¹⁸F-Fluorodeoxyglucose Positron Emission Tomography Predicts Survival of Patients with Neuroendocrine Tumors

Surgical resection is the only treatment that is able to cure patients with neuroendocrine (NE) tumors. However, for patients with unresectable disseminated disease, the treatment strategy and clinical outcome varies considerably. Based on the suspected aggressiveness of the disease, patients are given different treatment regimens. The medical approach is based on either symptomatic treatment with somatostatin analogues or α-IFN, which is preferred for the least aggressive tumors or a more aggressive approach for the rapidly proliferating tumors, generally using systemic chemotherapy. For patients with disseminated disease and a positive somatostatin receptor scintigraphy (SRS), peptide receptor radionuclide therapy based on somatostatin analogues labeled with a β-emitting isotope for local irradiation of the tumor is also an option for treatment. Furthermore, new therapy options under clinical evaluation, which are targeting angiogenesis, tyrosine kinases, or serine threonine kinases, seem promising for the treatment of NE tumors (1–3).

For the early selection of patients who will benefit from more aggressive treatment, validated prognostic factors are crucial. Currently, patients are selected based on the presence of metastases, elevated plasma chromogranin A (CgA) levels, localization of the primary tumor, and the high proliferation rate of the tumor as determined by immunohistochemical staining for Ki67. For staging of disease, determination of the proliferation rate is mandatory. However, in surgically unresectable patients, biopsy of the tumor is required for the determination of the Ki67 index, which is not always a trivial procedure, and the tissue obtained is not necessarily representative of the whole tumor in terms of aggressiveness.

Noninvasive functional imaging with ¹⁸F-fluorodeoxyglucose positron emission tomography (FDG-PET) seems promising as an alternative to tissue sampling for the determination of the aggressiveness of tumors (4) and has prognostic value in several cancer forms. Accordingly, it is now used in patients with pulmonary cancer (5–7), Hodgkin's lymphoma (8, 9), head and neck cancer (10), and colorectal cancer (11–13) for diagnosis; monitoring of treatment; and as a prognostic tool. However, FDG-PET is currently not used for NE tumor imaging on a routine basis because these tumors are generally slow growing and, accordingly, have low glycolytic activity compared with many other malignancies that therefore may lead to a low sensitivity for FDG-PET. Furthermore, the current gold standard for functional imaging, In-111-diethylenetriaminepentaacetic acid-octreotide, has a high sensitivity and specificity for visualization of NE tumors. Studies on small groups of patients with NE tumors suggest that FDG-PET may be of value for the imaging of the more aggressive tumors with high Ki67 indices (14, 15). The anticipated variable uptake of FDG in NE tumors may reflect differences in aggressiveness and, thus, FDG-PET may hold prognostic information for these tumors. To our knowledge, the prognostic potential of FDG-PET for NE tumors has never been evaluated.

The aim of the present study was, therefore, to investigate the prognostic value of FDG-PET in NE tumors. To do so, we performed a prospective study of FDG-PET and overall survival (OS) in a cohort of 98 consecutive patients with NE tumors.

From May 2007 to June 2008, 98 consecutive patients with NE tumors (47 men and 51 women; mean age, 60 y; range, 21-81 y) were prospectively enrolled in the study at the Department of Surgical Gastroenterology C, Copenhagen University Hospital, Rigshospitalet, which is a tertiary referral hospital and center for treatment of patients with NE tumors. The follow-up continued until February 2009. Inclusion criteria were (a) histologically verified NE tumors of gastroenteropancreatic origin or lung origin (typical and atypical bronchial carcinoids) and (b) the presence of primary, residual, or recurrent disease (primary tumor, metastases, or both) upon enrollment in the study.

The study population consisted of 45 intestinal carcinoid tumors, 30 functioning and nonfunctioning pancreatico-duodenal NE tumors, 8 typical or atypical broncho-pulmonary carcinoid tumors, 6 colonic NE tumors, and 2 poorly differentiated NE tumors with other primary localization (bile duct and stomach). In addition, seven patients had liver metastases with an unknown origin of the primary NE tumor. Patient characteristics are shown in Table 1.

Written informed consent was obtained from all participants and the study had been approved by the regional scientific ethical committee.

