Intracranial Pressure Monitoring in Children with Severe Traumatic Brain Injury National Trauma Data Bank–Based Review of Outcomes

Research

Original Investigation Intracranial Pressure Monitoring in Children With Severe Traumatic Brain Injury National Trauma Data Bank–Based Review of Outcomes

Fuad Alkhoury, MD; Tassos C. Kyriakides, PhD

IMPORTANCE The present study is the largest on the use and effect of intracranial pressure (ICP) monitoring in pediatric trauma patients.

OBJECTIVE To determine the effect of ICP monitoring on survival in pediatric patients with severe head injuries using the National Trauma Data Bank.

DESIGN, SETTING, AND PARTICIPANTS The National Trauma Data Bank was queried (version 6.2, 2001-2006) for information on patients younger than 17 years admitted to an intensive care unit with blunt traumatic brain injury (TBI), Injury Severity Score (ISS) greater than 9, and Glasgow Coma Scale (GCS) score less than 9. Patients with incomplete medical records and those with intensive care unit length of stay of less than 24 hours were excluded from the study.

MAIN OUTCOMES AND MEASURES Parametric comparisons (t tests and χ2 as appropriate) were performed to compare patients who received ICP monitoring with those who did not. Stepwise logistic regression methods were used to assess whether ICP monitoring in the presence of other variables (age, sex, ISS, Revised Trauma Score, and GCS score) was associated with survival.

RESULTS Monitoring of ICP was performed in only 7.7% of patients who met the monitoring criteria recommended by the Brain Trauma Foundation. There were no significant differences in age, sex, or GCS score. After adjustment for admission GCS score, age group, sex, Revised Trauma Score, and injury ISS, ICP monitoring was associated with a reduction in mortality only for patients with a GCS score of 3 (odds ratio, 0.64; 95% CI, 0.43-1.00). Comparison between the 2 groups showed that the ICP monitoring group had a longer hospital length of stay (21.0 days vs 10.4 days; P < .001), longer intensive care unit stay (12.6 vs 6.3 days; P < .001), and more ventilator days (9.2 vs 4.7; P < .001).

CONCLUSIONS AND RELEVANCE Despite current Brain Trauma Foundation guidelines, ICP monitoring is used infrequently in the pediatric population. The data suggest that there is a small, yet statistically significant, survival advantage in patients who have ICP monitors and a GCS score of 3. However, all patients with ICP monitors experienced longer hospital length of stay, longer intensive care unit stay, and more ventilator days compared with those without ICP monitors. A prospective observational study would be helpful to accurately define the population for whom ICP monitoring is advantageous.

Author Affiliations: Department of Pediatric Surgery, Joe DiMaggio Children’s Hospital, Hollywood, Florida. Corresponding Author: Fuad Alkhoury, MD, Department of Pediatric Surgery, Joe DiMaggio Children’s Hospital, 1150 N 35th Ave, JAMA Surg. 2014;149(6):544-548. doi:10.1001/jamasurg.2013.4329 Ste 555, Hollywood, FL 33021 Published online April 30, 2014. ([email protected]).

544 jamasurgery.com

Downloaded From: https://jamanetwork.com/ on 09/28/2021 Intracranial Pressure Monitoring in Children Original Investigation Research

