Neurocritical Care

, Volume 19, Issue 3, pp 311–319

Reduced Brain/Serum Glucose Ratios Predict Cerebral Metabolic Distress and Mortality After Severe Brain Injury

Authors

  • Pedro Kurtz
    • Neurological Intensive Care Unit, Departments of Neurology and NeurosurgeryColumbia University College of Physicians and Surgeons
  • Jan Claassen
    • Neurological Intensive Care Unit, Departments of Neurology and NeurosurgeryColumbia University College of Physicians and Surgeons
  • J. Michael Schmidt
    • Neurological Intensive Care Unit, Departments of Neurology and NeurosurgeryColumbia University College of Physicians and Surgeons
  • Raimund Helbok
    • Neurological Intensive Care Unit, Departments of Neurology and NeurosurgeryColumbia University College of Physicians and Surgeons
  • Khalid A. Hanafy
    • Neurological Intensive Care Unit, Departments of Neurology and NeurosurgeryColumbia University College of Physicians and Surgeons
  • Mary Presciutti
    • Neurological Intensive Care Unit, Departments of Neurology and NeurosurgeryColumbia University College of Physicians and Surgeons
  • Hector Lantigua
    • Neurological Intensive Care Unit, Departments of Neurology and NeurosurgeryColumbia University College of Physicians and Surgeons
  • E. Sander Connolly
    • Neurological Intensive Care Unit, Departments of Neurology and NeurosurgeryColumbia University College of Physicians and Surgeons
  • Kiwon Lee
    • Neurological Intensive Care Unit, Departments of Neurology and NeurosurgeryColumbia University College of Physicians and Surgeons
  • Neeraj Badjatia
    • Neurological Intensive Care Unit, Departments of Neurology and NeurosurgeryColumbia University College of Physicians and Surgeons
    • Neurological Intensive Care Unit, Departments of Neurology and NeurosurgeryColumbia University College of Physicians and Surgeons
Original Article

DOI: 10.1007/s12028-013-9919-x

Cite this article as:
Kurtz, P., Claassen, J., Schmidt, J.M. et al. Neurocrit Care (2013) 19: 311. doi:10.1007/s12028-013-9919-x

Abstract

Background

The brain is dependent on glucose to meet its energy demands. We sought to evaluate the potential importance of impaired glucose transport by assessing the relationship between brain/serum glucose ratios, cerebral metabolic distress, and mortality after severe brain injury.

Methods

We studied 46 consecutive comatose patients with subarachnoid or intracerebral hemorrhage, traumatic brain injury, or cardiac arrest who underwent cerebral microdialysis and intracranial pressure monitoring. Continuous insulin infusion was used to maintain target serum glucose levels of 80–120 mg/dL (4.4–6.7 mmol/L). General linear models of logistic function utilizing generalized estimating equations were used to relate predictors of cerebral metabolic distress (defined as a lactate/pyruvate ratio [LPR] ≥ 40) and mortality.

Results

A total of 5,187 neuromonitoring hours over 300 days were analyzed. Mean serum glucose was 133 mg/dL (7.4 mmol/L). The median brain/serum glucose ratio, calculated hourly, was substantially lower (0.12) than the expected normal ratio of 0.40 (brain 2.0 and serum 5.0 mmol/L). In addition to low cerebral perfusion pressure (P = 0.05) and baseline Glasgow Coma Scale score (P < 0.0001), brain/serum glucose ratios below the median of 0.12 were independently associated with an increased risk of metabolic distress (adjusted OR = 1.4 [1.2–1.7], P < 0.001). Low brain/serum glucose ratios were also independently associated with in-hospital mortality (adjusted OR = 6.7 [1.2–38.9], P < 0.03) in addition to Glasgow Coma Scale scores (P = 0.029).

Conclusions

Reduced brain/serum glucose ratios, consistent with impaired glucose transport across the blood brain barrier, are associated with cerebral metabolic distress and increased mortality after severe brain injury.

