CSA Home
Central Surgical Association

Back to 2022 Abstracts


Predicting Endocrine Function After Total Pancreatectomy And Islet Cell Auto Transplantation: A Novel Approach Utilizing CT Texture Analysis
*Kevin M Turner, *Shaun Wahab, *Aaron M Delman, *John Brunner, *Milton Smith, *Kyuran Choe, Sameer H Patel, Syed A Ahmad, Gregory C Wilson
University of Cincinnati, Cincinnati, OH

OBJECTIVE (S): Islet cell autotransplantation (IAT) is an effective method to minimize endocrine dysfunction after total pancreatectomy (TP). There is no effective method to predict long-term islet function after TPIAT. Therefore, the aim of this study was to create a model to predict long-term endocrine function after TPIAT
METHODS: All patients undergoing TPIAT from 2007-2020 who had a high-quality preoperative CT imaging available for texture analysis were included. The primary outcome was optimal long-term endocrine function, defined as stable glycemic control with <10 units of insulin per day. CT texture analysis was performed on pre-operative imaging including the following first order parameters: mean pixel intensity, standard deviation, mean positive pixel value, kurtosis, entropy, and skewness. Islet cell yield was evaluated by islet size and number.
RESULTS: 63 patients met inclusion criteria. Median islet yield was 6,111 islet equivalent (IEQ)/kg body weight. At a median follow-up of 64.2 months, 12.7% (n=8) of patients were insulin independent and 39.7% (n=25) demonstrated optimal endocrine function. Neither total IEQ nor IEQ/kg body weight alone were associated with optimal endocrine function. When islet yield was further evaluated by islet size, fewer small size (50-100 μm) islets was the best predictor of optimal endocrine function (51,000 IEQ vs. 71,250 IEQ, p=0.18). In order to improve endocrine function prediction, CT texture analysis parameters were analyzed, only identifying an association between kurtosis (OR: 2.32 95% CI: 1.08-4.80, p=0.02) and optimal endocrine function. Sensitivity analysis identified kurtosis of 0.60, with good endocrine function identified in 56.0% of patients with a kurtosis ≥0.60 compared to only 18.4% of patients with kurtosis <0.60 (p<0.01). Nominal logistic regression to predict optimal endocrine function utilizing factors associated on univariate analysis including kurtosis and operative gland texture, in addition to clinically important variables, such as islet yield 50-100 μm and pre-operative insulin use (AUC: 0.771), found only kurtosis ≥0.60 (OR: 5.07 95%CI: 1.33-19.33, p=0.01) predicted optimal long-term endocrine function.
CONCLUSIONS: Islet yield alone was not associated with endocrine outcomes. CT Texture analysis with islet yield data represents a novel approach to accurately predict long-term endocrine function after TPIAT.


Back to 2022 Abstracts