Introduction

The relationship between secondary brain injury and poor outcome in adults and children following traumatic brain injury (TBI) is well known [1–4]. Therefore, preventing, limiting, or managing secondary injuries like ischemia is crucial to improving outcome following TBI [5, 6]. Ischemia or hypoxia following TBI is significantly associated with an unfavourable outcome [7, 8]. Early identification and intervention is therefore critical to preventing secondary injury, and therefore [5, 9], methods of improving outcome are largely focused on preventing ischemic injury [10].

Prevention of ischemia following TBI traditionally involves methods aimed at improving cerebral perfusion pressure (CPP) and controlling intracranial pressure (ICP). Current TBI management focuses on maintaining ICP and CPP within recommended thresholds. However, the use/extrapolation of these thresholds in pediatric TBI is based on weak evidence [11]; maintaining ICP and CPP within these thresholds is not a strong indicator of acceptable levels of brain oxygenation [5, 12–16]. Up to one third of children with severe TBI may experience episodes of low brain oxygenation tension, even when recommended treatment targets for ICP and CPP are achieved [5].

In children, determining what represents adequate ICP and CPP control is more complex than that in adults, given that children of different ages have different and less well-established normative thresholds for ICP and blood pressure [5]. To date, there are no age-based recommendations for these thresholds, and so treatment and injury thresholds are extrapolated from adult studies [11, 17]. This strategy is suboptimal given that children are physiologically different from adults. Therefore, a measure of the adequacy of brain oxygenation, rather than using ICP or CPP values as a proxy measures is preferable [5].

Brain tissue oxygen tension (PbtO2) monitors have therefore been proposed as a complementary tool to ICP monitoring to detect the adequacy of brain perfusion and oxygenation. These monitors are used in both adults and children at risk of cerebral ischemia and so are being utilized increasingly in the management of patients with severe TBI [5, 9, 18–20]. The aim of PbtO2 monitoring is, ideally, to maintain PbtO2 values greater than 15–20 mmHg. PbtO2 values less than 20 mmHg suggest progressively increased tissue hypoxia or ischemia, and values less than 10 mmHg are deemed critical, as this appears to approximate an ischemic threshold [18].

Low PbtO2 post-TBI is common and is associated with increased mortality and morbidity after severe TBI in adults [9, 18]. PbtO2-directed treatment appears to be associated with reduced mortality in adults [20, 21]. Although fewer studies of this kind have been conducted with children [4, 5, 12, 15], emerging data are consistent with observations made in adult studies. In a study that included a large pTBI sample (N = 52), low PbtO2 was independently associated with poor outcome (as defined by the Glasgow Outcome Score [GOS] and Pediatric Cerebral Performance Category Scale [PCPCS]) and was a stronger predictor than other factors traditionally associated with outcome [13]. Furthermore, low PbtO2 was not predicted by measures of initial injury severity, suggesting that the contribution of low PbtO2 to poor outcome represents secondary brain injury that is, at least in theory, amenable to treatment.

There is some limited evidence from the adult literature that low PbtO2 is associated with poor performance in the domains of general intellectual functioning and memory [22], but to date, there have been no published studies on the relationship between PbtO2 and specific neuropsychological and behavioural outcomes in children. In fact, TBI-related neuropsychological outcome studies rarely consider neurosurgical monitoring variables such as PbtO2 levels, and neurosurgical outcome studies rarely include neuropsychological outcome variables. In this study, we aimed to investigate the relationship between PbtO2 levels and neuropsychological and behavioural functioning following severe pediatric TBI (pTBI). Specifically, we investigated whether PbtO2 levels that are maintained above the ischemic threshold (PbtO2 > 10 mmHg) are associated with more favourable outcomes for children who have sustained severe TBIs. We hypothesized that TBI patients who experienced at least one episode of brain hypoxia as measured by PbtO2 < 10 mmHg would perform more poorly on the administered tests than those who did not experience an episode of brain hypoxia.