Keywords

Brain injuryComaCerebral microdialysisGlucoseMetabolismGlucose transport

Introduction

Hyperglycemia has been associated with morbidity and poor outcome in patients with severe brain injury [14]. Tight glucose control with intravenous insulin has been shown to reduce mortality among surgical ICU patients, but not in mixed populations of critically ill patients [58]. For this reason, current guidelines recommend against tight glycemic control in most patients [9]. The impact of tight glycemic control in neurologic critically ill patients remains controversial [10, 11]. Recent data suggests that intensive insulin therapy not only fails to improve the outcome of neurologic patients, but may be deleterious due to an increased incidence of hypoglycemia and low brain tissue glucose levels [11, 12]. Microdialysis studies of cerebral metabolism after severe brain injury indicate that tight glucose control is associated with an increased risk of metabolic crisis, which is defined as an elevation of the lactate/pyruvate ratio (LPR) in the presence of low brain tissue glucose [1316].

Neuronal cellular energy production depends on an adequate supply of glucose and its metabolites. Serum glucose levels affect glucose availability to the brain and can impact cellular metabolism and energy production after severe brain injury [13, 14]. As a result of impaired glucose transport into the central nervous system (CNS), serum glucose levels considered to be normal may be relatively insufficient to meet the increased cerebral metabolic demand seen in patients with acute brain injury [1722]. The ratio between brain interstitial and serum glucose levels may reflect the adequacy of glucose transport across the blood–brain barrier into the CNS, glucose transport into the intracellular compartment, the adequacy of intracellular glucose and oxidative metabolism, and other factors. Low brain extracellular glucose levels and elevated LPRss are well-established predictors of mortality after severe brain injury [20, 23].

In this study, we sought to better understand the potential role of impaired glucose transport in producing secondary brain injury. Specifically, we hypothesized that low brain/serum glucose ratios are associated with cerebral metabolic distress and increased mortality in patients with severe brain injury.

Methods

Patients

We retrospectively analyzed prospectively collected physiologic and clinical data in 46 consecutive patients admitted to the neurological ICU at Columbia University Medical Center between May 2006 and January 2009 for severe brain injury caused by aneurysmal subarachnoid hemorrhage (SAH), intracerebral hemorrhage (ICH), traumatic brain injury (TBI), or cardiac arrest (CA). All patients underwent multimodality monitoring (MMM) with intracranial pressure (ICP), cerebral microdialysis, and brain tissue oxygen pressure (PbtO2) as part of their clinical care, and were enrolled in an observational MMM database. This study was approved by the Columbia University Institutional Review Board, and consent was provided by family members or legally authorized representatives. Twenty-eight of the SAH patients in this report were previously included in a separate analysis of serum glucose fluctuations and brain metabolic distress [16].

Clinical Management

Patient care for SAH and ICH conformed to guidelines established by the American Heart Association, and TBI management conformed to Brain Trauma Foundation guidelines. Hemodynamic and fluid management were targeted to maintain cerebral perfusion pressure (CPP) >60 mmHg and ICP <20 mmHg. Hemoglobin cutoff for blood transfusions was 7 mg/dL unless there was clinical, imaging, or laboratory evidence of active cerebral or myocardial ischemia. Therapeutic temperature modulation (fever control or induced hypothermia for CA or ICP control) using intravascular (Celsius Control System®, Innercool Therapies, Inc, San Diego, CA) or surface (Arctic Sun Cooling System®, Medivance Inc, Louisville, CO) cooling devices. Shivering was treated with buspirone, skin counterwarming, magnesium infusion, and analgo-sedation (dexmedetomidine, fentanyl or meperidine) according to a stepwise protocol [2426].

Serum Glucose Control

Serum glucose was measured with the Sure Step Flexx system (Lifescan, Milpitas, CA) and the target range was between 4.4 and 6.7 mmol/L (80–120 mg/dL) as part of a tight serum glucose control protocol using intravenous insulin infusion (Humulin). Enteral nutrition (Osmolite, Ross Nutrition, Abbott Laboratories, Columbus, OH) was provided via a naso-duodenal tube starting within the first 24 h of admission, aiming to 25 kcal/kg/day of ideal body weight. No parenteral nutrition was given. Values of serum glucose were categorized into four ranges: low range (<4.4 mmol/L or 80 mg/dL); tight range (4.4–6.7 mmol/L or 80–120 mg/dL); intermediate range (6.8–10.0 mmol/L or 121–180 mg/dl); high range (>10 mmol/L or >180 mg/dL) [13]. The majority of serum glucose measurements while patients underwent neuromonitoring were performed every 1 to 2 h; the maximum time between measurements was 4 h. To maintain consistency, only capillary finger stick glucose measurements were analyzed.

