The initial literature search yielded 991 potentially relevant articles, of which twenty-one unique studies involving 803 patients were included in this meta-analysis [6, 18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37]. All the selected studies were retrospective, with six [18, 19, 22, 29, 32, 33] being multicenter and the remaining fifteen [6, 20, 21, 23,24,25,26,27,28, 30, 31, 34,35,36,37] conducted at a single-center series. The reported mean age ranged from 22 to 50 years, and 62.2% of patients were females. A PRISMA flowchart of the study selection process is depicted in Supplementary Item V. The baseline characteristics of the included studies are described in Table 1.

The results of the risk of bias assessment of all studies included are shown in Supplementary Item II. Eight studies [6, 19, 22, 24, 28, 29, 31, 34] were judged to have low risk of bias. Eleven studies [20, 21, 23, 25,26,27, 32, 33, 35,36,37] were judged to have moderate risk of bias with potential confounding and selection bias due to inconsistent reporting of baseline characteristics, multicenter design, and prolonged study periods. Two studies [18, 30] were judged to have high risk of bias due to unclear inclusion criteria and lack of appropriate adjustment for confounding factors. Nonetheless, all the studies included were deemed adequate within the selection domain. The results demonstrating the certainty of evidence are summarized in a GRADE evidence table in Supplementary Table III.

Three studies [19, 34, 36] reported the preoperative body mass index (BMI) with a pooled mean of 26.3 kg/m2 (95% CI: 24.7–27.9, I2 = 54%). Fourteen studies [6, 18, 19, 22, 24,25,26, 28, 31, 33,34,35,36,37] reported the operative approach, consisting of open (69.8%, n = 369), laparoscopic (23.3%, n = 123), and robotic surgery (6.8%, n = 36). The mean preoperative ASA score was 2.1 in two studies [34, 36]. The pooled mean follow-up period was 74 months (95% CI: 65.9–83.6, I2 = 0%), with the longest follow-up period being 83 months. Clinical characteristics of our included studies are summarized in Table 2 and Fig. 1.

The pooled patient characteristics

Eleven [18, 19, 21, 24,25,26,27, 30, 33,34,35] studies reported anatomic location of the pancreatic insulinoma: head (45.0%, n = 136), body (30.8%, n = 93), tail (18.5%, n = 56), and neck (5.6%, n = 17). Ten studies [18, 19, 22, 25, 28, 29, 33, 35,36,37] reported tumor diameter, with a pooled mean diameter of 1.5 cm (95%CI: 1.3–1.6, I2 = 86%). The histological classification of insulinoma among three studies [31, 35, 36] consisted of G1 (63.0%, n = 80), G2 (36.2%, n = 46), and G3 (0.8%, n = 1). The tumor characteristics of included studies are comprehensively described in in Table 2.

Of 20 studies [6, 18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35, 37], overall postoperative morbidity after insulinoma EN was 37.3% (95%CI: 0.264–0.481, I2 = 92%, n = 277). The most commonly reported morbidities were POPF (27%, n = 176), abdominal collection (3.3%, n = 21), acute pancreatitis (1.7%, n = 11), and postoperative hemorrhage (1.7%, n = 10). Among 7 studies [6, 22, 24, 25, 30,31,32], 4 patients (1.5%) had immediate new-onset diabetes (95%CI: 0.000–0.030, I2 = 0%). Among the 6 studies [20,21,22,23, 27, 30] reporting on postoperative acute pancreatitis (POAP), 11 patients (5%) were diagnosed with POAP (95% CI: 0.020–0.074, I2 = 0%). Across 7 studies [22, 24, 30, 32, 34,35,36], 10 patients (3%) experienced postoperative hemorrhage (95% CI: 0.011–0.049, I2 = 0%). DGE was reported in 5 studies [19, 22,