Materials and methods

Research design

This was a case-control study. The study design was quantitative, retrospective and cross-sectional. It included two between-group comparisons. The first comparison was between a group of children who had sustained severe TBIs and who underwent PbtO2 monitoring and a healthy matched control group. The second between-group comparison involved dividing the pTBI group into two subgroups, one including those who had experienced at least one episode of PbtO2 lower than 10 mmHg (i.e. they reached the ischemic threshold) during monitoring (hypoxia group) and the other including those for whom PbtO2 had remained above 10 mmHg throughout the monitoring period (no-hypoxia group).

Participants

The patient sample included 11 children admitted to Red Cross War Memorial Children’s Hospital (RCCH), in Cape Town, South Africa, for severe TBI with an admission post-resuscitation Glasgow Coma Scale (GCS) score ≤8, or who deteriorated to this level after admission. All patients underwent PbtO2 monitoring and were managed according to institutional protocol described elsewhere [13].

Only children in whom monitoring was started within the first 24 h were considered for inclusion in the study. Of those for whom complete monitoring data were available, we selected for inclusion only those (a) who were English and/or Afrikaans speaking, (b) who were admitted for closedFootnote 1 severe TBI, (c) who were at least 1 year post-injury,Footnote 2 (d) who were aged 6–16 years at the time of assessment and (e) for whom informed consent and assent were granted.

We recruited a group of healthy controls (n = 11) against which to compare the performance of the TBI participants. Control participants were recruited by (1) identifying potential participants from the schools that the TBI participants attended and (2) consulting with other local researchers who were conducting pediatric neuropsychological studies and who had included healthy participants in their studies. These control participants were matched as closely as possible to TBI participants on a range of demographic variables, including age, sex, language, socioeconomic status (SES) and race. Exclusion criteria for all participants included previous head injuries that resulted in hospitalization or loss of consciousness and any formally diagnosed learning, psychiatric, neurological or developmental disorders.

Measures

The cognitive assessment tools were the Wechsler Abbreviated Scale of Intelligence [26] (WASI), the Rey Complex Figure Test [27] (RCFT), and selected subtests from the Children’s Memory Scale [28] (CMS), the Test of Everyday Attention for Children [29] (TEA-Ch), and the NEPSY-II [30]. Hence, the assessment battery covered a range of cognitive domains, including general intellectual functioning (verbal IQ [VIQ], performance IQ [PIQ], and full-scale IQ [FSIQ]), verbal and visual memory, attention, executive functions, and visuospatial functioning. The behavioural assessment tools were the Child Behaviour Checklist [31] (CBCL; Achenbach, 1991) and the Behaviour Rating Inventory of Executive Functions [32] (BRIEF; Gioia, Isquith, Guy and Kenworthy, 2000).

These tests and questionnaires were all originally published in English. To facilitate administration to those participants with a home language of Afrikaans, the test instructions and relevant stimuli were translated by the University of Stellenbosch Language Laboratory (Cape Town, South Africa) who carried out forward and back translations and authentication.

We also used a questionnaire designed to acquire socio-demographic information about the participants which captured details about parental education, occupation and income, as well as information about the home living environment (e.g. the type of dwelling and participants’ residence and neighbourhood).It also included an index of material resources in the household [33] as well as more traditional measures of SES.

Procedure

TBI participants: physiological monitoring

Following the local TBI management protocol, intracranial catheters for ICP (Codman, Raynham, MA, USA) and PbtO2 (Licox; Integra Neurosciences, Plainsboro, NJ) were inserted into normal appearing white matter in the right frontal lobe or on the side of the greatest cerebral swelling or most significant lesion (as per admission head computed tomography [CT] scan). The accurate positioning of the monitors was confirmed on follow-up CT scan. The treatment threshold for PbtO2 threshold was 20 mmHg [4, 13]. Physiological data were continuously recorded using a high-frequency computerized software recording system (ICMPlus®, Cambridge University, UK).

TBI and healthy control participants: neuropsychological and behavioural testing

Parents/caregivers were contacted via telephone and invited to participate in the study. Each participant was tested individually. The duration of testing was approximately 3 h. Parents completed the sociodemographic questionnaire and the BRIEF and CBCL forms during that time.