Multimodality Neuromonitoring

ICP, PbtO2, and microdialysis probes were placed via a triple lumen bolt at the bedside using full sterile technique. PbtO2 (mmHg) was measured with a flexible polarographic Licox Clark-type probe (Integra Neurosciences®, Plainsboro, NJ). ICP was measured using an intraparenchymal fiberoptic catheter (Camino System, Integra Neurosciences®, Plainsboro, NJ). Hourly microdialysis samples were obtained with a 10 mm membrane length CMA-70 microdialysis catheter (CMA Microdialysis®, Stockholm, Sweden). The probes were placed via a frontal approach into the hemisphere deemed at greatest risk for secondary injury (i.e., perihematomal or pericontusional tissue, or the ipsilateral anterior watershed zone in lateralized SAH), or in the right frontal lobe in patients with diffuse injury. Immediately after the procedure, a brain CT scan was performed in each patient to confirm the location of the microdialysis catheter.

Cerebral Microdialysis

A CMA 106 microdialysis perfusion pump (CMA Microdialysis®) was used to perfuse the interior of the catheter with sterile artificial cerebrospinal fluid (Na+ 148 mmol/L, Ca2+ 1.2 mmol/L, Mg2+ 0.9 mmol/L, K+ 2.7 mmol/L, Cl 155 mmol/L) at a rate of 0.3 μL/min. Samples were collected every 60 min into microvials, and immediately analyzed at the bedside for glucose, lactate, and pyruvate (mmol/L) with the CMA 600 analyzer (CMA Microdialysis®). At least 1 h passed between the insertion of the probe and the start of the sampling, to allow for normalization of changes due to probe insertion. The analyzer was automatically calibrated on initiation and every 6 h using standard calibration solutions from the manufacturer. Quality controls at three different concentrations for each marker were performed daily.

Brain/serum glucose ratios were calculated hourly by dividing glucose levels obtained through microdialysis by the most recent preceding serum glucose value (in mmol/L) within a maximum window of 2 h. Metabolic distress was defined as a LPR above 40. This threshold was defined based on previous reports demonstrating associations with cerebral metabolic disarray, cerebral ischemia, or poor clinical outcome [13, 19, 2733] .

Physiologic Variables

Physiological variables including heart rate, arterial blood pressure, respiratory rate, fraction of inspired oxygen (FiO2), and oxygen saturation (SpO2) were continuously monitored in all patients. Hourly ICP and mean arterial pressure (MAP) were prospectively recorded as part of the standard of care. CPP was calculated as CPP = MAP–ICP, with both MAP and ICP referenced to the level of the foramen of Monroe. A CPP threshold of 70 mm Hg was selected as a risk factor for metabolic distress (LPR elevation) based on prior studies showing that values below this level are associated with an increased risk of brain metabolic dysfunction [34, 35]. FiO2 was routinely maintained at 40 % according to our ICU standard management protocol.

Data Acquisition

A Solar 8,000i utilizing a General Electric Medical Systems Information Technologies’ Unity Network® was used as the patient physiologic monitor. A high resolution data acquisition system (BedmasterEX, Excel Medical Electronics, Jupiter, FL) using an open architecture of the Unity Network® automatically acquired vital signs, alarm, and waveform data from all the patient monitoring devices in the NICU. Digital data was acquired every 5 s and recorded in an SQL database. Waveform data was stored at a resolution of 240 Hz in binary files. LICOX® (Integra Neuroscience, Plainsboro, NJ, USA) and brain metabolism data were incorporated into the data acquisition system utilizing the communications (COM) port on the device which was plugged into a serial-to-TCP/IP interface device (Equinox ESP-8, Avocent, Sunrise, FL).

Statistical Analysis

We used the Kolmogorov–Smirnov test to evaluate MD lactate, pyruvate, LPR, and glucose for normality, and found that all were non-normally distributed (all P < 0.001). Due to the small sample of patients, large number of measurements, and lack of normality the data were analyzed using generalized estimating equations (GEE). Univariate analyses were used to test for associations between predictor and outcome variables. Variables with significant associations (P < 0.1) were considered candidates for the multivariable analyses. Multivariable models were constructed using a general linear model (GLM) with a logistic link function (logistic regression), extended by GEE to account for within-subject variation. The within-subject correlation structure was modeled using the autoregressor of the first order [36, 37]. Model building was performed with a stepwise procedure starting with the variable of interest. In order to assess the association between brain/serum glucose ratio and metabolic distress we considered the latter as a dichotomized outcome variable for every hour of measurement, brain/serum glucose ratio as the main predictor variable and other potential explanatory variables as covariates. Brain/serum glucose ratio was entered in the model as a binary variable dichotomized at the median.