All aspects of patient care and treatment were carried out at the discretion of the treating clinicians and according to routine procedures of the department. Treating clinicians were blinded to the FDG-PET results and the FDG-PET results were not used for treatment decisions. The treatment given during follow-up is summarized in Table 1. Patients were treated in accordance with European guidelines for NE tumors (16). Patients with hormonal symptoms and diarrhea received treatment with somatostatin analogues, and depending on proliferation index, tumor origin, and treatment tolerance, patients were treated either with α-IFN (least aggressive) or systemic chemotherapy (more aggressive). Patients with disseminated disease, normal kidney function, and a positive SRS were treated with peptide receptor radionuclide therapy. Patients with liver metastases were in selected cases treated with radiofrequency ablation or liver embolization.

As part of the routine follow-up, half-yearly or yearly, patients were followed by plasma CgA measurements and diagnostic computed tomography (CT) of the abdomen. Relevant CgA measurements were available for 97 patients and relevant CT images were available for 94 patients. These CT images were used for the analysis of the prognostic value of CT-verified liver metastases. In addition, the routine CT images were used for the evaluation of progression-free survival (PFS) according to the Response Evaluation Criteria in Solid Tumors criteria (17). Tumor burden was evaluated by an expert in radiology as part of the clinical routine and the CT image closest to and before the FDG-PET scan was used as baseline. Time to progression was calculated as the time from the FDG-PET scan until tumor progression.

Ki67 index was determined for patients whenever possible as part of the clinical routine either on resected tumors or from biopsies. Formalin-fixed, paraffin-embedded tissue samples were cut at 4-μm thick sections and were mounted on coated slides. Antigens were retrieved with DAKO Target Retrieval Solution, High pH for 20 min at 97°C (Code K8002) using the DAKO pretreatment link (DAKO,). After blocking the endogenous peroxidase activity with DAKO EnVision Flex+ (DAKO K8002) for 5 min, tissue sections were incubated with Monoclonal Mouse Anti-human Ki67 Antigen, DAKO Code M7240, at a dilution of 1:200 for 20 min at room temperature.

The reaction was visualized by using DAKO EnVision Flex+ mouselink for 15 min followed by DAKO EnVision Flex+/horseradish peroxidase for 20 min and finally DAKO EnVision Flex+ diaminobenzidine for 10 min. The sections were counterstained with hematoxylin for 1 min. A section of the human tonsil was used as a positive control. Reading of slides was done by counting positive tumor nuclei per 100 tumor cells.

PET/CT images were acquired 1 h postinjection of 342-467 MBq F-18-FDG. Blood glucose was measured before the FDG injection, and for all patients, the level was ≤8 mmol/l.

The PET/CT scans were done either by a GE Discovery LS PET/CT scanner or a Siemens Biograph 16 PET/CT scanner. The scanners used in the study were calibrated on a regular basis to ensure similar standardized uptake value (SUV) values on both scanners. Emission scan time was 3 min per bed position. The CT scans were done as low-dose CT scans with 10 mA for minimization of the radiation burden. The CT data were used for attenuation correction of the PET data. The PET and low-dose CT images were reconstructed in all three planes and were fused and analyzed on the GE eNTEGRA PET workstation and Siemens Leonardo workstation, respectively. The low-dose CT images were used as anatomic guides for the localization of the pathologic foci.

All patients were instructed to fast for at least 6 h before the F-18-FDG injection. All images were analyzed by an expert in nuclear medicine who also quantified the tracer accumulation [expressed as maximum SUV (SUVmax) calculated as radioactive concentration in target tissue/(injected dose/body weight)] in the primary tumor and the metastasis (if any) with highest FDG uptake in the PET-positive scans. The SUVmax was not calculated in PET scans without pathologic foci and these patients were included in the survival analyses with the SUVmax value zero.

The sample size was determined on basis of an ability to detect a hazard ratio (HR) of 3 or more with a significance level of 0.05, and a power of 0.8 with an assumed median time to death in the group with highest mortality of 30 mo. With 100 patients accrued over 12 mo, this would need a follow-up interval of 21 mo.

