Mesentery morphological features on computed tomography for preoperative prediction of tumor invasion and lymph node metastasis in colon cancer

Abstract

BACKGROUND

Accurate identification of tumor invasion depth and lymph node (LN) involvement in patients with colon cancer (CC) is critical for guiding treatment strategies. However, the preoperative prediction of tumor invasion depth and LN metastasis in CC remains challenging. As the intestinal tumor develops, the fat density in the mesentery increases.

AIM

To investigate the efficacy of computed tomography (CT) value change in the mesentery contributed by the tumor (CT-T value) for predicting tumor invasion depth and LN metastasis.

METHODS

Patients, who were diagnosed with CC and underwent surgery, were included and divided into the training and validation cohorts. CT-T values of the mesentery were extracted from the CT images. Cutoff points were determined using the receiver operating characteristic (ROC) curve, and the area under the ROC curve was employed to assess the performance of the CT-T value for tumor invasion depth and LN status prediction.

RESULTS

Cutoff values of 11.83 and 17.17 were identified to discriminate T1/2 vs T3/4 and N0 vs N1/2, respectively. With a cutoff CT-T value of 11.83, the total diagnostic accuracy for T stage was 83.1% (81.5% for the training cohort and 86.2% for the validation cohort). With a cutoff CT-T value of 17.17, the total diagnostic accuracy for N stage was 77.3% (75.8% for the training cohort and 80.1% for the validation cohort), which was higher than that of CT-reported LN metastasis.

CONCLUSION

In this study, we explored an efficient method for predicting preoperative T and N stages using the tumor-contributed CT value of the mesentery in CC, which displayed superior predictive accuracy.

Key Words: Mesentery; Computed tomography; T stage; N stage; Colon cancer

Core Tip: Preoperative prediction of tumor invasion depth and lymph node (LN) metastasis of colon cancer (CC) remains challenging. As the intestinal tumor develops, the fat density of the mesentery increases. Our study utilized computed tomography (CT) value change in the mesentery contributed by the tumors to discriminate tumor invasion depth and LN metastasis. The total diagnostic accuracy for T stage was 83.1%, and for N stage was 77.3%, which was higher than the CT-reported LN metastasis. This efficient method for predicting preoperative T and N stages using tumor-contributed CT value of the mesentery in CC displayed superior predictive accuracy.

INTRODUCTION

Colon cancer (CC) is a common malignant tumor of the digestive tract, with 1142222 newly-diagnosed cases and 2838138 deaths every year[1]. The prognosis of CC is mainly stage-dependent. The accurate identification of tumor invasion depth and lymph node (LN) involvement in patients with CC influences treatment strategy decisions. Different imaging modalities, including computed tomography (CT), magnetic resonance imaging (MRI), and positron emission tomography-CT scanning, are often used to predict colorectal cancer (CRC) stage. The type and modality of the tests used for the preoperative assessment of CC differ from those used for rectal cancer (RC). While MRI is the standard of care for preoperative assessment of RC[2], CT scanning with intravenous contrast is commonly used for the preoperative assessment of CC in clinical practice. However, preoperative prediction of tumor invasion depth and LN metastasis in CC remains limited. A retrospective cohort analysis with 125473 CC patients in stages I-III showed that the concordance between clinical and pathologic T and N stages was unsatisfactory. The Kappa coefficients were 0.606 for the T stage and 0.506 for the N stage[3].

The mesentery is an adipose structure that provides support to the intestines. Visceral obesity, contributed by mesenteric and omental adipose tissue, is a risk factor for the incidence of CC[4], through chronic inflammation[5], adipokine release (for example, adiponectin and leptin), and angiogenesis[6]. As a structure that directly connects with the intestines, mesenteric adipose tissue often develops secondary inflammation following intestinal inflammation, which may lead to enlargement of mesenteric fat, called ‘creeping fat’[7]. Compared with controls, CD68-positive and CD163-positive cells were significantly increased in the ‘creeping fat’ of CRC patients[8]. Our previous work demonstrated a reduction in anti-tumor lymphocytes (natural killer and tissue resident CD8⁺ T cells) and an increase in immunosuppressive lymphocytes (regulatory T cells) within the ‘creeping fat’[9]. Therefore, the mesentery may play an important role in the development and progression of intestinal tumors.

