The Effects of Intracranial Pressure Monitoring in Patients with Traumatic Brain Injury

Background Although international guideline recommended routine intracranial pressure (ICP) monitoring for patients with severe traumatic brain injury(TBI), there were conflicting outcomes attributable to ICP monitoring according to the published studies. Hence, we conducted a meta-analysis to evaluate the efficacy and safety of ICP monitoring in patients with TBI. Methods Based on previous reviews, PubMed and two Chinese databases (Wangfang and VIP) were further searched to identify eligible studies. The primary outcome was mortality. Secondary outcomes included unfavourable outcome, adverse events, length of ICU stay and length of hospital stay. Weighted mean difference (WMD), odds ratio (OR) and 95% confidence intervals (CIs) were calculated and pooled using fixed-effects or random-effects model. Results two randomized controlled trials (RCTs) and seven cohort studies involving 11,038 patients met the inclusion criteria. ICP monitoring was not associated with a significant reduction in mortality (OR, 1.16; 95% CI, 0.87–1.54), with substantial heterogeneity (I2 = 80%, P<0.00001), which was verified by the sensitivity analyses. No significant difference was found in the occurrence of unfavourable outcome (OR, 1.40; 95% CI, 0.99–1.98; I2 = 4%, P = 0.35) and advese events (OR, 1.04; 95% CI, 0.64–1.70; I2 = 78%, P = 0.03). However, we should be cautious to the result of adverse events because of the substantial heterogeneity in the comparison. Furthermore, longer ICU and hospital stay were the consistent tendency according to the pooled studies. Conclusions No benefit was found in patients with TBI who underwent ICP monitoring. Considering substantial clinical heterogeneity, further large sample size RCTs are needed to confirm the current findings.


Introduction
Traumatic brain injury (TBI) is the leading cause of death and disability after serious injury, an average of 235,000 hospitalizations and 50,000 deaths occurring each year in United States [1]. The damage in patients with TBI is not just due to direct consequences of the primary injury. Subsequently, traumatic space occupying lesions and cerebral edema accompanied by raised intracranial pressure (ICP) may lead to the hypoxic -ischaemic damage, which might result in herniation of brain tissue, inadequate cerebral perfusion, ischemia and death [2,3]. Theoretically, the management of patients with TBI would benefit from ICP monitoring [4]. The guideline from Brain Trauma Foundation (BTF) recommended ICP monitoring for patients with severe TBI (Glasgow Coma Scale (GCS) score #8 ) and an abnormal brain computerized tomography (CT) scan. Furthermore, ICP monitoring was also recommended for patients with severe TBI without CT abnormalities but with at least two of the following criteria: age .40 years, motor posturing, or systolic blood pressure , 90 mm Hg [5]. Lane et al. [6], Stocchetti et al. [7] and Mauritz et al. [8,9] confirmed the benefit of ICP monitoring. Conversely, Shafi et al. [10] and Griesdale et al. [11] reported ICP monitoring was associated with increased mortality. Biersteker et al. [12] and Thompson et al. [13] presented that ICP monitoring was not associated with mortality and unfavorable outcome, which was consistent with Cremer and colleagues [14]. Based on the published two randomized controlled trials (RCTs) [15,16], no significant difference was observed in the survival rate between ICP monitoring group and no ICP monitoring group. Up to date, the efficacy and safety of ICP monitoring following TBI still remains controversial.
Owning to the sample size (324 and 61 patients respectively) included in the two RCTs, the evidences from RCTs were not enough for the definite conclusion. Given no results from registered cochrane database systematic review [17], in our opinion, it would be interesting for us to conduct the first metaanalysis with respect to the efficacy and safety of ICP monitoring in the patients with TBI, which might be a beneficial complement to the present results from RCTs.

Search Strategy and Inclusion Criteria
Based on the previous registered cochrane database systematic review [17] and Mendelson et al. [18], two authors (S.-H.S and F. Y) further searched PubMed and two Chinese databases (Wangfang and VIP) for the relevant articles published up to March, 2013. Research works were examined with language restricted to English and Chinese, and were identified by using the following keywords: ''intracranial pressure monitoring'' or ''intracranial pressure monitor*'', and ''random'' or ''random*'' or ''case control'' or ''cohort'' or ''observational''. The references of all publications and reviews were then reviewed and re-searched to prevent missing any relevant publications.
The following inclusion criteria in PICOS order included: (i) population: patients with diagnosed TBI; (ii) intervention: ICP monitoring; (iii) comparisons: ICP monitoring group versus no ICP monitoring group (imaging or clinical examination); (iv) outcome measures: mortality, unfavourable outcome, length of ICU stay, length of hospital stay and adverse events, one of which should be mentioned in the studies; (v) study design: RCT, case control study and cohort study.

