1. Introduction
18F-fluorodeoxyglucose (
18F-FDG) positron emission tomography (PET)/computed tomography (CT) is a valuable imaging technique for managing oral cancer [
1].
18F-FDG PET/CT is commonly used not only for staging but also for assessing the therapeutic effect and prognosis of oral cancer [
1,
2,
3,
4]. The maximum standardized uptake value (SUV
max), generated during an integrated
18F-FDG-PET/CT scan, provides information on the metabolic activity of the tumor and is an index that is used to discriminate between malignant lesions and benign lesions. Generally, a high SUV
max indicates a high possibility of malignancy [
5]. Recently, dual-phase
18F-FDG PET/CT has been used as an alternative method when other types of preoperative imaging cannot clearly distinguish between benign and malignant lesions [
6,
7,
8,
9]. Thus, both early and delayed SUV
max values are important for diagnosing malignancy.
The Warburg effect, a key metabolic mechanism exploited by
18F-FDG PET/CT [
5,
10], is a cancer-specific metabolic shift manifesting as increased glucose absorption by aerobic glycolysis activation [
5,
11]. We have previously revealed that salivary metabolites have the potential to discriminate between patients with oral cancer and healthy controls [
12,
13,
14,
15], indicating a possible correlation between salivary metabolites and SUV
max. However, to the best of our knowledge, no studies have explored this relationship.
18F-FDG PET/CT is usually used for staging or evaluation of treatment response rather than for cancer screening. However,
18F-FDG PET/CT has also been used in Japan for cancer screening in people with no cancer symptoms, and accumulating evidence supports this application of
18F-FDG PET/CT [
16,
17,
18,
19]. Therefore, we believe that it is clinically meaningful to analyze the relationship between salivary metabolites and PET parameters such as SUV
max in patients with oral cancer. If salivary metabolites demonstrate a significant correlation with PET parameters such as SUV
max, they may have the potential to be used as a screening tool before PET/CT to identify patients with high SUV
max, i.e., the presence of oral cancer.
Hence, this study aimed to analyze the relationship between SUVmax and salivary metabolites in patients with oral cancer, which could reveal the utility and limitations of salivary metabolites as biomarkers.
3. Results
Patient characteristics are shown in
Table 1. The data on controls were the same as those reported in our previous studies [
12,
15].
Table 2 shows the salivary metabolites that were significantly and highly correlated with SUV
max (
p < 0.05, Spearman’s rank correlation). Eleven salivary metabolites were correlated with delayed-phase SUV
max—N-acetylneuraminate, pyruvate, hexanoate, homovanillate, 3-methylhistidine, 3-phenylpropionate, pipecolate, p-hydroxyphenylacetate, isethionate, crotonate, and o-phosphoserine. None of the metabolites showed a significant correlation with early-phase SUV
max.
Figure 1A shows the ROC curves for discriminating between patients with oral cancer and controls. The AUC was 0.738 (
p = 0.001, 95% confidence interval 0.619–0.857). In order to evaluate the discrimination ability of this MLR model for the early stages (T1 and T2) and advanced stages (T3 and Tt), ROC curves were depicted at
Figure 1B,C. Among the 11 metabolites,
N-acetylneuraminate and 3-phenylpropionate were selected as independent variables in the MLR model using the backward elimination method. The
p-values in the developed models were 0.018 and 0.018 for
N-acetylneuraminate and 3-phenylpropionate, respectively. To evaluate the bias caused by sex, we developed another MLR model including
N-acetylneuraminate, 3-phenylpropionate, and sex. The
p-values for these features in this MLR model were 0.018, 0.017, and 0.344, respectively.
4. Discussion
This study aimed to determine the relationship between salivary metabolites and SUVmax of 18F-FDG PET/CT. 18F-FDG PET/CT is a cancer screening tool, but it has several limitations. Thus, there is a need for another screening tool that is complementary to 18F-FDG PET/CT. We have evaluated the screening potential of saliva samples and the correlations between SUVmax and salivary metabolites in patients with oral cancer. To our knowledge, this is the first report to reveal an association between SUVmax and salivary metabolites in patients with oral cancer. All salivary metabolites showed significant positive correlations with SUVmax. Because high SUVmax values reflect increased anaerobic glycolysis metabolism in malignant tumors, the higher concentration of metabolites in patients with oral cancer in this study meant higher SUVmax values.
