Prediction of the aggressiveness of non-functional pancreatic neuroendocrine tumors based on the dual-tracer PET/CT

Background Predicting the aggressive behavior of non-functional pancreatic neuroendocrine tumors (NF-PNET) remains controversial. We wanted to explore, in a prospective setting, whether the diagnostic accuracy can be improved by dual-tracer functional imaging 68Ga-DOTANOC and 18F-FDG-PET/CT in patients with NF-PNETs. Methods Thirty-one patients with NF-PNET (90% asymptomatic) underwent PET-imaging with 18F-FDG and 68Ga-DOTANOC, followed by surgery (n = 20), an endoscopic ultrasonography and fine-needle biopsy (n = 2) or follow-up (n = 9). A focal activity on PET/CT greater than the background that could not be identified as physiological activity was considered to indicate tumor tissue. The imaging results were compared to histopathology. The mean follow-up time was 31.3 months. Results Thirty-one patients presented a total of 53 lesions (40 histologically confirmed) on PET/CT. Thirty patients had a 68Ga-DOTANOC-positive tumor (sensitivity 97%) and 10 patients had an 18F-FDG-positive tumor. In addition, one 68Ga-DOTANOC-negative patient was 18F-FDG-positive. 18F-FDG-PET/CT was positive in 19% (3/16) of the G1 tumors, 63% (5/8) of the G2 tumors and 1/1 of the well-differentiated G3 tumor. 68Ga-DOTANOC-PET/CT was positive in 94% of the G1 tumors, 100% of the G2 tumors and 1/1 of the well-differentiated G3 tumor. Two out of six (33%) of the patients with lymph node metastases (LN+) were 18F-FDG-positive. The 18F-FDG-PET/CT correlated with tumor Ki-67 (P = 0.021). Further, the Krenning score correlated with tumor Ki-67 (P = 0.013). 18F-FDG-positive tumors were significantly larger than the 18F-FDG-negative tumors (P = 0.012). 18F-FDG-PET/CT showed a positive predictive value of 78% in the detection of potentially aggressive tumors (G2, G3, or LN + PNETs); the negative predictive value was 69%. Conclusions 18F-FDG-PET/CT is useful to predict tumor grade but not the LN+ of NF-PNETs. Patients with 18F-FDG-avid NF-PNETs should be referred for surgery. The 68Ga-DOTANOC-PET/CT also has prognostic value since the Krenning score predicts the histopathological tumor grade. Trial registration The study has been registered at ClinicalTrials.gov; Non-functional Pancreatic NET and PET imaging, NCT02621541.


Background
Pancreatic neuroendocrine tumors (PNETs) constitute 3% of all pancreatic neoplasms and 60-80% of PNETs are defined as non-functional (NF-PNET) [1]. The incidence has increased in recent years due to the expanding use of imaging [2]. Despite the generally indolent nature, it has been recognized that the pathological potential of PNETs is highly variable and some NF-PNETs present at an advanced stage with local invasion and distant metastases [3,4]. Heterogeneity of PNETs makes therapeutic decisions difficult with no clear consensus.
The most powerful prognostic factors are the grade and distant metastases [4]. Although complete surgical resection is the only potentially curative treatment, a conservative approach seems to be safe for asymptomatic and stable sporadic NF-PNET≤ 2 cm [5,6]. However, a locally advanced disease with lymph node metastases (LN+) is rare but also possible on small (1-2 cm) NF-PNETs [7]. In a review of 136 surgical patients, Hashim and colleagues suggested a metastatic rate of 8% in PNETs as small as 1.5 cm [8]. In a retrospective analysis (n = 181), Partelli et al. [9] demonstrated that lymph node metastases decreased the 5-year disease-free survival in NF-PNETs (70% vs 97%, P < 0.001). Pancreatic resection is a high-risk operation with 20-40% morbidity and 1-2% mortality [10,11]. Knowing the risks of pancreatic surgery, it is challenging to decide between surgery and follow-up for these patients.
