OBJECTIVE: The objective of this study was to determine whether hyperglycemia during the first week of life in extremely preterm (EPT) infants was associated with increased mortality rates and with cerebral injury, as assessed with MRI of the brain, at term-equivalent age.
METHODS: All 143 EPT infants (gestational ages of <27 weeks) who were born at Karolinska University Hospital between January 2004 and December 2006 and were alive at 24 hours were eligible. Of the 118 surviving infants, 24 were excluded for various reasons. MRI was performed for the 94 included survivors at term age, with a 1.5-T system, and scans were scored for gray matter/white matter (WM) abnormalities. Of the 25 infants who died before term age, 6 were excluded because of missing glucose documentation and the remaining 19 were included. Hyperglycemia was defined as plasma glucose levels of >8.3 mmol/L.
RESULTS: Hyperglycemia occurring on the first day of life was identified as an independent risk factor for death (adjusted odds ratio: 3.7 [95% confidence interval: 1.3–10.6]; P = .01). Hyperglycemia occurring on the first day of life also was a risk factor for WM reduction, as determined through MRI, at term-equivalent age (adjusted odds ratio: 3.1 [95% confidence interval: 1.0–9.2]; P = .04).
CONCLUSION: In this population-based cohort of EPT infants, hyperglycemia on the first day of life was associated with increased mortality rates and brain damage, as reflected by WM reduction at term age.
WHAT'S KNOWN ON THIS SUBJECT:
Among preterm infants, hyperglycemia during the first week of life is recognized as a common condition. An association between hyperglycemia and increased morbidity/mortality rates has been shown in some studies, but results are conflicting.
WHAT THIS STUDY ADDS:
This study demonstrates a link between hyperglycemia in the first 24 hours of life and brain MRI abnormalities at term-equivalent age in EPT infants. It also supports previous findings that early hyperglycemia is associated with death.
The association of hyperglycemia with increased rates of morbidity and death is well described for adult patients undergoing intensive care.1,–,11 For critically ill children with trauma,12,–,15 sepsis,16 or burns,17 hyperglycemia has been shown to correlate with poorer outcomes.
Hyperglycemia is recognized as a common condition among preterm infants and is attributable to altered metabolism associated with immaturity, as well as the need for continuous parenteral nutrition.18 It can be triggered by respiratory distress,19 surgery,20 neonatal pain, sepsis, and other stressful events.21 Hyperglycemia in preterm infants has been associated with increased rates of death, intraventricular hemorrhage (IVH),22 sepsis,23 and retinopathy of prematurity24 and with increased lengths of stay,25 but results are conflicting. The use of different definitions of hyperglycemia, varying documentation of blood or plasma glucose levels, and different inclusion criteria complicate comparisons and interpretations. There also is a paucity of data concerning the effects of hyperglycemia on neurologic outcomes. The period between gestational ages (GAs) of 20 and 32 weeks is one of rapid brain growth and development.26 Infants born during this critical growth period are at particular risk for brain damage and for delayed maturation of the central nervous system.26 In this single-center study, we analyzed the relationship of hyperglycemia during the first week of life, in a cohort of extremely preterm (EPT) infants born at GAs of <27 weeks, with adverse outcomes, in terms of death and neurologic morbidity determined through MRI of the brain at term-equivalent age.
All EPT infants (GAs of <27 weeks) who were born alive at Karolinska University Hospital between January 1, 2004, and December 31, 2006, and survived ≥24 hours were eligible for the study (N = 143). After exclusion criteria were implemented, 113 infants were included in the study, including 94 survivors and 19 infants who died (Fig 1). Data on infant characteristics and on death and morbidities were obtained from hospital records. GA was determined through ultrasonography in early pregnancy.27 Small for GA was defined as birth weight >2 SDs below the mean for normal fetal growth.27 The study was approved by the Karolinska University Hospital ethics committee. Informed consent was obtained from all parents.
Blood and Plasma Glucose Measurements
Blood and plasma glucose level values were retrieved retrospectively from the clinical charts for the first week of life. The clinical protocol recommended glucose sampling several times per day for the first days of life. Urinary glucose levels were checked at each micturition. Glucose monitoring was less stringent on subsequent days if glucose levels were stable (3–8 mmol/L), if the infant was in clinically stable condition, and if there was no glycosuria. The number of glucose values for each day differed and, from 1 to 6 days of life, there were increasing numbers of infants without any measured values.