The salivary metabolites observed in this study may be used as a first-line screening tool to confirm the presence of oral cancer before PET/CT. MLR analyses with a backward elimination method identified that
N-acetylneuraminate and 3-phenylpropionate had the highest AUCs for discriminating between patients with oral cancer and controls (AUC = 0.738,
p = 0.001,
Figure 1A). Therefore, these metabolites could be used for detecting the pathological accumulation of
18F-FDG using PET/CT and discriminating between patients with oral cancer and controls. These findings are consistent with those reported in our previous studies, which showed that these metabolites could be reasonable choices for oral cancer screening. To evaluate the discrimination ability for the early stages and the advanced stages, we depicted the ROC curves (
Figure 1B,C). These curves also showed high AUC values 0.773 (
p = 0.005) and 0.712 (
p = 0.015). We also evaluated the effect of sex where the
p-value of sex was 0.344 in the MLR model including these two metabolites and sex. Therefore, only these metabolites provided enough evidence to discriminate between patients with oral cancers and healthy controls.
We speculated that 11 candidate salivary metabolites could be used to detect the pathological accumulation of
18F-FDG in oral cancer tissue using PET/CT. Some of our candidate metabolites were related to the glycolytic pathway or were a part of it, suggesting a Warburg effect as a form of cancer metabolism. Pyruvate is a final product of glycolysis. Pipecolate is a product of lysine degradation. Lysine is a ketogenic amino acid that is metabolized to acetyl-coenzyme A in the tricarboxylic acid cycle. These metabolites have been reported as screening biomarkers for various cancers [
14,
20]. In addition, in our previous study [
15],
N-acetylneuraminate, 3-methylhistidine, and pipecolate were significantly increased in oral cancer tissue. Thus, it is reasonable to suggest that these salivary metabolites are potential biomarkers that could be used to detect the pathological accumulation of
18F-FDG in patients with oral cancer.
In this study, 11 salivary metabolites showed significant correlations with delayed-phase SUV
max, but there was no significant correlation between any salivary metabolite and early-phase SUV
max. Dual-phase
18F-FDG PET/CT has recently been used to distinguish between benign lesions and malignant lesions [
6,
21,
22,
23,
24]. The SUV
max of malignant lesions increased in the delayed phase, whereas FDG uptake in most benign lesions decreased in the delayed phase [
21]. Erdem et al. concluded that delayed-phase
18F-FDG imaging increased the detectability of the primary lesion because of higher FDG uptake by primary tumors in the delayed phase compared to that in the early phase of imaging [
6]. Delayed-phase SUV
max has the potential to show the aggressiveness of the tumor better than early-phase SUV
max. These facts may partly explain the lack of a significant correlation between any salivary metabolite and early-phase SUV
max in our study. In contrast, 11 salivary metabolites showed significant correlations with delayed-phase SUV
max.
SUV
max is affected by several factors, including body weight, amount of time passed between the injection and the scanning, plasma glucose level, tumor size, and region of interest [
25,
26,
27,
28]. In this study, none of the patients were obese as their body mass index was 15.0–27.1 kg/m
2; only three patients had a body mass index of >25.0 kg/m
2. The time from FDG injection to scan was the same in all patients, and no patients had diabetes mellitus. Hence, the dispersion of the factors that could affect SUV
max values was minimal.
This study has several limitations. Firstly, SUV
max is occasionally high not only for malignant tumors but also for non-malignant lesions, such as lesions due to inflammation or benign tumors. In such cases, it may not be possible to distinguish a benign lesion from a malignant one based on SUV
max alone [
5]. Hence, there are concerns regarding the low specificity of the metabolites, which would be expected to be enhanced in the presence of both oral cancer and inflammation, such as periodontal disease. Our previous study, however, revealed differences between oral cancer and periodontal disease [
14], with periodontal disease having a lesser effect on the salivary metabolomics results than oral cancer [
15]. Secondly, SUV
max determined by PET/CT is not always high for low glycolysis malignant tumors, such as well-differentiated lung adenocarcinoma, hepatocellular carcinoma, prostate cancer, kidney cancer, and gastric cancer [
29,
30,
31]. It is unclear how salivary metabolites could change in the aforementioned lesions to produce a low SUV
max. Thirdly, our candidate salivary metabolites may not be biomarkers specifically related to oral cancer. Generally, most cancers are associated with high SUV
max.
Our previous salivary metabolomics studies have shown a large overlap of aberrant metabolites in oral, breast, and pancreatic cancer patients [
14]. We also analyzed the consistently elevated metabolites in saliva and oral cancer tissue [
15]. In the present study, we used a different approach to identify the metabolites showing the ability to discriminate between oral cancer and healthy controls based on the correlation between salivary metabolites and SUV
max of PET/CT. Evidence of the salivary metabolites associated with various cancers has recently been accumulated [
32]; therefore we have to validate the specificity of these metabolites with the saliva samples collected from other cancers.