The ‘creeping fat’ of the mesentery is edematous and densely infiltrated by inflammatory cells and hyperplastic adipocytes, leading to increased fat density on CT images[10]. In this study, we utilized CT value changes in the mesentery contributed by the tumors to discriminate tumor invasion depth and LN metastasis. Cutoff points were determined using the receiver operating characteristic (ROC) curve, which showed superior diagnostic accuracy. We also investigated the factors that influenced mesenteric CT values. This study highlights the clinical significance of the ‘creeping fat’ in CC and provides a new approach for the preoperative prediction of T and N stages.

MATERIALS AND METHODS

Patients

The study was performed in accordance with the guidelines outlined in the Declaration of Helsinki and approved by Guangdong Provincial People’s Hospital Ethics Review Committee. The informed consent requirement was waived in this retrospective study. The patients underwent surgery at Guangdong Provincial People’s Hospital between April 2020 and September 2023. The inclusion criteria were as follows: (1) Age ≥ 18 years with a diagnosis of CC based on the pathology of the tumors; (2) CT scan performed within 2 weeks before surgery; and (3) No neoadjuvant chemotherapy or radiotherapy. The exclusion criteria were as follows: (1) Presence of other cancers; (2) Unavailable preoperative CT data; (3) Receipt of neoadjuvant chemotherapy or radiotherapy; and (4) Concurrent Crohn’s disease (CD) or ulcerative colitis. All the clinical data were collected at the time of admission. Clinical information including sex, age, weight, height, glycosylated hemoglobin (GHb), carcinoembryonic antigen (CEA), carbohydrate antigen 19-9 (CA19-9), and blood lipid levels were collected from medical records. The abdominal area and CT values of the subcutaneous adipose tissues were measured on CT images. All the data were collected retrospectively adhering to the approved research protocol. A total of 348 patients with CC were enrolled in this study and divided into the training and validation cohorts according to the surgery time (the former two-thirds as the training cohort and the latter one-third as the validation cohort).

Tumor invasion and LN metastasis stages

In this study, T and N stages of CC referred to the 8^th edition of the American Joint Committee on Cancer Staging Manual[11]. We grouped T and N stages into binary outcomes (T1/2 vs T3/4, N0 vs N1/2) for predictive modeling.

Mesentery CT value measurement

To avoid the influence of contrast agents, non-contrast CT images were used to calculate the CT values of the mesentery. Tumor mesentery and non-tumor mesentery indicate the mesenteries that drain blood and lymph from the tumor and non-tumor tissues, respectively. Tissues in 2 cm sites from the colorectal tumors express more tumorigenesis and proliferation markers than those in 5 cm and 10 cm sites, which means the tissues in 2 cm away from the tumors can be influenced by the tumor microenvironment[12]. Since the mesentery can be more easily affected by the nearby tissues than the intestine due to its loose structure. So, we selected tissues ≥ 3 cm away from the tumor border. At the same time, non-tumor tissue needs to be located in the same segment with the tumor to make sure the basement microenvironment consistent in both tumor and non-tumor tissues. So, we selected the non-tumor tissue as 3-7 cm away from the tumor border. First, contrast-enhanced CT images were used to locate the tumor. Non-contrast images of tumors with the largest cross-sectional areas were then extracted for CT value measurement. The blood vessels on contrast CT images were used to determine the location and shape of the mesentery. The areas used for measuring CT values were within 3 cm of the intestinal wall border. Three points were chosen for CT value acquisition, and the mean value was calculated as the CT value of the tumor mesentery. Non-tumor mesentery CT values were acquired using the same method. The tumor-contributed CT value (CT-T value) = CT value of tumor mesentery-CT value of non-tumor mesentery. A diagram showing the procedure is presented in Supplementary Figure 1A. The CT images were reviewed by an experienced reader who was blinded to the clinical characteristics and postoperative pathological findings.

