SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. What this paper adds
  4. Method
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Aim  To determine risk factors for neurological sequelae following hypoglycemia.

Method  We analysed the neurological outcome in 164 patients (mean age 10y 10mo, SD 5.9) following hypoglycemia due to three diseases with various metabolic contexts, different ages at onset, and combinations with comorbidity (fever/infection, hypoxia/ischemia): glycogen storage disease type I (GSDI) (21 patients, mean age at first hypoglycemic episode 3.8mo, SD 3.5); fatty acid β-oxidation defects (FAOD) (29 patients, mean age at first hypoglycemic episode 14.8mo, SD 12.6); and hyperinsulinism (HIns) (114 patients, mean age at first hypoglycemic episode 2.3mo, SD 4.7).

Results  Risk factors of poor neurological outcome were aetiology (p<0.006), comorbidity (p<0.001), and prolonged convulsions (p<0.001). Ordinal logistic regression showed that comorbidity (p<0.001) and status epilepticus (p=0.002) were the main determinants of sequelae. Asymptomatic hypoglycemia did not lead to sequelae, whatever the aetiology. Age was not correlated to sequelae, whatever the aetiology. The highest prevalence of hypoglycemic sequelae was found in FAOD and HIns combined with comorbidity, the lowest in GSDI (p<0.001) in which hypoglycemia is often asymptomatic, associated with increased plasma lactate, and rarely combined with comorbidity.

Interpretation  Hypoglycemia is severely deleterious for the brain in the context of fever/infection and/or hypoxia/ischemia, and status epilepticus. The metabolic context providing alternative fuels may improve neurological outcome.


Abbreviations
FAOD

Fatty acid β-oxidation defect

GSDI

Glycogen storage disease

HIns

Hyperinsulinism

What this paper adds

  1. Top of page
  2. Abstract
  3. What this paper adds
  4. Method
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  •  In the context of inborn errors of metabolism causing hypoglycemia, comorbidity is the major determinant of poor neurological outcome, whatever the aetiology.
  •  The metabolic context in which hypoglycemia occurs and whether alternative fuels are available is crucial for outcome.
  •  Status epilepticus also contributes to the severity.

Hypoglycemia causes brain damage,1,2 particularly in the developing brain, but its deleterious consequences remain unpredictable. Although it can occur at any age, hypoglycaemia is particularly frequent in neonates and infants, within various aetiological contexts. Experimental studies point to the resistance of the neonatal brain to the deleterious effects of hypoglycemia.3,4 Hypoxia, seizures, and jaundice have been reported to contribute to brain injury in hypoglycemia, and duration of hypoglycemia plays a major role.3,5 However, the diversity of the aetiological and environmental context of hypoglycemia hinders efforts to determine their impact on the brain, and the deleterious consequences of hypoglycemia remain poorly predictable.

Various inborn errors of metabolism, namely fatty acid β-oxidation defect (FAOD), hyperinsulinism (HIns), and glycogen storage disease type I (GSDI), cause hypoglycemia early in life with variable incidence and severity of brain damage, and therefore constitute ‘experiments of nature’ to investigate the context of hypoglycemic brain damage.

To identify the risk factors of poor neurological outcome following hypoglycemia, we compared the outcome of three hypoglycemic conditions with different environmental and metabolic contexts: (1) GSDI in which hypoglycemia usually occurs after the neonatal period and provides alternative energy sources, namely lactic acid; (2) FAOD in which hypoglycemia occurs in both neonates and infants without alternative energy sources; and (3) HIns, which affects neonates and infants, and offers no alternative fuel to the brain.

Method

  1. Top of page
  2. Abstract
  3. What this paper adds
  4. Method
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

For the three aetiological groups we reviewed patients’ files and collected the age at first hypoglycemic episode, the clinical expression of first hypoglycemic episode, plasma glucose level at first episode, values of plasma lactate and the presence of ketonuria at the time of diagnosis of hypoglycemia, comorbidity (fever/infection, hypoxia/ischemia [i.e. shock, cardiorespiratory arrest, or severe bradycardia]), and neurological sequelae. Sequelae were divided into mild (speech delay, learning difficulties, and pharmacosensitive epilepsy) and severe (global psychomotor delay, microcephaly, motor deficit, lack of visual contact, and/or pharmacoresistant epilepsy). All patients with severe sequelae underwent magnetic resonance imaging, which disclosed brain lesions (Gataullina et al., unpublished material). Of the ten patients with mild sequelae who underwent magnetic resonance imaging, only one exhibited non-specific bilateral white-matter T2 hypersignal.

