Predictive Value of ¹⁸F-FDG PET and Somatostatin Receptor Scintigraphy in Patients with Metastatic Endocrine Tumors

Study Population

This prospective study was conducted between September 2003 and January 2006 among patients more than 18 y old and having a histologically proven well-differentiated metastatic neuroendocrine tumor (on initial data from the first biopsy, usually obtained at other centers) of the digestive tract or upper airways for whom our initially proposed management option was surveillance until tumor progression. Exclusion criteria were a high-grade poorly differentiated tumor requiring immediate treatment, pregnancy, and prior anticancer treatment other than surgery.

The therapeutic protocol, the information document, and the patient consent form were approved by the ethics committee of the Rennes University Hospital. All patients gave their written informed consent for inclusion in this study.

Examinations Performed and Follow-up

All patients underwent abdominothoracic CT, SRS, and ¹⁸F-FDG PET at inclusion. Histology slides were reread at inclusion to confirm the diagnosis of well-differentiated metastatic neuroendocrine tumor. All slides were also reexamined at the end of the study to obtain Ki67 immunostaining results needed for the grading system, which was published after the end of patient inclusion. A physical examination was performed at inclusion and 6 wk later. The standard follow-up was a physical examination and CT at 3 and 6 mo and then every 6 mo. Supplementary CT was performed if the clinical situation worsened. Anticancer treatment was initiated if tumor progression was identified on the CT scan (using the Response Evaluation Criteria in Solid Tumors). The investigator physician was free to use long-acting somatostatin for antisecretory purposes irrespective of the SRS results.

The SRS was performed 4 h and then 20 h after injection of 185 MBq of ¹¹¹In-pentetreotide (OctreoScan; Mallinckrodt). Planar and tomoscintigraphic images were acquired with a double-head γ-camera (Hawkeye; GE Healthcare). Ordered-subsets expectation maximization (2 iterations, 8 subsets) was used for reconstructions. Two investigators who were unaware of the ¹⁸F-FDG PET results interpreted the SRS findings as positive (tumor uptake > liver uptake, for liver lesions; tumor uptake > surrounding tissue, for lesions in other locations) or negative (the opposite) on the basis of visual inspection. After the patients had fasted for 4 h, imaging was performed on a dedicated ¹⁸F-FDG PET camera (Advance; GE Healthcare) for 19 patients and on the same camera integrated with a CT scanner (Discovery LS; GE Healthcare) for the other 19 patients. Images were acquired 45–73 min after an intravenous injection of 105–494 MBq (mean, 310 MBq) of ¹⁸F-FDG (Cisbio International). Ordered-subsets expectation maximization (32 iterations, 1 subset) was used for reconstruction without, and then with, attenuation correction. The ¹⁸F-FDG PET results were interpreted independently by 2 physicians unaware of the SRS results. A consensus was reached in cases of disagreement. According to the visual interpretation, the ¹⁸F-FDG PET results were considered to be negative (no ¹⁸F-FDG uptake or discrete uptake) or positive (overt ¹⁸F-FDG uptake). Uptake was standardized for injected activity and body weight to obtain the standardized uptake value (SUV) within a region of interest measuring 1 cm in diameter in a slice passing through the zone of greatest tumor uptake. The tumor-to-nontumor SUV ratio was determined from tumor uptake within a region of interest measuring 1 cm in diameter in a slice passing through the zone of greatest liver metastasis uptake and the average of 3 regions of interest measuring 1 cm within a slice of healthy liver tissue. The physician in charge of the patients did not know the SRS or ¹⁸F-FDG PET results.

The histology slides were reread by 1 operator to confirm the diagnosis of endocrine tumor and to stage it using the World Health Organization (WHO) criteria (4). Tissue blocks were fixed and embedded in paraffin for standard staining and immunostaining procedures (streptavidin biotin peroxidase revelation with diaminobenzidine) using the following antibodies: chromogranin A (1/400 F/MS382-P; MMFrance), synaptophysin (1/50 A0010; DAKO), Ki67 (1/100 M7240; DAKO), and p53 (1/25 M7247; DAKO). The tumors were classified as high-grade poorly differentiated or low-grade well-differentiated endocrine carcinoma on the basis of morphologic characteristics. For p53 and Ki67, the results were expressed as percentage of positive cells. The low-grade tumors were then staged using a new grading system (17) taking into account the percentage of Ki67-positive cells: grade 1, ≤2%; grade 2, 3%−20%; and grade 3, >20%.

