In this study, the diameter, histological type, and location of the primary tumor were not significantly different between N-negative (N0) or N-positive (N1–3) groups. The metabolic parameters, especially MTV-TLR2, differed more in the N-positive group than in the N-negative group. Primary carcinoma is known to infiltrate intratumoral lymphatic vessels and spread through lymph nodes, and these metabolic parameters are expected to be related to intratumoral lymphatic vessel invasion. In non-small lung cancer (NSCLC), the high FDG uptake of primary carcinoma is related to invasion and lymph node metastasis [
19], and both uptake and tumor size were significant factors [
19]. Therefore, MTV or TLG, signifying the spread of high-grade tumors, was strongly related to lymph node metastasis [
20,
21]. However, there are only a few reports on the correlation between metabolic parameters and intratumoral invasion of lymphatic vessels or lymph node metastasis in CRC [
13]. Previously, 40% of tumor SUVmax for TLG (TLG 40%), but not MTV, was significantly related with pathological T stage in CRCs on the left side of the large intestine [
22]. Therefore, TLG 40%, and not MTV 40%, was expected to be related to its invasion and node metastasis. Because of the different thresholds in MTV, MTV-TLR2 are expected to be more strongly related to T stage in CRC than other PET parameters. Further investigation is needed to explore whether MTV-TLR2 or TLG 40% are related to the depth of invasion and intratumoral vessel invasion in CRC. Based on our results, we expect MTV-TLR2 of the primary tumor to serve as a biomarker to predict its metastasis in patients with CRC. Further, we expect it to determine the suitability of neoadjuvant chemoradiation, neoadjuvant chemotherapy, or range of lymphadenectomy, even if the metastasis has low FDG uptake and poor morphological change. In CRC, the optimal threshold of metabolic parameters associated with prognosis was unknown, and MTV and TLG values change depending on the threshold. Therefore, we evaluated thresholds (SUVmax, MTV, and TLG) that were closely related to relapse—TLG 40% was the most common parameter related to recurrence within the first year after surgery, and MTV 3.5 and MTV-TLR2 performed better than MTV with other thresholds. A threshold of 36–44% was found to produce volumes similar to those measured from CT for lung cancer lesions > 4 mL [
16]. Consequently, thresholds of 40–42% are most commonly used to measure MTV, and we confirmed that TLG 40% was associated with prognosis. However, the prediction of prognoses was often limited by common underestimations of the tumor volume with heterogeneous uptake (such as necrotic cores) on relative thresholds. In addition, Erdi et al. [
23] reported that they overestimated small lesions with a low signal-to-background ratio using a fixed relative threshold. Figure
1 shows that VOIs in SUV %, in particular 50%, were smaller than other thresholds in a tumor with a high SUVmax; the inverse is shown in Fig.
2. The fixed relative method did not show the exact range of tumors with various sizes and signal-to-background ratios. However, the TLG threshold of 40% was observed to be the most prognostic parameter in this study. We therefore, speculated that TLG 40% could predict higher FDG uptake associated with poor prognosis in cases of high-SUVmax tumors, as well as tumors with low density and limited morphological changes with a tendency to be scattered throughout the body and are associated with poor prognosis in low-SUVmax tumors. In addition, their causes might predict distant relapse (e.g., in the lung, liver, peritoneum). Our findings indicated that MTV 3.5 and MTV TUR2 were more prognostic than were the other thresholds in MTV. Based on the results of earlier studies that used absolute cutoff values to distinguish benign from malignant lung nodules [
24‐
26], we selected an SUVmax threshold value of 2.5. A meta-analysis found that parameters with an SUVmax threshold of 2.5 were significantly related to prognosis [
27]. Various non-malignant lesions, such as inflammation and infection, may cause an increase in FDG uptake. In this study, MTV 3.5 was more prognostic than was MTV 2.5 in CRC; colon particular physiological uptake could be considered as a cause. Figure
1 shows that VOIs in SUV 3.5 were larger than the relative thresholds in a tumor with a high SUVmax; the inverse is shown in Fig.
