Daniel W Maxwell1, Mohammad Raheel Jajja2, Rodolfo J Galindo3, Chao Zhang4, Syed O Nadeem1, John F Sweeney1, Catherine M Blair1, Juan M Sarmiento5. 1. Department of Surgery, Emory University, Atlanta, GA. 2. Department of Surgery, Emory University, Atlanta, GA; Winship Cancer Institute, Emory University, Atlanta, GA. 3. Division of Endocrinology, Emory University, Atlanta, GA. 4. Winship Cancer Institute, Emory University, Atlanta, GA; Bioinformatics and Biostatistics Shared Resource, Emory University, Atlanta, GA. 5. Department of Surgery, Emory University, Atlanta, GA; Winship Cancer Institute, Emory University, Atlanta, GA. Electronic address: jsarmie@emory.edu.
Abstract
BACKGROUND: Literature is varied regarding risk factors associated with diabetes development after major pancreatic resection. The aim was to develop and validate a scoring index that preoperatively predicts the development of diabetes after pancreaticoduodenectomy and distal pancreatectomy. STUDY DESIGN: In this prospective study, perioperative fasting and postprandial (OGTT, oral glucose tolerance test) plasma glucose, glycated hemoglobin A1c (HbA1c), insulin, and c-peptide were measured in select consecutive patients undergoing pancreaticoduodenectomy and distal pancreatectomy by the senior author, from 2007 to 2018. American Diabetes Association definitions were used for glycemic classifications. Statistical analyses included multivariate generalized estimated equation for factor identification and variable weighting; area under the receiver operating curve (ROC) c-statistic for predictive ability, and survival analysis risk score grouping. RESULTS: Of 1,083 included patients with preoperative normoglycemia (253; 23.4%), prediabetes (362; 33.4%), and diabetes (468; 43.2%), the overall postoperative incidence of each diabetic class at 120 months was 152 (14.0%), 466 (43.0%), and 465 (42.9%), respectively. The development and validation groups included 1,023 and 60 patients, respectively. Five factors were identified predicting diabetes development, with a total possible score of 8. The C-statistics for development and validation groups were 0.727 (CI 0.696 to 0.759, p < 0.001) and 0.823 (CI 0.718 to 0.928, p < 0.001), respectively. At a cut point of 3 (sensitivity 0.691, specificity 0.644) the Post-pancreatectomy Diabetes Index (PDI) independently predicted diabetes in development (odds ratio [OR] 4.298, relative risk [RR] 2.486, CI 1.238 to 5.704, p < 0.001) and validation (OR 6.970, RR 2.768, CI 2.182 to 22.261, p < 0.001) groups. The PDI similarly predicted pre-diabetes in development (OR 1.961, RR 1.325, CI 1.202 to 2.564, p < 0.001) and validation (OR 4.255, RR 1.798, CI 1.247 to 14.492, p = 0.021) groups. CONCLUSIONS: The Post-pancreatectomy Diabetes Index predicts the development of diabetes and pre-diabetes in patients undergoing major pancreatectomy using routine endocrine laboratories and pre-surgical clinical data.
BACKGROUND: Literature is varied regarding risk factors associated with diabetes development after major pancreatic resection. The aim was to develop and validate a scoring index that preoperatively predicts the development of diabetes after pancreaticoduodenectomy and distal pancreatectomy. STUDY DESIGN: In this prospective study, perioperative fasting and postprandial (OGTT, oral glucose tolerance test) plasma glucose, glycated hemoglobin A1c (HbA1c), insulin, and c-peptide were measured in select consecutive patients undergoing pancreaticoduodenectomy and distal pancreatectomy by the senior author, from 2007 to 2018. American Diabetes Association definitions were used for glycemic classifications. Statistical analyses included multivariate generalized estimated equation for factor identification and variable weighting; area under the receiver operating curve (ROC) c-statistic for predictive ability, and survival analysis risk score grouping. RESULTS: Of 1,083 included patients with preoperative normoglycemia (253; 23.4%), prediabetes (362; 33.4%), and diabetes (468; 43.2%), the overall postoperative incidence of each diabetic class at 120 months was 152 (14.0%), 466 (43.0%), and 465 (42.9%), respectively. The development and validation groups included 1,023 and 60 patients, respectively. Five factors were identified predicting diabetes development, with a total possible score of 8. The C-statistics for development and validation groups were 0.727 (CI 0.696 to 0.759, p < 0.001) and 0.823 (CI 0.718 to 0.928, p < 0.001), respectively. At a cut point of 3 (sensitivity 0.691, specificity 0.644) the Post-pancreatectomy Diabetes Index (PDI) independently predicted diabetes in development (odds ratio [OR] 4.298, relative risk [RR] 2.486, CI 1.238 to 5.704, p < 0.001) and validation (OR 6.970, RR 2.768, CI 2.182 to 22.261, p < 0.001) groups. The PDI similarly predicted pre-diabetes in development (OR 1.961, RR 1.325, CI 1.202 to 2.564, p < 0.001) and validation (OR 4.255, RR 1.798, CI 1.247 to 14.492, p = 0.021) groups. CONCLUSIONS: The Post-pancreatectomy Diabetes Index predicts the development of diabetes and pre-diabetes in patients undergoing major pancreatectomy using routine endocrine laboratories and pre-surgical clinical data.