Prognostic Implications of 18-FDG Positron Emission Tomography/Computed Tomography in Resectable Pancreatic Cancer

There are currently no known preoperative factors for determining the prognosis in pancreatic cancer. The aim of this study was to examine the role of 18-fluorodeoxyglucose (18-FDG) positron emission tomography/computed tomography (18-FDG-PET/CT) as a prognostic factor for patients with resectable pancreatic cancer. Data were obtained from a retrospective analysis of patients who had a preoperative PET scan and then underwent pancreatic resection from January 2007 to December 2015. The maximum standardized uptake value (SUVmax) of 18-FDG-PET/CT was calculated. Patients were divided into high (>3.65) and low (≤3.65) SUVmax groups, and compared in terms of their TNM classification (Union for International Cancer Contro classification), pathological grade, surgical treatment, state of resection margins, lymph node involvement, age, sex, diabetes and serum Carbohydrate Antigen 19-9 (CA 19-9) levels. The study involved 144 patients, 82 with high SUVmax pancreatic cancer and 62 with low SUVmax disease. The two groups’ disease-free and overall survival rates were significantly influenced by tumor stage, lymph node involvement, pathological grade, resection margins and SUVmax. Patients with an SUVmax ≤ 3.65 had a significantly better survival than those with SUVmax > 3.65 (p < 0.001). The same variables were independent predictors of survival on multivariate analysis. The SUVmax calculated with 18-FDG-PET/CT is an important prognostic factor for patients with pancreatic cancer, and may be useful in decisions concerning patients’ therapeutic management.


Introduction
Pancreatic cancer is only the 12th most common cancer worldwide, but it is the 7th most common cause of cancer-related death [1]. The number of new cases of pancreatic cancer will continue to rise in future, largely due to population aging and growth. In the United States, pancreatic cancer was the second most common gastrointestinal malignancy in 2018 [2]. In the European Union (EU), it was estimated that deaths from pancreatic cancer surpassed those due to breast cancer in 2017, making the disease the third most important cause of cancer-related death in the EU, after lung and colorectal cancer [3]. The prognosis for pancreatic cancer is generally poor, with five-year pylorus-preserving pancreaticoduodenectomy (PD) for tumors of the head of the pancreas, and distal pancreatectomy with splenectomy for tumors of the body and tail. Total pancreatectomy was reserved for cases where the resection margin of the pancreas was involved by the tumor or when a pancreatic anastomosis was judged at high risk of leakage. All patients underwent standard lymph node dissection-5, 6, 8a, 12b1, 12b2, 12c, 13a, 13b, 14a and 14b right lateral side, 17a, 17b, [26] and para-aortic node sampling for pancreatic head carcinoma, and 8a, 9,10,11,18 for patients with pancreatic body and tail cancers. Para-aortic nodes were excised by harvesting the lymphocellular aortocaval tissue located below the left renal vein up to the origin of the inferior mesenteric artery (station 16b1). Resections were defined as curative (R0) when the pathology report confirmed negative resection margins, or R1 in the presence of tumor ≤ 1 mm from the resection margins, according to Leeds criteria [27]. Tumors were staged according to the Union for International Cancer Control (UICC)TNM classification [28]. Each patient's clinical and pathological records were reviewed, and the following characteristics were included in our analysis-age, sex, diabetes, type of surgery, preoperative serum CA 19-9 levels (RIA, Centocor Inc., Malvern, PA, reference: < 37 kU/L), tumor stage, lymph node status, pathological grade, R0 resection, disease-free survival and overall survival. Disease-free survival (DFS) was measured from the date of surgery to the date of radiologically detected recurrence or censoring. Overall survival (OS) was measured from the date of surgery to the date of death or censoring. All patients underwent regular follow-up, which included a physical examination, abdominal CT or US, and tumor marker assay every 3 months for the first 2 years, and every 6 months thereafter. Adjuvant gemcitabine-based chemotherapy was scheduled for all patients, whenever applicable.
Ethical approval: All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. Each scan was performed 50-70 min postinjection of 150-400 MBq of FDG in fasting patients (almost 6 hours), with serum levels of glucose < 110 mg/dL for nondiabetic patients and < 200 mg/dL in diabetic ones; in order to avoid interferences due to hyperglycaemia, blood glucose was checked just before the procedure.