Glucose infusion was started with a 100 mg/mL solution at 60 to 80 mL/kg per day, corresponding to 4 to 6 mg/kg per minute. The intravenous glucose infusion rate was increased daily by 1 to 2 mg/kg per minute, to achieve 10 to 12 mg/kg per minute (15–18 g/kg per day). When hypoglycemia was detected, the glucose infusion rate was increased more rapidly. If persistent hyperglycemia (plasma glucose levels of >11–14 mmol/L for ≥2 consecutive readings) with glycosuria was present, then the glucose infusion rate was reduced. If plasma glucose levels exceeded 14 to 17 mmol/L despite the reduced glucose infusion rate, then insulin therapy was considered. Insulin infusion was started at 0.03 IU/kg per minute and was increased to maintain euglycemia, with plasma glucose levels of 5 to 7 mmol/L. Enteral feedings were started on day 0 at 10 mL/kg per day and were increased by 10 to 20 mL/kg per day according to feeding tolerance. Parenteral nutrition was started on day 1 with amino acids and intralipids at 0.5 to 1 g/kg per day, and rates were increased daily to achieve 4 g/kg per day and 3.5 g/kg per day, respectively.
Glucose readings were performed and documented by the nursing staff by using the HemoCue 201 (HemoCue, Inc, Lake Forest, CA) glucose method.28,29 Blood samples were obtained from umbilical or peripheral arterial lines infused with saline solution only or from peripheral veins. During the 3-year study period, the unit changed from determining whole-blood glucose levels to determining plasma glucose levels (on January 20, 2006). Blood glucose levels were subsequently converted to plasma glucose levels (plasma level = whole-blood level × 1.12).
Hyperglycemia was defined as plasma glucose levels of >8.3 mmol/L and hypoglycemia as plasma glucose levels of <2.6 mmol/L (also see below).18 Infants were identified as having hyperglycemia, hypoglycemia, or both, according to these criteria, for each day of the week. To grade the hyperglycemic load for the first week of life, a scoring system was used.30 The infants were categorized according to the number of days with hyperglycemic levels as follows: group I, no day with plasma glucose levels of >8.3 mmol/L; group II, 1 to 3 days with plasma glucose levels of >8.3 mmol/L; group III, ≥4 days with plasma glucose levels of >8.3 mmol/L. To overcome the inconsistency of different numbers of documented values per patient per day, relative hyperglycemia was calculated as follows: (number of hyperglycemic plasma glucose levels/number of measurements obtained) × total number of measurements in the first week of life. The patients were then recategorized into new groups (groups IV–VI) on the basis of the relative number of plasma glucose levels of >8.3 mmol/L, as follows: group IV, no relative plasma glucose levels of >8.3 mmol/L; group V, 1 to 3 relative plasma glucose levels of >8.3 mmol/L; group VI, ≥4 relative plasma glucose levels of >8.3 mmol/L.
MRI of the Brain and Cranial Ultrasonography
MRI of the brain was performed for the 94 surviving infants at term-equivalent age (GA of 39–41 weeks).30 MRI was performed by using a 1.5-T magnetic resonance system (Philips Intera, Philips Medical, Best, Netherlands). Infants were fed and given chloral hydrate (30 mg/kg, administered orally or rectally) 15 to 30 minutes before the examination.
The scans first were analyzed by a single neuroradiologist experienced in pediatric MRI and then were evaluated by 3 observers blinded to the clinical course. The MRI findings were scored by using a combination of criteria for white matter (WM) and gray matter.31 WM was graded with scores between 1 and 3 (1 = normal, 2 = focal/mild, 3 = >1 region/extensive) for 5 variables, that is, WM signal abnormality (shortening in T1-weighted imaging), WM reduction (WMR) in volume, cystic abnormality, lateral ventricular size and myelination, and thinning of the corpus callosum. WM abnormalities were further classified on the basis of the composite scores in the following 5 categories (potential score range: 5–15): no WM abnormality (scores of 5 or 6), mild WM abnormality (scores of 7–9), moderate/severe noncystic WM abnormality (scores of 10–12), or moderate/severe cystic WM abnormality (scores of 13–15). Gray matter was graded similarly (score range: 1–3) for 3 variables, that is, abnormalities in cortical gray matter signal, maturity of cortical gyration rated with standard gyral models, and size of the subarachnoid space. Composite gray matter scores were then classified as normal (scores of 3–5) or abnormal (scores of 6–9). Interrater agreement was >95%, and discrepancies between observers were resolved through discussion. Results were analyzed for individual item scores as well as moderate/severe WM scores, total gray matter scores, total brain scores,31,32 and WMR coded as normal (scores of 5 or 6) or abnormal (scores of 7–15). Consensus was reached in all cases.