raumatic brain injury (TBI) is an important public health After excluding patients with incomplete records and those problem in the United States. In 2003, there were an es- with ICU lengths of stay less than 24 hours, we performed a T timated 1 565 000 TBIs in the United States: 1 224 000 univariate analysis to compare the 2 groups (n = 3107) using emergency department (ED) visits, 290 000 hospitalizations, unpaired, 2-tailed t tests for continuous variables and χ2 for and 51 000 deaths. Findings were similar to those from pre- categorical variables, as appropriate. Variables that appeared vious years in which rates of TBI were highest for young chil- to show a significant difference between the 2 groups were en- dren (aged 0-4 years) and men.1 However, intracranial hyper- tered as covariates into a logistic regression model. These vari- tension occurs in as many as 70% of patients and is responsible ables included the Injury Severity Score (ISS) (an anatomical for a substantial proportion of TBI-related deaths.2,3 Pub- measure of injury severity),11 the Revised Trauma Score (RTS) lished data4-7 and consensus practice since the late 1970s sug- (a physiological measure of injury severity),12 the GCS score gest that intensive management protocols may reduce the in- determined in the ED, and the probability of survival score cidence of secondary brain injury after severe TBI (Glasgow (Trauma and Injury Severity Score [TRISS] analysis). The ISS Coma Scale [GCS] score <9) and thus improve survival and out- is derived from the highest Abbreviated Injury Scale scores of come. The GCS is the most widely used assessment tool for pre- each of the 3 most severely injured of 7 defined regions of the dicting the severity of TBI. body (range, 1-75; the score increases with the severity of the In 1995, 2000, and 2007, the Brain Trauma Foundation injury). The RTS, derived from the first set of physiological data (BTF) published management guidelines8 that included indi- obtained from the patient, consists of the GCS score, blood pres- cations for intracranial pressure (ICP) monitoring. The goal of sure, and respiratory rate (range, 0-8; the score increases with ICP monitoring is to ensure adequate cerebral perfusion pres- the severity of the injury). The TRISS method uses a combi- sure through the placement of intraventricular catheters or fi- nation of anatomical (ISS) and physiological (RTS) indexes of ber optic monitors into the parenchyma of the brain. No ran- injury severity as well as coefficients derived from the Mul- domized clinical trials to evaluate the effect of the BTF tiple Trauma Outcome Study,11,12an age index, and coeffi- recommendation on the outcome of management of severe TBI cients for blunt and penetrating mechanisms to calculate prob- with or without ICP monitoring have been conducted. ability of survival. The TRISS ranges from 0 to 1, with 0 In 1991, a nationwide survey9 of TBI care in 219 trauma cen- indicating no probability of survival. ters across the United States documented that ICP monitor- Regression modeling was first applied to the entire study ing was used routinely in only 77 centers (35%) and not at all group to assess the relationship between ICP monitoring and in 16 centers. A subsequent national survey10 conducted in mortality after adjusting for other covariates. Subsequently, 2000 documented only a marginal increase in ICP monitor- regression models were completed for each GCS score group. ing, from 35% to 45%, with only 16% of centers in full compli- An α level of .05 was used to determine statistical signifi- ance with BTF recommendations. In light of this informa- cance. Data analysis was conducted using SAS, version 8.2 (SAS tion, we set out to examine the National Trauma Data Bank of Institute Inc). the American College of Surgeons between January 2001 and The primary outcome of the study was mortality. Odds ra- December 2006 to evaluate whether there has been any change tios (95% CIs) of mortality with ICP monitoring were calcu- in practice patterns and the outcomes of using ICP monitor- lated from the regression model variable estimates. Continu- ing in pediatric trauma patients. ous variables are summarized as mean (SD) and categorical variables as proportions; P = .05 was considered significant for all analyses. Methods

The National Trauma Data Bank of the American College of Sur- Results geons collates data from participating trauma centers through- out the United States. Institutional review board approval was Less than 15% of the study patients (318 of 4141 [7.7%]) under- waived for this study. During the study period (2001-2006), the went ICP monitoring despite meeting criteria defined by the National Trauma Data Bank contained information on more BTF.There were no significant differences between the 2 groups than 1 million injured patients, which constituted the study in univariate analysis (n = 3107) in age, sex, or the ED- universe. Inclusion criteria consisted of admission to a desig- assigned GCS score (Table 1). nated level I or II trauma center, blunt mechanism of injury, Monitored patients had a slightly higher ISS, a lower RTS, age younger than 17 years, admission to an intensive care unit, and similar TRISS-calculated probability of survival (Table 1). and an injury that met the following BTF criteria for ICP moni- Crude survival was similar in both groups. Patients in the monitoring: GCS score less than 9 determined in the ED and com- tored group had significantly longer hospital stays, ICU stays, puted tomography demonstrating TBI. ventilator days, and overall increase in hospital charges, with Based on these criteria, 4141 patients were identified for no improvement compared with the unmonitored group initial inclusion in the study. They were divided into 2 groups: (Table 2). those who underwent ICP monitoring (n = 318) and those who Monitoring of ICP was not associated with improvement did not (n = 3823). Intracranial pressure monitoring was iden- in survival compared with the unmonitored group after ad- tified using International Classification of Diseases, Ninth Edi- justing for age, sex, ISS, RTS, GCS score determined in the ED, tion, codes 02.2 and 01.18. and the probability of survival score (TRISS analysis) (Table 3).