18-FDG-PET/CT Imaging
The acquisitions were performed with standard modalities (scan length from base skull to 1/3 prox of legs, 6-7 beds, 2-3 min/bed); when necessary, a limited second scan of 2 beds with the same modalities was repeated on the hepato-pancreatic region at 90-100 min postinjection.
Images were reconstructed with standard algorithms, and the SUV value was calculated in the suspected neoplastic foci (SUV = tissue tracer concentration/injected dose/body weight); for the SUV analysis, a circular region of interest was placed over the area of maximal focal FDG uptake suspected to be a tumorous focus (SUVmax).
After acquisition, scan images were interpreted and referred by an experienced Nuclear Medicine physician, well-trained in PET/TC (almost five years).

Statistical Analyses
Statistical analyses were run using STATA, version 14.1 (4905 Lakeway Drive College Station, Texas, 77845, USA). Receiver operating characteristic (ROC) curve analysis was used to ascertain the optimal cut-off for predicting DFS and OS after pancreatectomy. The optimal cut-off was identified as the point of intersection nearest the top left-hand corner between the ROC curve and the diagonal line from the top right-hand corner to the bottom left-hand corner of the graph. For the univariate analysis, the patients were divided into two groups, with SUVmax (> vs. ≤ 3.65) as the cut-off. Differences between the characteristics of the patients in the two groups were tested for significance using the Mann-Whitney U test, chi-square test, Fisher's exact test or t-student test. Univariate and multivariate analysis were used to investigate the effect of the following variables on survival-age, sex, tumor stage, pathological grade, lymph node involvement, resection margins, diabetes, and serum CA 19-9 levels. Survival data were estimated with the Kaplan-Meier method and examined using the log-rank test. Multivariate analysis of survival was performed using Cox's proportional hazards model. Significance was set at p < 0.05. Table 1 shows the clinical and pathological details of the 144 patients. Fifty-three patients had diabetes, and 93 had preoperative serum CA 19-9 levels above the normal limit. The surgical procedure involved pylorus-preserving PD in 106 patients, distal pancreatectomy with splenectomy in 34, and total pancreatectomy in four. A segmental portal-mesenteric vein resection was included in 21 cases. The resection margins were positive (R1) in 38 patients (26.4%). Lymph node metastases (stage II1) were identified in 103 patients, 114 had stage I-II tumor, and 95 tumors (66%) were well-or moderately-differentiated. A total of 132 patients (92%) received gemcitabine-based adjuvant chemotherapy. The median SUVmax of the 144 patients was 4.0 (range 1.0 to 12.0). From the ROC analysis, the best cut-off was identified at 3.65. The area under the ROC curve (AUC) was 0.66 (95%CI 0.542-0.77) ( Figure 1). When patients were grouped by low SUVmax (≤ 3.65) versus high SUVmax (> 3.65), the two groups did not differ statistically in terms of age, sex, number of patients with of diabetes, tumor stage, type of treatment, or number of patients given adjuvant therapy. Median values of CA 19-9, numbers of patients with lymph node metastases and those with poorly-differentiated tumors were significantly higher in the high SUVmax group (Table 1). The 144 patients' median serum CA 19-9 level was 114 kU/L (range 1.0 to 6637 kU/L). Follow-up was available for all patients, and ranged from 6 to 152 months. stage, type of treatment, or number of patients given adjuvant therapy. Median values of CA 19-9, numbers of patients with lymph node metastases and those with poorly-differentiated tumors were significantly higher in the high SUVmax group (Table 1). The 144 patients' median serum CA 19-9 level was 114 kU/L (range 1.0 to 6637 kU/L). Follow-up was available for all patients, and ranged from 6 to 152 months.
