Risk factors affecting prognosis in metachronous liver metastases from WHO classification G1 and G2 gastroenteropancreatic neuroendocrine tumors after initial R0 surgical resection

Patients with hepatic metastatic GEP-NETs who were treated at Zhongshan Hospital, Fudan University (Shanghai, China) from August 2003 to July 2014 were consecutively enrolled in the study. Informed consent was obtained by all the patients. The study protocol followed the ethical guidelines of the Declaration of Helsinki and was approved by the Ethical Committee of Zhongshan Hospital of Fudan University. All the recruited patients underwent radical resection of primary lesion(s). Patients were excluded if any of following conditions were present: presence of distant metastasis at the initial diagnosis or at less than 6 months after the initial diagnosis [13], synchronous malignancy in other organs, removal of the tumor with endoscopic techniques, and incomplete medical records.

The diagnoses for the primary and liver metastatic lesions were based on pathological morphology and immunohistochemical assessment of a surgical specimen or of an intraoperative biopsy or needle biopsy performed by experienced pathologists, according to the site of origin and WHO criteria. Tumor grade was assigned according to the latest European Neuroendocrine Tumor Society classification guidelines for NET grading and staging [14]. The pathological characteristics that were recorded as present or absent in the tumors were tumor number, tumor necrosis and texture, and lymph-node (during the resection of primary lesion) and neural invasion (metastasis). The Ki-67 index and mitotic count results were obtained from the pathological examination records for primary and metastatic lesions; the examinations were performed on formalin-fixed, paraffin-embedded tissues. The Ki-67 proliferative index was expressed as the percentage of Ki-67 positive cells in 2000 tumor cells within areas of the highest immunostaining using the MIB1 antibody (Dako, Denmark). The increase in value of the Ki-67 indices between the primary and metastatic lesions was also recorded and was regarded as an elevation in the Ki-67 index [15]. Histopathology slides were stained using the Bond-Max Leica autostainer (Leica Biosystems, United Kingdom). Antibody detection was performed using the biotin-free Bond Polymer Refined Detection System (DS9800; Leica Microsystems).

The first follow-up was performed within 2 months after surgery. Subsequent follow-up cycles ranged from 3 to 4 months or shorter periods of time depending on clinical stage and whether tumor relapse or metastasis occurred. The treatment modalities for liver metastases depended on the liver nodule patterns. For the surgically resectable nodules, surgical resection and RFA were the first treatment choices. TACE was used for patients with unresectable nodules and confirmed liver metastasis. Systemic therapy was also used for patients with liver metastasis only and concurrent metastasis in other organs [16]. The overall survival (OS) and metastasis time data were also recorded. OS duration was calculated from the date of diagnosis of GEP-NETs until tumor-specific death or was censored at the patient’s last follow-up visit, metastatic survival was estimated from the date of hepatic metastasis to tumor-specific death and the last follow-up [17]. The time to relapse (TTR) duration was the interval from the date of the primary site operation until the diagnosis of liver metastasis [18]. Follow-up management was performed via outpatient clinics and telephone interviews. The interval of imaging examination such as CT or MRI was 6 months. The duration of follow-up was estimated from the date of diagnosis of liver metastasis to the patients last follow-up date.

A predictive nomogram was formulated based on the results of a multivariable analysis and using the package ‘rms’ in R version 3.3.0 (http://​www.​r-project.​org/​). Validation and discrimination of the nomogram was determined using the Harrell concordance index (C-index) as an index of model performance. Higher C-index values indicated better discrimination. A value of 0.5 defined no predictive discrimination; a value of 1.0 defined perfect discrimination of individuals with different outcomes.

Table 1 presents the results for the baseline characteristics for the entire population. A total of 108 patients were enrolled in the study. The median age was 54.0 years (range, 36–81 years); there were 57 (60.6%) male patients. All patients were diagnosed with GEP-NETs. There were 17 patients with gastric NETs, 4 with duodenal NETs, 3 with small intestinal NETs, 5 with ascending colon NETs, 24 patients with rectum NETs in the rectum, and 55 patients with pancreatic NETs (pNETs). All the metastatic diseases were metachronous. All of the recruited patients experienced recurrent or metastatic disease, or both, after R0 resection of primary lesion. The results for the WHO grade classification for the primary site indicated that 16 (14.8%) patients had grade-1 disease and 92 (85.2%) patients had grade-2 disease. Besides, only 5 (4.6%) NETs patients was found coexisted extra-hepatic metastasis during the postoperative follow-up, including 3 patients of lung metastasis and 2 patients of bone metastasis.

All the patients underwent R0 resection (R0 resection was characterized as the resection performed with curative intent when no evidence of metastasis was identified) of the primary tumor. The types of surgical resections performed to remove the primary tumors were pancreatico-duodenectomy (27 patients), distal pancreatectomy (25 patients), resection of the rectum (24 patients), partial gastrectomy (13 patients), right colectomy (5 patients), and total gastrectomy (4 patients).