jamasurgery.com JAMA Surgery June 2014 Volume 149, Number 6 545

Downloaded From: https://jamanetwork.com/ on 09/28/2021 Research Original Investigation Intracranial Pressure Monitoring in Children

Table 1. Demographics and General Characteristicsa Table 2. Hospital Outcomes

ICP No ICP Mean (SD) Variable Monitoring Monitoring P Value ICP No ICP Sex, No. (%) Variable Monitoring Monitoring P Value Male 166 (5.4) 1762 (56.8) Ventilator days 9.2 (8.5) 4.7 (16.7) <.001 .20 Female 117 (3.8) 1055 (34.0) ICU LOS 12.6 (10.3) 6.3 (8.7) <.001 Age, mean (SD), y 8.8 (5.9) 8.4 (6.0) .29 Hospital LOS 21.0 (19.5) 10.4 (14.9) <.001 GCS score, No. (%) Hospital charges, $ 49 000 (34 000) 34 000 (25 000) .001

3 172 (5.5) 1770 (57.1) Abbreviations: ICP, intracranial pressure; ICU, intensive care unit; 4 22 (0.7) 145 (4.7) LOS, length of stay. 5 30 (1.0) 139 (4.5) .34 6 23 (0.7) 263 (8.5) randomized trial. Thus, it is difficult to separate the benefi- 7 22 (0.7) 282 (9.1) cial or deleterious effects of ICP monitoring—or other efforts 8 16 (0.5) 223 (7.2) to manage ICP—from the many other developments in the treat- RTS, mean (SD) 2.4 (1.8) 2.8 (2.0) .01 ment of severe TBI. ISS, mean (SD) 31.5 (12.7) 27.6 (12.6) <.001 Our study confirms that ICP monitoring is undertaken in TRISS, mean (SD) 0.56 (0.27) 0.56 (0.31) .89 few pediatric patients with severe TBI who meet the current Vital status, No. (%) criteria for monitoring. When it is used, ICP monitoring is as- Alive 207 (6.7) 1946 (62.8) sociated with a decrease in mortality rate in only a small sub- .2 Dead 78 (2.5) 876 (28.2) set of the targeted population. In the present study, children

