Anthony C. Wang, MD; Khoi D. Than, MD; Hugh J.L. Garton, MD, MHSc
Department of Neurosurgery, University of Michigan, Ann Arbor, MI
Journal Club Article: Oddo M, Levine J, Mackenzie L, et al. Brain hypoxia is associated with short-term outcome after severe traumatic brain injury independently of intracranial hypertension and low cerebral perfusion pressure. Neurosurgery. 2011;69:1037-1045
Significance/ context and importance of the study
Oddo et al. find that brain hypoxia, determined by regional brain oxygen tension (PbtO2), independently predicts worse short-term outcome after severe traumatic brain injury (TBI). PbtO2 measurement is one of a number of modalities attempting to monitor brain oxygenation, and due to its relevant purported mechanism, safety, and ease of use, is an attractive candidate for study.
Originality of the work
This adds to four published studies that examine treatment of severe TBI in adults based upon PbtO2, as opposed to intracranial pressure (ICP) and cerebral perfusion pressure (CPP), which, as mentioned by Oddo et al., yield conflicting results (citations 37-40). Two of these four studies show an advantage to the use of PbtO2 as a treatment parameter, while 2 do not. This study is methodologically unique in that treatment based on a multi-modal paradigm including PbtO2, ICP, and CPP is not compared against another treatment algorithm.
Appropriateness of the study design
The study by Oddo et al. is a single-institution, retrospective cohort study with a primary aim of analyzing risk factors that predict outcome in patients with severe TBI. The hypothesis, inclusion and exclusion criteria, studied variables, and measured endpoints are clearly defined. The inclusion and exclusion criteria are reasonable, and the measured endpoint well-validated and reproducible. The patient population appears typical for trauma, and we identify no likely biases to affect the risk factors studied. The study applies univariate and multivariate analyses, with modeled risk factors. The authors use linear interpolation to calculate the duration of monitored conditions. With approximately 17 patients per risk factor in multivariate analysis, the number of patients studied is sufficient for the methods used. Outcomes are determined by Glasgow Outcome Score within 30 days of trauma.
This study reports results of multi-modal monitoring including PbtO2, ICP, and CPP, without data from a control group; thus, this study cannot inform treatment due to its design, and the authors make no claims to the contrary. However, the statistical methodology does not support the claim in the discussion section that PbtO2 may be a better prognostic marker than ICP or CPP. The utility of prognostic markers is more properly judged on the basis of sensitivity and specificity, and receiver operating characteristic curve analysis. The data presented in Table 4 do not allow such calculations to be made, since data on PbtO2 are not provided per patient for patients without ICP elevations or CPP depressions.
Adequacy of experimental techniques
Criteria used to define and report TBI in this study differs slightly from that used in much of the existing literature. Rather than the usual requirement of Glasgow Coma Scale (GCS) score ≤ 8 at presentation after trauma, the authors include patients suffering non-penetrating TBI with GCS ≤ 8 at any time during their hospitalization, with radiographic exclusion of alternative causes of coma. As outcome after TBI is strongly correlated with pupillary reactivity and motor score at presentation,3 including patients with initial GCS > 8 who later decompensated introduces a patient population not examined in other studies on this topic. Causes of delayed decline in GCS would be important to know, as toxicity, substance withdrawal, metabolic derangement, and delayed ischemia could explain such a decline. Though the authors mention radiographic exclusion of other causes of coma, they do not specify the modalities used. Specifically, some instances of anoxic brain injury or blunt cerebrovascular injury are impossible to identify without diffusion-weighted magnetic resonance imaging and angiographic studies, respectively.
Hypotension and hypoxia are very strong drivers of head injury outcome.1 According to Figure 2, only 10% of low PbtO2 episodes were coincident with systemic hypoxia. However, up to 25% of episodes were related to systemic hypotension and resultant low CPP. Neither hypoxia nor hypotension was included in the multivariate model. It would be important to ascertain to what extent the success of prediction by PbtO2 is due to these variables. In addition, whether there is any overlap between the Low MAP/Low CPP and High ICP/Low CPP groups in Figure 2 is uncertain—that 75% of brain hypoxia attributed to low CPP is a significantly higher rate than has been previously noted in the literature, though, is not unreasonable.