The aim of this study was to determine the role of dualtracer functional imaging in predicting aggressive behavior in NF-PNETs in a prospective setting with histopathological references. The hypothesis was that the higher the maximum standardized uptake value (SUV max ) of 18 F-FDG-PET/CT then the higher the Ki-67 of the tumor would be and the higher the SUV max of 68 Ga-DOTA-NOC-PET/CT the lower the Ki-67 of the tumor would be.

Study design
The study was a prospective, multicenter clinical trial at Turku and Helsinki University Hospitals in Finland.
From January 2016 to January 2018, a total of 35 patients suspected of having NF-PNET on a primary CT were prospectively imaged using 68 Ga-DOTANOC-PET/CT and 18 F-FDG-PET/CT. Four patients were excluded; one patient diagnosed with a pancreatic adenocarcinoma, another with a functional duodenal neuroendocrine tumor (gastrinoma), one with mixed neuroendocrine non-neuroendocrine neoplasm and one patient with an uncertain diagnosis lacking pathology. A total of 31 patients (age range 20-83 years; mean age 60 ± 18 years) were enrolled in the study. Patient characteristics are given in Table 1. Twenty patients were operated. Four patients underwent endoscopic ultrasonography and fine-needle biopsy (EUS-FNB), but two of these were found to be non-specific. One patient with Von Hippel-Lindau (VHL) syndrome had a pancreatic intraepithelial neoplasia lesion in the histopathological analysis and two NF-PNETs (Ø 13 and 10 mm on MRI) were followed up. Twenty-two patients had a histopathological confirmation (20 resection specimens and two EUS-FNB) and nine patients with a positive 68 Ga-DOTANOC PET/ CT finding were followed up (33.5 ± 6.2 months). Thirtyone patients had a total of 53 lesions on the PET/CT imaging. The median PET/CT imaging interval was 34 days (range 2-164, IQR 12-62 days) in 29 patients. The PET/ CT imaging interval was delayed in two patients, in 1 to 360 days due to logistical reasons and in another to 244 days due to a more urgent operation. Since there were two methods to measure chromogranin A (CgA) in use at the laboratories during the study period, the CgA was reported as three subgroups. For an accurate diagnosis of PNETs every attempt was made to establish a tissue sample by means of an operation or an EUS-FNB. Follow-up time was measured from the date of the first PET/CT scan to the review time. Patients were treated in accordance with the routine procedures of the departments and the European Guidelines [19]. The study has been registered at Clinical Trials.gov (NCT02621541). Patients underwent a whole-body PET/CT scan from the level of the skull base to the mid-thigh starting 64 ± 13 min after the injection of 68 Ga-DOTANOC and 55 ± 9 min after 18 F-FDG. The mean dose of intravenous 68 Ga-DOTANOC was 143.8 ± 17.1 MBq and 18 F-FDG was 321.9 ± 67.3 MBq. The patients fasted for 6 h before the study. Blood glucose levels were checked before any 18 F-FDG-PET/CT for patients with diabetes or previous history of glucocorticoids use (range 4.6-8.3 mmol/l). A low-dose PET/CT was followed with a whole-body diagnostic CT scan after automated intravenous injection of the contrast agent, either with 68 Ga-DOTANOC-or 18 F-FDG-PET/CT. Attenuation correction was performed using a low-dose ultrafast CT protocol (80 mAs, 140 kV, 0.3 mSv per field of view). Images were reconstructed full width at half maximum and fully three-dimensional maximum-likelihood ordered-subset expectation maximization (OSEM). Data was corrected for dead time, decay, and photon attenuation and was reconstructed to a 128 × 128 matrix.

Data analysis
The diagnostic accuracy of the PET/CT studies was assessed by comparing the PET-images and the histopathological reports (n = 22). When proper histology was not available (n=9), the consensus was based on the sum of the laboratory tests and imaging procedures. This information was used for the interpretation of the lesion analysis (Figs. 1 and 2). Data on the primary tumor site and the diameter were collected from the pathology reports and from other imaging studies (contrast CT or MRI). The histopathological analysis was blind and conducted by two experienced pathologist (J.A and J.Su). Clinical TNM and grade classification was based on the 2017 World Health Organization (WHO) classification of PNETs [5].