The diagnosis of grade I to IV IVH and periventricular leukomalacia was determined through cranial ultrasonography (ACUSON Sequoia [Siemens, Erlangen, Germany]) performed, according to the clinical protocol, within the first days of life, at 1 to 2 weeks, and subsequently at 2- to 4-week intervals. All ultrasonography was performed by the same neonatologists.
Continuous variables are presented as means and SDs or medians and interquartile ranges and categorical variables as numbers and proportions. Group differences were examined by using Student's t test for continuous variables and χ2 test for categorical variables. Mean glucose levels for all glucose values, hyperglycemic values, and hypoglycemic values obtained for each day of the week were calculated. Repeated-measures analysis of variance was used to determine the differences in glucose variability for the whole week among survivors and deceased infants.
Multivariate linear regression was performed to examine the associations between the outcome variables (primary outcome of death and secondary outcome of MRI abnormalities) and clinical and demographic explanatory variables selected on the basis of earlier research and results of univariate analyses. Age and hyperglycemia were included as explanatory variables. The crude association of each explanatory variable was determined, and then variables were entered into a multivariate logistic model. Interactions between independent variables were examined, and the Wald test was used to test the associations of the variables and interactions. The Hosmer-Lemeshow test was used to examine the overall fitness of the model. SPSS 17.0 (SPSS Inc, Chicago, IL) was used for all data analyses. The level of significance was specified at .05.
There are different definitions of hyperglycemia in the literature, that is, plasma glucose levels of >7.6 mmol/L,11,18 >8.3 mmol/L,22,33 or >10 mmol/L.23,34 We tested these 3 cutoff levels for sensitivity and specificity, with death as the outcome, by using receiver operating characteristic curve analysis.35 The definition of hyperglycemia during the first 24 hours as a plasma glucose level of >8.3 mmol/L was more sensitive and specific for death than were the other cutoff levels (7.6 or 10.0 mmol/L); therefore, that definition was used for coding of hyperglycemia (Fig 2).
Perinatal characteristics and morbidities for survivors and deceased infants are presented in Table 1. Survivors had significantly greater GAs and birth weights and significantly lower Clinical Risk Index for Babies (CRIB) scores than did infants who died before term-equivalent age. Moreover, deceased infants suffered from higher rates of major neonatal morbidities, such as grade III or IV IVH and necrotizing enterocolitis. The perinatal characteristics and morbidities of included infants did not differ from those of surviving or deceased infants who were excluded from the study. The median age of death was 9 days (interquartile range: 3–30 days). Eight infants died during the first week of life. The major causes of death were grade IV IVH (12 of 19 infants), sepsis (9 of 19 infants), and respiratory/cardiac failure (10 of 19 infants), with overlap in 8 of 19 cases.
Hyperglycemia and Hypoglycemia
The mean ± SD plasma glucose level for all 1354 values obtained was 7.4 ± 4.0 mmol/L (range: 0.1–24.9 mmol/L). Hyperglycemia was seen in 81% of the infants and was twice as common as hypoglycemia, which was seen in 41% of infants during the first week of life. Only 10 infants (9 survivors and 1 deceased infant) maintained euglycemia during the first week of life. Infants with euglycemia did not differ from the other infants with regard to perinatal characteristics, except for the fact that none was born small for GA. Mean daily glucose values obtained from all glucose levels, mean hyperglycemic levels, and mean hypoglycemic levels, stratified according to whether infants survived or died, are presented in Table 2. Mean glucose values for all infants increased from 5.8 to 10.3 mmol/L during the first week of life (P < .01) (Table 2). According to our clinical protocol, infants with persistently high plasma glucose levels were treated with decreases in glucose infusion rates. Only 6 infants (5 survivors and 1 deceased infant) were treated with insulin infusions started between the 4th and 6th days of life.