Abdominal area and CT value of the subcutaneous adipose tissues measurement

We measured the abdominal area on the CT scan without contrast using the transverse section. A region was initially delineated around the umbilicus (Supplementary Figure 1B), and the area was calculated using software in the picture archiving and communication system (Carestream, Canada). The CT value of the subcutaneous adipose tissue was randomly measured around the umbilicus three times, and the mean value was adopted (Supplementary Figure 1C).

Determination of CT-reported LN status

The CT images were reviewed by two experienced radiologists who were blinded to the clinical characteristics and postoperative pathological findings. An irregular border, larger than 1 cm, central necrosis, and a tendency to adhere to each other were signs of suspected LN metastasis. Disagreements were resolved through consultation.

Correlation analysis

The correlations between mesentery CT-T values and other clinical characteristics, which may influence the prognosis of CC, were analyzed. We included basic clinical characteristics, such as age, weight, height, body mass index (BMI); laboratory tests parameters, such as GHb, CEA, CA19-9, and blood lipid. Since the subcutaneous adipose tissue radiodensities were linearly associated with all-cause mortality of CRC[13], the CT values of the subcutaneous adipose tissues were also included. Those variables with more than 10% missing values were excluded. Spearman analysis was applied. False discovery rate (FDR) correction (q < 0.05) was applied to all Spearman tests.

Statistical analysis

Statistical analyses were conducted using the R software (version 4.2.3). Continuous variables were expressed as mean ± SD or median (interquartile range) for normally or non-normally distributed values, respectively. For continuous variables, the t-test or Mann-Whitney U test was used. Categorical variables were compared using the χ² test or Fisher’s exact test. The ROC curve and the area under the ROC curve (AUC) were employed to assess the performance of the CT-T value for tumor invasion depth and LN status prediction according to the method described by DeLong et al[14]. The coords function in R was used to determine the optimal threshold on the ROC curve. Statistical significance was set at P < 0.05.

RESULTS

Clinical characteristics

In total, 348 patients diagnosed with CC were included in the study. Patient characteristics in the training and validation cohorts are presented in Table 1. In the entire cohort, 84.2% of patients were T3/T4 stage and 45.9% were confirmed to have LN metastases. The percentages of patients in the T3/T4 stage were 81.9% and 88.8% in the training and validation cohorts, respectively. LN metastasis was identified in 43.5% of patients in the training cohort and in 50.9% of patients in the validation cohort. There were no significant differences in clinical and pathological characteristics between the training and validation cohorts (all P > 0.05). We also displayed the baseline characteristics stratified by the dichotomized outcome (T1/2 vs T3/4; N0 vs N1/N2) in the Supplementary Tables 1 and 2. Compared with T1/2 and N0, higher levels of CEA and CA19-9 were observed in T3/4 and N1/N2 (P < 0.05, Supplementary Tables 1 and 2). Higher weight, BMI and lower periumbilical CT were observed in T1/2 patients (P < 0.05, Supplementary Tables 1 and 2).

Table 1 Patient characteristics in the training and validation cohorts, n (%).