Only the patients and the tumors with a histopathological confirmation were included in the correlation analysis. Ki-67-labeling indexes were studied from the whole series of tumors using MIB-I antibody (Dako, Agilent Pathology Solutions, Santa Clara, CA, USA) and automated staining instrument (BenchMark Ultra, Ventana Medical Systems, Inc., Tucson, AZ, USA). All stainings were prepared in the clinical pathology laboratory (HUSLAB, Helsinki University Hospital) under standardized conditions. Stained slides were digitized with a Panoramic scanner (3DHISTECH, Budapest, Hungary) and Ki-67 values were calculated by T.V from hot spot areas comprising of at least 2000 cells. A publicly available application that was shown suitable for pancreatic NETs, was used for quantitative image analysis (Immu-noRatio) [20,21].
Analyses of PET/CT-images were interpreted by a dedicated nuclear medicine physician (J.K), with short referral information but blinded to histopathological  report. The SUV max -values were determined for every tumor or abnormal anatomical region on both 68 Ga-DOTANOC and 18 F-FDG-PET/CT. For the PET/CT-studies areas with a focal activity greater than the background that could not be identified as physiological activity were considered to indicate tumor tissue. Lesions were graded on 68 Ga-DOTANOC-PET/CT with the Krenning score, by semi-quantitatively comparing SUV max of the tumors to the reference organs such as liver and spleen. Krenning scoring was performed as follows: score 1: uptake < normal liver, 2: uptake = normal liver, 3: uptake > normal liver and 4: uptake > spleen or kidneys [22]. Further, by using the dualtracer PET/CT the NETPET-score (grades P1-5) was defined. Grade P1 indicated purely somatostatin analog avid lesion without 18 F-FDG-uptake and P5 indicated the presence of a significant 18 F-FDG-positive and somatostatin analog-negative diseas e [23]. The imaging analysis was performed using ADW 4.4 workstation.

Statistical analysis
Normally distributed variables were expressed as means and standard deviations (SD) variables not following a normal distribution as medians and interquartile ranges (IQR) and categorical variables as frequencies and proportions. The Shapiro-Wilk test was used to test deviations from a normal distribution. Spearman's rank correlation was used to test the relationship between Ki-67 and SUV max due to a lack of normally distributed data. The Mann-Whitney or the Kruskal-Wallis tests were used to discover the differences between the groups in continuous variables; Fisher's exact test was used for binary variables and the linear by linear association test for ordinal variables. A P value of < 0.05 was considered statistically significant, and two-tailed tests were used. The data analysis was performed using commercially available software (Statistical Package for Social Sciences, version 24, IBM Corp., Armonk, NY, USA).

Features of NF-PNETs and histopathological grading
The type of surgery and histopathological finding of the tumors are presented in Table 1. Patients with G1 NF-PNET were more commonly asymptomatic (92%) compared to G2 patients (75% asymptomatic), but the only patient with a G3 tumor was also asymptomatic (P = 0.595). One symptomatic patient had jaundice and two had upper abdominal pains. All the symptomatic patients had sporadic NF-PNET. Multiple endocrine neoplasia 1 (MEN1) patients (n = 7) more commonly had a G1 (71%) rather than a G2 (29%) tumor and none had a G3 tumor, but the difference was not statistically significant (P = 0.470). There was no correlation between the tumor size and the grade: G1 (34 ± 27 mm), G2 (36 ± 28 mm), and G3 (n = 1, 45 mm) (P = 0.800). However, all the operated primary tumors ≤ 2 cm (n = 6) were G1 tumors without lymph node metastases. The most common location for the G1 tumor was the tail of the pancreas (54%). Three out of eight (38%) of the G2 NF-PNETs were located in the head and the same amount; 3/8 (38%) in the tail of the pancreas. The G3 NF-PNET was located in the head of the pancreas. There was no statistical significance between the location of a tumor and the grade of a tumor (P = 0.554). There was a trend between the grade and CgA (P = 0.108).