2. If a tumor had an intense FDG uptake (such as SUV of > 15) according to an absolute method, their parameters could easily be overestimated by the spillover effect. Figure
1 also shows the overestimation of the absolute method. The degree of the overestimation was comparatively small and might not influence the prognosis. As a result, we considered MTV 3.5 to be more prognostic than MTV with the other thresholds, except for MTV TLR2. The most prognostic threshold in MTV was different in TLG. MTV was related more strongly related with regional lymph node metastasis than with TLG, and TLG was related more strongly related with relapse than MTV. These findings informed the conclusion that MTV was related more strongly with the depth of invasion or intratumoral invasion of lymphatic vessels than was TLG and that TLG was related more strongly with distant relapse than was MTV. These reasons might lead to the differences in the most prognostic threshold between MTV and TLG. Further investigation is needed to explore whether MTV or TLG is related to the depth of invasion, intratumoral vessel invasion, and distant relapse. To measure SUV, previous investigators placed a region of interest in the liver or mediastinal blood pool, and the mean SUV plus one or two standard deviations was then used as the background threshold [
16]. Lesions with heterogeneous tracer uptake are often underestimated using the relative threshold; however, the liver can be subtracted as the background-based threshold, and its tumor voxels can be included in the MTV or TLG [
28]. As seen in Fig.
1, a VOI of the liver background threshold method resulted in greater overestimation than did the relative method. However, a VOI of the liver background threshold method, in particular TUR2, showed greater underestimation than did that of the relative method (Fig.
2). The degree of the overestimation and underestimation were also comparatively small, and we therefore, considered MTV TUR2 to be more prognostic than the others in MTV, except for MTV 3.5. However, this method is more time consuming, and its reproducibility was lower than that at other thresholds. We also evaluated the SUL of the lesion at baseline. If FDG was distributed evenly inside the body, the SUV would be calculated as 1.00. However, since there are physiologically high and low metabolic organs, the volume of FDG distribution and SUV can be overestimated with increased body fat. Hence, the use of SUL has previously been advocated [
28]. There were few obese patients with CRC, and SUL might be not more strongly associated with prognosis than are other thresholds in this study.
Our study had some limitations. We studied retrospectively in a single center and there could be a patient selection bias. Although we performed the clinical assessment and laboratory tests including CEA CA19-9 and CT scans based on our hospital’s protocol, MRI and chest X-rays might also be performed for follow-up after colorectal cancer surgery, and follow-up including those tests could detect more cases of recurrence. We defined prognosis based on recurrence within the first year after surgery in CRC. Recurrence might be observed one year after colorectal cancer surgery, numerous new medications have been developed, recurrence might be delayed, and a longer follow-up period might be required. The relationships between metabolic parameters using original thresholds were estimated. As the background threshold, mediastinal blood pool, or adipose tissue near the tumor might be used, or another numerical value such as 2.0 might be used as the fixed relative threshold, so research including those thresholds is needed. Depending on the various sizes and signal-to-backgrounds, results could be greatly misleading by when using a fixed relative threshold. In such cases, it cannot be ruled out that other thresholds, such as the background threshold, might be more prognostic than a fixed relative threshold. In the future, the connection between the parameters and overall survival should be examined over a longer period with a multicenter common protocol, including MRI and chest X-ray in addition to blood and CT tests, as new medications have decreased mortality rates. The relationship between the optimal threshold of glucose metabolic factors among more thresholds, associated with prognosis and conventional high-risk factors, including lymph node metastasis, should be examined in a prospective multicenter study with a larger patient population exhibiting high mortality and a longer follow up of these patients will provide more definitive insight into the realistic value of MTV and TLG. Subgroup analysis will be performed for each size and signal-to-background, and it will be necessary to evaluate how the analysis will change in relation to prognosis.