For the study of the prognostic value of FDG-PET, PFS, and OS were chosen as end points. PFS was defined as the time from FDG-PET acquisition to progression or disease-related death. Overall survival was defined as the time from FDG-PET acquisition to death by any cause. In one case, the patient committed suicide and was censored from the PFS analysis at the time of death but was maintained in the analysis of OS. There were no other deaths unrelated to the NE tumor diagnosis.

Survival probability and PFS was estimated by the method of Kaplan and Meier, and significance was tested by the log-rank test. The assumption of proportional hazard was tested by means of plotting partial residuals (Schoenfeld residuals) against time. The plots were inspected visually following line fitting using the locally weighted scatterplot smoothing (LOESS) iterative algorithm. The multivariate analyzes were carried out using the Cox proportional hazard regression analysis.

The Ki67 index was categorized according to the WHO 2000 criteria, in which NE tumors are divided into well-differentiated endocrine tumors (Ki67 < 2%), well-differentiated endocrine carcinomas (Ki67 between 2% and 15%), and poorly differentiated endocrine carcinomas (Ki67 > 15%; ref. 18). CgA levels were dichotomized by the median value to minimize the likelihood of inadvertent bias, which could be the result of data-determined cutoff values. For 2 × 2 tables, a χ² test was used. All data analyses were carried out using the SPSS version 16.0 (SPSS, Inc.). P < 0.05 was considered significant.

A total of 98 consecutive patients with histologically verified NE tumors were enrolled in the study and all had FDG-PET imaging performed. The mean follow up time after the FDG-PET scan was 11.5 months (range, 4 days to 21.1 month). One patient, with a newly diagnosed pancreatic NE tumor (glucagonom), a Ki67 proliferation index of 100%, and progressive disease died 4 days after the FDG-PET scan due to his NE tumor. Fourteen patients had died at the end of the follow-up period. Of the 14 deaths in the survival analysis, 13 were FDG-PET positive and 1 was FDG-PET negative (small intestinal carcinoid; χ²: P = 0.007). The mortality group consisted of three small intestinal carcinoids, five pancreatic NE tumors, one bronchial carcinoid, and four colonic NE tumors. During the follow-up period, 12 patients had surgery.

Of the 98 included patients, FDG-PET was positive in 58% (n = 57) and negative in 42% (n = 41) of the cases. The overall risk of death was significantly higher in the FDG-positive group than the FDG-negative group with a HR of 10.3 [95% confidence interval (CI), 1.3-78.7]. The Kaplan-Meier curve of FDG-PET is shown in Fig. 1A (log rank P = 0.005. For comparison, PFS for the FDG-PET–positive group and the FDG-PET–negative group is shown in Fig. 1B. Patients in the FDG-PET–positive group had a significantly lower PFS compared with the FDG-PET–negative group with a HR of 9.4 (95% CI, 2.9-30.8; log rank: P < 0.001). Quantification of the tracer accumulation calculated as SUVmax revealed a higher risk of death for patients with a SUVmax above 9 with a HR of 8.8 (95% CI, 2.7-28.7). The Kaplan-Meier curve is shown in Fig. 2A (log rank P ≤ 0.001). The analysis of PFS revealed a lower PFS for patients with SUVmax above 3 with a HR of 8.9 (95% CI, 2.7-29.5; P < 0.001). The results of the other cutoff values of SUVmax for overall and PFS are listed in Table 2.

The Ki67 index was available in 88 patients (12 of 13 in the mortality group), of which 43 were derived from resected specimens and 45 from biopsy. The median (interquartile range) time interval from Ki67 index examination to the FDG-PET scanning was 7.7 (range, 1.1-47.9) months. Using the proliferation stratification values of 2% and 15% according to the WHO criteria (18), there was a significantly higher risk of death with increasing Ki67 index with a HR of 2.6 (95% CI, 1.3-5.1). A Kaplan-Meier curve with the OS based on Ki67 categorization according to the WHO criteria is shown in Fig. 3A (log rank P < 0.001). Using a cutoff value of 15% for the Ki67 index, there was a significantly higher risk of death for patients with a Ki67 index above 15% with a HR of 7.4 (95% CI, 2.5-22.1; log rank P < 0.001). Additionally, PFS decreased with higher Ki67 proliferation index with a HR of 2.7 (95% CI, 1.7-4.4; P < 0.001; Fig. 3B).