Scoring procedures and statistical analyses

Identifying and measuring episodes of low PbtO2

Patient physiological data were examined to identify episodes of low PbtO2 including (a) the lowest PbtO2 reading that persisted for at least 30 min during the entire monitoring period and (b) the cumulative time that PbtO2 was less than thresholds of 20, 10 and 5 mmHg. Data from the first 2 h of PbtO2 catheter stabilization were excluded from analyses. Although PbtO2 was treated at 20 mmHg, this represented a ‘softer’ target for interventions with more aggressive interventions being used when PbtO2 fell below 10 mmHg. Because adult and pediatric data suggest a stronger association with outcome when PbtO2 is below 10 mmHg, the 10–20 mmHg range likely represents a region of oligemia rather than ischemia if cerebral blood flow restriction is the cause of the decreased PbtO2. For this reason, the 10 mmHg threshold was used for analysis in the current study.

Scoring procedures for neuropsychological and behavioural data

For each (sub)test in the battery, we followed the conventional scoring procedures described in the respective test administration manuals. We converted all raw scores to age-adjusted scaled scores.

Statistical procedures

We used SPSS version 22.0 and set the threshold for statistical significance (α) at 0.05. For each analysis, we calculated the appropriate effect size estimate.

TBI cases vs. controls

For demographic data, we used one-way ANOVAs or Mann-Whitney U tests to assess between-group differences on continuous variables, depending on whether or not assumptions of normality and homogeneity were upheld, and chi-square or Fisher’s exact test to assess between-group differences on categorical variables. We used the latter statistical procedure in instances where the sample was small and where the cells of the variables in the analyses had expected counts of less than 5.

For the neuropsychological data, there were a large number of dependent variables (32) in proportion to the sample size (N = 22) for the comparison between TBI and control groups. Therefore, we used a standardized set of procedures to reduce the number of dependent variables. The resulting ten outcome variables included three IQ measures, WASI nerbal IQ, performance IQ and full-scale IQ, and seven composite measures covering the domains of basic and higher-order attention, verbal and visual memory, executive function, visuospatial ability and expressive language. We created these composites using a hybrid method [34, 35].

We then used one-way ANOVAs or Mann-Whitney U tests to investigate between-group differences in neuropsychological test scores and behavioural measures. Despite the fact that we conducted multiple comparisons, we did not apply the Bonferroni (or similar) correction to the results of these analyses. Although one might typically control for the risk of type I error using a conservative measure such as this, in other public health research contexts (e.g. pediatric exposure to neurotoxins), researchers are more concerned about missing important effects (type II errors) than about the strict control of alpha values [36]. This concern might also be extrapolated to TBI research: employing an adjustment to control for type I error may result in an underestimation of the effects of TBI on neuropsychological and behavioural outcomes.

Hypoxia vs. no-hypoxia

We compared outcome on the (a) demographic and injury variables, (b) SES data, (c) physiological variables and (d) neuropsychological and behavioural variables for the hypoxia and no-hypoxia groups. We repeated the steps outlined above in terms of checking assumptions, deriving composites, between-group comparisons of demographic, neuropsychological and behavioural data and non-use of Bonferroni adjustment.

Ethical considerations

Ethical approval for this study was obtained from the University of Cape Town’s Faculty of Health Sciences Human Research Ethics Committee. Permission to include the school learners and to use the school facilities for testing was obtained from the Western Cape Education Department. This study was performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments. The parents of all participants and controls provided informed consent for participation in this study.

Results

TBI cases vs. controls

There were 16 boys and 6 girls in the sample, with a mean age at assessment of 10.87 years (SD = 31.49). Most participants (n = 20) were mixed race, with the rest (n = 2) being White. Most (n = 16) reported they spoke English and Afrikaans equally well; the rest (n = 6) spoke Afrikaans only. The two groups did not differ significantly in terms of age at assessment (p = 0.936) and were evenly matched in terms of sex, race and home language. There were also no significant between-group differences for any of the SES measures.

Table 1 presents the results of the between-group comparisons on measures of IQ and neuropsychological test performance. The control group performed significantly better than the pTBI group on the following: PIQ, FSIQ, basic attention composite, higher-order attention composite, visual memory composite, executive functions composite, visuospatial skills composite and expressive language composite. Moderate-to-large effect sizes were associated with each of these comparisons.