Statistical Analysis

Our main objective was to determine whether one of the nuclear medicine techniques could be used to identify patients whose tumor would progress early. For this purpose, we defined 2 groups of patients according to the pattern of tumor growth: early progressive disease (within 6 mo) or stable disease. We chose the 6-mo cutoff as the most clinically pertinent time point for early progression: 3 mo might have been clinically useful but would probably have corresponded more to poor initial staging than to actual rapid progression and 12 mo would have been somewhat long for early progression. In a retrospective analysis (unpublished) of 45 new cases diagnosed between 1998 and 2002 that were proposed to initially undergo surveillance, we found that 18 (40%) were cases of early progressive disease. We decided to consider as useful an examination that would correctly predict early progressive disease in at least 90% of the cases, with an α-risk of 5% and a β-risk of 20%. To achieve this statistical power, the prospective study would have to include 34 patients. Taking into account a possible dropout rate of 10%, we decided to include 38 patients. Secondary objectives were overall survival and progression-free survival as determined using the Kaplan–Meier method and compared with the log-rank test. Survival rates ± SD are given, and nondiscrete values are expressed as mean ± SD. Receiver-operating-characteristic analysis was used to determine a positive threshold for SUV and tumor-to-nontumor SUV ratio. Comparisons between SUV or tumor-to-nontumor SUV ratio and early progressive disease or stable disease used the Mann–Whitney test; correlation between ¹⁸F-FDG PET results (positive or negative), WHO classification, Ki67 results (<15% vs. ≥15%), and P53 results (<15% vs. ≥15%) used the Fisher exact test.

Univariate analysis with the χ² test (or Fisher exact test if necessary) was applied to search for factors predictive of rapid progression. Univariate analysis to search for predictive factors of overall survival and progression-free survival used the Kaplan–Meier method, the curves being compared by a log-rank test. The variables tested included visual interpretation of the ¹⁸F-FDG PET results, SUV (<4.5 vs. ≥4.5; threshold determined by receiver-operating-characteristic analysis), tumor-to-nontumor SUV ratio (<2.5 vs. ≥2.5; threshold determined by receiver-operating-characteristic analysis), SRS results (positive vs. negative), WHO stage (high grade vs. low grade), Ki67 results (<15% vs. ≥15%), and p53 results (<15% vs. ≥15%). Multivariate analysis using linear regression and a Cox model (forward Wald method) was performed with the following variables: visual interpretation of the ¹⁸F-FDG PET results, SRS results, WHO stage, and Ki67 and p53 results (variables with P < 0.05 at univariate analysis); for ¹⁸F-FDG PET evaluations, only visual interpretation was analyzed because the three ¹⁸F-FDG PET parameters were autocorrelated.

SPSS was used for statistical analysis, and P < 0.05 was considered significant.

Study Population

From September 2003 through January 2006, 38 consecutive patients (24 men and 14 women; mean age, 60 ± 15 y; range, 23–68 y) were included prospectively. At inclusion, 31 patients had incident cases of metastatic neuroendocrine tumor and 7 were undergoing surveillance for an untreated metastatic neuroendocrine tumor; in 5 of these 7 patients, the diagnosis was established less than 6 mo before inclusion, and in the other two, the diagnosis was established less than 1 y before inclusion. Twenty-nine patients were in excellent general health (performance status, 0), and 9 had minor symptoms related to their disease (performance status, 1). The site of the primary tumor was unknown in 12 patients; localized foci were in the pancreas (n = 9), small intestine (n = 11), colon (n = 3), lung (n = 2), and gallbladder (n = 1). All patients had multiple metastases. Metastases were noted in the liver (n = 34 patients), lymph nodes (n = 15), peritoneum (n = 8), bones (n = 3), lungs (n = 1), and other locations (3). Among the 34 patients who had liver metastases, 18 had more than 10 metastases and 16 had fewer than 10; 1 patient had only 3 liver metastases but had metastases in several other locations. After multidisciplinary assessment, it was decided that the liver involvement in these patients was too advanced to allow curative surgery.

