Data search and study selection
In this retrospective study, 110 consecutive patients who underwent surgery for CRC at Toho University Hospital between April 2015 and April 2019 after examination with FDG-PET/CT were enrolled. The exclusion criteria included the following: (1) underwent neoadjuvant chemotherapy or radiation therapy before the operation (n = 14) and/or did not have pathologically proven primary CRCs, (2) had distal metastasis (n = 3; M1), (3) had any other malignant tumor (n = 3), (4) were not evaluated for recurrence (n = 3), and (5) were recommended adjuvant chemotherapy (n = 3). Consequently, in the study, 84 patients were included.
The enrolled patients were classified into two groups based on the pathological N stage (Japanese Classification of Colorectal Carcinoma, 8th edition) [14]—N0 (no lymph node metastasis) and N1–N3. The histological types were classified into two groups—one group consisted of poorly differentiated carcinoma and mucinous adenocarcinoma and the other group consists of well-defined adenocarcinoma, adenocarcinoma with moderate differentiation, papillary adenocarcinoma, poorly differentiated carcinoma, and mucinous adenocarcinoma. The T stage was categorized into ≤ T2 or ≥ T3 based on criteria by the Japanese Association of Clinical Cancer Centers (5-year survival rate was > 95% in stages Tis, T1, and T2, and < 95% in T3 and T4 without metastasis) [15]. Accordingly, stage II or stage III were defined (5-year survival rate was > 95% in Stage 0, Stage I, and Stage II, and < 95% in Stage III and Stage IV).
Acquisition of PET/CT image
Patients fasted for at least 5 h before undergoing the imaging test. For routine PET/CT (Biograph mCT flow20, Siemens Healthcare, Tokyo, Japan), 185 MBq FDG was administered. The CT machine was set at the following settings: voltage, 120 kV; rotation time, 0.5 s/rotation; pitch, 1.2; and automatic adjustment of electric current with automatic exposure control. PET data from the skull base to the ilium and the ilium to the knee were acquired at 1.1 cm/s and 1.3 cm/s, respectively, using continuous bed motion mode. We used the three-dimensional ordered subset expectation maximization algorithm (iterations, 2; subsets, 21) in combination with point spread function (PSF) and time of flight (CT attenuation correction; matrix size, 200 × 200; pixel size, 3.54; and Gaussian filter [full width at half-maximum = 6 mm]) for reconstruction of the PET image. The image was harmonized without PSF, and SUVs were measured. We used EQ PET software (Siemens Healthcare; filter = 5.8 m) to smoothen the digital input image data using a Gaussian filter. This process yielded an accurate SUV without degradation of image quality. Given that we analyzed the results with EQ PET software, which conformed to an EARL reconstruction, the outcomes should be comparable to those assessed with other scanners.
Measurement of FDG metabolic parameters
A focus with 18F-FDG uptake greater than in the surrounding tissue, excluding any physiologically increased accumulations, was defined as a positive PET/CT finding. In axial, coronal, and sagittal sections, sufficiently wide volume boundaries were drawn to include the target lesion. Two nuclear medicine physicians independently defined the voxel of interest for target lesions and measured metabolic parameters on a three-dimensional image viewer (syngo.via, Siemens Healthcare). As the rate of concordance between the two reviewers was approximately 90%, we considered the results reproducible. In the case of multiple CRCs, we selected the VOI such that it included the region with the most advanced pathological T and N stages and measured its metabolic parameters. The margins of the target lesion inside the VOI were automatically delineated. Tumor segmentation was performed using relative thresholds of 30%, 40%, and 50% of the SUVmax (SUVmax of 2.5, 3.0, and 3.5, respectively). The threshold of 40% of the SUVmax of 2.5 was chosen because of its common use, while the others were selected based on their proximity to it. A VOI was placed in the liver to measure the background SUV. The colon close to lesions is engaged in physiological uptake that is expected to fluctuate more greatly than that of the liver. We suspected that the colon’s physiological uptake might affect metabolic parameters as a result. The criteria of PERCIST 1.0 concerning the use of FDG PET-CT for predicting treatment response recommended obtaining volume-based parameters with the use of liver SUV as a threshold to minimize the influence of inter-study variability of tumor SUV; the SUV mean plus 1 or 2 standard deviations (SD) of the background has been acknowledged as useful in determining parameter thresholds [16].