Abbreviations: GCS, Glasgow Coma Scale; ICP, intracranial pressure; ISS, Injury who received a monitoring device in accordance with the BTF Severity Score; RTS, Revised Trauma Score; TRISS, Trauma and Injury Severity guidelines had a longer hospital stay, longer ICU stay, and more Score. ventilator days. These findings suggest that the BTF criteria a Percentages are of the total study population. Data on sex were missing for 2 for ICP monitoring do not identify patients who are most likely patients with ICP monitoring and 5 patients with no ICP monitoring. to benefit from it. The potential reason that the BTF recommendations for The primary goal of our analysis was to determine when ICP ICP monitoring have not been widely adopted is the lack of vali- monitoring would be best used for patients with severe TBIs. dation through prospective measures, and neurosurgeons may When analyzed by each ED-assigned GCS score, ICP monitor- think that the level of evidence used in formulating these rec- ing was associated with a reduction in mortality only for pa- ommendations is insufficient and inconclusive. Although the tients with a GCS score of 3 (OR, 0.64; 95% CI, 0.43-1.00) BTF criteria have been promoted as evidence based, there is (Table 4). There was no effect of monitoring on mortality for no class I evidence to address this. any other GCS group. Numerous studies5,9,13-18 have suggested that therapies that lower ICP, including sedation, chemical paralysis, hyperventilation, cerebrospinal fluid drainage, osmotherapy, and pen- Discussion tobarbital-induced coma, reduce mortality and improve the likelihood of recovery when effective. However, it remains un- Monitoring of ICP has been used since the 1970s in the man- clear whether the association between elevated ICP and mor- agement of severe TBI and has been included in evidence- tality is a result of the severity of the underlying injury. Only based practice guidelines for nearly a decade. However, no ran- one prospective clinical trial has examined this association.14 domized clinical trial to evaluate the effect of ICP monitoring Seventy-three patients whose ICP could not be controlled by on outcomes in the management of severe TBI has been con- conventional means were randomly assigned to receive high- ducted. The obstacles to performing such a study include the dose pentobarbital or a placebo. The outcome of patients in ethical concern of not monitoring ICP in patients serving as a either group whose ICP could be maintained below 20 mm Hg control group and the widely accepted use of ICP monitoring was better than in those whose ICP could not be controlled. A in major pediatric centers involved in TBI research. lower ICP, or one that can be lowered with therapy, might sim- Although modern, ICP-focused, intensive management ply identify patients with less severe injuries who are likely to protocols have almost unquestionably improved outcomes, do well regardless of therapeutic interventions. This could ex- these protocols have involved the simultaneous changes of sev- plain why ICP monitoring in our study was not associated with eral other major variables: improved prehospital care, cranial an improved mortality rate in patients with a GCS score other computed tomographic imaging for accurate diagnosis of mass than 3. lesions, increased use of tracheal intubation and controlled ven- Another possible explanation for an association between tilation, more aggressive use of enteral and parenteral nutri- ICP monitoring and increased morbidity (increased length of tion, more active medical management of acute injury, and stay and ventilator days) in our study is that the interven- more widespread availability of formal rehabilitation pro- tions designed to reduce ICP are misapplied, harmful, or as- grams. During this period of change, ICP monitoring has sociated with complications. Hyperventilation has been emerged as an accepted practice without being evaluated in a shown19,20 to decrease cerebral perfusion and cause cerebral

546 JAMA Surgery June 2014 Volume 149, Number 6 jamasurgery.com

Downloaded From: https://jamanetwork.com/ on 09/28/2021 Intracranial Pressure Monitoring in Children Original Investigation Research

Table 3. Logistic Regression Results for Mortality Outcome Table 4. Stepwise Logistic Regression Results for Mortality Outcome of ICP Monitoring vs No ICP Monitoringa Variable OR (95% CI) ICP, all patients 0.837 (0.638-1.099) GCS Score OR (95% CI) Stepwise model 3 0.64 (0.43-1.00) ICP 0.716 (0.514-0.998) 4 1.09 (0.75-1.60) Sexa 5 0.59 (0.22-1.56) RTSa 6 0.31 (0.78-1.00) Age, y 7 0.41 (0.16-1.06) 2-9 vs <2 0.330 (0.253-0.431) 8 0.20 (0.05-1.73) 9-14 vs <2 0.199 (0.149-0.267) Abbreviations: GCS, Glasgow Coma Scale; ICP, intracranial pressure; >14 vs <2 0.313 (0.243-0.404) OR, odds ratio. a ISS All variables were entered into all models: ICP, sex, age category, Revised Trauma Score category, and Injury Severity Score category; ICP was the only 26-50 vs 10-25 2.637 (2.162-3.217) one that remained. >50 vs 10-25 11.346 (6.690-18.499)