Using univariate analysis, we first searched the whole series for an association between the occurrence of sequelae – mild or severe – and age, clinical expression of hypoglycemia, the presence of comorbidity, aetiology, and plasma glucose level. We then performed an ordinal logistic regression, using the proportional odds model, to test the specific contribution of each variable. This allowed identification of which variables contributed to the item studied, here the occurrence of neurological sequelae, when all other variables were fixed.

To discard the role of aetiology, we then concentrated on each etiological group, searching for associations within the same list of items. Because there were neonatal cases of HIns with perinatal comorbidity, we separated patients with HIns alone from those with comorbidity. Values of plasma lactate and the presence of ketonuria at the time of diagnosis of hypoglycemia were compared within aetiology groups. We then performed an ordinal logistic regression restricted to the HIns group.

Participants

We included in the study all consecutive patients with GSDI, FAOD, and HIns referred for hypoglycemia to the Reference Centre of Inherited Metabolic Diseases in Necker-Enfants Malades Hospital, Paris, from 1971 to 2007 for GSDI and FAOD, and from 1989 to 2007 for HIns (because the number of patients with HIns was much higher). Patients followed-up for less than 18 months were not included.

The 164 selected patients had the following distribution: 21 with GSDI types Ia or Ib (nine females, 12 males); 29 with FAOD (16 females, 13 males) including 13 patients with medium-chain acyl-CoA dehydrogenase deficiency, six with long-chain acyl-CoA dehydrogenase deficiency, five with very long-chain acyl-CoA dehydrogenase deficiency, four with glutaric aciduria type 2, and six with carnitine palmitoyl transferase-1 deficiency; 114 with HIns (55 females, 59 males) when excluding patients with syndromic HIns (congenital defects of glycosylation and Beckwith–Wiedemann, Costello, Sotos, and Kabuki syndromes), HIns associated with hyperammonaemia, and exogenous insulin administration (Münchausen by proxy syndrome).

Hypoglycemia was defined as a plasma glucose level below 2.8 mmol/l (glucose oxidase measure eventually preceded by Dextrostix®). Because many patients exhibited recurrent hypoglycemia (many were probably overlooked, especially for HIns and GSDI), the first identified episode was analysed. Routine measurement of plasma glucose level in the absence of any clinical symptom related to hypoglycemia was usual in neonates with HIns and in infants with GSDI as hepatomegaly led to the diagnosis. In the other patients, symptoms of hypoglycemia were mild (tremor, sweating, pallor, hypotonia), or comprised brief (less than 15min duration) or prolonged convulsions (more than 15min) with coma. We called these four modes of expression respectively ‘asymptomatic’, ‘mildly symptomatic’, ‘brief’ convulsion, and ‘coma/status epilepticus’.

GSDI was diagnosed when hypoglycemia followed brief fasting, without response to glucagon, usually associated with hepatomegaly, increased plasma lactate, and hypertriglyceridemia, and was confirmed by measurement of hepatic enzyme in earlier cases and by molecular genetics in the more recent patients.6 FAOD was suspected in the presence of severe hypoglycemic attacks (mostly coma) after prolonged fasting, usually without ketosis. Diagnosis was confirmed by plasma acylcarnitine and urinary organic acid profiles, fatty acid oxidation studies, and molecular genetics.7 In HIns, hypoglycemia occurred in both fasting and postprandial states, was associated with inappropriate secretion of insulin, and responded well to subcutaneous or intravenous glucagon; 61 (53.5%) were resistant to diazoxide. The diagnosis was confirmed by molecular genetics.8