On univariate Cox regression analysis (Table 2), lymph node metastases, pathological grade, resection margins, tumor stage, and SUVmax correlated significantly with DFS, while diabetes and serum CA 19-9 levels did not. Multivariate Cox regression analysis (Table 2) showed that the same parameters were independent predictors of DFS. Patients with a preoperative SUVmax > 3.65 had a significantly shorter DFS than patients with a SUVmax ≤ 3.65 (p = 0.001) ( Figure 2). When the patients grouped by SUVmax were stratified by stage of disease, 18-FDG-PET/CT uptake correlated with survival even among patients in stage I-II, with a better survival for patients with SUVmax ≤3.65 (p = 0.0004) ( Figure 3)
On univariate Cox regression analysis (Table 2), lymph node metastases, pathological grade, resection margins, tumor stage, and SUVmax correlated significantly with DFS, while diabetes and serum CA 19-9 levels did not. Multivariate Cox regression analysis (Table 2) showed that the same parameters were independent predictors of DFS. Patients with a preoperative SUVmax > 3.65 had a significantly shorter DFS than patients with a SUVmax ≤ 3.65 (p = 0.001) ( Figure 2). When the patients grouped by SUVmax were stratified by stage of disease, 18-FDG-PET/CT uptake correlated with survival even among patients in stage I-II, with a better survival for patients with SUVmax ≤3.65 (p = 0.0004) ( Figure 3)

Overall Survival
With a median follow-up of 100.8 months (range 6-152), 125/144 patients (87%) died of pancreatic cancer, and another two patients died of causes unrelated to their pancreatic disease. The median OS for the whole cohort was 22.4 months (range [19][20][21][22][23][24][25][26][27]. At univariate Cox regression analysis (Table 3) lymph node metastases, pathological grade, resection margins, tumor stage, and SUVmax correlated significantly with OS. Multivariate Cox regression analysis (Table 3) identified the same variables as being significantly associated with OS. As in the case of DFS, diabetes and CA 19-9 serum levels were not independent predictors of OS. Survival analysis with the Kaplan-Meier method showed a significantly lower OS for patients with a preoperative SUVmax > 3.65 than for those with a SUVmax ≤ 3.65 (p < 0.001) (Figure 4). When the patients grouped by SUVmax were stratified by tumor stage, 18-FDG uptake correlated with OS among patients with stage I-II (better survival for patients with SUVmax ≤ 3.65, p = 0.0002), but not for those with stage III-IV tumors (p = 0.71). The survival curves for patients with stage I-II and SUVmax > 3.65 did not differ statistically from those of patients with stage III-IV and SUVmax ≤ 3.65. (Figure 5). At latest follow-up, 17 patients were alive (16 disease-free): 13 in the group with SUVmax ≤ 3.65, and 4 in the group with SUVmax > 3.65 (one with recurrent cancer).  (Table 3) lymph node metastases, pathological grade, resection margins, tumor stage, and SUVmax correlated significantly with OS. Multivariate Cox regression analysis (Table 3) identified the same variables as being significantly associated with OS. As in the case of DFS, diabetes and CA 19-9 serum levels were not independent predictors of OS. Survival analysis with the Kaplan-Meier method showed a significantly lower OS for patients with a preoperative SUVmax > 3.65 than for those with a SUVmax ≤ 3.65 (p < 0.001) (Figure 4). When the patients grouped by SUVmax were stratified by tumor stage, 18 -FDG uptake correlated with OS among patients with stage I-II (better survival for patients with SUVmax ≤ 3.65, p = 0.0002), but not for those with stage III-IV tumors (p = 0.71). The survival curves for patients with stage I-II and SUVmax > 3.65 did not differ statistically from those of patients with stage III-IV and SUVmax ≤ 3.65. (Figure 5). At latest follow-up, 17 patients were alive (16 disease-free): 13 in the group with SUVmax ≤ 3.65, and 4 in the group with SUVmax > 3.65 (one with recurrent cancer).