We placed a VOI approximately 3 cm (14 mL) in diameter on the normal right hepatic dome to determine the mean liver SUV (SUVliver). The tumor SUV normalized to the SUV of the reference liver (tumor-to-liver ratio [TLR]) was computed as the ratio of the maximum lesion SUV to SUVliver. Therefore, we used TLRs of 1, 1.5, and 2 as the background threshold. Lean body mass (LBM)-corrected SUV (SUL) was found to be a better quantitative method for obese patients than using body weight (BW) or body surface area. Therefore, we used SUL values of 2.0, 2.5, and 3.0 as thresholds. SUL was measured using James’ formula which relied on sex, height (cm), and total BW (kg) to estimate LBM—Men = 1.1 × BW − 128 × (BW/Height); Women = 1.07 × BW − 148 × (BW/Height). We defined MTV as the sum of voxels and calculated TLG as MTV multiplied by the SUV mean (the average SUV within a VOI). For metabolic evaluation, we also measured SUVmax, calculated using maximum activity values in a VOI placed manually over the visible area of the target lesion on each PET image. MTV and TLG values are significantly affected by the segmentation methods used. As the optimal prognostic thresholds of MTV and TLG in CRC have yet to be identified, we estimated them at our hospital and, based on our findings, determined whether other hospitals should evaluate their own standard optimal thresholds.
Statistics
Identifying the optimal threshold for the association of MTV and TLG with recurrence within one year after surgery
Recurrence within the first year after surgery was assessed in patients with CRC by analyzing receiver operating characteristic (ROC) curves for MTV and TLG. PET parameters’ optimal cut-off values were defined to maximize sensitivity and specificity. In our hospital, clinical assessment and laboratory tests, including CEA・CA19-9, are performed every three months within the first two years following the surgical resolution of Stage II or III CRC, and CT scans of the chest, abdomen, and pelvis with IV contrast are conducted every three to six months. Alternatively, the abovementioned assessments and tests are performed every six months during the first two years following the surgical treatment of Stage I CRC, and the CT every six to twelve months within the same period. If the levels of CEA or CA19-9 exceeded the upper limits of their normal ranges (5.0 ng/ml, 37.0 U/ml) and were increasing, an additional CT was performed. According to the 2005 ASCO guidelines, patients with CRC who are at risk of recurrence after surgery and who could be candidates for curative-intent surgery, are recommended to undergo chest and abdomen CT. Additionally, a pelvic CT scan should be considered for rectal cancer surveillance, especially for patients who have not been treated with radiotherapy. Hence, these guidelines informed our selection of CT scans of the chest, abdomen, and pelvis with IV contrast. The most common site of local and distant recurrence is reportedly the liver [17], and many liver metastases in Japan reportedly occur within a year of surgery [18]. Early prediction of recurrence may make resection possible; hence, we limited our investigation of recurrence to the scope of a year. We confirmed the lack of the appearance of a tumor immediately after surgery and defined recurrence as the new appearance of a tumor, as confirmed on a follow-up CT scan performed according to schedule or after abnormal tumor markers (CEA or CA19-9) exceeded normal limits.
Based on the optimal cut-off value for MTV and TLG with the threshold value for each, obtained using ROC analysis, patients were separated into two groups (N0 and N+). We used the product-limit method and comparative test log-rank test to analyze the progression-free survival curves of these groups.
Association between PET/CT parameters and pathological N stage
We compared the increase in SUVmax, MTV, and TLG with pathological N stage (N0 or N1–3) with Wilcoxon signed-rank test. We used JMP 13 (SAS Institute, Inc., Cary, NC) to conduct all statistical analyses. For analysis, p < 0.05 was considered statistically significant.
If patients with CRC are suspected of having regional lymph node metastasis without distant metastasis before surgery or change in clinical stage from I or II to III, we considered them to be at high risk of recurrence. Determining the risk of recurrence helped to inform our decision to administer neoadjuvant chemotherapy or perform lymphadenectomy to avoid the recurrence.