Objective To investigate the construction and application value of a predictive model for prolonged surgical duration in Da Vinci robotic radical gastrectomy for gastric cancer.

Methods The retrospective cohort study was conducted. The clinicopathological data of 534 patients who underwent Da Vinci robotic radical gastrectomy for gastric cancer in the Fujian Medical University Union Hospital from August 2016 to August 2021 were collected. There were 389 males and 145 females, aged (60±11)years. All 534 patients were randomly divided into the training dataset of 374 cases and the validation dataset of 160 cases with a ratio of 7∶3 based on random number method in the SPSS 25.0 software. Observation indicators: (1) incidence of prolonged surgical duration; (2) intraoperative and postoperative conditions in patients with prolonged surgical duration and without prolonged surgical duration; (3) complications in patients with prolonged surgical duration and without prolonged surgical duration; (4) analysis of risk factors influencing prolonged surgical duration; (5) construction and evaluation of an artificial neural network predictive model for pro-longed surgical duration. Measurement data with normal distribution were represented as Mean±SD, and comparison between groups was conducted using the independent sample t test. Measurement data with skewed distribution were represented as M(Q₁,Q₃), and comparison between groups was conducted using the Mann‑Whitney U test. Count data were described as absolute numbers or per-centages, and comparison between groups was conducted using the chi‑square test or Fisher exact probability. Comparison of ordinal data was analyzed using the nonparametric test. Univariate and multivariate analyses were conducted using the Logistic regression model. Based on the results of univariate analysis, a multilayer perceptron was employed to train an artificial neural network pre-dictive model for prolonged surgical duration. The receiver operating characteristic (ROC) curve was drawn, and the area under curve (AUC), the accuracy, sensitivity, specificity, positive predictive value, and negative predictive value were used to assess the model′s performance.