GCS score national database, we have no way of validating the informa- 4 vs 3 0.612 (0.403-0.928) tion provided by individual trauma centers, including exper- 5 vs 3 0.527 (0.347-0.799) tise in coding procedures and ability to capture all ICP- 6 vs 3 0.332 (0.229-0.481) monitored patients within their facility. However, given the 7 vs 3 0.215 (0.142-0.324) sample size and methods used in the present study to control 8 vs 3 0.151 (0.090-0.253) the comparisons, this limitation is unlikely to invalidate the Abbreviations: GCS, Glasgow Coma Scale; ICP, intracranial pressure; ISS, Injury finding that there was no improvement in the mortality rate Severity Score; OR, odds ratio; RTS, Revised Trauma Score. in ICP-monitored patients. In addition, this was an observa- a Variable excluded for model (criterion for entry into the model, P = .05). tional study, with no control during therapeutic interventions. Finally, we used strict inclusion criteria; hence, the re- ischemia. Osmotic diuresis with mannitol may cause hypovo- sults may not be extrapolated to the entire spectrum of patients lemia and result in episodes of hypotension, which have been with head injuries. We believe that such questions can be an- shown21,22 to significantly increase mortality in patients with swered only by a prospective randomized clinical trial. Our data head injuries. There is also a small amount of risk associated provide a scientific basis for conducting such a trial. with the placement of an ICP monitor.23 Other interventions to reduce ICP or to increase cerebral perfusion pressure may result in fluid overload, inappropriate use of vasopressors, ex- Conclusions cessive use of paralytics and sedatives, and delay in libera- tion from the ventilator.24 In the present study, ICP monitor- Monitoring of ICP in accordance with current BTF criteria is ing was associated with an increase in hospital length of stay, associated with limited survival benefit in pediatric patients ICU length of stay, ventilator days, and overall hospital charges. with TBI. A prospective randomized clinical trial of ICP- There are several potential limitations of the study; the guided therapy may provide further evidence for when ICP most important is the retrospective design. As with any large monitoring may be beneficial.

ARTICLE INFORMATION Services, Memorial Regional Hospital, Hollywood, head injury with early diagnosis and intensive Accepted for Publication: August 1, 2013. Florida, assisted in editing and revising the management. J Neurosurg. 1977;47(4):491-502. manuscript. No financial compensation was 5. Marshall LF, Smith RW, Shapiro HM. The Published Online: April 30, 2014. provided. doi:10.1001/jamasurg.2013.4329. outcome with aggressive treatment in severe head injuries, part II: acute and chronic barbiturate Author Contributions: Dr Alkhoury had full access REFERENCES administration in the management of head injury. to all the data in the study and takes responsibility 1. Rutland-Brown W, Langlois JA, Thomas KE, Xi J Neurosurg. 1979;50(1):26-30. for the integrity of the data and the accuracy of the YL. Incidence of traumatic brain injury in the United data analysis. 6. Miller JD, Butterworth JF, Gudeman SK, et al. States, 2003. J Head Trauma Rehabil. 2006;21(6): Further experience in the management of severe Study concept and design: Alkhoury. 544-548. Acquisition, analysis, or interpretation of data: All head injury. J Neurosurg. 1981;54(3):289-299. authors. 2. Miller JD, Becker DP, Ward JD, Sullivan HG, 7. Ritter AM, Muizelaar JP, Barnes T, et al. Brain Drafting of the manuscript: Alkhoury. Adams WE, Rosner MJ. Significance of intracranial stem blood flow, pupillary response, and outcome Critical revision of the manuscript for important hypertension in severe head injury. J Neurosurg. in patients with severe head injuries. Neurosurgery. intellectual content: All authors. 1977;47(4):503-516. 1999;44(5):941-948. Statistical analysis: All authors. 3. Saul TG, Ducker TB. Effect of intracranial 8. Bratton SL, Chestnut RM, Ghajar J, et al. Administrative, technical, or material support: pressure monitoring and aggressive treatment on Guidelines for the management of severe traumatic Alkhoury. mortality in severe head injury. J Neurosurg. 1982; brain injury, VII: intracranial pressure monitoring Study supervision: Alkhoury. 56(4):498-503. technology. J Neurotrauma. 2007;24(suppl 1):S45- Conflict of Interest Disclosures: None reported. 4. Becker DP, Miller JD, Ward JD, Greenberg RP, S54. Additional Contributions: LeAnne S. Kerr, MSN Ed, Young HF, Sakalas R. The outcome from severe 9. Ghajar J, Hariri RJ, Narayan RK, Iacono LA, Firlik RN, CPN, CPEN, Office of Human Research/Trauma K, Patterson RH. Survey of critical care

