Radiographic Sarcopenia and Self-reported Exhaustion Independently Predict NSQIP Serious Complications After Pancreaticoduodenectomy in Older Adults.

Abstract

Sarcopenia is linked to poor outcomes after abdominal surgery. We hypothesized that radiographic sarcopenia metrics enhance prediction of complications after pancreaticoduodenectomy (PD) when combined with clinical and frailty data. Preoperative geriatric assessments and CT scans of patients undergoing PD were reviewed. Sarcopenia was assessed at L3 using total psoas area index (TPAI) and weighted average Hounsfield units (HU), i.e., estimates of psoas muscle volume and density. Outcomes included 30-day American College of Surgeons National Surgical Quality Improvement Program (NSQIP) serious complications, Clavien-Dindo complications, unplanned intensive care unit (ICU) admission, hospital length of stay (LOS), non-home facility (NHF) discharge, and readmission rates. Low HU score correlated with NSQIP serious complications (r = -0.31, p = 0.0098), Clavien-Dindo complication grade (r = -0.29, p = 0.0183), unplanned ICU admission (r = -0.28, p = 0.0239), and NHF discharge (r = -0.25, p = 0.0426). Controlling for a "base model" of age, body mass index, American Society of Anesthesiologists score, and comorbidity burden, Fried's exhaustion (odds ratio [OR] 4.72 [1.23-17.71], p = 0.021), and HU (OR 0.88 [0.79-0.98], p = 0.024) predicted NSQIP serious complications. Area under the receiver-operator characteristic (AUC) curves demonstrated that the combination of the base model, exhaustion, and HU trended towards improving the prediction of NSQIP serious complications compared with the base model alone (AUC = 0.81 vs. 0.70; p = 0.09). Additionally, when controlling for the base model, TPAI (β-coefficient = 0.55 [0.10-1.01], p = 0.018) and exhaustion (β-coefficient = 2.47 [0.75-4.20], p = 0.005) predicted LOS and exhaustion (OR 4.14 [1.48-11.6], p = 0.007) predicted readmissions. When combined with clinical and frailty assessments, radiographic sarcopenia metrics enhance prediction of post-PD outcomes.