Results (1) Incidence of prolonged surgical duration. Of 534 patients, 284 cases underwent total gastrectomy, and 250 cases underwent distal gastrectomy, with operation time of (206±42)minutes and (187±36)minutes, res-pectively. Cases with prolonged surgical duration and without prolonged surgical duration who under-went total gastrectomy were 41 and 243, and cases with prolonged surgical duration and without prolonged surgical duration who underwent distal gastrectomy were 40 and 210. The gender (male, female), age, body mass index (BMI), tumor diameter, tumor location (upper stomach, middle stomach, lower stomach, mixed type), cases with neoadjuvant therapy, cases with preoperative American Society of Anesthesiologists (ASA) score as 1, 2, 3, cases with clinical T staging as stage T1, stage T2, stage T3, stage T4a, cases with clinical N staging as stage N0, stage N1, stage N2, stage N3, cases with clinical TNM staging as stage Ⅰ, stage Ⅱ, stage Ⅲ, cases with surgical resection scope as total gastrec-tomy or distal gastrectomy, cases with digestive tract reconstruction method as Billroth‑Ⅰ anasto-mosis, Billroth‑Ⅱ anastomosis, Roux‑en‑Y anastomosis, cases with surgeon experiences as ≤20 cases or >20 cases were 61,20, (61±9)years, (24±3)kg/m², 4.0(2.5, 5.0)cm, 34, 10, 33, 4, 1, 3, 73, 5, 3, 6, 26, 46, 14, 41, 19, 7, 5, 13, 63, 41, 40, 1, 33, 47, 5, 76 in the 81 patients with prolonged surgical duration, versus 328, 125, (60±11)years, (23±3)kg/m², 3.5(2.0, 5.0)cm, 129, 71, 227, 26, 6, 45, 382, 26, 73, 100, 118, 162, 211, 180, 52, 10, 138, 108,207, 243, 210, 13,200, 240, 15, 438 in the 453 patients without prolonged surgical duration, showing significant differences in the BMI, clinical T staging, clinical N staging, clinical TNM staging (t=-3.68, Z=-4.63, -5.53, -5.56, P<0.05), and no significant difference in the gender, age, tumor diameter, tumor location, preoperative ASA score, surgical resec-tion scope, digestive tract reconstruction method, and surgeon experiences (χ²=0.29, t=-0.95, Z=-1.27, χ²=5.92, Z=-1.46, χ²=0.25, 1.35, 0.87, P>0.05). There was no significant difference in cases with neoadjuvant therapy between them (P>0.05). (2) Intraoperative and postoperative conditions in patients with prolonged surgical duration and without prolonged surgical duration. The operation time, volume of intraoperative blood loss, the number of lymph nodes dissected, time to postopera-tive first ambulation, time to postoperative anal exhaust, time to postoperative first intake of liquid diet, time to postoperative first intake of semi‑liquid diet, duration of postoperative hospital stay were (261±34)minutes, 50(30, 50)mL, 39±15, (2.3±0.6)days, (3.4±0.9)days, (4.1±1.2)days, (5.7±1.2)days, 8.0(7.0, 9.0)days in the 81 patients with prolonged surgical duration, versus (186±29)minutes, 30(20,50)mL, 42±14, (2.2±0.6)days, (3.4±0.8)days, (4.1±1.1)days, (5.7±1.4)days, 8.0(7.0, 9.0)days in the 453 patients without prolonged surgical duration, showing significant differences in operation time, volume of intraoperative blood loss (t=-20.46, Z=-3.32, P<0.05), and no significant difference in the number of lymph nodes dissected, time to postoperative first ambulation, time to postopera-tive anal exhaust, time to postoperative first intake of liquid diet, time to first intake of semi‑liquid diet, duration of postoperative hospital stay (t=1.87, -0.87, -0.16, 0.28, 0.03, Z=-1.45, P>0.05). (3) Complications in patients with prolonged surgical duration and without prolonged surgical duration. The overall incidence of complications, incidence of surgical complications (abdominal infection, anastomotic fistula, abdominal bleeding, incision‑related complications, intestinal obstruction, lymphatic fistula), incidence of medical complications (pulmonary infection, liver‑related complications) were 22.22%(18/81), 0, 0, 2.47%(2/81), 0, 8.64%(7/81), 1.23%(1/81), 12.35%(10/81), 1.23%(1/81) in the 81 patients with prolonged surgical duration, versus 13.47%(61/453), 2.65%(12/453), 0.44%(2/453), 1.77%(8/453), 0.44%(2/453), 3.31%(15/453), 0, 7.28%(33/453), 1.55%(7/453) in the 453 patients without prolonged surgical duration, showing a significant difference in the overall incidence of complications (χ²=4.18, P<0.05), and no significant difference in the incidence of abdo-minal infection, anastomotic fistula, abdominal bleeding, incision‑related complications, intestinal obstruction, lymphatic fistula, liver‑related complications (P>0.05). There was no significant difference in the incidence of pulmonary infection between them (χ²=2.38,P>0.05). (4) Analysis of risk factors influencing prolonged surgical duration. Results of univariate analysis showed that BMI ≥25 kg/m², tumor located in the lower stomach, clinical T3-T4a stage, clinical N1-N3 stage were correlated factors influencing prolonged surgical duration in Da Vinci robotic radical gastrectomy for gastric cancer (odds ratio=1.88, 0.40, 6.24, 6.51, 3.08, 3.39, 17.15, 95% confidence interval as 1.03-3.42, 0.21-0.76, 1.40-27.76, 1.50-28.30, 1.43-6.60, 1.29-8.92, 4.84-60.74, P<0.05). Results of multivariate analysis showed that BMI ≥25 kg/m², clinical T3 stage, clinical N3 stage were independent risk factors influencing prolonged surgical duration in Da Vinci robotic radical gastrectomy for gastric cancer (odds ratio=2.31, 4.97, 11.08, 95% confidence interval as 1.19-4.46, 1.05-23.55, 2.72-45.13, P<0.05). (5) Construction and evaluation of an artificial neural network predictive model for pro-longed surgical duration. The BMI, tumor location, clinical T staging, and clinical N staging were incorporated into a multilayer perceptron to construct an artificial neural network predictive model for prolonged surgical duration. Results of ROC curve showed that the AUC, accuracy, sensitivity, specificity, positive predictive value, negative predictive value of the predictive model in the training dataset were 0.73 (95% confidence interval as 0.68-0.78), 91.4%, 68.1%, 94.8%, 65.3%, 95.4%. The above indicators of the predictive model in the validation dataset 0.72 (95% confidence interval as 0.65-0.79), 88.1%, 67.6%, 93.7%, 74.2%, 91.5%.

Conclusions BMI ≥25 kg/m², clinical T3 stage, clinical N3 stage are independent risk factors influencing prolonged surgical duration in Da Vinci robotic radical gastrectomy for gastric cancer. The artificial neural network predictive model con-structed based on BMI, tumor location, clinical T staging, and clinical N staging can effectively predict patients at high risk of prolonged surgical duration in Da Vinci robotic radical gastrectomy for gastric cancer.