	Training cohort (n = 232)	Validation cohort (n = 116)	P value
Age (years)	65 (18)	67 (15)	0.255
Gender			0.402
Male	131 (56.5)	60 (51.7)
Female	101 (43.5)	56 (48.3)
Height (cm)	163.00 (13.0)	162.00 (14.0)	0.553
Weight (kg)	60.00 (14.40)	58.75 (16.78)	0.116
Body mass index (kg/m2)	22.84 (4.18)	22.24 (3.37)	0.185
Carbohydrate antigen 19-9 (U/mL)	11.59 (13.29)	12.44 (20.23)	0.206
Carcinoembryonic antigen (ng/mL)	3.32 (6.61)	4.08 (9.81)	0.074
Total cholesterol (mmol/L)	4.72 (1.28)	4.62 (1.45)	0.747
Triglyceride (mmol/L)	1.16 (0.61)	1.25 (0.58)	0.257
High-density lipoprotein cholesterol (mmol/L)	1.02 (0.34)	1.10 (0.29)	0.075
Low-density lipoprotein cholesterol (mmol/L)	3.07 (1.00)	2.99 (1.26)	0.651
Glycosylated hemoglobin	5.80 (0.70)	5.70 (0.70)	0.237
Periumbilical area	547.5 (157.4)	535.00 (153.53)	0.72
Periumbilical computed tomography (HU)	-110.33 (8.33)	-111.67 (7.67)	0.755
Diabetes mellitus			1.0
Yes	40 (17.2)	20 (17.2)
No	192 (82.8)	96 (82.8)
Dyslipidemia			1.0
Yes	173 (77.2)	84 (77.1)
No	51 (22.8)	25 (22.9)
Primary site			0.29
Cecum-ascending colon	68 (29.3)	43 (37.1)
Transverse colon	30 (12.9)	14 (12.1)
Descending colon	34 (14.7)	10 (8.6)
Sigmoid colon	100 (43.1)	49 (42.2)
Pathology			0.196
Adenocarcinoma	219 (94.4)	108 (93.1)
Mucinous adenocarcinoma	12 (5.2)	5 (4.3)
Signet ring cell carcinoma	1 (0.4)	3 (2.6)
T stage			0.326
T1	13 (5.6)	4 (3.4)
T2	29 (12.5)	9 (7.8)
T3	167 (72.0)	87 (75.0)
T4	23 (9.9)	16 (13.8)
N stage			0.33
N0	131 (56.5)	57 (49.1)
N1	69 (29.7)	37 (31.9)
N2	32 (13.8)	22 (19.0)
Mesentery location			0.703
Mesenteric	164 (70.7)	79 (68.7)
Antimesenteric	68 (29.3)	36 (31.3)
Lymph node metastasis report			0.819
Yes	129 (55.6)	66 (56.9)
No	103 (44.4)	50 (43.1)

Cutoff mesenteric CT-T value to predict T and N stages

In the training cohort, the CT-T value increased as the tumor invaded deeper and metastasized to the LNs (Figure 1A and B). The ROC curve was used to identify the cutoff point of the CT-T value to predict the tumor invasion depth (T1/2 vs T3/4 stage) and LN metastasis (N0 vs N1/2 stage) in the training cohort. To discriminate T1/2 vs T3/4 stages, the cutoff CT-T value was 11.83, which yielded an AUC of 0.86 (95%CI: 0.80-0.92) (Figure 1C). For predicting LN metastasis (N0 vs N1/2), the cutoff CT-T value was 17.17, and the AUC was 0.8 (95%CI: 0.74-0.85) (Figure 1D).

Figure 1 The Box-whisker plots and receiver operating characteristic plots for tumor-contributed computed tomography value to discriminate T and N stages in the training cohort. A: The Box-whisker plot for tumor-contributed computed tomography value (CT-T value) in patients with T1/2 vs T3/4 stages; B: The Box-whisker plot for CT-T values in patients with N0 vs N1/2 stages; C: The receiver operating characteristic (ROC) plot for CT-T value to discriminate T1/2 vs T3/4 stages; D: The ROC plot for CT-T value to discriminate N0 vs N1/2 stages (^aP < 0.01). AUC: Area under the receiver operating characteristic curve; CT-T: Tumor-contributed computed tomography.

