[¹⁸F]FMISO PET/CT as a preoperative prognostic factor in patients with pancreatic cancer

Our institutional review board approved the protocol of this study (Registration #14-110). From March 2015 to March 2018, we enrolled 26 patients with pancreatic cancer who had been initially planned for surgery. Inclusion criteria are the following: a pancreatic cancer patient who was referred to our surgery department, was 20-years old or older, and agreed to attend this study by written informed consent. We excluded a patient who was pregnant or who had the possibility of pregnancy and patient who was difficult to acquire PET/CT due to the physical condition such as unbearable pain. Patients were followed-up until November 2018 as much as possible. Recurrences after operations were evaluated by conventional methods including CT, MRI or [¹⁸F]FDG PET/CT.

Based on the previously described procedure [20], [¹⁸F]FMISO was synthesized in our hospital. Precisely 150 min after intravenous administration of 7.4 MBq/kg [¹⁸F]FMISO, PET/CT images were acquired by Biograph 16 (Siemens, Erlangen, Germany). The dose of syringes before and after [¹⁸F]FMISO administration was counted, and the correct injection dose was calculated. Images including entire pancreas were acquired for 20 min without any bed movement and reconstructed in a 168 × 168 × 32 matrix with 3.0 × 3.0 × 5.0 mm voxel size using ordered subset expectation maximization with iteration 3 subset 8 with Gaussian filter of 8-mm full width at half maximum. The attenuation correction was performed by transmission CT using the following parameters: slice thickness, 5 mm; tube rotation, 0.5 s; table speed, 10 mm/rotation; pitch, 1:1; 120 keV; and 50 mAs.

Two experienced nuclear medicine physicians (T.Y. and I.K.) evaluated [¹⁸F]FMISO uptake on tumor which can be visually identified and decided as positive or negative by discussion. The tumor that showed higher uptake than the surrounding normal pancreatic tissue was considered as positive. Three-dimensional spheroidal regions of interest (ROI) were drawn to cover the pancreatic tumor, and three-dimensional spherical ROI of a 10-mm diameter were drawn in the descending aorta on the same axial image to evaluate each standardized uptake value (SUV). SUVpeak of the tumor was divided by SUVpeak of the aorta, and we defined tumor blood ratio using SUVpeak (TBRpeak) as the quantitative value. SUVpeak was defined as the mean value of the voxels within a fixed sphere of 1-cm³ volume (approximately 22 voxels) centered on the hottest area of the ROI over the tumor and the aorta. Method for ROI placement was illustrated in Additional file 1: Figure S1. In case the tumor was difficult to identify on [¹⁸F]FMISO PET/CT alone, we set the ROI of the tumor by visually comparing side by side with the other modalities of enhanced CT, enhanced MRI, or [¹⁸F]FDG PET/CT. We used Syngo.via VA30A (Siemens, Erlangen, Germany) for the image analysis.

Based on the thorough discussion of patients’ information including enhanced CT and the other examinations except for [¹⁸F]FMISO PET/CT, surgeons finally decided their operability.

The expression of HIF-1α in the tumor was measured by immunohistochemistry for the patients who underwent tumor resection. For the patients who did not undergo surgery, tissue sections after biopsies were also used for HIF-1α immunohistochemistry when the tissues are available.

After deparaffinization, antigen retrieve was performed by boiling the section for 20 min at 98 °C in EnVision FLEX Target Retrieval Solution, High pH (Dako, Glostrup, Denmark) using PT Link (Dako). Subsequently, the tissues were incubated for 60 min with a monoclonal antibody for HIF-1α (ab210073, Abcam plc, Cambridge, United Kingdom) diluted to 1:100 in REAL Antibody Diluent (Dako). The endogenous enzyme was blocked with FLEX Peroxidase Block (Dako) for 10 min. For visualization, the sections were incubated for 30 min with FLEX/HRP (Dako) followed by a 10-min treatment by FLEX DAB+ (Dako). Furthermore, the sections were counterstained with Mayer’s hematoxylin for 3 min.

HIF-1α was evaluated by an experienced pathologist (M.Y.) without information of PET/CT and the other clinical information. The intensity of HIF-1α staining in the nuclei was semi-quantitatively categorized into no staining (score 0), weak staining (score 1), moderate staining (score 2), and strong staining (score 3). These scores were dichotomized into low HIF-1α expression (score 0 and 1) and high HIF-1α expression (score 2 and 3).

