The aim of this study is to assess whether a complete analysis of all early cortical somatosensory-evoked potentials (SEPs) components and computed tomography (CT) scan features can provide a better prognostic measure than the early cortical component N20/P25 alone, in patients with severe head injury.
We studied 81 consecutive patients admitted to intensive care unit with diagnosis of severe head injury. All patients underwent neurophysiological assessment with SEPs and electroencephalography within the first 6 days after trauma. The marginal effect of each variable on Glasgow Outcome Scale score was evaluated by using univariate measures of association. We fit a cumulative logit model by maximum likelihood, and the partial effect of each variable was assessed by likelihood ratio test. We performed variable selection by forward stepwise, according to the Akaike information criterion.
Our final cumulative logit model including SEPs primary complex (pN20/fP20/cP22), SEPs middle latency (N30/P45/N60), and CT scan hypodensity values showed a significantly increased predictive power of Glasgow Outcome Scale, compared with pN20 alone (P<0.0001).
Statistical analysis revealed a highly significant (P<0.0001) improvement in outcome prediction when the model includes a pool of amplitudes and latencies referred to different early-evoked components pN20, pP25, fP20, cP22, N30, P45, and N60, associated to CT scan hypodensity values, compared with the use of the cortical parietal N20/P25 alone.
Supplemental Digital Content is available in the text.
*Anaesthesiology and Intensive Care Unit
∥Tissue Bank of Treviso, Treviso Hospital, Treviso
†FBOV—Fondazione Banca Occhi, Zelarino, Venezia
‡Department of Neurosurgery, Treviso Hospital, University of Padova, Treviso
§Department of Statistics, University of Padova, Padova, Italy
Supplemental Digital Content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal's Website, www.jnsa.com.
Author Disclosure Statement: An application for electronic devices has been created for a rapid use of the prediction model. The authors have no funding or conflicts of interest to disclose.
Reprints: Alberto Feletti, MD, PhD, Department of Neurosurgery, Treviso Hospital, University of Padova, Piazzale Ospedale, 1, Treviso 31100, Italy (e-mail: firstname.lastname@example.org).
Received September 16, 2012
Accepted December 16, 2013