Diagnostic validation of the mesenteric CT-T value

The validation cohort was used to test the prediction performance of the cutoff CT-T values determined by the training cohort. In the validation cohort, the CT-T value also increased as the tumor invaded deeper and metastasized to the LNs, consistent with the findings in the training cohort (Figure 2A and B). An AUC of 0.79 was obtained for T stage (T1/2 vs T3/4 using the cutoff of 11.83) and an AUC of 0.8 was obtained for N stage (N0 vs N1/2 using the cutoff of 17.17), respectively (Figure 2C and D). With a cutoff CT-T value of 11.83, the total diagnostic accuracy for T stage of the CT-T value was 83.1% (81.5% for the training cohort and 86.2% for the validation cohort) (Supplementary Table 3). With a cutoff CT-T value of 17.17, the total diagnostic accuracy for N stage of the CT-T value was 77.3% (75.8% for the training cohort and 80.1% for the validation cohort) (Supplementary Table 3).

Figure 2 The Box-whisker plots and receiver operating characteristic plots for tumor-contributed computed tomography value to discriminate T and N stages in the validation cohort. A: The Box-whisker plot for tumor-contributed computed tomography value (CT-T value) values in patients with T1/2 vs T3/4 stages; B: The Box-whisker plot for CT-T values in patients with N0 vs N1/2 stages; C: The receiver operating characteristic (ROC) plot for CT-T value to discriminate T1/2 vs T3/4 stages; D: The ROC plot for CT-T value to discriminate N0 vs N1/2 stages (^aP < 0.01). AUC: Area under the receiver operating characteristic curve; CT-T: Tumor-contributed computed tomography.

Diagnostic accuracy of the CT report for LN metastasis

As shown in Supplementary Table 3, the CT report accuracy for discriminating LN involvement was 58.4% for the entire cohort (58.6% for the training cohort and 57.8% for the validation cohort). We then compared the diagnostic accuracy of the CT report and mesenteric CT-T value in predicting LN metastasis. The cutoff CT-T value of 17.17 yielded a diagnostic accuracy of 77.3% in the entire cohort, which was higher than that of the CT report (Supplementary Table 3).

Correlation between clinical characteristics and mesenteric CT-T value

The mesentery status of the human body depends on many factors such as weight, abdominal circumference, and blood lipid levels. We investigated the relationship between clinical characteristics and mesenteric CT-T values and found that the level of CEA was positively correlated with the CT-T value (r = 0.23, FDR adjusted P < 0.001) (Figure 3A and B).

Figure 3 Correlation between clinical characteristics and mesenteric tumor-contributed computed tomography value. A: Correlation analysis between clinical characteristics and mesenteric tumor-contributed computed tomography value (CT-T value) value; B: Correlation plot between carcinoembryonic antigen and mesenteric CT-T value. BMI: Body mass index; CA19-9: Carbohydrate antigen 19-9; CEA: Carcinoembryonic antigen; CT: Computed tomography; CT-T: Tumor-contributed computed tomography; FDR: false discovery rate; GHb: Glycosylated hemoglobin; HDL-C: High-density lipoprotein cholesterol; LDL-C: Low-density lipoprotein cholesterol; TC: Total cholesterol; TG: Triglyceride.

DISCUSSION

Non-invasive methods to accurately predict the tumor invasion depth and LN status before surgery are urgently needed to optimize individualized therapy for CC. The development of ‘creeping fat’ in the mesentery depends on the severity of intestinal inflammation[15]. In our previous study, intestinal tumors were found to modify the immune microenvironment of the ‘creeping fat’, shifting it towards a tumor-promoting state[9]. In the present study, we used the tumor-contributed CT value of the ‘creeping fat’ to predict the T and N stages. Cutoff values of 11.83 and 17.17 were identified to discriminate T1/2 vs T3/4 and N0 vs N1/2, respectively, which displayed satisfactory predictive accuracy in both the training and validation cohorts.