Treatment was aimed to maintain PbtO2 ≥ 20 mm Hg and ICP < 20 mm Hg. Brain hypoxia was defined as PbtO2 < 15 mm Hg, and elevated ICP defined as ICP > 20 mm Hg for > 2 minutes. Use of CPP as a treatment parameter and as a measure of treatment failure is confusing, however. Though the authors define CPP treatment parameter < 60 mm Hg, their described methods only indicate that CPP was actively “increased” when brain hypoxia occurred independent of, or refractory to, elevated ICP. As high CPP is known to cause pulmonary complications,2 this could have increased the predictive value of low PbtO2 at the expense of the predictive value of low CPP. In addition, in their assessment of causes of secondary brain hypoxia, the authors define low CPP as a decrease of ≥ 10 mm Hg, and do not define a specific threshold. Existing evidence suggests that maintenance of CPP within 50 to 70 mm Hg yields nearly equivalent outcomes, and a threshold CPP of 60 mm Hg could potentially diminish the predictive value of “low CPP”. The authors discuss the apparent variability in CPP threshold that occurs among individual patients, but it is unclear which definition of low CPP was used for which analysis.
The median time to ICP monitoring was 6 hours, which is quite fast; however, cerebral blood flow is known to fluctuate, particularly in the first 12 hours after TBI.4 Any number of causes of variability in time-to-monitoring can occur, affecting the uniformity of the presentation of these patients. Most importantly, how many of these patients initially presented at an outside hospital and were subsequently transferred to the study institution? Because the initial hours after trauma are so importantly representative of long-term outcomes,3 even small variations in time-to-monitoring could exact a large toll on the validity of the patient population studied. Other sources of delay, such as airway management or circulatory resuscitation, could affect brain oxygenation as well, and represent potential influences on outcomes that are not controlled in this study. In effect, though the study team has monitored these patients as well as could be expected, patients could easily have presented for monitoring at very different points in their disease process.
Average duration of PbtO2 monitoring was 5 days, with standard deviation 3 days—whether any patients decompensated after monitor removal, perhaps owing to cerebral edema or vasospasm, is unknown. Only 1-month follow-up is reported, though 6-month functionality is commonly reported in the TBI literature. As the authors explain, only short-term outcomes are assessed in their ICU-based database, which is typically adequate in the assessment of ICU processes. However, this timeframe might represent abnormalities in PbtO2, ICP, and CPP differently, since outcome from low PbtO2 is determined so early in the disease course.
The mean age of patients in this study was 43 years with a standard deviation of 19 years, but whether pediatric patients were included is not specified. A small number of studies have explored the use of PbtO2 monitoring in pediatric patients separately, but it is uncertain as to the generalizability of this study’s conclusions if children were included.
Clarity of writing, strength, and organization of the paper
Oddo et al. have produced a well-organized manuscript describing a logical statistical analysis of a useful dataset. Discussion regarding the role of ICP and CPP monitoring might precede explanation of the need for PbtO2 monitoring, given that a need for PbtO2 monitoring implies inadequacy of ICP and CPP monitoring.
Number and quality of figures, tables, and illustrations
The manuscript provides nine high-quality data summary tables to illustrate their results. However, their methodology is complex, and the data they include is incomplete. Inclusion of individual patient data would be necessary for the reader to better analyze the authors’ interpretation of the events leading to drops in PbtO2.
Economy of words
Wording is concise, and data tables are used effectively to summarize the methods and large amounts of data. The statistical methods utilized are explained succinctly.
Relevance of discussion
In their discussion, the authors present their study’s limitations with exceptional quality. They interpret their results, and give appropriate contextual reference for their work. Discussion regarding ICP and CPP monitoring alone is somewhat irrelevant to this study, because it does not examine such an algorithm, while other studies in the literature do.
Relevance, accuracy, and completeness of bibliography
The authors have identified each of the four important studies examining PbtO2-based treatment algorithms published in the literature to this point. They cite appropriate references to support their factual claims.