In an exploratory analysis, we compared pooled mean LPR, lactate, pyruvate, and glucose levels grouped by quartile of concurrent measures of serum glucose and brain/glucose ratio. Analysis of variance (random effects model) was performed to identify metabolites that were significantly different between all four quartiles, with Bonferroni correction (P < 0.0083 for all comparisons).

In order to identify independent associations between the variables of interest and mortality, we fitted a multivariable logistic regression model with hospital mortality as the binary outcome. Brain/serum glucose ratios were entered as the primary predictor variable and adjusted for other potential predictors of mortality (age, GCS, and APACHE-2 score). Goodness of fit was assessed with the Hosmer–Lemeshow test.

Adjusted odds ratios (OR) and 95 % confidence intervals (CI) were reported for all significant predictor variables. All statistical analyses were performed using SPSS 15 software (SPSS Inc., Chicago, IL, USA). A P value <0.05 was considered statistically significant.

Results

Baseline Characteristics

Patients’ baseline characteristics are listed in Table 1. All 46 patients included in the study were mechanically ventilated and had an admission GCS less than or equal to 10. The most frequent diagnosis was SAH, followed by ICH, TBI, and CA. All results were pooled for analysis. Sixty-three percent (29/46) had an external ventricular catheter placed, and 40 % (17/46) underwent a craniotomy prior to monitoring. Of the 28 SAH patients, 9 were coiled and 11 were clipped. No patients underwent decompressive hemicraniectomy prior to monitoring, and none were treated with pentobarbital. During the study period, 5,187 hourly microdialysate samples and serum glucose measurements were collected (median per patient 99 hourly samples, IQR 66–146). The median duration from admission to the start of neuromonitoring was 2 days and the median duration of monitoring was 5 days.
Table 1

Baseline characteristics (N = 46)

Variable

Median or n

IQR or %

Age, years

55

42–64

Female

27

59

Diagnosis, n (%)

  

 Subarachnoid hemorrhage

28

61

 Intracerebral hemorrhage

10

22

 Traumatic brain Injury

6

13

 Cardiac arrest

2

4

Diabetes mellitus

7

15

APACHE-II score

22

18–27

Admission glasgow coma scale

7

5–9

Days from admission to neuromonitoring

2

1–4

Days with neuromonitoring

5

3–8

Mortality rate

13

28

 Subarachnoid hemorrhage

7/28

25

 Intracerebral hemorrhage

3/10

30

 Traumatic brain Injury

3/6

50

 Cardiac arrest

0/2

0

Data are reported as n (%) or median (interquartile range) unless otherwise indicated

Serum Glucose and Brain Metabolism

Median values for all hourly microdialysis samples, serum glucose measurements and other physiological variables are shown in Table 2. Across all patients median LPR was found to be high, and PbtO2 and MD glucose were in the low range. Forty-one percent of MD glucose values were below 0.7 mmol/L, 26 % of LPR were above 40 and 31 % of PbtO2 measurements were below 20 mm Hg. Three percent of serum glucose values were in the low range, 33 % in the tight range, 51 % in the intermediate range, and 13 % in the high glucose range. Eight percent of hemoglobin values measured were below 8 mg/d, 19 % between 8.1 and 9 mg/dL, 28 % between 9.1 and 10 mg/dL, and 46 % above 10 mg/dL.
Table 2

Multimodality monitoring

Variable

Median

IQR

Normal values

Cerebral perfusion pressure (mmHg)

91

79–104

>70

Serum glucose (mmol/L)

7.4

6.3–8.3

4.4–6.6

Microdialysis

   

 Lactate (mmol/L)

3.9

2.9–4.8

<3

 Pyruvate (μmol/L)

123

92–160

>160

 Glucose (mmol/L)

0.8

0.4–1.3

>2

 LPR

30

26–42

<20

Brain/serum glucose ratio

0.12

0.06–0.18

>0.4

PbtO2 (mmHg)

22

18–35

>25

Normal values are from Ref. [47]