Analysis of the FDG-PET results according to the WHO classification revealed that 19 of 47 patients (40%) with a proliferation index below 2% were FDG positive, 19 of 27 patients (70%) with proliferation index between 2% and 15% were FDG positive, and 13 of 14 patients (93%) with proliferation index above 15% were FDG-PET positive.

In 94 patients, diagnostic CT imaging of the abdomen was available. Liver metastases were detected in 62 of 94 patients, but the survival analysis revealed no difference in the mortality rate for patients with and without CT-verified liver metastases (log rank P = 0.255; Fig. 4A). Additionally, when using the number of liver metastases (≥5 versus <5), no difference in survival was found in our material. The analysis of PFS supported the analysis of OS and did not reveal any significant difference in PFS between patients with or without CT-verified liver metastases (log rank P = 0.406; Fig. 4A).

Plasma CgA at the time of the FDG-PET scan was available for 97 patients. The median plasma CgA level was 460 pmol/L for all patients. In the univariate analysis, there was no significant difference in mortality for plasma CgA levels above or below 460 pmol/L (HR, 2.0; P = 0.192; 95% CI, 0.7-6.1). Other cutoff values of CgA up to 5,000 pmol/L were also tested, but there were no significant differences in mortality regardless of the cutoff value. In addition, there was no significant difference in PFS between patients with CgA levels above or below 460 pmol/L (HR, 1.6; P = 0.154; 95% CI, 0.8-3.3).

Due to the few events, multivariate analysis could not be applied for the OS. In a multivariate Cox proportional hazard regression analysis applied to PFS (n = 84; 28 events), with a median CgA level of 460 pmol/L, Ki67 index categorized according to the WHO criteria (<2%, between 2% and 15%, and >15%), and FDG-PET result entered in the model, FDG-PET result and Ki67 index remained statistically significant predictors of PFS with HRs of 1.8 and 6.0 (95% CI, 1.7-21.1 and 1.04-3.0; P = 0.005 and P = 0.035, respectively). When SUVmax with a cutoff value of 3 replaced FDG-PET result in the model, only SUVmax remained as a statistically significant predictor of PFS with a HR of 8.4 (95% CI, 2.5-27.9; P = 0.001.

The variable aggressiveness and clinical outcome of NE tumors make the selection of an optimal treatment strategy challenging. Good and validated prognostic factors are needed to predict the aggressiveness of the tumor and thereby the optimal treatment for each patient. With the results of this prospective study we show that there is a strong prognostic value of FDG-PET for prediction of the OS of patients with NE tumors in which a HR of 10 for risk-of-death was found for patients with FDG-PET–positive foci compared with FDG-PET–negative patients.

In contrast to other and generally more aggressive cancer forms, in which FDG-PET is gaining widespread use for staging of disease, for treatment monitoring, and as a prognostic tool, the imaging modality of FDG-PET is at present not used clinically for NE tumors. The assumption has been that due to the low proliferation rate and low metabolic activity generally seen in these tumors, FDG-PET would have a low sensitivity and not provide additional information to conventional CT and SRS (15, 19–22). Studies on small groups of patients have confirmed this hypothesis, although FDG-PET tended to be positive for the low-differentiated NE tumors (14, 19). The low diagnostic sensitivity of FDG-PET for NE tumors was confirmed in the present study in which we found an overall sensitivity of 58%. In comparison, SRS is known to have a sensitivity of ∼90% for NE tumor diagnostics (23–25).

Despite a high diagnostic sensitivity of SRS, the imaging modality gives limited information about the aggressiveness of the NE tumor, which can vary considerably. With this study, we have shown that although the diagnostic sensitivity of FDG-PET is low for NE tumors, the prognostic value is high. This makes the low diagnostic sensitivity less important because a negative FDG-PET result is predictive of low aggressiveness and a high survival rate (6, 7, 9). This study shows, both in univariate analysis and multivariate analysis, that FDG-PET, both in terms of positive/negative and quantified by SUVmax, was an independent prognostic factor for the prediction of survival for NE tumor patients, exceeding the prognostic value of traditionally used parameters such as Ki67, CgA level, and the presence of liver metastases. The analysis of PFS confirmed the survival data with a statistically significant difference in PFS between the FDG-positive and the FDG-negative group.

