Predictive value of F-FDG PET/CT metabolic heterogeneity parameter combined with peritumoral adipose tissue abnormality and cN stage in perineural invasion of colorectal cancer: a retrospective analysis.

Introduction
Colorectal cancer (CRC) is among the most prevalent cancers and holds the second-highest mortality rate among all malignant tumors (1). Common clinical treatment methods include surgery, chemotherapy, immunotherapy, and targeted therapy (2). Perineural invasion (PNI) refers to the invasion of tumor cells into nerves, followed by their spread along the nerve sheath (3). PNI represents a significant mechanism of cancer metastasis, contributing to poor prognosis and elevated local recurrence rates (4,5). In recent decades, PNI has been recognized as an indicator of tumor aggressiveness, with its prognostic value established across various malignancies, including CRC, gastric cancer, and pancreatic cancer (6,7). The presence of PNI is significantly correlated with factors such as tumor diameter, lymph node metastasis, tumor, node, metastasis (TNM) staging, and histological type (8). The presence of PNI is associated with lower overall survival (9). Neoadjuvant therapy could sufficiently reduce the PNI-positive rate (10). Some studies have suggested that radiomics may hold potential in predicting PNI status (11,12). However, the radiomics process is relatively complex and has not yet been widely implemented in clinical practice.
18F-fluorodeoxyglucose (18F-FDG) positron emission tomography/computed tomography (PET/CT) is a functional imaging technique that not only facilitates disease diagnosis and staging but also offers various metabolic parameters to evaluate tumor heterogeneity and predict latent cancer metastasis (13,14). Several metabolic parameters associated with 18F-FDG PET/CT can reflect tumor heterogeneity; however, their applicability may vary depending on the specific context. Some studies have indicated that peritumoral adipose tissue (PAT) plays a significant role in promoting tumor growth, angiogenesis, and modulating the tumor microenvironment (15). The imaging features of PAT are linked to the prognosis of CRC (16). However, limited research has examined the relationship between PAT and PNI. In this retrospective study, we investigated the predictive value of 18F-FDG PET/CT metabolic heterogeneity parameters for PNI. Additionally, we developed a nomogram model that integrates the coefficients of variation (CV), PAT grade, and cN stage, which exhibited strong predictive capability for PNI. We present this article in accordance with the TRIPOD reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-708/rc).

Methods

Patient selection
The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Institutional Review Board of Nanjing Drum Tower Hospital, Affiliated Hospital of Medical School, Nanjing University (No. 2024-425-01), and informed consent was provided by all patients. A total of 116 patients with CRC, treated and diagnosed between January 2016 and July 2024, were included in this study on a retrospective basis. The inclusion criteria were as follows: (I) pathologically confirmed diagnosis of CRC; (II) underwent 18F-FDG PET/CT examination before treatment, with acceptable image quality. The exclusion criteria were as follows: (I) recurrent CRC; (II) presence of other malignant tumors; (III) pregnant or lactating patients; (IV) incomplete clinical or pathological data. PNI was determined as tumor cell invasion in, around, and through the nerves (3). The detailed steps of the study design are illustrated in Figure 1.

PET/CT scanning
The patients were examined using a Philips GXL16 PET/CT scanner (Philips Healthcare, Amsterdam, Netherlands). The radiochemical purity of 18F-FDG was determined to be greater than 95%. All patients were instructed to fast for a minimum of six hours prior to the examination, and their fasting blood glucose levels were monitored and maintained below 11.1 mmol/L. The radiopharmaceutical 18F-FDG was injected intravenously at a dose of 5.18 MBq/kg, adjusted according to the patient’s weight. Following the injection, the patients were required to rest quietly for a period of 50–60 minutes prior to undergoing PET/CT scanning. The scanning range extended from the top of the skull to the upper thigh, with each bed position scanned for a duration of two minutes. The CT scanning parameters were as follows: tube voltage of 120 kV, tube current of 100 mA, and slice thickness of 4 mm. The CT data were used for attenuation correction, and the corrected PET images were reconstructed using the ordered subset expectation maximization (OSEM) method.