LPR lactate/pyruvate ratio, PbtO2 partial pressure of brain tissue oxygen

Brain/Serum Glucose Ratio and Metabolic Distress

Brain/serum glucose ratio was divided into quartiles for exploratory purposes (Fig. 1) and analyzed in the multivariable model as a dichotomized variable at the median value of 0.12. The proportion of hourly measurements with metabolic distress was progressively higher with decreasing brain/serum glucose ratios. As shown in Fig. 2, brain tissue hypoglycemia was more tightly linked with reduced brain/serum glucose ratios than with low serum glucose levels, and the stepwise increase in mean LPR that occurred with decreasing brain/serum glucose ratios was associated with a statistically significant reduction in pyruvate (rather than an increase in lactate). In addition to low CPP (<70 mmHg) and GCS score, having a brain/serum glucose ratio below the median of 0.12 was independently associated with an increased risk of metabolic distress (Table 3).
https://static-content.springer.com/image/art%3A10.1007%2Fs12028-013-9919-x/MediaObjects/12028_2013_9919_Fig1_HTML.gif
Fig. 1

Relative frequency of metabolic distress (LPR >40) across the quartiles of brain/serum glucose ratios. The multivariable general linear model (GLM) with a logistic link function using GEE showed a independent association between brain/serum glucose ratio and metabolic distress

https://static-content.springer.com/image/art%3A10.1007%2Fs12028-013-9919-x/MediaObjects/12028_2013_9919_Fig2_HTML.gif
Fig. 2

Mean (±standard deviation) brain interstitial LPR, lactate, pyruvate, and glucose values obtained by microdialysis (MD), grouped by quartile of concurrent serum glucose levels (left) and brain/serum glucose ratios (right). Group 1 indicates the lowest quartile and 4 the highest. Brain glucose levels were more closely related to brain/serum glucose ratios than to absolute serum glucose levels, implicating impaired glucose transport as a potentially important cause of brain hypoglycemia. *Denotes metabolites that were significantly different between all four quartiles, according to a random effects ANOVA with Bonferroni correction (P < 0.0083 for all pairwise comparisons)

Table 3

Predictors of metabolic distress (LPR >40)

Variable

Threshold

Adjusted OR

CI

P value

Brain/serum glucose ratio

<0.12

1.4

1.2–1.7

<0.001

Serum glucose

Every 1 mmol/L

1.02

0.99–1.06

0.1

Cerebral perfusion pressure

<70 mmHg

1.2

1.0–1.4

0.05

Glasgow coma scale

Every 1 point

0.7

0.6–0.8

<0.001

Hemoglobin

<9 g/dL

1.16

0.9–1.5

0.26

Multivariable logistic regression model accounting for between-subject and within-subject variations over time using generalized estimating equations (GEE) adjusted for the variables listed. Brain/serum glucose ratio was treated as a binary variable dichotomized at the median (0.12). Serum glucose was entered in the model as a negative continuous variable to assess the effect of serum glucose reductions in the outcome. CPP was treated as a dichotomous variable with a cutoff of 70 mmHg. Hemoglobin was treated as a dichotomous variable with a cutoff of 9 g/dL

OR odds ratio, CI confidence interval

Brain/serum Glucose Ratio and Outcome

Patients with an average brain/serum glucose ratio below the median (0.12) showed increased hospital mortality as compared to those with higher average ratios (Fig. 3). In a univariate analysis, mortality was also associated with increased age, higher APACHE-II scores, and lower GCS scores. After adjustment for these variables and mean serum glucose levels in a multivariable logistic regression model, mortality was independently associated with low GCS score and a reduced brain/serum glucose ratio (Table 4).
https://static-content.springer.com/image/art%3A10.1007%2Fs12028-013-9919-x/MediaObjects/12028_2013_9919_Fig3_HTML.gif
Fig. 3

Hospital mortality of patients with daily brain/serum glucose ratio below and above the median value of 0.12. Multivariable logistic regression demonstrated that brain/serum glucose ratio was independent associated with hospital mortality

Table 4

Predictors of hospital mortality

Variable

Threshold

Adjusted OR

CI

P value

Brain/serum glucose ratio

<0.12

6.7

1.2–38.9

0.032

Serum glucose

Every 1 mmol/L

2.4

0.9–6.3

0.084

Age

Every 1 year

1.00

0.99–1.10

0.12

GCS score

Every 1 point

0.61

0.39–0.95

0.029

APACHE-II score

Every 1 point

0.95

0.84–1.10

0.46

Multivariable logistic regression model with hospital mortality as the binary outcome adjusted for the variables listed. The OR, CI, and p value reported for brain/serum glucose ratio are adjusted for the variables listed, which were the predictor variables for the best fit final model

OR odds ratio, CI confidence interval, NA not applicable

Discussion

In this study, we demonstrated an association between reduced brain/serum glucose ratio with cerebral metabolic distress and mortality after severe brain injury.

