151751-Najiba-Chargi

145 Surgery: skeletal muscle mass and oral cavity cancer the first CT-slice at C3 level to entirely depict the vertebral arc, the transverse and the spinous process(es) was selected. Skeletal muscle was characterized by Hounsfield units (HU) ranging from -29 to +150 HU 23 and identified using Slice-O-Matic software V4.3 (Tomovision, Montreal, Quebec, Canada). By delineating the paravertebral and the sternocleidomastoid muscles at C3 level, the cross-sectional area of the muscles was calculated in cm 2 using Slice-O-Matic software. The cross-sectional muscle area at L3 level was then estimated using Equation (1) as described by Swartz et al. 26 This result was normalized by dividing by the squared body height inm 2 to calculate the lumbar skeletal muscle index (LSMI). Sarcopenia was then defined as the lowest quartile LSMI of the patient group. 31,32 STATISTICAL METHODS Statistical analyses were performed using SPSS (IBM Corp. Released 2017. IBM SPSS Statistics for Windows, Version 25.0. Armonk, NY: IBM Corp.). A summary of the patient group and tumor description was generated. The perioperative complications were modeled with sarcopenia and the other potential risk factors. Firstly, univariate analysis was performed for potential risk factors to examine the association of potential risk factors with presence of perioperative complications. Potential risk factors with p value <0.25 in the univariate analysis and sarcope - nia were then entered in the multivariate analysis using logistic regression for the presence of perioperative complications and negative binomial regression for the number of perioperative complications. Secondly, univariate linear regression was performed for potential risk factors to examine the association of potential risk factors with length of hospital stay. Potential risk factors with p value <0.25 in the univariate analysis and sarcopenia were then entered in the multivariate analysis using normal linear regression for length of hospital stay. For the linear regression, length of hospital stay was transformed using a Log transformation. After the Log transformation, continuity and normality was assumed. All hypotheses in the multivariate analysis were tested two-sided with a statistical level of α=0.05. RESULTS PATIENT CHARACTERISTICS The study group consisted of 259 patients. 33 of these patients were excluded because of un - available CT/MRI scans or poor image quality. The remaining group of 226 patients consisted of 110 males and 116 females, with median age 67 years. 51 patients had developed one or multiple complications. The patient characteristics are listed in Table 1. Because just one patient had ASA score 4, three ASA groups were made: ASA 1, 2 and 3+. The tumor variables are shown in Table 2. 8