Imaging quantitative analysis
A nuclear medicine physician with five years of experience analyzed PET/CT imaging data using the LIFEx software (version 7.5.15; https://www.lifexsoft.org/), and a senior physician with 15 years of experience was responsible for the review. The primary CRC lesions were delineated using a threshold of 40% of the maximum standardized uptake value (SUVmax). Subsequent to this, the traditional metabolic parameters were obtained, including SUVmax, the mean standardized uptake value (SUVmean), the minimum standardized uptake value (SUVmin), the metabolic tumor volume (MTV), and the total lesion glycolysis (TLG). Furthermore, the SUVmax threshold was set to 60% and 80%, respectively, in order to obtain the corresponding MTV and TLG. The 18F-FDG PET/CT metabolic heterogeneity parameters included the area under the cumulative standardized uptake value (SUV) histogram curve (AUC-CSH), CV, heterogeneity index (HI), and heterogeneity factor (HF). The SUV volume histogram function in the LIFEx software was utilized to analyze the corresponding volume of each SUV in the lesion, generate a CSH curve, and calculate the AUC-CSH. Figure 2A shows the measurement method of AUC-CSH. The CV was defined as the ratio of the standard deviation of SUV to SUVmean, multiplied by 100. The HI was defined as the ratio of the maximum to the mean SUV. The slope of the line between the different thresholds and MTV was calculated using the least squares method, and its absolute value represented the HF. An example of the measurement of HF is presented in Figure 2B. The tumor diameter was measured in the range where the intestinal wall thickened due to tumor accumulation in the intestine, and the range of increased radioactive uptake on PET images could also play an auxiliary role in the assessment. Independent measurements were made by two physicians with five or more years of experience in nuclear medicine, and the final diameter number was taken as the average of the two measurements. Lymph nodes with short diameters of ≥1.0 cm (17) or SUVmax ≥2.5 (18) were considered metastatic lymph nodes.
The PAT evaluation comprised the following: (I) Severity. No discernible anomaly in the PAT or a slight stripe shadow in the PAT was assigned a score of 0, whereas the degree of PAT was deemed to be comparable to the “stain sign” of peritoneal metastasis, receiving a score of 1. (II) The extent of the horizontal abnormal range of PAT. The objective was to ascertain whether the extent of PAT abnormalities in the cross-sectional images reached or exceeded the circumferential extent of tumor invasion of the intestinal wall. A score of 0 was not reached, whereas a score of 1 was reached or exceeded. (III) The extent of longitudinal abnormal range of PAT. The objective was to ascertain whether the range of PAT abnormalities in coronal or sagittal images reached or exceeded the diameter of the tumor, with a score of 0 not reached and a score of 1 reached or exceeded. The PAT grades were divided according to the cumulative scores of the aforementioned three items, with 0–1 point equating to PAT grade 1, 2 points equating to PAT grade 2, and 3 points equating to PAT grade 3. Examples of the PAT score of CRC are presented in Figure 3.

Statistical analysis
Statistical analysis was performed using the software SPSS 27.0 (IBM Corp., Armonk, NY, USA) and MedCalc version 20.0.27 (MedCalc Software, Ostend, Belgium). The normality of all continuous variables was evaluated through the implementation of the Shapiro-Wilk test. Univariate analysis was conducted using the two-independent-samples t-test for normally distributed continuous variables, Mann-Whitney U test for non-normally distributed continuous variables, and χ2 test for categorical variables to identify variables with statistically significant differences between the PNI and non-PNI groups. A multivariate logistic regression analysis was conducted to identify independent risk factors associated with PNI. Based on these factors, a nomogram model was constructed using the rms package in R 4.1.3 (R Foundation for Statistical Software, Vienna, Austria). The predictive efficacy of each independent risk factor and the nomogram model for PNI was assessed through the use of receiver operating characteristic (ROC) curve analysis. The internal validation was conducted using the bootstrap method with 1,000 resamples. The calibration curves and decision curve analysis (DCA) were applied to evaluate the accuracy and practical utility of the model, respectively. A P value of less than 0.05 was deemed statistically significant.

Results

Patient characteristics and baseline PET/CT quantitative parameters
Table 1 summarizes the baseline clinical and PET/CT parameters of the patients. Among the 116 CRC patients (male 66 cases, female 50 cases), 50 were in the PNI group and 66 in the non-PNI group. The median age was 56.9 years (range 25–100 years).
The univariate analysis demonstrated that MTV40% (Z=−2.21, P=0.027), TLG40% (Z=−2.21, P=0.027), TLG60% (Z=−2.14, P=0.033), CV (t=−4.18, P<0.001), and HF (Z=−2.31, P=0.021). Significant differences were observed between groups with regard to PAT grade (χ2=15.12, P=0.002), cN stage (χ2=18.18, P<0.001), and cM stage (χ2=5.77, P=0.016).

