Patients and study design
We evaluated surgical patients admitted to a tertiary-level ICU between June 2014 and December 2018.
Inclusion criteria were the following: age ≥ 18 years; ICU admission after urgent abdominal surgery or reoperation for complicated elective major abdominal surgery (gastrointestinal, gynecologic or urologic procedures); patients who underwent an abdominal CT within 30 days before and 48 hours after the admission in intensive care. We excluded patients admitted to ICU following trauma, intraoperative medical complications (e.g. cardiac arrhythmia) or patients with planned ICU admission for monitoring after elective surgery.
The study was reviewed and approved by the Institutional Review Board (Comitato Etico ASST Monza). Written informed consent was waived. Results are reported according to Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement.
CT scan and Body Composition measurement
Body composition indexes were measured and calculated from the CT scan closest in time to the admission (from 72 hours before to 24h after the admission to the ICU). A multidetector CT scan was performed with 256-slice CT scanners (Brilliance iCT or iCT-elite, Philips Medical Systems, Eindhoven, Netherlands) and an unenhanced scan was acquired in every study protocol. The raw data were reconstructed with Hybrid-Iterative Reconstruction algorithm (iDose4), in order to obtain the same image quality between the two different CT scanners; subsequently, all the scans were transferred to an image workstation (Intellispace portal 8.0; Philips Medical Systems) to evaluate, select, anonymize and save the image for the analysis in DICOM format. The analysis was performed on the axial CT 5mm-image through L3 with the open-source image analysis software ImageJ (developed by the National Institutes of Health; available from http://rsbweb.nih.gov/ij/download.html), which produces comparable results to other software for body composition analysis .
Two different radiologists, blinded to patient information and using the selected CT image, drew multiple regions of interests (ROI) in the outer and inner perimeter of abdominal muscles, and analyzed pixels with densities between -20 Hounsfield units (HU) to +150 HU for muscles and with densities in the -190 HU to -30 HU range for fat-tissue. The radiologists then calculated total muscle area (TMA, which estimates the total muscle mass) , total fat area (TFA), visceral fat area (VFA) and intramuscular fat area (IMFA). Subcutaneous fat area (SFA) was obtained by subtracting VFA from TFA and intramuscular fat area (IMFA) was obtained subtracting VFA from the ROI of the outer abdominal muscle perimeter.
Body composition indexes were normalized for height in meters squared [1,18], and expressed as cm2/m2. We used TMA and VFA to calculate Skeletal Muscle Index (SMI=TMA/m2) and Sarcopenic Obesity (SO=VFA/TMA). We also determined the grade of MyoSteatosis through the intramuscular adipose tissue content (MS=IMFA/TMA) . Tertiles were estimated for each index according to sex [21,22].
The following parameters were retrieved from the medical records: demographics (age, sex, height, weight, and body mass index [BMI]), Charlson Comorbidity Index (CCI), Simplified Acute Physiology Score (SAPS) II, Sequential Organ Failure Assessment (SOFA) score on ICU admission, clinical data (use of vasopressor, use of mechanical ventilation, diagnosis of abdominal urgency or type of complication after elective surgery), length of ICU and hospital stay, duration of mechanical ventilation, outcome (hospital discharge or death).
The primary aim of the present study was to investigate the association of body composition parameters, specifically skeletal muscle index, myosteatosis and sarcopenic obesity, with 90-day mortality.
Secondary aims included exploring the association of the same parameters with ICU, in-hospital and 1-year mortality, mechanical ventilation days and ICU length of stay
Continuous variables were described as mean ± standard deviation or median [interquartile range] depending on their distribution, categorical variables as absolute (relative) frequency. The normality of distribution was assessed using the Shapiro–Wilk test.
Differences in 90-day mortality were tested using a non-parametric test for trend across tertiles of body composition indexes. If the trend was statistically significant, multiple pairwise comparisons among tertiles of indexes were explored using the Chi-square test.
The Kaplan Meier approach was applied to assess the unadjusted probability of survival at 90-days and 1-year follow up. Log-rank test was used to compare curves between 3 groups defined by tertiles of body composition indexes. Mortality at 90-day follow up was then tested into a multivariate Cox-regression model using a stepwise selection approach (cut-off p <0.20 for selection at the univariate analysis).
The association of patient characteristics with outcomes was assessed with univariate logistic regression (for ICU, in-hospital and 1-year mortality) and linear regression (duration of mechanical ventilation and ICU length of stay) analyses. Variables with a p<0.20 were included into a multivariable logistic and linear regression model, respectively, using a stepwise selection approach.
To further explore the interaction of low skeletal muscle index and myosteatosis - we explored the role of combination of low/high levels of SMI with low/high levels of myosteatosis on 90-day hospital mortality. Patients with low or high SMI were defined as the groups of patients with SMI level below or above the 50th percentile, respectively. Equally, patients with low or high myosteatosis were defined as the groups of patients below or above the 50th percentile of myosteatosis, respectively. Furthermore, to evaluate the improvement in the discrimination performance of the multivariate Cox model on the primary end-point (i.e. 90 days mortality) when myosteatosis and skeletal muscle index are combined, we compared its Harrel’s C-index to alternate Cox models where only one of the 2 indexes is included.
Statistical significance was considered with a p <0.05 (two-tailed). Statistical analyses and graphs were performed using STATA-16/MP (StataCorp LP, College Station, TX, USA) and GraphPad Prism 8.0.2 (GraphPad Software, San Diego, CA, USA).