At the 6-mo follow-up, 16 patients (42%) had early progressive disease and 22 (58%) had stable disease. During the surveillance period (follow-up endpoint, September 2006; median follow-up of surviving patients, 20 mo), 19 patients (50%) received anticancer treatments and 11 (29%) died. The survival rates at 1 and 2 y were 86% ± 5.8% and 73% ± 7.9%, respectively, for overall survival and 55% ± 8.1% and 45% ± 8.5%, respectively, for progression-free survival.

Pathology Findings

The diagnosis of metastatic neuroendocrine tumor was confirmed in all patients. Ki67 and p53 immunostaining tests could not be performed for 4 patients (lack of sufficient material). The WHO stage (not taking into account grading based on Ki67 immunostaining) was high-grade in 4 patients and low-grade in 34. In these patients, the diagnosis of high-grade tumor was made some months after inclusion in the study, and the disease of all these patients had already progressed.

The percentage of Ki67-positive cells in the tumor samples was less than 2% in 14 patients, 3%−20% in 13, and more than 20% in 7. The percentage of p53-positive cells in the tumor samples was 0% in 15 patients, less than 10% in 5, 15%−20% in 4, and massive (>50%) in 4. Among the 34 low-grade tumors, 14 were grade 1, 13 were grade 2, and 4 were grade 3. This grading system could not be applied for 3 of the low-grade tumors.

Nuclear Medicine Studies

The SRS results were negative (Fig. 1) in 11 patients (29%) and positive (Fig. 2) in 27 (71%).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1. 

A 63-y-old patient who has liver metastases of pancreatic low-grade endocrine tumor. (A) SRS shows no uptake. (B) PET shows intense uptake in pancreatic tumor (SUV, 14.6; tumor-to-nontumor ratio, 6.3) and in liver metastases (SUV, 9.9; tumor-to-nontumor ratio, 4.3). Ki67 immunostaining was less than 2%; p53 immunostaining was 18%. Disease progressed at 3 mo.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2. 

A 74-y-old patient who has low-grade ileal endocrine tumor with multifocal liver metastases. (A) SRS shows intense uptake in ileal tumor and in liver metastases. (B) PET shows no liver uptake and faint ileal uptake. Ki67 immunostaining was less than 2%; p53 immunostaining was 0%. Tumor was stable after 2.5 y of follow-up.

At visual inspection, the ¹⁸F-FDG PET results were considered positive (Fig. 1) in 15 patients (39%) and negative (Fig. 2) in 23 (61%). SUV ranged from 35 to 1.4 (median, 3.7). The tumor-to-nontumor SUV ratio ranged from 0.7 to 15.2 (median, 1.9). The ¹⁸F-FDG PET results correlated strongly with the WHO stage (P < 0.001) and with the Ki67 (P < 0.001) and p53 (P < 0.05) immunostaining results.

Factors Associated with Early Progression

Table 1 correlates the early follow-up with the results of pathology, PET, and SRS. WHO staging predicted tumor progression (early progressive disease vs. stable disease) (P < 0.001): tumor size increased within 6 mo in the 4 high-grade cases and in 12 of the 34 low-grade cases. The efficacy of the grading system was less clear: all 4 grade 3 tumors progressed within 6 mo, but 3 of 14 grade 1 tumors and 3 of 13 grade 2 tumors also did so.

Fourteen of 15 ¹⁸F-FDG PET–positive patients had early progressive disease, and 21 of 23 ¹⁸F-FDG PET–negative patients had stable disease (P < 0.001). When one considers solely the 34 patients with low-grade tumors, ¹⁸F-FDG PET distinguished those with early progression: 10 of 11 patients who were ¹⁸F-FDG PET–positive and only 2 of 23 who were ¹⁸F-FDG PET–negative were in the early progressive disease group (P < 0.001). When the analysis is limited to tumors that were low-grade or WHO grade 1 or 2 (excluding high-grade and grade 3 tumors), the same observations are made: 6 of the 7 patients who were ¹⁸F-FDG PET–positive but only 2 of the 23 who were ¹⁸F-FDG PET–negative were in the early progressive disease group (P < 0.001), and 3 of the 23 patients who were SRS-positive and 5 of the 7 who were SRS-negative had early progressive disease (P < 0.001). SUV was higher in patients with early progressive disease than in patients with stable disease (9.6 ± 7.9 vs. 2.8 ± 1.2, P < 0.001). The same was true for tumor-to-nontumor SUV ratio (5.4 ± 3.6 vs. 1.4 ± 0.6, P < 0.001). At receiver-operating-characteristic analysis, early progressive disease was predicted most accurately for the following thresholds: 4.5 for SUV and 2.5 for tumor-to-nontumor SUV ratio.