In this study, five of the tumors in the mortality group (38%) had a proliferation index below 2% and four of these had FDG-PET–positive foci. FDG-PET is considered to have a limited value in particular for slowly proliferating NE tumors with corresponding lower metabolic activity. In accordance with this, we found that 40% of patients with proliferation index below 2% were FDG positive compared with 70% of patients with proliferation index between 2% and 15%, and 93% of patients with proliferation index above 15%. However, this study shows that even for patients with a low Ki67 proliferation index, FDG-PET may be positive and indicate poor survival. Because the proliferative activity is currently determined through tissue collected by tumor resection or ultrasonographically guided needle biopsy, repeated measurements over time is not practically feasible and the proliferation index will only be determined on a randomly selected part of a tumor not necessarily representative of the patient's whole tumor burden. In contrast, FDG-PET is a noninvasive, whole-body imaging technique visualizing all aggressive tumors and metastases simultaneously. The discrepancy between a positive FDG-PET scan and a low proliferation index for four of the cases in the mortality group could be due to the dedifferentiation of the tumor from the time of Ki67 evaluation to the FDG-PET scan. However, in the four cases in the mortality group with positive FDG-PET scans and low proliferation index, the median time from Ki67 evaluation to FDG-PET examination was 4.9 months, which was shorter than the median time interval of 8.3 months for the rest of the patients.

Despite the high OS for patients with NE tumors, the cancer often metastasize to the liver over time, and the 5-year survival rate for patients with liver metastases has been reported to be between 50% to 85% depending on treatment, with the presence of liver metastases being associated with decreased 5- and 10-year survival rates (26–29). Whereas the diagnostic sensitivity of CgA is well validated (30), the prognostic value of CgA for prediction of survival in a cross-sectional measurement is less evident, despite some findings to the contrary (31–33). In this study, with a follow-up of ∼1 year, we did not find a significantly higher mortality or lower PFS for the patients with CT-verified liver metastases or for patients with markedly elevated plasma CgA. The relatively short follow-up in this study may partly explain this, and in this respect, it seems that FDG-PET at a much earlier stage can better predict survival than the presence of liver metastases or elevated CgA levels. For a validation of these findings, longer follow-up is however needed because it is possible that it may not be applicable to all patients with NE tumors and the survival data may differ with a longer follow-up.

Patients with NE tumors remain a very heterogeneous group, both in terms of symptoms, clinical course, treatment strategy, and OS. The fact that FDG-PET overall predicts survival for NE tumors points to the high clinical effect this imaging modality could have in the future selection of appropriate treatment for each patient. Perhaps FDG-PET should be included in the follow-up and initial staging in this patient group to reveal high aggressiveness at an earlier stage and thereby be able to initiate a correspondingly more aggressive treatment approach for the most malignant NE tumors. FDG-PET may very well be the desired prognostic factor, which has been lacking for NE tumor staging. Because Ki67 index, as a marker of proliferation rate in the tumor, also predicted survival in the univariate analysis and due to the central role of Ki67 evaluation in the treatment planning of NE tumors, it is possible that the proliferation tracer F-18-fluoro-thymidine could be a useful tracer for staging of NE tumors and for treatment evaluation. Whether FLT-PET adds information or performs better than FDG-PET has yet to be determined in a comparative study in this patient population.

In conclusion, this study for the first time shows a strong prognostic value of FDG-PET for NE tumors, which exceeds the prognostic value of traditional markers such as Ki67, CgA, and liver metastases. Despite the convincing results, further studies are needed with longer follow-up for validation of these findings. If confirmatory, we suggest that FDG-PET may become an important routine-imaging modality for NE tumors.

No potential conflicts of interest were disclosed.

Grant Support: Danish Cancer Society (A. Kjaer and T. Binderup), the Danish Medical Research Council, the Ministry of the Interior and Health, the National Advanced Technology Foundation, the John and Birthe Meyer Foundation, the A P Møller Foundation, the Lundbeck Foundation, the Novo Nordisk Foundation and the Svend Andersen Foundation. (A. Kjaer).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.