Data Extraction and Outcome Measures
Two authors (S.-H.S and Y.-F.W) independently screened studies. For each study, we recorded the first author, year of publication, the sample size of population, patients characteristics, patients selection criteria, definitions of outcomes, etc. Any disagreements were resolved by discussion and consensus. A third investigator (F.W) was consulted in case of disagreement to improve accuracy. The analytical data missing from the primary reports were requested from their authors. When the same population was reported in several publications, we retained only the most informative article or complete study to avoid duplication of information.
The primary outcome was mortality. Secondary outcomes included unfavourable outcome, adverse events, length of ICU stay and length of hospital stay.

Quality Assessment
Cochrane risk of bias assessment [19], which consists of seven items including random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting and other bias, was used to evaluate the methodologic quality of RCTs. Newcastle-Ottawa quality assessment scale (NOS) [20], which includes three questions in selection, one question in comparability and three questions in outcome, was applied to assess the methodologic quality of cohort studies. Two authors (J. H and Y.-H. Z) subjectively reviewed all studies and assigned a value of low risk, high risk and unclear risk to the RCTs, and awarding  points for cohort studies (points were then added up and used to compare quality of each study).

Statistical Analysis
Meta-analysis was carried out by using Cochrane RevMan (version 5.1) software. Continuous data presented as median and interquartile range were transformed to the data with mean 6 standard deviation (SD) [21]. For continuous and dichotomous outcomes, differences were calculated using weighted mean difference (WMD) or odds ratio (OR), 95% Confidence Interval (CI) respectively. Heterogeneity for each pooled summary was estimated using Cochran's Q statistic and the I 2 statistic. Substantial heterogeneity will be considered to exist with I 2 . 50% and Chi square test P,0.1. Fixed-effects model was used if the number of studies included in the meta-analysis was less than 5, while random-effects model were used if the number of studies included in the meta-analysis was more than 5. Because patients characteristics, clinical center, types of ICP monitoring used, definitions of outcomes, and other confounding factors were not consistent among studies, we further conducted sensitivity analyses to verify the results or explore possible explanations for heterogeneity or examine the influence of various inconsistent criteria on the overall pooled estimate. We also investigated the influence of a single study on the overall pooled estimate by omitting one study in each turn. If the same directional tendency of outcome was found among studies, meta-analysis would not be applied. Potential publication bias was assessed visually with funnel plot.

Study Identification and Selection
The combined search strategy identified 139 papers (92 in English, 47 in Chinese). After careful screening, two RCTs and seven cohort studies satisfied all the inclusion criteria. An additional cohort study was identified by hand searching. One article was excluded for no available data. Thus, eventually nine studies were included in the present meta-analysis. We only received the missing analytical data for meta-analysis from one correspondence author of the included studies [9]. The selection process for studies included in the meta-analysis is shown in Figure 1.
The quality of the included RCTs was assessed by Cochrane risk of bias assessment. If no specific descriptions were found in studies, we tended to choose the answer of unclear risk (Table 2). The quality of the included cohort studies was evaluated by NOS ( Table 1). The results only reflected our views.

Primary Outcome
Mortality was observed in eight studies [6,[8][9][10][11][12]15,16], which occurred in 944/2925 (32%) patients with ICP monitoring and 1862/7258 (26%) patients with no ICP monitoring respectively. Six-months mortality was shown in two RCTs and one cohort study [12,15,16] and hospital mortality was used in three cohort studies [8,9,11], while no specific time of mortality evaluation was found in two cohort studies [6,10]. ICP monitoring was not associated with a significant reduction in mortality (OR, 1.16; 95% CI, 0.87-1.54) (Figure 2). However, there was evidence of substantial heterogeneity (I 2 = 80%, P,0.00001). Further exclusion of any single study was used to verify the result, which did not materially alter the overall combined OR, with a range from 1.05 (95% CI, 0.81-1.37) to 1.27 (95% CI, 0.96-1.68). Moreover, the sensitivity analyses were also performed to examine the influence of various criteria on the combined estimates, which also showed that our result was reliable (Table 3).
Length of ICU stay was observed in five studies [6,8,11,12,16]. The same directional tendency was found in all the studies that the days of ICU stay were longer in ICP monitoring patients.
Length of hospital stay was presented in four studies [6,10,12,16]. Due to no data for comparison in one RCT [16] and data only presented as mean in one cohort study [6], hence, two cohort studies included in the final meta-analysis. ICP monitoring had significant impact on length of hospital stay (WMD, 6.32 days; 95% CI, 4.90-7.75), with substantial heterogeneity (I 2 = 99%, P,0.00001).

Outcomes from RCTs or Cohort Studies
RCTs and cohort studies are two different types of studies, which may enhance the methodological heterogeneity if they were used together in the meta-analysis. Thus, we further conducted the meta-analysis using RCTs or cohort studies respectively. Outcomes from RCTs or cohort studies are shown in Table 4. According to meta-analysis using cohort studies, the incidence of unfavourable outcome, adverse events and longer hospital stay were significant higher in patients with ICP monitoring, while mortality were not associated with ICP monitoring. Based on the meta-analysis using RCTs, no difference was found for mortality, unfavourable outcomes and adverse events between patients with ICP monitoring and patients without ICP monitoring.

Publication Bias
No obvious evidence of publication bias was found from funnel plots ( Figure 4).