Table 1 IQ variables and neuropsychological composites: between-group comparisons, TBI vs. controls (N = 22)
Full size table

Table 2 presents results from the between-group comparisons of behavioural data. The groups differed significantly on all of the BRIEF indices, with the controls reportedly scoring better than the pTBI participants. Large effect sizes were associated with these comparisons.

Table 2 BRIEF indices and CBCL syndrome profiles: between-group comparisons, TBI vs. controls (N = 22)
Full size table

The groups also differed significantly on the CBCL anxious/depressed, withdrawn depressed, rule-breaking and aggression syndrome scales and on the externalizing behaviour syndrome grouping, with the controls reportedly scoring better than the pTBI participants. Again, moderate-to-large effect sizes were associated with these comparisons.

Hypoxia vs. no-hypoxia

Table 3 shows that there were no significant differences between the hypoxia and the no-hypoxia groups in terms of the demographic or injury related variables (admission GCS, mechanisms of injury). The only significant between-group difference was that the hypoxia group had experienced a longer time since injury than those in the no-hypoxia group. There was a large effect size associated with this finding (r = 0.68).

Table 3 Demographic characteristics and injury variables (N = 11): hypoxia vs. no-hypoxia
Full size table

There were no significant between-group differences on any of the SES measures or for the categorical physiological variables (pupils on admission, initial systemic hypoxia, initial SBP < 90 mmHg, polytrauma, and ICU risk of mortality score ≥0.50 (p = 0.286–1.00).

Table 4 presents the results of between-group comparisons for the continuous physiological variables. Besides the significant difference between the groups in terms of the number of episodes of PbtO2 < 10 and the lowest PbtO2 value reached during the monitoring period, the groups also differed significantly on the following variables: mean ICP > 20, highest ICP, lowest CPP and lowest PaO2. The effect size estimates suggest, however, that PbtO2 accounted for greater between-group variance than the ICP, CPP and PaO2 variables.

Table 4 Descriptive statistics for continuous physiological variables (N = 11): hypoxia vs. no-hypoxia
Full size table

Table 5 presents the results for the between-group comparisons on measures of IQ and neuropsychological test performance. There were significant between-group differences, in favour of the no-hypoxia group, on the following: VIQ, FSIQ, basic attention composite, higher-order attention composite, verbal memory composite, visual memory composite, executive functions composite, visuospatial skills composite and expressive language composite. There were large effect sizes associated with each of these comparisons.

Table 5 Between-group comparisons for general intellectual functioning and neuropsychological c: hypoxia vs. no-hypoxia groups (N = 11)
Full size table

Table 6 shows that the hypoxia and the no-hypoxia groups did not differ significantly on any of the BRIEF indices, on any of the CBCL syndrome scales or on the internalizing and externalizing syndrome groupings.

Table 6 BRIEF indices: between-group comparisons, hypoxia vs. no-hypoxia groups (N = 11)
Full size table

Discussion

Although a growing body of research demonstrates the association between decreased brain oxygenation and increased mortality and morbidity in both adults and children, there is a dearth of investigations of this nature in children. The few pediatric studies of PbtO2 that have been conducted previously have not included neuropsychological outcome measures. The inclusion of both neuropsychological and behavioural outcome measures is therefore novel in this study.

Summary of results

TBI cases vs. controls

There were no significant between-group differences on any of the sociodemographic variables. These factors can impact on neuropsychological test performance [e.g. see 23, 37]. Hence, it is important that the groups were matched as closely as possible.

The groups did, however, differ significantly on measures of IQ, and on composite indices of basic attention, expressive language, visual memory, visuospatial skills, higher-order attention and executive functions; in each case, the TBI participants performed more poorly than their matched controls. These outcomes are consistent with published literature describing expected neuropsychological sequelae following pTBI [38–47].