Correlation between 68 Ga-DOTANOC-PET/CT and grade 68 Ga-DOTANOC-PET/CT was positive in 30 patients and thus the sensitivity of the 68 Ga-DOTANOC-PET/CT detecting of NF-PNETs was 97% and specificity was 100%. The only 68 Ga-DOTANOC-negative tumor (G1) was 18 F-FDG-positive and the patient had multiple (11/13) lymph node metastases. Thirty-one study patients had a total of 53 lesions, 45 of these were 68 Ga-DOTANOC-positive (32/45 histologically confirmed lesions). Eight lesions were not detected by PET/CT at all and were false negative (FN) lesions (Fig. 1); one patient had negative 68 Ga-DOTANOC-PET/CT and histologically confirmed G1 NET and seven FN lesions were detected on the same MEN1 patient who underwent a total pancreatectomy. He had a total of 13 PNETs (five G2 T1N0 tumors and eight G1 T1N0 tumors). Six of these lesions were detected on 68 Ga-DOTANOC-PET/CT and one of these tumors was also 18 F-FDG-positive. Three lesions were > 10 mm and 10 were 5-10 mm in size. There were also multiple microadenomas (NET< 5 mm). Due to several lesions, it was impossible to combine the information concerning the 13 tumors on a histopathological report and the six lesions on 68 Ga-DOTANOC-PET/CT. Only the largest primary tumor was included in the lesion analysis.

Local and distant metastases
Six patients had histopathologically confirmed lymph node metastases (mean 4.8 ± 3.4 nodes). Three LN+ were in patients with G1 NF-PNETs and three in patients with G2 NF-PNETs. All the LN+ patients were asymptomatic and sporadic. Two of these six patients with LN+ (33%) had an 18 F-FDG-positive tumor and four tumors with LN+ were 18 F-FDG-negative (67%). Thus, 18% (2/11) of 18 F-FDG-patients had LN+ vs. 20% (4/20) of 18 F-FDGnegative patients had LN+ (Fig. 5). In patients with LN+, the mean primary tumor size was 54 ± 25 mm (range 24-90 mm). In five patients, preoperative 68 Ga-DOTANOC-PET/CT did not reveal LN metastases. One patient had several LN metastases but 68 Ga-DOTANOC-PET/CT showed only one metastatic lesion. Two patients were diagnosed with a NF-PNET in the tail of the pancreas and liver metastasis (T3N1M1/G2/Ki-67 10% and T2N1M1/ G2/Ki-67 3%). Liver resection is arranged for another patient but to another no liver resection was performed due to progression of the disease.

Prognosis and PET outcome
During the mean follow-up of 31.3 ± 6.2 months, one patient died (due to long-term surgical complications). He had a 68 Ga-DOTANOC-negative and 18 F-FDG-positive G1 T3N1 tumor. No recurrence had been detected during the follow-up time. One patient with 68 Ga-DOTANOC-positive and 18 F-FDG-negative primary tumor and liver metastases underwent a laparoscopic distal pancreatectomy and splenectomy. Histopathological analysis confirmed a G2 T3N1M1 tumor. Due to the progression of the disease no liver resection was arranged and capecitabine and temozolomide combination therapy and previously Lutetium-177octreotate therapy was performed. One MEN1 patient with 68 Ga-DOTANOC-positive and 18 F-FDG-negative primary  Currently, a liver resection is planned due to the residual liver metastasis. One patient underwent robotic distal pancreatectomy and splenectomy due to two sporadic T1N0 G1 tumors. Stable residual tumor at the body of the pancreas is followed up. Among the follow-up group the VHL patient's tumor has enlarged and uptake of 18 F-FDG increased and surgery is considered. No disease progression has been detected in other patients belonging to the followup group (9 patients, 11 lesions).

