Table 3 presents the final models of linear multiple regression analysis, illustrating the demographic, clinical, radiological,

and neurosurgical variables independently Selleckchem Roscovitine associated with cognitive test performance of the TBI patients. Education was independently associated with better scores in all the studied scores, except for LM II and VP Rec. Older age was independently associated with lower scores in all the applied tests, except for the vocabulary test. Admission glucose serum levels of 150 mg/dL or higher were independently associated with lower scores in the RAVLT-retention test. Absence of SAH at admission CT was independently associated with higher scores in the LM II. The Marshal Class IV or worse in the admission CT revealed a non-significant trend for independent association with lower performance in the vocabulary (p = .08) and LM 1st tests (p = .10). The linear multiple logistic regression analysis models showed a moderately strong linear relationship (60

(i) there were no outliers on the residuals, (ii) residual data points were independent, and (iii) the residual distribution was normal (Table 4). Our initial hypothesis was that variables FG-4592 order classically associated with TBI prognosis could be useful to predict the cognitive performance of severe

TBI patients. However, our findings suggest that hospitalization selleckchem variables had limited capacity to predict the long-term cognitive outcome of those patients. The low association among variables well known to be predictive of mortality or morbidity, such as admission pupils’ size, GCS, serum glucose levels, Marshal CT classification (Gullo et al., 2011; Hohl et al., 2012; Martins et al., 2009; Roozenbeek et al., 2011) was unexpected. This may indicate that in our sample of patients with severe TBI, the mechanisms involved in mortality and morbidity may have differed at least in part from those involved in global cognition recovery. We cannot exclude the possibility of a type II error, because the sample size was in the limit for the inclusion of 3–4 variables in a multiple regression analysis. However, the p level of significance for association between these variables and almost all the analysed cognitive tests became higher than .5 when they were individually included in the regression model together with age and education level (data not shown). As expected, older age and lower education level were also predictors of worse long-term cognitive impairment. These findings are in agreement with the results of previous retrospective studies of severely impaired TBI patients (Chu et al., 2007; Sidaros et al., 2008; Vanderploeg, Curtiss, Luis, & Salazar, 2007).