Statistical analysis

Data were analysed using R software9 for univariate analysis and for ordinal logistic regression with proportional odds. For the latter, the outcome was y=0, 1, 2, for ‘no sequelae’, ‘mild sequelae’, and ‘severe sequelae’. There was one continuous covariable, the glucose level. Factors with k levels (i.e. aetiology three levels, comorbidity two levels, age group at diagnosis four levels, clinical expression four levels) were taken into account by k − 1 indicator variables. Because of the need of k − 1 indicator variables for a k-level factor, nine variables corresponded to these four factors; thus there was a total of 10 variables in the full model. The proportional odds model is logit p(Yi)=αi+βx for i=1, 2, where α1, α2 and β=(β1,..., β10) are the 12 model parameters and x=(x1, …, x10) is the vector of variables.

Threshold for significance was 0.05. All tests were two tailed.

Results

  1. Top of page
  2. Abstract
  3. What this paper adds
  4. Method
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Determinants of neurological sequelae

Whole series

For the 164 patients, follow-up ranged from 18 months to 30 years (mean 10y 5mo). The occurrence of sequelae was strongly correlated with aetiology (p=0.006), the mildest being GSDI and the worst FAOD (Table I). Sequelae also correlated with the presence of comorbidity (p<0.001) and the occurrence of status epilepticus (p<0.001) for the whole series. By contrast, the prevalence of mild and severe sequelae did not correlate with age at diagnosis of hypoglycemia. For plasma glucose levels, although there was a trend, it did not quite reach significance. Ordinal logistic regression was performed on the whole series of 164 patients. As Table II shows, only comorbidity (p<0.001) and level 4 of clinical expression corresponding to coma/status epilepticus (p=0.002) were significantly related to the sequelae. The above result remained unchanged when a classical logistic regression was performed where sequelae were considered in two levels, severe versus no or mild sequelae.

Table I.   Features associated with sequelae in the whole series
 No sequelae (n=110)Mild sequelae (n=38)Severe sequelae (n=16) p a
  1. aχ2 or Fisher’s exact test for percentages, or analysis of variance for means. bOne patient with HIns with white Dextrostix® never had plasma glucose below 2.9 mmol/l after immediate correction. GSDI, Glycogen storage disease; FAOD, Fatty acid β-oxidation defect; HIns, Hyperinsulinism.

Age at diagnosis
 Mean (SD), mo4.8 (8.4)4.6 (8.2)4 (6.7)0.93
 Range<1–48.6<1–36.6<1–24.3 
 Patients (%) <1d303719 
 1d to 1mo181838 
 1mo to 1y423431 
 >1y1011120.59
Clinical expression (percentage of patients)
 Asymptomatic27260 
 Mildly symptomatic15130 
 Brief convulsions432912 
 Coma/status epilepticus153288<0.001
 Comorbidity (percentage of patients)143987<0.001
Aetiology (percentage of patients)
 GSDI15116 
 FAOD122150 
 HIns7368440.006
Plasma glucose level
 Mean (SD), mmol/l1.2 (0.8)1.3 (0.8)0.8 (0.9)0.08
 Range0–2.9b0–2.80–2.4 
Table II.   Ordinal logistic regression (total sample, n=164)
CoefficientEstimateStandard error p
  1. *p values are for a two-sided Wald test. The ‘omnibus test’ leads to a χ2 value of 57.28 with 10 df; the corresponding p value is 10−8.

α1−2.0800.7890.008
α2−4.2250.873<0.001
Aetiology=2−0.7130.9310.444
Aetiology=3−0.1840.6480.776
Comorbidity=11.9860.479<0.001
Age at diagnosis=20.6850.5580.219
Age at diagnosis=30.1930.5910.744
Age at diagnosis=4−1.1350.8600.187
Clinical expression=20.1840.6700.784
Clinical expression=3−0.0350.5790.951
Clinical expression=42.2650.7400.002
Plasma glucose level0.2730.2480.271
Within etiological groups

In FAOD, all patients with severe sequelae except one presented with coma/status epilepticus, and 66% had comorbidity, mainly fever and viral infection (most often gastroenteritis) and cardiorespiratory failure. Only one neonatal case without comorbidity had severe sequelae following hypoglycemia revealed by brief seizures. All five other hypoglycemic events in the context of FAOD causing severe sequelae were triggered by infection. There was no association between the occurrence of sequelae and plasma glucose level or the age of occurrence.

