Article
Development of a predictive model for sustained elevations in intracranial pressure following traumatic brain injury
Search Medline for
Authors
-
Julie Woodfield
- Western General Hospital & University of Edinburgh, Department of Clinical Neurosciences, Edinburgh, United Kingdom
-
John A. Emelifeonwu
- Western General Hospital & University of Edinburgh, Department of Clinical Neurosciences, Edinburgh, United Kingdom
-
Aryelly Rodriguez
- The University of Edinburgh, Centre for Population Health Sciences, Edinburgh, United Kingdom
-
Andreas K. Demetriades
- Western General Hospital & University of Edinburgh, Department of Clinical Neurosciences, Edinburgh, United Kingdom
-
Randall Chesnut
- University of Washington Mailstop, Harborview Medical Center, Neurological Surgery , Seattle, Washington, United States
-
Peter J. D. Andrews
- Western General Hospital & University of Edinburgh, Department of Clinical Neurosciences, Edinburgh, United Kingdom
| Published: | June 9, 2017 |
|---|
Outline
Text
Objective: To develop and validate a model that uses clinical variables to predict the risk of sustained elevations in intracranial pressure (ICP) following traumatic brain injury (TBI) using data from two large TBI studies.
Methods: Logistic regression models were fitted to prospectively collected clinical data from the Eurotherm3235 trial [1]. Included participants that had sustained elevations in ICP (defined as ICP>20mmHg for at least 5 minutes) were compared with patients who were excluded from the trial because they did not develop elevated ICP. Models were compared for model fit and predictive function. The model with the highest area under the receiver operating curve (AUC) was tested using data from the Benchmark Evidence from South American Trials: Treatment of Intracranial Pressure (BESTTRIP) trial [2].
Results: There were 387 randomised participants in the Eurotherm3235 trial with sustained elevations in ICP and 867 patients with clinical and demographic data who were screened but did not develop sustained elevations in ICP. BESTTRIP included 157 ICP monitored patients, and 111 had sustained elevations in ICP. The logistic regression model with the highest predictive function for raised ICP in the Eurotherm3235 trial data set included age, GCS motor score, pupil reactivity, and CT Marshall grade. In the Eurotherm3235 trial data, this model had an AUC of 0.68. When applied to the BESTTRIP dataset, the AUC was 0.53.
Conclusion: Clinical variables can predict the likelihood of a sustained rise in ICP following TBI. The model developed with the Eurotherm3235 trial dataset has lower discrimination in the BESTTRIP data set. This is likely to be due to the different clinical characteristics of the patients included in the two trials. Further training of the model on additional datasets may improve predictive ability and generalisability.
References
- 1.
- Andrews PJ, Sinclair HL, Rodriguez A, Harris BA, Battison CG, Rhodes JK, Murray GD; Eurotherm3235 Trial Collaborators. Hypothermia for Intracranial Hypertension after Traumatic Brain Injury. N Engl J Med. 2015 Dec 17;373(25):2403-12. DOI: 10.1056/NEJMoa1507581
- 2.
- Chesnut RM, Temkin N, Carney N, Dikmen S, Rondina C, Videtta W, Petroni G, Lujan S, Pridgeon J, Barber J, Machamer J, Chaddock K, Celix JM, Cherner M, Hendrix T; Global Neurotrauma Research Group. A trial of intracranial-pressure monitoring in traumatic brain injury. N Engl J Med. 2012 Dec 27;367(26):2471-81. DOI: 10.1056/NEJMoa1207363