Hyperglycemia and Death Before Term-Equivalent Age
Repeated-measure analysis of variance was used to determine the differences in glucose variability for the whole week in the 2 groups (survivors and deceased infants). We could not demonstrate statistical differences in glucose variability (F1,111 = 0.345; P = .58). At 24 hours, however, the mean glucose values differed significantly (P = .001) between survivors and deceased infants (Table 2). The proportion with hyperglycemia was significantly greater among infants who subsequently died (10 [53%] of 19 infants), compared with survivors (20 [21%] of 94 infants; P = .03) (Table 2). Mortality rates did not differ in relation to the absolute number of days with hyperglycemic episodes during the first week of life for groups I to III. After adjustment for the unequal number of hyperglycemic episodes, an increase in hyperglycemic episodes showed weak evidence of a trend to increased mortality rates, from 6% for group IV to 16% for group V and 27% for group VI (P = .06, by χ2 test for trend). Multivariate logistic regression analysis revealed that hyperglycemia during the first 24 hours of life remained a risk factor for death (Table 3).
Hyperglycemia and Brain Abnormalities on MRI Scans at Term-Equivalent Age
Multivariate linear regression analyses revealed that GA remained statistically significant as a predictor of WM scores, with controlling for demographic and clinical factors, CRIB scores, gender, and hyperglycemia (odds ratio [OR]: −0.19 [95% confidence interval [CI]: −0.35 to −0.02]; P < .04). Variance inflation factor values ranged between 1.0 and 1.3, which indicated the absence of a multicolinearity problem. No statistical association between hyperglycemia and gray matter abnormalities on MRI scans was observed (OR: 1.6 [95% CI: 0.13–18.2]; P = .7).
Multivariate logistic regression analysis revealed that hyperglycemia during the first 24 hours of life remained a risk factor for WMR (Table 3). In addition, male gender and high CRIB scores were observed to be associated with WMR (Table 3). No interaction or multicolinearity problems were observed for the final model. None of the independent variables had a Cook's distance of >1. The Hosmer-Lemeshow test result was P = 0.31, which indicated the fitness of the overall model.
No statistically significant association between hyperglycemia and IVH was found (OR: 1.4 [95% CI: 0.63–3.4]; P = .37). With adjustment for gender and CRIB scores, the association remained statistically nonsignificant (OR: 1.7 [95% CI: 0.58–5.01]; P = .33). When the cutoff point for hyperglycemia was increased from 8.3 to 10 mmol/L, however, the OR was 2.7 (95% CI: 0.99–37.34; P = .052). The CI is very wide, reflecting low power and hence lack of statistical significance.
The results from logistic regression analyses revealed that hypoglycemia was not significantly associated with WMR (OR: 0.57 [95% CI: 0.22–1.47]; P = .24), even after adjustment for gender and hyperglycemia (OR: 0.61 [95% CI: 0.22–1.68]; P = .34). We did not demonstrate statistically significant associations between prenatal (P = .57) or postnatal (P = .77) steroid treatment and WMR. WMR is illustrated in Figure 3.
Hyperglycemia in the first 24 hours after birth among EPT infants was identified as a risk factor for death and for WMR in the brains of surviving infants, as seen on MRI scans (Fig 3), at term-equivalent age. To our knowledge, this is the first report of WMR after early hyperglycemia in EPT infants.
The strengths of the study are that it is population-based and involves a well-defined group of EPT infants. Blood and plasma glucose levels were well defined. By using receiver operating characteristic curves, we were able to test for the best diagnostic cutoff value to define hyperglycemia during the first 24 hours of life, for prediction of the risk for death in this particular population. The weaknesses of the study include its retrospective nature, with inconsistent numbers of glucose measurements, and the clinical nature of the study. Statistical analyses, however, took into consideration the variability in glucose measurements as a confounder or bias. The scoring system used for the relative hyperglycemic events is not an established scoring system. The difference between arterial, venous, and capillary glucose values also was not taken into consideration.