GSDI disclosed the lowest prevalence of severe sequelae: only one patient remained with severe brain injury following hypoglycemia with prolonged convulsions and fever.

In HIns, 24 patients had a neonatal onset with perinatal comorbidity. They consisted of asphyxia (seven patients), maternal infection with respiratory distress (six patients), asphyxia with preterm birth, and maternal infection (three patients each), low birthweight at term delivery (two patients), preterm birth alone or with jaundice or maternal infection (one case each). Among the patients with HIns without comorbidity, 48 had a neonatal onset, and 42 patients had infantile disclosure of hypoglycemia. Table III shows that comorbidity, the occurrence of coma/status epilepticus, and plasma glucose level were correlated with sequelae. The occurrence of severe sequelae was significantly higher in the group with neonatal onset and comorbidity compared with the group with neonatal onset but no comorbidity and with the group with infantile onset (p<0.001; Fig. 1a). Results were similar when excluding the 30 asymptomatic patients. In the HIns group, ordinal logistic regression showed that, as for the whole series, only comorbidity (p=0.001) and level 4 of clinical expression corresponding to coma/status epilepticus (p<0.001) were related to the sequelae. The above result remained unchanged when a classical logistic regression was performed where sequelae were considered in two levels, severe versus no or mild sequelae.

Table III.   Features associated with sequelae in patients with hyperinsulinism
 No sequelae (n=81)Mild sequelae (n=26)Severe sequelae (n=7) p a
  1. aχ2 or Fisher’s exact test for percentages, or analysis of variance for means.

Age at diagnosis
 Mean (SD), d84.2 (163.5)35.8 (54.6)2.0 (1.6)0.14
 Range0.01–10950.01–1800.17–5.0 
 Patients (%) <1d374243 
 1d to 1mo212757 
 1mo to 1y38310 
 >1 y4000.25
Clinical expression (percentage of patients)
 Asymptomatic28270 
 Mildly symptomatic15190 
 Brief convulsions543929 
 Coma/status epilepticus31571<0.001
 Comorbidity (percentage of patients)113586<0.001
Plasma glucose level
 Mean (SD), mmol/l1.2 (0.8)1.5 (0.7)0.6 (0.6)0.01
 Range0–2.90–2.80–1.65 
image

Figure 1.  (a) Prevalence of sequelae in hyperinsulinism (HI). Prevalence of sequelae after neonatal or infantile hyperinsulinaemic hypoglycemia with or without associated factors. Light grey bar, no sequelae; dark grey bar, mild sequelae; black bar, severe sequelae; NN, neonates; Inf, infants. Fisher’s exact test, p<0.001. (b) Prevalence of sequelae according to aetiology. Sequelae according to aetiology and associated factors. Light grey bar, no sequelae; dark grey bar, mild sequelae; black bar, severe sequelae. Fisher’s exact test, p<0.001. GSDI, Glycogen storage disease; FAOD, fatty acid β-oxidation defect.

Download figure to PowerPoint

As shown in Figure 1b, the prevalence of sequelae was the same in FAOD and in HIns with comorbidity, whereas it was low in GSDI and HIns without comorbidity (p<0.001).

Alternative sources of energy

Plasma lactate was normal in all patients with HIns. Its level did not correlate with plasma glucose level (data not shown) in GSDI and FAOD. Increased plasma lactate was concomitant with hypoglycemia in 20 of 21 (95.2%) patients with GSDI, but did not differ significantly between patients with convulsions and those without. Only 24.1% (7/29) of patients with FAOD exhibited moderate increased plasma lactate (mean 6mmol/l) during hypoglycemic coma. Ketonuria, assessed using reactive strips at the time of hypoglycemia, was present in 23.8% of patients with GSDI and 24.1% of patients with FAOD (Glutaric aciduria type 2 and medium-chain acyl-CoA dehydrogenase), whereas it was absent in those with HIns.