jamasurgery.com JAMA Surgery June 2014 Volume 149, Number 6 547

Downloaded From: https://jamanetwork.com/ on 09/28/2021 Research Original Investigation Intracranial Pressure Monitoring in Children

management of comatose, head-injured patients in control of elevated intracranial pressure in patients tissue oxygen in patients with severe head injury. the United States. Crit Care Med. 1995;23(3):560- with severe head injury. J Neurosurg. 1988;69(1):15- Neurosurgery. 1996;38(1):21-31. 567. 23. 20. Dings J, Meixensberger J, Amschler J, Roosen

10. Hesdorffer DC, Ghajar J, Iacono L. Predictors of 15. Vukic M, Negovetic L, Kovac D, Ghajar J, Glavic K. Continuous monitoring of brain tissue PO2: a new compliance with the evidence-based guidelines for Z, Gopcevic A. The effect of implementation of tool to minimize the risk of ischemia caused by traumatic brain injury care: a survey of United guidelines for the management of severe head hyperventilation therapy. Zentralbl Neurochir. 1996; States trauma centers. J Trauma. 2002;52(6):1202- injury on patient treatment and outcome. Acta 57(4):177-183. 1209. Neurochir (Wien). 1999;141(11):1203-1208. 21. Manley G, Knudson MM, Morabito D, Damron S, 11. Baker SP, O’Neill B, Haddon W Jr, Long WB. The 16. Lundberg N, Troupp H, Lorin H. Continuous Erickson V, Pitts L. Hypotension, hypoxia, and head Injury Severity Score: a method for describing recording of the ventricular-fluid pressure in injury: frequency, duration, and consequences. Arch patients with multiple injuries and evaluating patients with severe acute traumatic brain injury: Surg. 2001;136(10):1118-1123. emergency care. J Trauma. 1974;14(3):187-196. a preliminary report. J Neurosurg. 1965;22(6):581- 22. Jeremitsky E, Omert L, Dunham CM, Protetch 12. Champion HR, Sacco WJ, Copes WS, Gann DS, 590. J, Rodriguez A. Harbingers of poor outcome the day Gennarelli TA, Flanagan ME. A revision of the 17. Johnston IH, Johnston JA, Jennett B. after severe brain injury: hypothermia, hypoxia, and Trauma Score. J Trauma. 1989;29(5):623-629. Intracranial-pressure changes following head injury. hypoperfusion. J Trauma. 2003;54(2):312-319. 13. Colohan AR, Alves WM, Gross CR, et al. Head Lancet. 1970;2(7670):433-436. 23. North B, Reilly P. Comparison among three injury mortality in two centers with different 18. Bulger EM, Nathens AB, Rivara FP, Moore M, methods of intracranial pressure recording. emergency medical services and intensive care. MacKenzie EJ, Jurkovich GJ; Brain Trauma Neurosurgery. 1986;18(6):730-732. J Neurosurg. 1989;71(2):202-207. Foundation. Management of severe head injury: 24. Barrientos-Vega R, Mar Sánchez-Soria M, 14. Eisenberg HM, Frankowski RF, Contant CF, institutional variations in care and effect on Morales-García C, Robas-Gómez A, Cuena-Boy R, Marshall LF, Walker MD. High-dose barbiturate outcome. Crit Care Med. 2002;30(8):1870-1876. Ayensa-Rincon A. Prolonged sedation of critically ill 19. van Santbrink H, Maas AI, Avezaat CJ. patients with midazolam or propofol: impact on Continuous monitoring of partial pressure of brain weaning and costs. Crit Care Med. 1997;25(1):33-40.

548 JAMA Surgery June 2014 Volume 149, Number 6 jamasurgery.com

Downloaded From: https://jamanetwork.com/ on 09/28/2021