In terms of the behavioural measures, the groups also differed significantly on all of the BRIEF indices, with the parents of controls reporting better functioning than the parents of pTBI participants. Regarding the CBCL, the groups differed significantly on the externalizing behaviour syndrome grouping. They also differed on both of the syndrome profiles (rule-breaking and aggressive behaviour) included in this grouping, as well as on the anxious/depressed and withdrawn/depressed profiles of the internalizing behaviour syndrome grouping. The pTBI group reportedly showed more problems in these domains than the controls.

These results from the BRIEF are consistent with those from the neuropsychological tests. Executive functions are subserved primarily by the frontal lobes, and in particular, by the prefrontal cortex. The frontal lobes are especially vulnerable to the effects of TBI owing to their anatomical position and the kinds of biomechanical forces acting on the skull in many TBIs [48]. In terms of the CBCL, behavioural and emotional problems (particularly internalizing and externalizing problems) are reported to be the main reason that children who have sustained TBIs are referred to mental health and rehabilitation professionals [49]. Hence, the syndrome profiles on which significant between-group differences were detected are commonly reported behavioural sequelae following pTBI [49–53].

Although the scores of the pTBI group were not significantly different from controls on the internalizing behaviour syndrome grouping, their mean score fell within the clinical range, whereas the mean score for the healthy controls fell within the normal range. Although there is a small effect size associated with this comparison, these results show a trend in the expected direction. Somatic complaints are not necessarily suggestive of emotionally based problems. Therefore, a non-significant finding here does not conflict with reports of emotional and behavioural problems commonly associated with TBI.

In summary, the current data confirmed that participants who have sustained severe TBI perform more poorly than matched healthy controls on neuropsychological and behavioural measures.

Hypoxia vs. no-hypoxia groups

Although participants in the hypoxia group had had a longer time since injury to evaluation, the significant difference in time since injury is not expected to have an effect on the outcome. Even if the overall time since injury differed significantly between the two groups, all of the participants in both groups were at least 1 year post-injury. It has long been reported that improvement in outcome post-TBI is more limited and plateaus after 1-year post injury, particularly in children who have sustained severe TBIs [24, 25, 54].

Between-group differences in neuropsychological outcomes

The hypoxia group performed significantly more poorly on measures of IQ, as well as on the composite measures of basic and higher-order attention, verbal and visual memory, executive functioning, visuospatial ability and expressive language. These findings suggest that secondary injury effects, such as exposure to episodes of hypoxia, create further unfavourable outcomes in children who have sustained severe TBI. Overall, these data are consistent with literature on neuropsychological sequelae following hypoxia-ischemia [23, 55, 56].

Between-group differences in behavioural outcomes

The results show that the experience of one or more hypoxic episodes does not, however, seem to be directly related to outcome on any of the behavioural measures. The hypoxia and no-hypoxia groups did not differ significantly on any of the BRIEF or CBCL outcome variables. Despite this pattern of data, there were moderate effect sizes associated with the MI index of the BRIEF, and the withdrawn depressed, rule-breaking and aggressive behaviour syndrome scales and externalizing behaviour syndrome grouping of the CBCL, suggesting that with a larger sample size, these comparisons could potentially reach significance.

The literature on predictors of cognitive and behavioural outcomes post-TBI suggests a divide in terms of these two areas of outcome. Although cognitive outcomes are strongly determined by injury-related variables, a combination of injury-related factors (e.g. severity) and environmental factors (e.g. family functioning and psychosocial adversity), rather than injury-related factors on their own, is strongly predictive of behavioural outcome [23, 57–60]. Hence, there is a wider range of predictors for behavioural outcomes than for cognitive outcomes. Researchers view behavioural outcomes following TBI as complex and as a product of a range of interacting factors that are not only limited to injury severity but also extend to the family environment and to resources both prior to and after the injury [61–63]. There is a dose-response relationship between the predictors of behaviour and associated outcome, such that more marked and persistent post-injury behavioural difficulties are associated with more severe TBI and poorer family environments [64].

