BACKGROUND: To propose a combination of blood biomarkers for the prediction of hospital-acquired pneumonia (HAP) and for the selection of traumatic brain-injured (TBI) patients eligible for corticosteroid therapy for the prevention of HAP. METHODS: This was a sub-study of the CORTI-TC trial, a multicenter, randomized, double-blind, controlled trial evaluating the risk of HAP at day 28 in 336 TBI patients treated or not with corticosteroid therapy. Patients were between 15 and 65 years with severe traumatic brain injury (Glasgow coma scale score ≤ 8 and trauma-associated lesion on brain CT scan) and were enrolled within 24 h of trauma. The blood levels of CRP and cortisoltotal&free, as a surrogate marker of the pro/anti-inflammatory response balance, were measured in samples collected before the treatment initiation. Endpoint was HAP on day 28. RESULTS: Of the 179 patients with available samples, 89 (49.7%) developed an HAP. Cortisoltotal&free and CRP blood levels upon ICU admission were not significantly different between patients with or without HAP. The cortisoltotal/CRP ratio upon admission was 2.30 [1.25-3.91] in patients without HAP and 3.36 [1.74-5.09] in patients with HAP (p = 0.021). In multivariate analysis, a cortisoltotal/CRP ratio > 3, selected upon the best Youden index on the ROC curve, was independently associated with HAP (OR 2.50, CI95% [1.34-4.64] p = 0.004). The HR for HAP with corticosteroid treatment was 0.59 (CI95% [0.34-1.00], p = 0.005) in patients with a cortisoltotal/CRP ratio > 3, and 0.89 (CI95% [0.49-1.64], p = 0.85) in patients with a ratio < 3. CONCLUSION: A cortisoltotal/CRP ratio > 3 upon admission may predict the development of HAP in severe TBI. Among these patients, corticosteroids reduce the occurrence HAP. We suggest that this ratio may select the patients who may benefit from corticosteroid therapy for the prevention of HAP.