Discussion
To our knowledge, this is the first prospective study evaluating the impact of combined 68 Ga-DOTANOC and 18 F-FDG-PET/CT in predicting the aggressive behavior of NF-PNET. There is some previous data of the dual-tracer imaging with series of different gastroenteropancreatic neuroendocrine tumors [18,24] and retrospective series of PNETs [17,25]. The main interest of this study was to evaluate if the malignant potential of NF-PNETs could be predicted with dual-tracer PET/CT imaging. We found that there was a statistically significant positive correlation between 18 F-FDG-uptake and the proliferation index of the tumor (P = 0.021). However, there was no negative correlation between 68 Ga-DOTANOC-uptake and the Ki-67index (P = 0.190). Both the Krenning score and NETPET score correlated statistically significantly with the Ki-67. In this patient cohort, the lower 68 Ga-DOTANOC uptake, expressed as SUV max values, was not directly associated to more aggressive phenotype unless the comparison was made to physiological organ activity by using Krenning score or combined to 18 F-FDG-imaging in NETPET score. We also studied the ability of the dual-tracer PET/CT to assess LN+ but no effect was seen in order to enable a diagnosis of a local metastatic disease. 18 F-FDG-PET/CT imaging has traditionally been used to assess adenocarcinoma and poorly-differentiated G3  showed two different uptake intensity areas around a large cystic component (SUV max 29 g/ml and 10.6 g/ml) in the pancreatic tumor. 18 F-FDG-PET/CT (c) was negative (SUV max 3.2 g/ml). She underwent a distal pancreatectomy and splenectomy and histopathological analysis revealed a G1 PNET (Ki-67 < 2%) (d, image magnification × 40) with five lymph node metastases neuroendocrine carcinomas (NECs). However, recent data has shown that 18 F-FDG-PET/CT has prognostic value in patients with well-differentiated NET, showing that high uptake is associated with increased risk of early progression [5,26]. In retrospective series the prognostic value of 18 F-FDG-PET/CT imaging was even more powerful than the grade of the tumor [15,16]. A study by Chan et al. [23] of 62 metastatic NET patients combined the results of somatostatin analogue and 18 F-FDG-PET/CT to obtain an imaging biomarker, a NETPET-score (grades P1-5), and a classification system reflecting the burden of the disease. Grade P1 indicated purely somatostatin analogue avid lesion without 18 F-FDG-uptake and P5 indicated the presence of a significant 18 F-FDG-positive and somatostatin analogue-negative disease. The NETPET-score correlated positively with the histopathological grade and offered a better prediction of overall survival. In concordance with this we found a positive correlation between the NETPET score and the proliferation index. Further there was a correlation between the SUV max of 18 F-FDG-PET/CT and the proliferation index, and this correlation maintained significance when the patients were divided into WHO grades by Ki-67. Sharma et al. [27] concluded that SUV max on 68 Ga-DOTANOC-PET/CT was an independent, positive prognostic factor in patients with well-differentiated NET in a two-year median follow-up time. However, in light of this observation, we did not find significant prognostic role for 68 Ga-DOTANOC-uptake in our study due to short followup time.
SSTR-based functional imaging with 68 Ga-labelled peptides, 68 Ga-DOTATATE, 68 Ga-DOTATOC, and 68 Ga-DOTANOC, has recently become the gold standard in the diagnosis of the NETs due to better sensitivity compared to 111 In-DTPA-octreotide (Octreoscan®) [28] and lower patient radiation and higher spatial resolution than SPECT/CT [29]. 68 Ga-DOTATATE has the strongest SSTR2-binding affinity. However, 68 Ga-DOTATOC and especially 68 Ga-DOTANOC have wider affinity profiles, including SSTR2, SSTR3 and SSTR5. The three analogs have shown no differences in clinical practice.
The sensitivity of the 68 Ga-DOTANOC PET/CT for detecting NF-PNETs is high and SSTR-based imaging is a mandatory procedure to guide treatment planning [30]. In our series, the sensitivity was 97% and the smallest 68 Ga-DOTANOC-avid and histologically confirmed tumor was 6 mm. Lesion-based analysis sensitivity was 85% (45/53). Seven FN lesions occurred in the same MEN1 patient who had a total of 13 PNETs (smallest 5-10 mm in size) in the histopathological analysis. The fact that not all of these multiple lesions of this MEN1 patient were detected with the 68 Ga-DOTANOC should not be considered a diagnostic failure.
