Body composition as a predictor of cancer-related death in colon cancer: an AI-based volumetric analysis.
Authors
Affiliations (9)
Affiliations (9)
- Department of Medical-Surgical Sciences and Translational Medicine, School of Medicine and Psychology, Sapienza University of Rome - Sant'Andrea University Hospital, Via Di Grottarossa, 1035-1039, 00189, Rome, Italy.
- School in Traslational Medicine and Oncology, Department of Medical-Surgical Sciences and Translational Medicine, Faculty of Medicine and Psychology, Sapienza University of Rome, Rome, Italy.
- Department of Medical-Surgical Sciences and Translational Medicine, School of Medicine and Psychology, Sapienza University of Rome - Sant'Andrea University Hospital, Via Di Grottarossa, 1035-1039, 00189, Rome, Italy. [email protected].
- Department of Surgery, Policlinico Umberto I, Sapienza University of Rome, Via Giovanni Maria Lancisi, 2, 00161, Rome, Italy.
- Department of Emergency Radiology, Azienda Ospedaliera S. Camillo-Forlanini, Rome, Italy.
- Department of Radiology, Seoul National University College of Medicine, Seoul, Republic of Korea.
- AI Center, Medical IP, Co., Ltd., Seoul, Republic of Korea.
- Department of Biomedical Sciences, Humanitas University, Via Rita Levi Montalcini 4, 20072, Pieve Emanuele, Milan, Italy.
- Department of Diagnostic Imaging, IRCCS Humanitas Research Hospital, Via Manzoni 56, 20089, Rozzano, Milan, Italy.
Abstract
To investigate body composition as a predictive biomarker of cancer-related death in patients with non-metastatic colon cancer (CRC). Patients with CRC (stages II-III) treated with upfront surgery, with availability of baseline CT, clinico-histological and survival data were retrospectively enrolled. Patients with stage IV or CT unavailability were excluded. Body composition parameters derived from baseline abdominal CT using AI-based automatic segmentation software. Seventy-six parameters regarding adipose visceral fat(AVF), subcutaneous fat(SF), bone density, liver density, and fat-fraction were automatically extracted from both whole segmentation volume and multislices region. According to the CRC-related death (CRC-related death), the population was divided into Group 1 (CRC-related death) and Group 2 (non-CRC-related death). Body composition features were compared between two groups. Predictive model and survival analysis were performed with ROC curves, Cox regression, and Kaplan-Meier method. P < 0.05 was considered significant. A total of 293 patients were included, 101/293(34.5%) with CRC-related deaths. MeanHU of AVF and SF resulted directly correlated with CRC-related death (P = 0.004 and HR = 1.03, P = 0.002 and HR = 1.04, respectively) for the multislice analysis. MeanHU of AVF resulted in direct correlation with CRC-related deaths, also for the volumetric analysis (P = 0.04 and HR = 1.02). In multivariable Cox regression analysis, MeanHU AVF was confirmed as an independent predictor of CRC-related death in both multislice and volumetric analyses. SF HU remained significant in multislice analysis. In Kaplan-Meier analysis, AVF and SF for the multislice analysis resulted statistically significant (P = 0.033 and < 0.001, Chi-square = 4.56 and 11.7, respectively). In conclusion, our study demonstrated that body composition metrics of visceral and subcutaneous fat were significantly associated with cancer-related death in non-metastatic CRC patients.