For treatment of hepatic metastatic disease, 48 patients received liver-directed local treatment including TACE, RFA and metastasectomy, 15 patients received systemic treatment only (i.g., IFN, somatostatin analogs, and chemotherapy), and 45 patients received both systemic and local treatment. Sixty-one (56.5%) patients underwent surgical resection of metastatic lesion(s). Among the 108 patients, 65 (60.2%) patients received TACE treatment and 55 (50.9%) underwent surgical resection followed by TACE treatment; this group included 30 patients with gastrointestinal NETs (GI-NETs) and 25 patients with pNETs; 32 (29.6%) patients received RFA therapy; this group included 22 patients with GI-NETs and 10 patients with pNETs. Medical treatment was generally based on the use of somatostatin analogs, which were administered in 15 patients with pNETs and 24 patients with GI-NETs. Systemic chemotherapy and molecular targeted treatment were performed in 35 patients and 37 patients, respectively. Seven patients with pNETs were treated with everolimus; two were treated with sunitinib.

The median follow-up duration was 36.8 months (range, 22.0–145.2 months, 95% CI: 30.2–50.6 months). Survival duration from diagnosis of GEP-NETs disease to tumor-specific death was 42.6 ± 38.1 months. Figure 1a presents the results for total duration of OS time for the 108 patients. The median survival duration was 5.15 years. Table 2 presents the results for metastatic survival duration according to clinicopathological characteristics. Survival was not significantly different by gender (P = 0.287), age (P = 0.818), primary tumor origin (P = 0.273), or longest hepatic tumor diameter (P = 0.343). OS was associated with number of liver metastases (P < 0.001) and lymph-node metastasis (P = 0.035). OS was also associated with primary tumor grade (Fig. 1b, grade-1 vs grade-2, 5.82 years versus 2.33 years, P = 0.017). Treatment modality (systemic treatment only, vs local treatment only vs combined treatments) was associated with OS (Fig. 1c, P = 0.009).

The median TTR duration for the total population was 28 months (95% CI, 24.1–35.8 months). The difference in TTR duration for gender (P = 0.514) and for tumor origin (P = 0.178) was not statistically significant. The association between WHO grade and TTR duration was statistically significant (Median survival duration: 2.83 years versus 1.86 years; P = 0.012; Fig. 1d).

Samples of the hepatic lesions and related primary tumor tissues were taken from the excised specimen for analysis. Immunohistochemical staining was performed on primary tumor tissue samples from 92 patients. The Ki-67 index values were obtained from pathological examination of both the hepatic and related primary tumor tissues. The median ki-67 value of metastatic and primary was 22.8 and 5.9%, respectively. The positive rates of synaptophysin (Syn), neuron-specific enolase (NSE), and chromogranin A (CgA) were 72.3, 37.7, and 81.5%, respectively. The results for a comparison of immunohistochemical results for the primary site specimens between the GI-NETs group and pNETs group patients were summarized. The expressions of NSE and CgA were different between the two cohorts (P < 0.001 and P = 0.002, respectively). Our results also indicated that positive expression of CgA, NSE, or Syn in the primary tumor was not associated with OS time (P > 0.05).

The results for the univariate and multivariable analyses of metastatic survival are presented in Table 3. The multivariable analysis of OS duration revealed that number of liver metastases (P < 0.001, hazard ratio [HR] = 0.07; 95% confidence interval [CI], 0.023–0.214), treatment modality (P = 0.045, HR = 0.801; 95% CI, 0.684–0.187, and HR = 0.922; 95%CI: 0.781–0.998), and elevation of the Ki-67 index (P = 0.027, HR = 1.396; 95% CI, 1.174–3.483) were statistically significant independent risk indicators for patient OS.

The median Ki-67 index value was 5.0 (95% CI, 4.8–6.9) for the primary site and 20.0 (95% CI, 19.1–30.2) for the metastatic masses. Elevation of the Ki-67 index between the hepatic mass and the original mass was found in 57 (60.6%) patients. The GI-NET patient group had a higher elevated Ki-67 index incidence than the pNET patient group (41/53 vs 16/55, respectively, P = 0.015). The median OS time for the group of patients with no Ki-67 elevation was significantly longer compared with the group of patients with an elevated Ki-67 index (4.50 years vs 1.92 years, respectively, P = 0.022, Fig. 2).

An accurate prognostic nomogram consisting of all significant independent prognostic factors (i.e., hepatic tumor number, treatment modality, and elevation of the Ki-67 index) for OS was constructed according to multivariable cox regression model results (Fig. 3). The C-index for OS prediction using the formulated nomogram was 0.63. A higher C-index value indicated better predictive accuracy for OS. The calibration plot for the probability of OS rates at 1-, 2-, and 3- years in our patients revealed good consistency between the observed OS and the nomogram-predicted OS (Fig. 4).