Discussion

  1. Top of page
  2. Abstract
  3. What this paper adds
  4. Method
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

We determined in neonates and young children the risk factors associated with hypoglycemic neurological sequelae in various hypoglycemic situations differing by the environmental and metabolic context in three etiological groups: GSDI, FAOD, and HIns. Age at hypoglycemia and plasma glucose level do not influence neurological outcome. Aetiology and combination of hypoglycemia with status epilepticus, as well as with hypoxia–ischemia, fever, or infection significantly correlate with neurological sequelae, with comorbidity and status epilepticus being the main determinants of sequelae.

Doubtful effect of age

Although most hypoglycemic injuries are described in neonates, age at hypoglycemia does not seem to be a determinant of sequelae.10 HIns is a main cause of hypoglycemia in neonates. Because neonates are mostly predisposed to hypoxia, jaundice, and other factors known to aggravate the outcome of hypoglycemia, a higher prevalence of neurological sequelae in neonates could be assumed. However, experimental studies show that the immature rat brain is resistant to the deleterious effects of hypoglycemia alone. Resistance of the newborn brain to hypoglycemia could result from preferential use of alternative fuels, mainly lactic acid and ketone bodies, whereas low glucose uptake reflects a low expression of glucose transporters.3,11,12 Later in life, although ketone bodies remain an alternative source for the developing brain, glucose becomes an essential fuel.3 The precise age at which the human brain becomes glucose dependent is unknown. The progressive expression of glucose transporter proteins corresponds to maturational increase in cerebral glucose uptake and use, whereas monocarboxylic acid transporter expression peaks during the suckling period and declines with maturation in the rat brain.13 Although age does not contribute to the occurrence of neurological sequelae, we could show that it determines the topography of brain lesions.

The impact of aetiology

Because neither ketone bodies nor lactate are available in hypoglycemic states due to HIns and FAOD, it is not surprising that these conditions are the most deleterious to the neonatal brain. HIns inhibits lipolysis and the production of ketone bodies, and sustains a low level of plasma lactate.14 In FAOD, there is a combination of energetic Acyl-CoA enzyme defect by muscle tissue, including the heart, and lack of production of ketone bodies which are essential fuels for normal cardiac function in addition to the brain.15 By contrast, the lactate level is elevated in GSDI owing to a block in the last step of gluconeogenesis. Studies in animals show that exogenous and endogenous lactate can replace glucose in the brain.16 This could explain why GSDI, which provides lactate as an alternative fuel to glucose, is characterized at the time of diagnosis (within the first year of life) by a low incidence of neurological symptoms. The ability to produce lactate to feed the brain could account for the well-tolerated neonatal fasting. However, lactate supply may not be sufficient several months or years later, when hypoglycemia is combined with infection, as observed in one of our patients.

Conflicting impact of glucose blood level

Brain lesions are believed to be caused by severe and prolonged hypoglycemia, duration being stated as a major feature.5 However, duration and number of episodes of hypoglycemia are difficult to quantify in clinical studies because vigorous treatment is administered at the occurrence of hypoglycemia, which might be frequent and unexpected. In contrast with studies in preterm infants whom values are monitored prospectively, the context of occurrence of hypoglycemia in the present study could only provide the blood level at the time of diagnosis. Although association between severe sequelae and glucose blood level reached significance in patients with HIns, this was not the case for the two other aetiologies.

Impact of status epilepticus

We showed that status epilepticus in combination with comorbidity (namely hypoxia–ischemia, fever, infection) is the most highly correlated with severe neurological sequelae. Although status epilepticus could be the mere expression of brain damage itself (as for hypoxic–ischemic encephalopathy or bacterial meningitis) and therefore more likely a consequence than the cause of brain damage, prolonged epileptic activity per se could also contribute to the damage in case of hypoglycemia. Experimentally, long-lasting seizures are known to generate brain damage in early life, particularly in the context of inflammation.17 Long-lasting convulsions enhance energy needs, especially in the context of hypoglycemia or hypoxia, and contribute to severe neurological sequelae.18 Decoupling of enhanced energy requirements and the ability to provide it in a particular metabolic context may contribute to such damage. This explains why the occurrence of epilepsy in the case of mitochondrial disease, a major cause of energy failure, precipitates deterioration and death.19 FAOD and neonatal HIns with comorbidity, the two aetiological conditions with the highest incidence of sequelae, are also those in which status epilepticus is the most frequent.