We studied a heterogeneous population of neurocritical care patients who received continuous insulin infusion targeted to maintain tight glycemic control (4.4–6.7 mmol/L). Although specific disease processes may differ, all patients included were comatose and at risk for cerebral energy failure and secondary brain injury [19, 29, 32, 33, 38]. Moreover, the hypothesized mechanism through which a low brain/serum glucose ratio might contribute to secondary injury—reduced substrate availability and oxidative stress—should be common to all patients after acute severe brain injury.

In our study, we used metabolic distress to evaluate energy failure. Metabolic distress, defined as an elevated LPR above 40, has been reported in the absence of ischemia, and may also possibly be caused by mitochondrial dysfunction, seizures or reduced substrate availability [19, 22, 27, 29, 30]. LPR elevation is a well-studied phenomenon that is correlated with poor outcome [20, 23].

The ratio between brain interstitial and serum glucose levels presumably reflects glucose transport across the blood brain barrier, which is dependent mainly on GLUT1-type transporter proteins in the brain tissue. More specifically the high-density 55-kDa GLUT1 transporter is thought to be responsible for the passage of glucose through the endothelial cells of the blood–brain barrier; the 45-kDa GLUT1 transporter is probably responsible for the transport of glucose into glial cells; and the GLUT3 type mediates the uptake of glucose into neurons [3941]. Ineffective upregulation or dysfunction of GLUT1 transporters can potentially limit glucose utilization by the neuron and astrocyte. In a high-energy demand environment, this may lead to energy failure. A low brain/serum glucose ratio may thus reflect inadequate glucose availability relative to cerebral metabolic demand.

Previous studies have consistently shown that low levels of serum glucose (<6 mmol/dL) are associated with reductions in interstitial brain glucose levels and metabolic crisis [1315, 42, 43]. In TBI patients, impaired oxidative metabolism as evidenced by a LPR >25 has been associated with loss of the normal parallel increase in brain and serum glucose that occurs with glycemic surges, and with lower brain glucose levels in general [22]. On the other hand, there is conflicting evidence associating serum hyperglycemia with elevated brain LPRss after severe TBI [44, 45]. In our study, reduced brain/serum glucose ratios were more strongly associated with metabolic distress (isolated LPR ≥40) than were reductions in brain or serum glucose alone. These findings suggest that impaired glucose transport across the BBB may play a key role in the pathogenesis of disturbed energy metabolism after severe brain injury.

In addition to impaired glucose transport, reductions in brain/serum glucose ratio may also result from ischemic cellular damage, mitochondrial failure, or hyperglycolysis. These processes can contribute to LPR elevation. Low CBF causes reduced oxygen and glucose delivery leading to reduced brain glucose and increased anaerobic metabolism and overproduction of lactate [28, 31, 45]. This phenomenon is also known as a type 1 LPR elevation. Low glucose availability from impaired glucose transport with preserved oxidative metabolism primarily causes decreased production of pyruvate without high lactate production, leading to a type 2 increase in LPR. Hyperglycolysis reduces interstitial glucose levels by overconsumption of substrate, and presumably might result in reduced brain/serum glucose ratios and a parallel increase in lactate and pyruvate, without affecting the LPR [21, 46]. This notion is supported by studies comparing cerebral metabolism measured by microdialysis with FDG-PET-derived glucose uptake that have failed to show a consistent relationship between glucose utilization and LPR or MD glucose levels [19].

