabolism disorders (excluding glycogen myopathies).  Handb Clin Neurol. 2013;113:1689-1694.PubMedGoogle Scholar
8.
Leen  WG, Klepper  J, Verbeek  MM,  et al.  Glucose transporter-1 deficiency syndrome: the expanding clinical and genetic spectrum of a treatable disorder.  Brain. 2010;133(pt 3):655-670.PubMedGoogle ScholarCrossref
9.
Tzadok  M, Nissenkorn  A, Porper  K,  et al.  The many faces of Glut1 deficiency syndrome.  J Child Neurol. 2014;29(3):349-359.PubMedGoogle ScholarCrossref
10.
Pearson  TS, Akman  C, Hinton  VJ, Engelstad  K, De Vivo  DC.  Phenotypic spectrum of glucose transporter type 1 deficiency syndrome (Glut1 DS).  Curr Neurol Neurosci Rep. 2013;13(4):342.PubMedGoogle ScholarCrossref
11.
Haspel  HC, Stephenson  KN, Davies-Hill  T,  et al.  Effects of barbiturates on facilitative glucose transporters are pharmacologically specific and isoform selective.  J Membr Biol. 1999;169(1):45-53.PubMedGoogle ScholarCrossref
12.
Kim  SK, Yang  H, Pascual  JM, De Vivo  DC.  Valproic acid enhances glucose transport in the cultured brain astrocytes of glucose transporter 1 heterozygous mice.  J Child Neurol. 2013;28(1):70-76.PubMedGoogle ScholarCrossref
13.
Klepper  J, Fischbarg  J, Vera  JC, Wang  D, De Vivo  DC.  GLUT1-deficiency: barbiturates potentiate haploinsufficiency in vitro.  Pediatr Res. 1999;46(6):677-683.PubMedGoogle ScholarCrossref
14.
Wong  HY, Chu  TS, Lai  JC,  et al.  Sodium valproate inhibits glucose transport and exacerbates Glut1-deficiency in vitro.  J Cell Biochem. 2005;96(4):775-785.PubMedGoogle ScholarCrossref
15.
clinicaltrials.gov. The Glucose Transporter Type I Deficiency (G1D) Registry. NCT02013583. https://clinicaltrials.gov/ct2/show/NCT02013583. Accessed March 10, 2017.
16.
Klepper  J, Flörcken  A, Fischbarg  J, Voit  T.  Effects of anticonvulsants on GLUT1-mediated glucose transport in GLUT1 deficiency syndrome in vitro.  Eur J Pediatr. 2003;162(2):84-89.PubMedGoogle Scholar
17.
Pascual  JM, Van Heertum  RL, Wang  D, Engelstad  K, De Vivo  DC.  Imaging the metabolic footprint of Glut1 deficiency on the brain.  Ann Neurol. 2002;52(4):458-464.PubMedGoogle ScholarCrossref
18.
Akman  CI, Yu  J, Alter  A, Engelstad  K, De Vivo  DC.  Diagnosing glucose transporter 1 deficiency at initial presentation facilitates early treatment.  J Pediatr. 2016;171:220-226.PubMedGoogle ScholarCrossref
19.
Pascual  JM, Wang  D, Yang  R, Shi  L, Yang  H, De Vivo  DC.  Structural signatures and membrane helix 4 in GLUT1: inferences from human blood-brain glucose transport mutants.  J Biol Chem. 2008;283(24):16732-16742.PubMedGoogle ScholarCrossref
20.
Kass  HR, Winesett  SP, Bessone  SK, Turner  Z, Kossoff  EH.  Use of dietary therapies amongst patients with GLUT1 deficiency syndrome.  Seizure. 2016;35:83-87.PubMedGoogle ScholarCrossref
×