In CD, mesenteric changes, such as hypervascularity, edema, fibrofatty proliferation, and increased fat density, identified on CT enterography, provide a gauge of intestinal disease manifestations[10,16-18]. As a conduit for blood and lymphatic vessels and the enteric nervous system of the intestine, the mesentery can act as a potential site for residual tumors, leading to local recurrence of CC after surgery[19,20]. The technique of complete mesocolic excision (CME), which was proposed by Hohenberger et al[21] in 2009, introduced a further concept: The importance of preserving mesocolic integrity and achieving complete removal. The largest meta-analysis of 26640 patients found that both overall and disease-free 5-year survival rates were higher in patients who underwent CME resection compared with conventional resection[22]. Furthermore, tumor circumferential localization on the colonic wall [mesenteric (M) or antimesenteric (AM)] can influence the prognosis of patients with CC. M-located tumors are associated with higher numbers of metastatic LNs than AM tumors[23]. Mesenteric tumors appear to have a shorter survival time[24]. Thus, future evaluations of CC must consider mesenteric-based events.

Locoregional staging for CC has become more important, as neoadjuvant therapy has the potential to downsize locally advanced tumors. CT is the most frequently used method to evaluate patients for CC screening and staging due to its good clinical performance, safety profile, and cost-effectiveness. It has been reported that the intestinal wall can be classified by degree and is associated with T stage. According to the Utano’s study, intestinal wall deformities are classified into arc, trapezoid, and apple-core types, which are primarily associated with T1, T2, and T3/T4, respectively. The overall accuracy for T stage was 79%[25]. The use of both transverse and multiplanar reconstructed images extracted from contrast-enhanced CT improves the diagnostic accuracy to 83%[26]. In the present study, we explored an efficient method to predict preoperative T1/2 vs T3/4 using the tumor-contributed CT value of the mesentery, and achieved an accuracy of 83.1%.

Compared with predicting the T stage, discriminating LN status before surgery is more challenging. An irregular border, larger than 1 cm, central necrosis, and a tendency to adhere to each other are signs of suspected LN metastasis[27-30]. The overall accuracy of predicting LN status based on contrast-enhanced CT has been reported to range from 59% to 71%[26,31]. As a conduit for lymphatic vessels of the intestine, the mesentery acts as an extension of local intestinal inflammation. We used the CT-T value of the mesentery for the prediction of N0 vs N1/2 and obtained an accuracy of 77.3%, which was higher than the CT-reported accuracy. However, the baseline CT-T values differed in the training and validation cohorts, which may have been caused by an imbalance in the sample size, as well as heterogeneity in the subjects. Cohort with larger sample size is needed to establish a more stable model.

In this study, we investigated the factors that can influence the CT-T value. The most positively related parameter was CEA. It is reported that high preoperative CEA levels predict poor disease-free survival, overall survival, and increased risk of recurrence and metastasis of CRC patients and correlated with poor prognosis[32,33]. The most negatively related parameter was high-density lipoprotein cholesterol (HDL-C), however, without significance. Epidemiological studies have reported a protective role of HDL-C in cancers, which is inversely associated with the risk of lung, colorectal, and liver cancers[34-36]. The inverse correlation between the CT-T value and HDL-C level indicated that the protective mechanism of HDL-C in CC may depend on its influence on the mesentery.

However, there are several drawbacks to the mesenteric CT-T value in predicting T and LN status. This method is not applicable in thin people with less visceral fat, as it is difficult to discriminate between the mesentery and colonic wall in these individuals. Tumors with poor differentiation, such as signet ring cell carcinoma, tend to have LN metastasis before changing the mesentery. This may decrease the prediction accuracy. In addition, circumferential localization of the tumor on the colonic wall (M or AM) may also influence diagnostic efficacy. Finally, the measurement error or inconsistent sensitivity of radiologists may also be a confounding factor. Therefore, our results should be further validated using a larger population and multicenter prospective research.

CONCLUSION

Preoperative prediction of tumor invasion depth and LN metastasis of CC remains challenging. In this study, we explored an efficient method to predict preoperative T1/2 vs T3/4 and N0 vs N1/2 stages using the tumor-contributed CT value of the mesentery in CC, which displayed superior predictive accuracy.