A few studies showed associations between hyperglycemia and death and/or morbidity among preterm infants but, in those studies, different glucose cutoff levels were used, different populations were studied, and different outcomes were analyzed. Most of those studies also concentrated on hyperglycemia during the first week of life. Kao et al23 examined the impact of hyperglycemia (blood glucose levels of >10 mmol/L) in preterm infants born at GAs of <31 weeks and found increases in the rates of death and sepsis. Hall et al25 found longer hospital stays for infants born at GAs of <29 weeks who were admitted with necrotizing enterocolitis and had blood glucose levels of >8.0 mmol/L. In a study of infants born at GAs of <27 weeks, mortality rates were found to correlate with increasing glucose levels and repeated (≥4) incidents of blood glucose levels of ≥9.4 mmol/L.33 Hays et al22 found that hyperglycemia (blood glucose levels of >8.3 mmol/L) in infants with birth weights of <1000 g was associated with increased mortality rates and increased incidence of severe IVH. Our study could not demonstrate an association between hyperglycemia and sepsis or IVH but found associations of hyperglycemia during the first 24 hours of life with death and with cerebral WM changes seen on MRI scans at term age.
Glucose levels are recognized as a predictive factor in some illness-severity scoring systems, such as the Score for Neonatal Acute Physiology Perinatal Extension II (simplified newborn illness severity and mortality risk score) and the Neonatal Therapeutic Intervention Scoring System.36 The results of this study, demonstrating the relationship between hyperglycemia during the first 24 hours of life and adverse outcomes, support the recommendation to use glucose values in these scoring systems. Hyperglycemia probably represents perinatal stress in this context.22 A catecholamine surge attributable to a generalized perinatal stress response could be triggering the high glucose levels.22
The relationship between hyperglycemia and WMR likely represents an association and not a causal effect of hyperglycemia on brain cells. However, experimental studies suggested that hyperglycemia might be acting as an oxidative factor, with subsequent negative effects on brain cells. In this context, the pattern of WM abnormalities may be related to the maturation-dependent vulnerability of developing oligodendrocytes,37,–,39 with more-diffuse, oligodendroglia-specific injury occurring in extremely immature infants. Oligodendrocytes are vulnerable to excitotoxicity. At an early stage of energy deprivation, neurons and oligodendrocytes are fatally injured, mainly as a result of compensatory mechanisms during reperfusion. Hyperglycemia magnifies the compensatory reactions, leading to pronounced astrocytic acidosis during ischemia.40 Therefore, oxidative stress damage might be occurring in part because of high glucose levels.
This is the first study demonstrating a link between hyperglycemia during the first 24 hours of life and brain MRI abnormalities at term-equivalent age in EPT infants. It supports previous findings that early hyperglycemia is associated with death. Future prospective studies are needed to evaluate whether hyperglycemia later in postnatal life has a detrimental effect in EPT infants.
The study was supported by Jerringfonden, Barnforskningen at Astrid Lindgren Children's Hospital, Mjölkdroppen, Sällskapet Barnavård, and Foundation Samariten.
We thank the infants and their families, the neonatal units, the MRI unit, and Brigitte Vollmer, for her valuable insight.
- Accepted October 21, 2009.
- Address correspondence to Mireille Vanpée, MD, PhD, Neonatal Unit, Astrid Lindgren Children's Hospital/Karolinska University Hospital, SE-171 77 Stockholm, Sweden. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
- CRIB =
- Clinical Risk Index for Babies •
- EPT =
- extremely preterm •
- GA =
- gestational age •
- IVH =
- intraventricular hemorrhage •
- WM =
- white matter •
- WMR =
- white matter reduction •
- OR =
- odds ratio •
- CI =
- confidence interval
- Capes SE,
- Hunt D,
- Malmberg K,
- Pathak P,
- Gerstein HC
- Sala J,
- Masiá R,
- González de Molina FJ,
- et al
- Weir CJ,
- Murray GD,
- Dyker AG,
- Lees KR
- Hays SP,
- Smith EO,
- Sunehag AL
- Colvin M,
- McGuire W,
- Fowlie PW
- Bergmeyer HU
- Gray JE,
- Richardson DK,
- McCormick MC,
- Workman-Daniels K,
- Goldmann DA
- Back SA,
- Gan X,
- Li Y,
- Rosenberg PA,
- Volpe JJ
- Back SA,
- Han BH,
- Luo NL,
- et al
- Back SA,
- Luo NL,
- Borenstein NS,
- Levine JM,
- Volpe JJ,
- Kinney HC
- Copyright © 2010 by the American Academy of Pediatrics