Discussion
An accurate pretreatment prognosis for patients with pancreatic cancer would be very helpful for the purpose of tailoring their treatment (either surgery or multimodality clinical management). This is particularly true for apparently localized, resectable carcinoma of the pancreas because several authors have recommended neoadjuvant therapy for such patients rather than upfront surgery, the benefits of which have yet to be definitely established. The rationale for using PET/CT preoperatively for prognostic purposes in cases of pancreatic cancer stems from evidence of an accelerated glucose transport rate and increased rate of glycolysis being among the most characteristic biochemical markers of malignant transformation. Overexpression of glucose transporter 1 (Glut-1) [29] and glycolytic enzymes [30] has been demonstrated in human pancreatic adenocarcinoma. 18-FDG is a glucose analog actively taken up into the pancreatic cell by Glut-1 and phosphorylated by hexokinase in the first step of glycolysis. Its accumulation thus reflects the rate of carbohydrate metabolism and the malignant activity of a pancreatic cancer [22]. The standardized uptake value (SUV), a semiquantitative parameter of glucose consumption that enables a quantitative estimation of 18-FDG accumulation, can easily be obtained preoperatively on PET/CT. 18-FDG-PET/CT is therefore useful for distinguishing benign from malignant tumors, for diagnosing tumor recurrences, and for assessing the effects of neoadjuvant chemoradiation therapies [20,31,32]. Preliminary evidence of the correlation between 18-FDG uptake and prognosis for pancreatic adenocarcinoma has been reported in small series of patients [21,25,33]. Nakata et al. [33] introduced 18-FDG PET and SUV as metabolic prognostic factors in patients with pancreatic carcinoma. In a small series of 14 patients, they found survival significantly shorter in the high SUV group (>3.0) than in the low SUV group (<3.0) (p < 0.05). These results were only partially confirmed, however, by the same authors four years later [19] in 37 patients with histologically-confirmed pancreatic cancer. While SUV was unable to predict survival for patients with resectable tumor, among those with unresectable disease, patients with a low SUV survived significantly longer than those with a high SUV (p = 0.03); furthermore, multivariate analysis confirmed tumor SUV as an independent prognostic indicator for patients with unresectable tumors.
In the present study, we analyzed 18-FDG uptake in a cohort of patients (n = 144) with histologically-confirmed pancreatic cancer. When grouped by high (> 3.65) versus low (≤ 3.65)

Discussion
An accurate pretreatment prognosis for patients with pancreatic cancer would be very helpful for the purpose of tailoring their treatment (either surgery or multimodality clinical management). This is particularly true for apparently localized, resectable carcinoma of the pancreas because several authors have recommended neoadjuvant therapy for such patients rather than upfront surgery, the benefits of which have yet to be definitely established. The rationale for using PET/CT preoperatively for prognostic purposes in cases of pancreatic cancer stems from evidence of an accelerated glucose transport rate and increased rate of glycolysis being among the most characteristic biochemical markers of malignant transformation. Overexpression of glucose transporter 1 (Glut-1) [29] and glycolytic enzymes [30] has been demonstrated in human pancreatic adenocarcinoma. 18-FDG is a glucose analog actively taken up into the pancreatic cell by Glut-1 and phosphorylated by hexokinase in the first step of glycolysis. Its accumulation thus reflects the rate of carbohydrate metabolism and the malignant activity of a pancreatic cancer [22]. The standardized uptake value (SUV), a semiquantitative parameter of glucose consumption that enables a quantitative estimation of 18-FDG accumulation, can easily be obtained preoperatively on PET/CT. 18-FDG-PET/CT is therefore useful for distinguishing benign from malignant tumors, for diagnosing tumor recurrences, and for assessing the effects of neoadjuvant chemoradiation therapies [20,31,32]. Preliminary evidence of the correlation between 18-FDG uptake and prognosis for pancreatic adenocarcinoma has been reported in small series of patients [21,25,33]. Nakata et al. [33] introduced 18-FDG PET and SUV as metabolic prognostic factors in patients with pancreatic carcinoma. In a small series of 14 patients, they found survival significantly shorter in the high SUV group (>3.0) than in the low SUV group (<3.0) (p < 0.05). These results were only partially confirmed, however, by the same authors four years later [19] in 37 patients with histologically-confirmed pancreatic cancer. While SUV was unable to predict survival for patients with resectable tumor, among those with unresectable disease, patients with a low SUV survived significantly longer than those with a high SUV (p = 0.03); furthermore, multivariate analysis confirmed tumor SUV as an independent prognostic indicator for patients with unresectable tumors.