Subgroup analysis of PNI for CRC with M0-stage
Radical surgery is a common treatment for M0-stage CRC. Subgroup analysis was performed on these patients. Univariate analysis demonstrated that the differences in CV (t=−2.59, P=0.012), PAT grade (χ2=13.80, P=0.003), and cN stage (χ2=11.78, P<0.001) remained statistically significant between the PNI group and the non-PNI group (Table 2).

Multifactor logistic regression analysis
The variance inflation factor (VIF) was calculated to account for the multicollinearity problem. The variables MTV40% (VIF =409.69) and TLG60% (VIF =44.22) with VIF >5 were excluded, and the remaining variables with statistical differences in univariate analysis were analyzed by bivariate multivariate logistic regression. The results of the multivariate logistic regression analysis revealed that CV [odds ratio (OR) =3.128, 95% confidence interval (CI): 1.476–6.628, P=0.003], PAT grade (PAT grade 2: OR =12.016, 95% CI: 2.859–50.509, P<0.001; PAT grade 3: OR =22.417, 95% CI: 4.291–117.104, P<0.001), and cN stage (OR =4.769, 95% CI: 1.636–13.900, P=0.004) were independent risk factors for predicting PNI (Table 3).

Building the nomogram model and evaluation
A nomogram model was developed based on the CV, PAT grade, and cN stage (Figure 4). The area under the curve (AUC) values for CV, PAT grade, and cN stage were 0.705 (95% CI: 0.611–0.798), 0.702 (95% CI: 0.605–0.798), and 0.694 (95% CI: 0.605–0.798), respectively, as calculated by the ROC analysis. The nomogram model predicted the AUC of PNI to be 0.893 (95% CI: 0.837–0.949) (Figure 5). The DeLong test demonstrated that the AUC value of the nomogram model was significantly higher than that of CV (Z=4.17, P<0.001), PAT (Z=4.15, P<0.001), and cN (Z=5.05, P<0.001). The internal validation concordance index (C-index) was 0.861. The calibration curve demonstrated that the predicted probability of PNI by this model was in close alignment with the actual probability, and the calibration degree was high (Figure 6). The DCA curve demonstrated that the majority of the model’s curves were distant from the two extreme curves, and the probability range of the optional domain was considerable, which was better than that of CV, PAT grade, and cN stage (Figure 7).

