Neuroendocrine tumours are a heterogeneous group of neoplasms that originate from a common precursor cell population that shares a number of antigens with nerve elements, such as neuron-specific enolase and chromogranins (11
). PNETs are uncommon tumours of the pancreas, accounting for only 1% to 2% of all primary pancreatic neoplasms. The incidence of PNETs has been estimated to be approximately 0.4 to 1.0 in 100,000 people. However, in autopsy and surgical series, up to 15% of pancreatic neoplasms have been identified as PNETs (11
Most PNETs are well- to moderately differentiated, and can be classified as those associated with a clinical syndrome caused by excessive hormone production (functional or syndromic PNETs) or those without such an association (nonfunctional or nonsyndromic PNETs) (13
). Between 70% and 85% of PNETs are functional, with insulinomas accounting for 40% to 60% and gastrinomas accounting for 20% to 30%.
PNETs vary in size, ranging from smaller than 1 cm to 5 cm or larger, and are benign in up to 40% of cases. Differentiation between benign and malignant PNETs can be difficult, although the presence of local invasion of adjacent organs or distant metastasis usually indicates malignant behaviour (14
). Even in the face of metastatic disease, the prognosis of PNETs is more favourable than the more common pancreatic adenocarcinoma. Hence, a stepwise preoperative evaluation for PNET localization is very important for potentially curative surgery.
Despite the advances in imaging modalities, up to 30% of PNETs can be missed during a preoperative assessment. The sensitivity of transabdominal ultrasound for detecting PNETs ranges from 20% to 86% and increases with tumour size (16
). Similarly, the sensitivity of nonhelical CT of the abdomen is reported to be 30% if the size of the primary tumour is between 1 cm and 3 cm, and 95% if it is larger than 3 cm, although a primary tumour smaller than 1 cm is rarely detected. The sensitivity of a CT scan can be enhanced using a multiphase and multidetector CT scanner. MRI is just as accurate as a CT scan for localizing PNETs. As with ultrasound and CT, tumour detection using MRI increases with tumour size. The overall sensitivity of MRI is between 85% and 94%, with a specificity of 78% to 100% (16
). Because 80% to 90% of neuroendocrine tumours have somatostatin receptors, octreotide scintigraphy could potentially be the initial imaging procedure of choice. However, there are pitfalls in localizing small tumours and tumours that lack somatostatin receptors (19
). Because these modalities are not accurate enough for preoperative visualization and identification of PNETs in the pancreas, which is paramount in planning the extent of surgery, intraoperative ultrasound (IOUS) has been used for direct examination of the pancreas (20
). In this study (20
), IOUS was found to localize 96% of PNETs and 58% of non-PNETs. The authors concluded that IOUS altered surgical management in 11% of gastrinomas, mainly by identifying additional gastrinomas or determining that the gastrinoma was malignant. However, this modality requires laparoscopy or laparotomy, and decisions regarding the extent of surgical intervention required cannot easily be made and discussed with patients preoperatively.
EUS enables a high-frequency ultrasound probe to be placed in close proximity to the pancreas. Compared with other imaging modalities, EUS was more accurate in detecting and localizing PNETs, especially those smaller than 2.5 cm, with an overall accuracy between 89% and 97% (21
). Varas Lorenzo et al (6
) compared preoperative EUS with transabdominal ultrasound, CT, MRI, angiography and OctreoScan (Mid-South Imaging and Therapeutics, USA) in 37 patients suspected to have gastrointestinal neuroendocrine tumours. The sensitivity and specificity of EUS was 78% and 80%, respectively. EUS detected three PNETs (all insulinomas) that were smaller than 1 cm in size, which were missed by ultrasound, CT and MRI (6
). In most cases, adding EUS-FNA to a basic EUS examination enhances the sensitivity of EUS by providing a cytological diagnosis without the risk of significant complications or the need for exploratory surgery (25
). Jani et al (9
) recently reported 41 patients with PNETs diagnosed by EUS-guided FNA. Interestingly, 85% of the tumours were nonfunctional and all of these nonfunctional tumours were discovered incidentally on CT scan. Surgical resection was performed in 78% of cases. Precise localization of the tumours in the body or the tail of the pancreas by EUS led to laparoscopic resection of these tumours in 34% of patients. Nine patients (22%) did not undergo surgery because of tumour metastasis, significant medical comorbidity or patient refusal (9
Our study has some obvious limitations. It was a single-centre, retrospective review of a relatively small number of patients. However, our results suggest that preoperative EUS had a significant impact on the possible surgical management of 50% of patients and on definite surgical management in 36% of patients suspected of having PNETs. Patients with PNETs identified by EUS were then scheduled for surgery, and those who were found to have multiple and multifocal lesions underwent total pancreatectomy, instead of distal pancreatectomy, which was planned originally based on standard radiographic imaging. We currently do not routinely perform EUS-guided FNA in PNETs. This procedure is indicated only for those patients in whom a neuroendocrine tumour or syndrome, such as MEN1, is in doubt or when a CT scan has failed to localize a mass but one is suspected based on biochemical abnormalities.
Our results suggest that EUS with or without FNA is a very helpful diagnostic tool in the preoperative assessment of patients suspected of having PNETs, particularly in patients at risk for multifocal disease such as those with MEN1. The information obtained is important for surgeons to plan surgery in advance and preoperatively discuss the appropriate procedure with the patients.