Discussion
ICP monitoring allows early detection of pressure changes and can guide treatment of elevated ICP [22,23], which has been recommended by international guideline in the treatment of severe TBI [5,24,25]. Nevertheless, owning to the definitions of severe TBI, the types of ICP monitor used, and the levels of intervention, etc, there were conflicting outcomes attributable to ICP monitoring in published studies. The effects of ICP monitoring still remain controversial. In our study, two RCTs and seven obersevational  cohort studies with available crude data were firstly pooled to evaluate the efficacy and safety of ICP monitoring in adult patients with TBI. Restricting meta-analysis only to RCTs, which could ensure that confounders are balanced between different treatment groups, would be more accurate to speculate the effects of treatment. However, case control studies or cohort studies are also used for meta-anlysis in recent years. Heterogeneity, which consists of clinical heterogeneity, methodological heterogeneity and statistical heterogeneity, can not be actually eliminated in the process of meta-analysis. If substantial heterogeneity is found in the meta-analysis, sensitivity analysis or stratified analysis could be used to verify the results reliable or find the probable explanations of heterogeneity. Hence, it may be a deserved choice to conduct this meta-analysis to investigate the effects of ICP monitoring in patients with TBI under the current studies.
In our study, we found ICP monitoring did not significantly decrease mortality. Due to the inconsistent baseline of patients characteristics and various clinical interventions, substantial heterogeneity was presented in the analysis. Nevertheless, exclusion of any single study did not materially alter the pooled results. In addition, sensitivity analyses based on different categories of included studies were used to verify the pooled results, suggestive of reliable result. Moreover, the subgroup meta-analysis using RCTs or cohort studies also showed that ICP monitoring was not associated with mortality. With respects to unfavourable outcome and advese events, we only chose the data with consistent inclusion criteria for meta-analysis to reduce the clinical heterogeneity. No significant difference was found in the occurrence of unfavourable outcome and advese events. However, We should be cautious to the result of adverse events because of the substantial heterogeneity in the comparison. Meta-analysis using RCTs confirmed the above results, whereas the meta-analysis with only cohort studies found ICP monitoring was related to the higher incidence of unfavourable outcome and advese events. More aggressive interventions (osmotherapy, hypothermia, cerebrospinal fluid [CSF] drainage, hyperventilation, craniotomy, etc) were found in the patients with ICP monitoring in the cohort studies [9,11,12], in which two studies [9,12] were included in the metaanalysis of unfavourable outcome using only cohort studies. Hence, more aggressive interventions might be responsible for the unfavourable outcome following TBI. Huge difference in the number of patients (1646 patient in Shafi 2008, 265 patients in Biersteker 2012) exactly existed, which may be the reason of substantial heterogeneity in the meta-analysis of the length of hospital stay. Although no futher meta-analysis could be conducted because of the missing data from the included studies, we could speculate longer days in hospital for patients underwent ICP monitoring through these incomplete data.
ICP monitoring is only the first step in ICP/cerebral perfusion pressure (CPP) -based therapy, subsequent therapeutic strategies including efficient interventions (analgesia, sedation, barbiturates, steroids, mannitol, hypertonic saline, hyperventilation, hypothermia, CSF drainage, etc), mechanical ventilation strategies (peak inspiratory pressure, positive end-expiratory pressure, and pO 2 / FiO 2 ratio), neurosurgical procedures (intra-and extracranial surgery), and ICP treatment thresholds also played important roles in the management of TBI [9]. Different cut-off point of ICP (18 mmHg or 20 mmHg) oriented therapy, different types of ICP monitor used (intraventricular, intraparenchymal or non-invasive) and different therapeutic strategies following ICP monitoring might result in different outcomes, which did not achieve consensus at present. Nevertheless, the articles comparing the above aspects were scarce. Furthermore, we found adverse events (such as infection, nervous system events, cardiovascular system events,etc) seldom mentioned in the published studies, which could be the important risk factor of mortality and poor prognosis in TBI patients who underwent ICP monitoring. Apparently, the need for such further studies should be stressed.
One potential limitation of the present meta-analysis is the various diagnostic or inclusion criteria for ICP monitoring and different levels of interventions used among each studies. With special respect to mortality, the data without scaling the mortality into the same time interval were pooled together for the metaanalysis. Although sensitivity analyses and further exclusion of any single study were used to verify that our result was reliable, we should be very cautious to treat this result. Another limitation is that RCTs and cohort studies were used together in this metaanalysis, which could enlarge potential methodological heterogeneity. The clinical and methodological heterogeneity in the discussed studies may be resposible for the lack of clear evidence to support our results. Finally, missing data in these studies might influence the overall results and should be taken into account. Therefore, our current data need to be substantiated by adequate prospective studies.
In summary, our meta-analysis suggested that no benefit was found in patients with TBI who underwent ICP monitoring. Considering substantial clinical heterogeneity, further large sample size RCTs are needed to confirm our current findings. Hopefully, clinicians may be able to elicit indications and benefit from ICP monitoring by refining and optimizing the use of ICP monitors in the future.