In the present study, we analyzed 18-FDG uptake in a cohort of patients (n = 144) with histologically-confirmed pancreatic cancer. When grouped by high (> 3.65) versus low (≤ 3.65) SUVmax, patients did not differ statistically in terms of age, sex, tumor stage, pathological grade, serum CA 19-9 levels, diabetes, or type of treatment. DFS and OS were significantly influenced by SUVmax, however, being 20 and 28 months, respectively, for low-SUVmax patients as opposed to 9 and 19 months for high-SUVmax patients (p = 0.001). Among the clinicopathological variables considered, tumor stage, pathological grade, lymph node involvement, and resection margins correlated significantly with both DFS and OS after univariate analysis. Multivariate analysis confirmed SUVmax, tumor stage, grade, resection margins and lymph node status as independent predictors of DFS and OS.
Interestingly, when patients in the two SUVmax groups were stratified by tumor stage, 18-FDG uptake significantly influenced survival for cases in stage I-II, but not for those in stage III-IV. Serum CA 19-9 levels and diabetes had no influence on survival. The different biological aggressiveness of the tumor indicated by the SUVmax may explain the different survival rates after potentially curative resection with otherwise similar prognostic variables. 18-FDG-PET/CT is known to be a less accurate indicator in diabetic patients, and may be unable to predict their survival adequately. Some authors [34,35] recently reported that preoperative SUVmax and serum CA 19-9 independently predicted pathological stages and OS. However, it is hard to establish an optimal cut-off value of CA 19-9 as a reproducible preoperative prognostic factor, because 10-15% of the population does not express CA 19-9 and because the levels of such tumor markers are notoriously influenced by liver and renal insufficiency [36].
Our results confirm previous evidence [22,23,[37][38][39][40][41][42] of SUVmax (measured in terms of the tumor's uptake of 18-FDG) being a simple and reliable pretreatment prognostic parameter, as in other malignancies. A summary of the results of other studies on SUVmax as a prognostic factor in cases of resectable pancreatic cancer is given in Table 4. Including ours, seven studies have been published [22,23,[38][39][40][41], all retrospective, concerning a total of 658 patients. SUVmax cut-offs vary greatly, but all the studies report a significantly longer DFS for patients with a low SUVmax, and 4 studies also describe a significantly better OS [23,38,40,41]. Since SUVmax only gives an indication of peak metabolic activity, not of tumor burden, some authors have explored the value of metabolic tumor volume (MTV) and total lesion glycolysis (TLG) as predictors of pancreatic cancer outcome [40,43,44]. Xu et al. [44] found that MTV and TLG independently predicted OS and DFS, and did so better than CA 19-9 levels, SUVmax, or tumor size. These findings were confirmed by Lee et al. [40] in 87 patients with resectable carcinoma of the pancreas (30 treated with neoadjuvant therapy)-MTV and TLG were independent prognostic factors irrespective of neoadjuvant therapy. On the other hand, SUVmax is less time-consuming and easier to calculate, and in our and others' experience, it provides the same important information.
Several previous studies found the tumor's histological characteristics important in establishing the prognosis for pancreatic cancer patients [8][9][10][11][12][13][14], but most of them are only available after surgery. The great advantage of the SUVmax calculated on 18-FDG-PET/CT is that it can be obtained before any treatment is undertaken. As the prognostic value of SUVmax is equivalent to that of tumor staging, stratifying patients by extent of disease on multidetector CT scans and SUVmax may improve our understanding of the actual effect of different treatments.
There is evidence to suggest that glycolytic activity as measured from 18-FDG uptake gives an indication of a tumor's growth and biological behavior, enabling a prediction of patients' DFS and OS. 18-FDG-PET/CT might therefore be used to identify patients with resectable pancreatic cancer at higher risk of early recurrence and shorter survival who could benefit from neoadjuvant therapy. The feasibility and clinical usefulness of this approach would need to be confirmed in prospective trials.
Another topic of interest could be the evaluation of SUVmax with 18-FDG PET/CT measured before and after chemotherapy in those patients scheduled for neoadjuvant therapy and its association with their survival.
There are some limitations of our study to mention. First, it was a retrospective study conducted at a single institution. Second, various drugs were used for adjuvant therapy during the study period, and this may have influenced the results. The significant number of patients and PET findings considered nonetheless sufficed to show statistically significant and clinically relevant differences.

Conclusions
The SUVmax calculated on 18-FDG-PET/CT provides useful prognostic informations in patients with pancreatic cancer before any surgical or medical treatment is administered, and may therefore help stratify patients for prospective studies comparing different treatment options (surgery versus chemotherapy).