Negative findings on SRS were also significantly associated with early progression (P < 0.03): 20 of the 27 SRS-positive patients had stable disease, and 9 of 11 SRS-negative patients had early progressive disease.

Table 2 summarizes the diagnostic performance (sensitivity, specificity, negative and positive predictive values, and accuracy) of ¹⁸F-FDG PET, SRS, WHO classification, and p53 and Ki67 immunostaining for the detection of rapid progressive disease. The best imaging test was ¹⁸F-FDG PET; in cases of ¹⁸F-FDG PET–positive findings, the relative risk of early progressive disease was 10.7 (95% confidence interval, 2.8–40.6).

Survival Analysis

Survival was better among patients with a low-grade tumor than among patients with a high-grade tumor (P < 0.006 for overall survival and P < 0.003 for progression-free survival), and survival was better among ¹⁸F-FDG PET–negative patients than among ¹⁸F-FDG PET–positive patients (P < 0.001 for both overall survival and progression-free survival) (Fig. 3). Overall survival (±SD) was 95% ± 5% and 95% ± 5% at 1 and 2 y, respectively, for ¹⁸F-FDG PET–negative patients, versus 72% ± 12% and 42% ± 13% at 1 and 2 y, respectively, for ¹⁸F-FDG PET–positive patients. Progression-free survival was 87% ± 7% and 75% ± 10% at 1 and 2 y, respectively, for PET–negative patients, versus 7% ± 6% and 0% at 1 and 2 y, respectively, for ¹⁸F-FDG PET–positive patients. Similar results were observed for SRS: compared with negative SRS results, positive SRS results were associated with better overall survival (1- and 2-y survivals of 92% ± 5% and 70% ± 13%, respectively, for SRS-positive patients, vs. 60% ± 15% and 48% ± 17%, respectively, for SRS-negative patients; P < 0.03) and better progression-free survival (1- and 2-y survivals of 70% ± 9% and 61% ± 10%, respectively, for SRS-positive patients, vs. 18% ± 12% and 0%, respectively, for SRS-negative patients; P < 0.001) (Fig. 4).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3. 

Progression-free survival (PFS) and overall survival (OS) are significantly (P < 0.001) better in PET-negative patients than in PET-positive patients; n = 38.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4. 

Progression-free survival (PFS) and overall survival (OS) are significantly (P < 0.001) better in SRS-positive patients than in SRS-negative patients; n = 38.

Among the 34 patients with a low-grade tumor, progression-free survival was significantly better (P < 0.001) for the 23 who were ¹⁸F-FDG PET–negative than the 11 who were ¹⁸F-FDG PET–positive and for those who were SRS-positive than those who were SRS-negative (P < 0.001). Overall survival was similar: better overall survival (P < 0.007) was found for ¹⁸F-FDG PET–negative than ¹⁸F-FDG PET–positive patients and for the 25 SRS-positive than the 9 SRS-negative patients (P < 0.01). Among the 15 patients who were ¹⁸F-FDG–positive, those with positive SRS results had better progression-free survival (P < 0.05), but not better overall survival, than those with negative SRS results. When we excluded from the analysis the 4 patients who had a tumor initially considered to be low-grade but later graded as grade 3, the results remained unchanged except for progression-free survival, which was no longer significantly associated with SRS results (P = 0.09).

At univariate analysis, factors associated with progression-free survival and overall survival (P < 0.05) were ¹⁸F-FDG PET result, SUV, tumor-to-nontumor SUV ratio, SRS result, WHO stage, and Ki67 and p53 results. At multivariate analysis, only the ¹⁸F-FDG PET result (P = 0.001) was associated with progression-free survival. No independent factor predictive of overall survival could be identified.