The prognostic value of 18 F-FDG-PET/CT in functional imaging of NETs has been a target for recent research. Garin et al. [24] investigated prospectively well-differentiated metastatic NETs and reported that 18 F-FDG-positivity correlated with decreased progression free survival and overall survival. Cingarlini et al. [17] compared retrospectively G1 and G2 PNETs and concluded that 18 F-FDG-PET/CT was positive in a smaller proportion of G1 tumors (20%) compared with G2 tumors (76%). In our study 34% NF-PNET patients had a positive 18 F-FDG-PET/CT and the 18 F-FDG-PET/CT showed a PPV of 78% in the detection of potentially aggressive tumors (G2, G3, LN+ or M+ PNETs), and the NPV was 69%. 18 F-FDG-PET/CT should have some relevance when treatment options are discussed but the possibility of false imaging findings, especially false negative, should be kept in mind. In this prospective study there was a statistically significant correlation between the SUV max of 18 F-FDG-PET/CT and the Ki-67 of the primary tumor.
In the diagnostic workup of NF-PNETs, an endoscopic ultrasonography-guided fine-needle aspiration (EUS-FNA) could be an option, but it does not necessarily represent the whole tumor in terms of aggressiveness and the tissue material may be insufficient. In the French study [31], 30% of tumor grading was up-scaled on the resected tissue. Additionally, the heterogeneity in the tumor tissue may interfere the assessment of Ki-67-labeling [32]. However, in the prospective, multicenter, randomized controlled trial the EUS-FNB, which also was used in our study, produced a more accurate diagnosis than EUS-FNA [33].
Boninsegna et al. [34] found that Ki-67 > 5% and the ratio between the number of metastatic LNs above that of the examined LNs were better predictors of the recurrence than tumor size. Further, in our series there was no correlation between the tumor size and Ki-67-index, but the 18 F-FDG-positive tumors were significantly larger than the 18 F-FDG-negative tumors.
Our study results show that dual-tracer PET/CT imaging has a complementary role and it enables the detection of a potentially aggressive disease during the diagnostic workup on well-differentiated NF-PNETs. Due to this we recommend 18 F-FDG-PET/CT to be a part of a systematic diagnostic work-up of asymptomatic NF-PNETs. Especially if a consensus regarding surgical therapy cannot be reached, dual-tracer PET/CT could supplement the evaluation of the nature of the tumor. Due to our results 18 F-FDG-avid tumors should be considered for surgery if there are no contraindications. In a case of a 18 F-FDG-negative lesion follow-up could be considered if all the other aspects of the diagnostic work-up also support this treatment strategy.
The strengths of the present study were histopathological re-evaluation of all the tumors as well as the immunohistochemical Ki-67 staining and automated image analysis. The major limitation of the present study is the small number of patients, as the disease is rare. In addition, the lack of histopathological confirmation of the follow-up patients limits the analysis. For this subgroup of patients, the collection of samples was not technically or ethically feasible. However, the correlation analyses were performed with histopathologically confirmed cases. Since pancreatic neuroendocrine carcinomas are rare, the study population consists of well-differentiated PNETs, which impairs the sensitivity and specificity of 18 F-FDG-PET/CT to detect potentially aggressive tumors. In two patients the interval between PET imaging was delayed (244 and 360 d), however, due to the indolent nature of NF-PNETs the long dual PET/CT interval is likely to be of no great importance and we preferred not to exclude these patients. Another limitation is the relatively short follow-up time (mean 31.3 mo, SD 6.2 mo) which is insufficient to estimate a definitive prognosis. There is a possibility that during a longer follow-up time the 18 F-FDG-positive tumors in this study would prove to be progressive, and our intent is to analyze long time prognosis. In addition, it would be essential to detect which biological and immunohistochemical features are common for these tumors.
Conclusion SUV max measured on 18 F-FDG-PET/CT was significantly associated with the proliferation index, Ki-67, and will serve as a surrogate measure of tumor aggressiveness in patients with asymptomatic NF-PNET. Further, the 68 Ga-DOTANOC-PET/CT has notable prognostic value since the Krenning score predicts the histopathological tumor grade.
To conclude, we recommend 18 F-FDG-PET/CT to be a part of a systematic diagnostic work-up of asymptomatic NF-PNETs. This prospective study shows that dual-tracer PET/CT imaging, using 18 Ga-DOTANOC and 18 F-FDG, has a complementary role and it enables the detection of a potentially aggressive disease during the diagnostic workup on well-differentiated NF-PNETs.