Soundness of conclusions and interpretation
Overall, the authors can draw one main conclusion from their data—that PbtO2 is independently associated with short-term outcome after severe TBI. However, the application of their results is an extremely complex topic. The authors regularly infer causation where only association is demonstrated by their methodology and data. As represented in Figure 2, the cause of a drop in PbtO2 was diagnosed and categorized; however, the co-incidence of low PbtO2 with any of these other factors does not necessarily imply causation. For example, the authors implicate hemoglobin < 9 mg/dL to have caused low PbtO2 in 2% of patients; however, evidence that transfusion improves tissue oxygenation is severely lacking. Whether PbtO2 consistently recovered as a direct result of treatment is unknown. As noted by Dr. Maa in his comment, the results presented are unclear as to whether patients with poor short-term outcomes are simply non-responders to PbtO2-based therapy.
In the end, we still do not understand exactly what PbtO2 measures. PbtO2 is a quantification of oxygen diffusion across a membrane covering a Clark-type electrode inserted into brain parenchyma. In theory, it reflects concentration of oxygen in the tissue surrounding the electrode. Many factors affect oxygen transfer from the hemoglobin that carries it to the cytoplasmic mitochondria that use it. Oxygen tension gradient drives diffusion, while perivascular, interstitial, and intracellular edema increase the distance and resistance against which oxygen diffuses. In addition, mitochondrial function is impaired in head injury5 and thus, adequate oxygen delivery to supply metabolic demand provides no guarantee of adequate cellular function.
Optimal PbtO2 monitor placement, timing and duration, and multi-modal treatment algorithm have yet to be elucidated. In addition, the relationship between CPP and PbtO2 is still indeterminate. Whether local PbtO2 monitoring can be generalized to the entire cerebral hemisphere or entire brain, even when cerebral edema is the suspected underlying pathology, remains questionable. It is unproven that PbtO2 monitoring in an unaffected area accurately reflects brain oxygenation of distant regions affected by ischemia (or trauma), as can occur with blunt cerebrovascular injury, mild anoxic brain injury, and vasospasm. It also remains indeterminate whether PbtO2-based treatment algorithms can improve outcomes in severe TBI, compared to ICP- and CPP-based treatment.
In summary, the authors’ primary conclusion that PbtO2 correlates with short-term outcome is sound. However, the clinical utility of their results is unclear, and important questions regarding PbtO2 remain. Statistically, the inter-related variables make this article extremely complex. If the goal of the article is to support the use of PbtO2 monitoring, then the methodology represents only level III evidence, as a retrospective cohort study.
Given the relatively narrow statistical significance reported and the biases inherent in retrospective studies of this nature, the potential for type I error demands caution in interpreting these results. As the study team included members of the Speaker’s Bureau of Integra LifeSciences, manufacturers of the Licox monitor used in this study, the unblinded, retrospective data analysis holds the potential for bias. External validity at multiple centers will also be important to confirm. Without blinding or a control group, this study is also unable to support the assertion that PbtO2-based treatment affects neurological outcome—this will require testing in a randomized fashion before being widely adopted.
1. Butcher I, Maas AI, Lu J, et al. Prognostic value of admission blood pressure in traumatic brain injury: results from the IMPACT study. J Neurotrauma. 2007;24(2):294-302.
2. Contant CF, Valadka AB, Gopinath SP, Hannay HJ, Robertson CS. Adult respiratory distress syndrome: a complication of induced hypertension after severe head injury. J Neurosurg. 2001;95(4):560-568.
3. Steyerberg EW, Mushkudiani N, Perel P, et al. Predicting outcome after traumatic brain injury: development and international validation of prognostic scores based on admission characteristics. PLoS Med. 2008;5(8):e165.
4. Kety SS, Schmidt CF. The effects of altered arterial tensions of carbon dioxide and oxygen on cerebral blood flow and cerebral oxygen consumption of normal young men. J Clin Invest. 1948;27(4):484-492.
5. Verweij BH, Muizelaar JP, Vinas FC, Peterson PL, Xiong Y, Lee CP. Impaired cerebral mitochondrial function after traumatic brain injury in humans. J Neurosurg. 2000;93(5):815-820.