Mann-Whitney-Wilcoxon test was used to compare 2 groups of TBRpeak on visual assessment, operability, and HIF-1α expressions. Kruskal-Wallis log-rank test was used to evaluate the difference of 4 categories of HIF-1α expressions. After dichotomization into positive or negative by the visual assessment of the nuclear medicine physicians, the results were compared with TBRpeak. The cutoff TBRpeak of visibility was determined at the maximum value of the sum of sensitivity and specificity on receiver operating characteristics (ROC) curve. An area under the curve (AUC) of ROC analysis was also calculated, and 0.9 or more of AUC was defined as high accuracy. The patients were dichotomized into high- or low-TBRpeak groups by the cutoff. For the patients who received operation, recurrence-free survivals (RFS) of the 2 groups were analyzed using Kaplan-Meier method, respectively. RFS was calculated by the days from [¹⁸F]FMISO PET/CT was performed to recurrence was diagnosed. Moreover, overall survivals (OS) were also analyzed for all patients and for patients who received operation only. OS were calculated by the days from [¹⁸F]FMISO PET/CT was performed to the patient died. For patients who are alive at the time of last follow-up, RFS and OS were censored of the day of the last follow-up. Median survivals and the 95% confidence interval (CI) were calculated, and these survival difference between high and low TBRpeak were compared by log-rank test.

One patient out of 26 participants was finally diagnosed as solid-pseudopapillary neoplasm. Therefore, we excluded the patient and totally analyzed 25 patients with pancreatic adenocarcinoma (11 women and 14 men, median age, 73 years; range, 58–81 years). Observation period was 39–1101 days (median, 369 days). The tumor size ranged from 15.5 to 43.0 mm (median, 21.2 mm). Actual injected dose of [¹⁸F]FMISO ranged from 265 to 579 MBq (median, 417 MBq). Patient characteristics are shown in Additional file 2: Table S1 and flow diagram of the patients is shown in Additional file 3: Figure S2.

After evaluation by the 2 nuclear medicine physicians, 9 cases (36.0%) were considered visually positive. Mean TBRpeak of the negative cases was significantly lower than that of the positive cases (median 1.08, IQR 1.02–1.15 vs median 1.50, IQR 1.25–1.73, p < 0.001), and the cutoff TBRpeak was calculated as 1.24 by the results of ROC analysis with the AUC of 0.99 (high accuracy) (Fig. 1a, b).

As a result of thorough clinical examinations other than [¹⁸F]FMISO PET/CT, five patients were finally considered inoperable. There was no significant difference between inoperable and operable patients (median 1.48, IQR 1.06–1.98 vs median 1.12, IQR 1.05–1.21, p = 0.10) (Fig. 1c).

We evaluated HIF-1α expressions of 20 patients who underwent surgery plus 2 additional inoperable patients whose tissues of biopsy were available for immunohistochemistry. While there was a significant difference of TBRpeak between category 0 and 3 (p = 0.04), no significant difference was observed among all the groups by Kruskal-Wallis log-rank test (p = 0.32). Although the median of TBRpeak in the low-HIF-1α expression group (category 0 and 1) was lower than in the high-HIF-1α expression group (category 2 and 3), there was no significant difference (p = 0.22). Figure 1d, e shows the boxplots of TBRpeak by HIF-1α expression, and Fig. 2 shows the images of [¹⁸F]FMISO PET/CT and HIF-1α immunohistochemistry of representative cases.

The patients were dichotomized at the cutoff TBRpeak of 1.24, and the high-TBRpeak group showed significantly shorter RFS than the low-TBRpeak group (median survival, 218 (IQR 60–246) vs 441 (IQR 309–) days, p = 0.002) (Fig. 3a).

As for OS of all 25 cases, the high-TBRpeak group showed significantly shorter OS than low-TBRpeak group (median survival, 346 (IQR 208–415) vs > 1000 (IQR 587–) days, p < 0.001) (Fig. 3b). As for 20 cases who received operation, the high-TBRpeak group also showed shorter OS than low-TBRpeak group (Median survival, 415 (IQR 369–437) vs > 1000 (IQR 587–) days, p = 0.04) (Fig. 3c).