Discussion
As the global burden of cancer continues to increase on an annual basis, the importance of cancer control is becoming increasingly evident. As one of the most prevalent malignant neoplasms, CRC remains the primary cause of mortality among cancer patients, and its prognosis has become a significant area of concern for clinicians (19). PNI is an important prognostic factor for CRC. Neoadjuvant therapy can reduce the positive rate of PNI (10), and postoperative chemotherapy can prolong the survival of CRC patients (20). Accurate assessment of PNI status before treatment is critical for effective identification of high-risk patients, as it can help clinicians to assess prognostic risk early and prepare for further intensive treatment (9).
The metabolic heterogeneity parameters of 18F-FDG PET/CT encompass AUC-CSH, CV, HI, and HF, which are correlated with tumor recurrence and metastasis (21). In comparison to radiomics features, these parameters are straightforward to obtain, safe, and effective, and easier to implement in clinical practice. At present, there have been studies using CT images to predict PNI, but few studies have investigated the predictive value of 18F-FDG PET metabolic heterogeneity parameters on PNI. In this study, it was found that CV and HF demonstrated statistically significant differences in the univariate analysis between PNI and non-PNI, whereas AUC-CSH and HI did not. This suggests that the metabolic heterogeneity parameters of 18F-FDG PET/CT may possess potential predictive value for PNI, although different parameters may be more applicable in different scenarios. The CV, defined as the ratio of the mean value to the standard deviation of the SUV, reflects the degree of variation of the SUV. An increase in SUV variability is indicative of a more aggressive tumor phenotype, which may facilitate tumor cell invasion into surrounding tissues, including vessels, lymphatics, and nerves. The results of the multivariate logistic regression analysis indicated that CV was an independent risk factor for predicting PNI, with an AUC value of 0.705 (95% CI: 0.611–0.798). This suggests that CV is the optimal parameter for predicting PNI. As documented in research reports, CV is associated with the CRC immune microenvironment, and the inflammatory response within the immune microenvironment is significantly correlated with patients’ relapse-free survival (22). Pellegrino et al. found that CV, which reflected the heterogeneity of glycolytic phenotype, was an independent prognostic factor for predicting the survival rate of patients with advanced non-small cell lung cancer (23). In this study, HF was defined as the absolute slope of the least square equation based on 40%, 60%, and 80% of MTV, which reflected the spatial heterogeneity of tumor metabolism. The correlation between HF and PNI status was identified, yet HF was not determined to be an independent risk factor.
SUVmax is the most commonly utilized parameter in PET/CT examinations, reflecting the degree of metabolic activity exhibited by tumor cells (24). SUVmax is a commonly utilized metric for differentiating between benign and malignant lesions and assessing the efficacy of therapeutic interventions (25). It is noteworthy that the univariate analysis of this study revealed no statistically significant difference in SUVmax between the PNI and non-PNI groups. One reason is that SUVmax only represents the specific value of the highest point of glucose metabolic activity in the tumor, and thus cannot reflect the overall metabolic state of the lesion. Another potential explanation is that glucose metabolism in colorectal lesions is influenced by a multitude of factors, including intestinal flora, inflammatory processes, and physiological uptake. Accordingly, the SUVmax in the lesion may be a superimposed state resulting from the influence of multiple factors, potentially leading to discrepancies in the representation of the tumor’s intrinsic glucose metabolism (26). In certain fields, the 18F-FDG PET/CT metabolic heterogeneity parameter is more effective than traditional metabolic parameters.
The low-dose CT images obtained in PET/CT examinations lack the tissue contrast of enhanced CT, which limits their utility in evaluating the T-stage of CRC. Fortunately, 18F-FDG PET/CT demonstrates high sensitivity for the N stage, with a superior diagnostic yield compared to that of enhanced CT and magnetic resonance imaging (MRI) (27,28). The presence of lymph node metastasis is indicative of the tumor’s high level of activity and aggressiveness, which may suggest an increased likelihood of PNI. This study’s findings align with previous research in that CRC with lymph node metastasis is more prone to developing PNI. The cN stage, when considered an independent risk factor for predicting PNI (29,30), demonstrated a sensitivity of 60.0% and a specificity of 78.79%.
A substantial body of evidence from numerous studies has demonstrated that PAT exerts a crucial influence on the regulation of the tumor microenvironment (31,32). A relationship exists between PAT and tumor progression, which involves crosstalk between the tumor and adjacent tissues, as well as their continuous metabolic exchange and interaction (15). Zoico et al. discovered a correlation between the morphological alterations observed in PAT and an increase in lipolysis and adiponectin production, as well as a heightened repair response to inflammatory macrophage infiltration (33). Conti et al. found that PAT is exposed to the mesenchymal transformation process by using MRI and optical imaging in vivo techniques. The mesenchymal transition is associated with obtaining a more aggressive tumor phenotype (15). These pathological alterations result in a range of morphological modifications in PAT. Although the utility of low-dose CT for T staging is limited, it does not impact the imaging of PAT. The PAT grade was constructed through an evaluation of the abnormal form representation of PAT. The univariate and multivariate analyses demonstrated that PAT grading is an effective method for reflecting PNI status. Furthermore, the grading system is straightforward to implement and readily comprehensible to clinicians.
In this study, the nomogram model that we established, based on CV, PAT grade, and cN stage, demonstrated a high degree of predictive value for PNI. The calibration curves and DCA demonstrated that the nomogram model exhibited high accuracy and clinical practicability. Radical surgical resection is the standard treatment for M0-stage CRC (34). Therefore, a subgroup analysis was conducted on M0-stage patients. The findings demonstrated that CV, PAT grade, and cN stage could effectively identify PNI status, suggesting a broad range of applications for these parameters.
This study was not without limitations. As a single-center study with a limited sample size, our findings may be constrained in their generalizability. The study relied on retrospective data collection, which may have introduced selection bias and incomplete information, potentially affecting the interpretation of the results. The efficacy of other metabolic parameters, excluding CV, as predictors of PNI of CRC with M0 stage, is limited, consequently restricting their practical application.

Conclusions
This study examined the capacity of 18F-FDG PET/CT to predict PNI in CRC. The findings indicate that CV, PAT grade, and cN stage are independent risk factors for PNI. Moreover, the nomogram constructed based on these factors demonstrated enhanced predictive accuracy for PNI. This study lays the foundation for the following clinical translations: (I) screening of high-risk patients: identifying the population that needs intensive treatment and avoiding under-treatment; (II) focusing on the risk of recurrence: strengthening the emphasis on follow-up; and (III) stratification of precise clinical trials: accelerating the accumulation of evidence for PNI-directed treatments.

Supplementary
The article’s supplementary files as