In our study, brain tissue hypoglycemia was more closely related to reduced brain/serum glucose ratios than low serum glucose levels per se (Fig. 2), implicating impaired glucose transport as a potentially important causative factor. In addition, the stepwise increase in LPR that occurred with lower brain/serum glucose rations was associated with a statistically stepwise reduction in pyruvate, but not a significant stepwise increase in lactate. In sum, the data indicate that type 2 LPR elevation, related to reduced glucose substrate availability, is the main driver of metabolic distress in the setting of low brain/serum glucose ratios. It is important to note that our patients were treated with a tighter serum glucose target range (4.4–6.7 mmol/L) than is currently recommended. We also found a trend toward higher LPR levels in the highest quartile of serum glucose (Fig. 2), confirming the findings of larger studies of TBI patients [42, 43], but this was not statistically significant.

We found that mortality was independently associated with low GCS score and a reduced brain/serum glucose ratio (<0.12). Presumably, the mechanism by which a low brain/serum glucose ratio contributes to secondary injury is brain tissue hypoglycemia and inadequate energy to support neuronal metabolism. It may be possible, however, that impaired intracellular glucose transport is a nonspecific marker of some other form of cellular dysfunction. Larger studies within specific diagnostic groups are needed to evaluate the relative importance of MD glucose levels versus brain serum glucose ratios.

We used robust statistical methodology to analyze non-normally distributed MD data with repeated measurements within individual patients [36]. Univariate and multivariable analyses were undertaken accounting for between-subject and within-subject variation by using GEE and GLM models, which used an autoregressive of the first order as the correlation structure. This structure assumes that the closer the observations, the larger the magnitude of the correlation. The processes that affect cerebral metabolism after brain injury are multiple and complex, thus we believe such statistical approach and multivariable analyses are essential for the correct interpretation of the data.

Our study has a number of important limitations and caveats. First, because of its retrospective observational design, patients were monitored at different time periods after injury, and for varying amounts of time. Thus, the degree of metabolic crisis that we observed may have been affected by variations in the timing and duration of monitoring. Further studies are needed to specifically analyze the time course of serum/glucose ratios and their relative importance with regard to other MMM-derived parameters within specific diagnostic groups. Second, we analyzed a heterogeneous cohort with four principle diagnoses, which prevented probe location from being standardized, and raises the possibility that our findings may have been primarily driven by processes that are specific to one or more diagnostic groups. Third, glucose, lactate and pyruvate are both produced and consumed in multiple biochemical pathways. This limits straightforward interpretation of their concentration, especially as microdialysis only measures the extracellular pool. Moreover, we could not assess mitochondrial dysfunction, which can cause abnormal oxidative metabolism in the presence of adequate oxygen and substrate delivery, or the potential impact of variations in cerebral glucose metabolism. Fourth, we did not evaluate the use of insulin or status as a diabetic patient and its impact in brain/serum glucose ratio. Fifth, our patients were treated prior to publication of the NICE-SUGAR study results in 2009, and thus were exposed to tighter glycemic control than is generally practiced in ICU’s today. In all likelihood, this practice heightened exposure of our patients to critical brain hypoglycemia. Finally, although we found an association with hospital mortality, we did not prospectively evaluate long-term outcomes, and we cannot rule out the possibility the MMM data may have influenced decisions to withdraw care.

Our findings are hypothesis generating but may have important clinical implications. Monitoring brain/serum glucose ratio closely in combination with regional cerebral blood flow and tissue oxygenation may provide prognostic information, and allow an even more individualized approach to insulin therapy and serum glucose management. Further studies are needed to confirm our findings within specific diagnostic groups, and to determine the potential benefits of taking brain/serum glucose ratios into account when defining optimal serum glucose targets and implementing glucose control protocols. As MMM becomes increasingly integrated into clinical practice, randomized clinical trials will be needed to assess the effect of goal-directed interventions aimed at improving cerebral metabolic profiles on long-term outcomes of patients with severe brain injury. Until that point, additional studies are needed to better understand how to incorporate monitoring of brain/serum glucose levels into clinical practice.

Acknowledgments

This work was supported in part by a grant from the Charles A. Dana Foundation. The project described was also supported by Grant UL1 RR024156 from the National Center for Research Resources (NCRR), a component of the National Institutes of Health (NIH) and NIH Roadmap for Medical Research, and its contents are solely the responsibility of the authors and do not necessarily represent the official view of NCRR or NIH. The authors were solely responsible for the study design, data collection, analysis and interpretation of the data, writing the manuscript, and in the decision to submit for publication.

Conflict of interest

The authors have no conflicts of interest to declare.

Copyright information

© Springer Science+Business Media New York 2013