Brain hypoxia-ischemia is a secondary injury-related factor and not an environmental factor. In line with the argument above, it is not surprising that PbtO2 on its own would not predict behavioural functioning, at least not independently of environmental factors (e.g. constraints that our children with TBI face with poor schooling, low parental education levels, poor rehabilitation facilities and special schooling post-injury) or certainly not as strongly as the cognitive outcomes. That being the case, in light of the aforementioned literature on predictors of cognitive and behavioural outcomes, the fact that episodes of PbtO2 < 10 mmHg may be associated more strongly with cognitive rather than behavioural outcomes is consistent with the literature.

An alternative explanation might be that the patients from both groups came from lower SES backgrounds, and the two groups did not differ significantly in terms of sociodemographic factors. Therefore, the same factors that could contribute to poor behavioural outcomes were present in both groups.

Finally, one might also consider how behaviour is typically measured (i.e. via self- or other-report) in this field. In contrast, there are objective measures for cognition. Hence, reports on behaviour might be less accurate than data collected for cognition.

Significant differences on other physiological parameters

Besides significant differences between the groups on the PbtO2 variables (number of episodes when PbtO2 < 10 mmHg and lowest PbtO2 value), the basis on which the groups were formed, the hypoxia and no-hypoxia groups also differed significantly on variables relating to raised ICP, low CPP and lowest PaO2. The hypoxia group experienced higher ICP values and lower CPP and PaO2 values. Hence, perturbations in variables other than PbtO2 might also contribute to the differences found between hypoxia and no-hypoxia groups on measures of neuropsychological test performance.

There was a large effect size associated with the between-group comparison on lowest CPP. This suggests that this physiological parameter accounts for a substantial proportion of the total variance in neuropsychological test performance between the two groups. Very limited literature exists exploring the relationship between CPP and cognitive outcomes. Lannoo et al. [65] did not find a definite association between CPP and ICP in combination and cognitive outcomes. However, Lannoo et al. included both CPP and ICP measures in combination and focused on an older sample (15–65 years). Hence, follow-up studies are required to investigate the specific relationship between CPP and cognition.

Limitations and directions for future research

The small sample size limits the strength of the conclusions that can be drawn and the generalizability of these results. However, effect sizes are large and suggestive of real between-group differences. We will aim to increase the sample size in follow-up studies.

Implementing a three-group comparison (i.e. a pTBI/Hypoxia group, a pTBI/no-hypoxia group and a healthy control group) might have been most ideal for the questions we attempted to answer in this study. The ultimate aim in implementing this design would be to tease apart TBI and hypoxic effects, both independently and combined. One way to do this legitimately, however, would be in a regression model where one could partial out the two effects and look at an interaction effect. This design was not implemented, however, due to the limited sample of eligible participants.

The sample’s broad age range might be interpreted as another limitation, due to the fact that a great degree of neurodevelopment can occur during the years covered by that range. An increased sample size would not only give the study better power but would also allow the detection of developmental trends across more age bands (7–8, 9–10, 11–12 etc.).

A final possible limitation is that the measures of behavioural outcomes included in this study were all self-report measures, and hence, the fidelity of the behavioural results depends solely on reliable reporting by parents. There are obvious limitations to using these self-report measures, including (a) the possibility of social desirability biases and, with that, (b) under- or over-reporting of behaviours, (c) problems related to accessing data on moods and behaviours retrospectively, which can lead to inaccurate reporting, or (d) potential lack of information from respondents on the wide range of behaviours surveyed in the questionnaires, rendering the data incomplete [66–68]. Administering teacher, as opposed to just parent versions, of each of the behavioural measures would have strengthened the power of these results.

Summary and conclusion

The data reported here suggest that reaching a critical PbtO2 threshold of ≤10 mmHg may be detrimental to cognitive outcomes following pTBI. Therefore, over and above the effects of the TBI, which lead to poor neuropsychological and behavioural outcomes, there may also be additional post-TBI hypoxic effects that contribute to even worse cognitive outcomes.

In spite of the outlined limitations associated with this study, the findings presented here and their potential implications warrant attention and further inquiry. This study is an important first step in discerning the prognostic value of low PbtO2 in determining neuropsychological outcomes post-pTBI. However, although the conclusions that may be drawn from these results are noteworthy and could have important implications, they are tentative at this stage, requiring replication in studies with larger samples.