Impact of comorbidity

Several perinatal factors, namely hypoxia–ischemia, are known to exacerbate brain injury in the context of hypoglycemia.3–5 We show here that neurological sequelae are caused mainly by the combination of hypoglycemia and comorbidity, whereas patients of our series with recurrent hyperinsulinaemic hypoglycemia but no comorbidity had a favourable outcome. Interestingly, comorbidity exhibits a higher link with sequelae than glucose blood level, whether dealing with the whole series or restricting to HIns. The occurrence of sequelae also depends on the aetiology of hypoglycemia, being significantly higher in FAOD in which comorbidities such as infection and cardiorespiratory failure are more frequently observed than in the other aetiologies. However, the relative impact of the various comorbidities, i.e. hypoxia–ischemia and infection, cannot be established because they are often combined.

Birth-related hypoxia–ischemia in neonatal HIns, and cardiac failure in FAOD, cause energy failure leading to impaired brain ionic homeostasis and neuronal transporter function, resulting in convulsions and glutamate excitotoxicity.20 Cerebral glucose use is increased in situations of hypoxia–ischemia,21 causing plasma glucose levels to be particularly low in FAOD where hypoglycemia occurs in the context of cardiorespiratory failure. The combination of hypoxia–ischemia with hypoglycemia is more deleterious to the immature brain than either condition alone3–5, whereas uncomplicated hypoglycemia in the newborn of a diabetic mother does not cause severe neurological sequelae, although learning problems are on record.22

In our series, infection was a clear precipitating factor for FAOD and for several patients with neonatal HIns. Although acute brain damage in the context of FAOD is usually considered to result from hypoglycemia due to prolonged fasting caused by anorexia and vomiting related to acute infection, this series shows that fever/infection itself may determine the severity of hypoglycemia and neurological sequelae. In GSDI, infection triggered severe brain damage in the only patient who suffered an acute event with viral infection. Therefore fever/infection contributes to neurological sequelae following hypoglycemia, whatever the aetiology. Inducers of inflammation decrease the expression of various enzymes involved in fatty acid β-oxidation, orienting energy supply towards exclusive glucose consumption.23 This could aggravate energy failure and explain respiratory and especially cardiac distress observed in critical infectious situations. Inflammation exacerbates energy depletion causing ischemic sequelae.24 It is also a trigger for convulsions in the developing brain and is responsible for neurological sequelae.19,25

On the other hand, hypoglycemia might reinforce the deleterious neuro-inflammatory response induced by infection. Thus, it has been shown that glycemic dysregulation is a potent activator of key triggering receptors of innate immune response, namely nod-like receptors. In addition to pathogenic components and glycemic variations, nod-like receptors also sense many other slight defects of cellular integrity – through the so-called danger-associated molecular patterns – which might well make, from multiple-hit-induced and nod-like receptor-driven massive release of deleterious neuroinflammatory molecules, one of the major detrimental pathways in hypoglycemic patients.26

Conclusions

  1. Top of page
  2. Abstract
  3. What this paper adds
  4. Method
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

The risk of poor neurological outcome in the context of hypoglycemia is mainly determined by the presence of comorbidity such as hypoxia–ischemia, fever and infection, as well as status epilepticus. These energy-depleting conditions could, in combination with hypoglycemia, determine energy failure, resulting in severe neurological sequelae. The main aetiologies that determine such a context are FAOD and neonatal HIns with perinatal comorbidity because severe hypoglycemia in both instances is accompanied by energy-depleting factors. This should be taken into account in the care of patients.

Acknowledgements

  1. Top of page
  2. Abstract
  3. What this paper adds
  4. Method
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

We are grateful to M Brivet, G Scher, C Chiron, R Nabbout, and G Sebire, who reviewed the manuscript, and to the French Neuropediatric Society, which supported this study.

References

  1. Top of page
  2. Abstract
  3. What this paper adds
  4. Method
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References