Characterising the dynamics of cerebral metabolic dysfunction following traumatic brain injury: A microdialysis study in 619 patients

Abstract

Traumatic brain injury (TBI) is a major cause of death and disability, particularly amongst young people. Current intensive care management of TBI patients is targeted at maintaining normal brain physiology and preventing secondary injury. Microdialysis is an invasive monitor that permits real-time assessment of derangements in cerebral metabolism and responses to treatment. We examined the prognostic value of microdialysis parameters, and the inter-relationships with other neuromonitoring modalities to identify interventions that improve metabolism. This was an analysis of prospective data in 619 adult TBI patients requiring intensive care treatment and invasive neuromonitoring at a tertiary UK neurosciences unit. Patients had continuous measurement of intracranial pressure (ICP), arterial blood pressure (ABP), brain tissue oxygenation (PbtO2), and cerebral metabolism and were managed according to a standardized therapeutic protocol. Microdialysate was assayed hourly for metabolites including glucose, pyruvate, and lactate. Cerebral perfusion pressure (CPP) and cerebral autoregulation (PRx) were derived from the ICP and ABP. Outcome was assessed with the Glasgow Outcome Score (GOS) at 6 months. Relationships between monitoring variables was examined with generalized additive mixed models (GAMM). Lactate/Pyruvate Ratio (LPR) over the first 3 to 7 days following injury was elevated amongst patients with poor outcome and was an independent predictor of ordinal GOS (p<0.05). Significant non-linear associations were observed between LPR and cerebral glucose, CPP, and PRx (p<0.001 to p<0.05). GAMM models suggested improved cerebral metabolism (i.e. reduced LPR with CPP >70mmHg, PRx <0.1, PbtO2 >18mmHg, and brain glucose >1mM. Deranged cerebral metabolism is an important determinant of patient outcome following TBI. Variations in cerebral perfusion, oxygenation and glucose supply are associated with changes in cerebral LPR and suggest therapeutic interventions to improve cerebral metabolism. Future prospective studies are required to determine the efficacy of these strategies.

Fig 1

Boxplots of individual patients’ mean lactate-pyruvate ratio (LPR) over the first 72h following injury, by (A) Glasgow Outcome Score (GOS) and (B) dichotomised GOS.

Fig 2. Fitted population mean time courses for microdialysis analytes using Generalised Additive Mixed Models (GAMM), by dichotomised glasgow outcome score group.

Lactate (A), Pyruvate (B), Lactate-Pyruvate Ratio (LPR, C), Glucose (D), Glutamate (E), and Glycerol (F). Shaded areas indicated 95% confidence intervals.

Fig 3. Fitted population mean time courses for intracranial pressure derived parameters and brain tissue oxygenation.

Temporal course fitted with generalised additive mixed models (GAMM), by dichotomised Glasgow Outcome Score group. ICP (A), CPP (B), PRx (C), and PbtO₂ (D), Shaded areas indicated 95% confidence intervals.

Fig 4. Relationship between lactate-pyruvate ratio (LPR) and other physiological parameters.

In each panel a GAMM fit of glucose (A), CPP (B), PRx (C), and PbtO₂ (D) against LPR is plotted. On the left axis LPR is scaled to zero mean without the intercept term; on the right axis the scale includes the overall intercept. Shaded region indicates 95%CI for the fit. The regions of each curve highlighted in yellow indicate non-zero first derivative i.e. where the is significant change. Alongside each plot is the distribution of random effects for intercept and slope/coefficient. Rug plot within in each panel indicates density of data.

Fig 5. LPR elevation and associated physiological derangements.

Survival based curves (A) indicating likelihood of LPR >25 and the cumulative incidence of at least one epoch of LPR>25. Euler plot (B) showing prevalence of abnormalities in brain glucose, CPP, PRx, and PbtO₂ (defined by the thresholds described in the text) during episodes of elevated LPR. Alternative representation as ‘Upset’ plot (C) showing differing frequency of permutations of physiological abnormalities (as indicated in the panel beneath the x-axis) associated with elevated LPR.