Oncological and prognostic impact of lymphovascular invasion in Colorectal Cancer patients

Objectives: Lymphovascular invasion (LVI) is correlated with unfavorable prognoses in several types of cancers. We aimed to identify the informative features associated with LVI and to determine its prognostic value in colorectal cancer (CRC) patients. Methods: We retrospectively analyzed 1,474 CRC patients admitted in Wuhan Union Hospital between 2013 and 2017 as the development cohort and 549 CRC patients from The Cancer Genome Atlas (TCGA) database as the validation cohort. Logistical and Cox regression analyses were conducted to determine the oncological and prognostic significance of LVI in CRC patients. A survival nomogram based on LVI status was established using the Wuhan Union cohort and validated using TCGA cohort. Results: The LVI detection rates were 21.64% in the Wuhan Union cohort and 35.15% in TCGA cohort. LVI was closely correlated with advanced T stage, N stage, and TNM stage. LVI positivity was an independent biomarker for unfavorable overall survival (hazard ratio [HR]=2.25, 95% confidence interval [CI]=1.70-2.96, P<0.0001) and worse disease-free survival (HR=2.34, 95% CI=1.76-3.12, P<0.0001) in CRC patients. The survival nomogram incorporating LVI exhibited good predictive performance and reliability in the Wuhan Union cohort and TCGA cohort. Conclusion: LVI is a significant indicator of advanced stage and is remarkably correlated with worse prognosis in CRC patients. The survival nomogram incorporating LVI may assist clinicians to better strategize the therapeutic options for patients with CRC.


Introduction
Colorectal cancer (CRC) is one of the most common types of malignant tumors occurring in the gastrointestinal tract and the main cause of cancer-related death [1][2][3]. Despite the great advances in surgical and targeted therapies [4], the long-term survival of CRC patients with metastasis is far from our expectation [5]. A specific pathological variable associated with CRC metastasis is lymphovascular invasion (LVI) [6]. LVI is defined as the presence of tumor cells within the lymphatic or vascular channels [7], which is a common histopathological finding in CRC. LVI is an early sign of lymph node metastasis and increases the risk for micrometastasis in patients with localized CRC [8]. Hence, clarifying the oncological impact and prognostic significance of LVI is of great significance for patients with CRC [9,10].
Although a panel of clinical studies have explored the association between LVI and survival outcomes in patients with CRC, the results were inconsistent [7,11,12]. No study has established a survival nomogram based on LVI for accurate stratification of CRC patients with high risk for poor outcomes. More importantly, the clinical features closely related to the incidence of LVI are still unknown. In our previous study [13], we investigated the prognostic value of another ominous pathologic feature, perineural invasion, of CRC. In this study, we aimed to identify the critical clinical characteristics Ivyspring International Publisher associated with LVI and to create a survival nomogram based on LVI status for patients with CRC.

Study population from Wuhan Union cohort
Newly diagnosed patients with confirmed CRC admitted in Wuhan Union Hospital between July 2013 and September 2017 were enrolled in this study [13]. Data on demographic characteristics, tumor markers, staging, pathology, treatment, and survival outcomes (overall survival [OS] and disease-free survival [DFS]) were retrospectively collected from each patient. As for follow-up, the frequency was twice a half year after surgical resection, and then the frequency was twice a year. OS refers to the period from the date of surgical resection to the time of death by any cause. DFS refers to the period from the time of surgical removal to the date of recurrence or death by CRC. Staging was implemented based on the 8 th edition of the American Joint Committee on Cancer TNM staging system. Patients with histologically diagnosed CRC, patients whose LVI was diagnosed based on a postoperative histopathology report, and CRC patients who underwent surgical treatment were included in the study. By contrast, patients whose CRC was complicated with other malignant tumors and CRC patients with missing critical information were excluded. Moreover, children patients and patients without surgical resection were also excluded from this study. A total of 1,474 CRC patients with intact data were included in this study. Informed consent was obtained from all participants prior to the commencement of the study. The study was reviewed and approved by the ethics Committee of Wuhan Union Hospital (no. 2018-S377).

CRC patients from The Cancer Genome Atlas (TCGA) database
For independent validation, TCGA database [14] was searched to screen CRC patients who fulfilled the inclusion criteria. Over 549 CRC patients with complete clinical information and follow-up data were identified from the database. A total of 1,474 CRC patients from Wuhan Union Hospital were utilized as the development cohort (Wuhan Union cohort), while 549 CRC patients from TCGA database were used as the validation cohort (TCGA cohort).

Development of the survival nomogram
The clinical features, which were significantly correlated (P<0.05) with the OS of CRC patients, were identified by conducting a multivariable Cox analysis in the development cohort. These characteristics were selected for the construction of an OS nomogram. The predictive discrimination of the OS nomogram was determined using the receiver operating characteristic (ROC) analysis. The calibration ability of the OS nomogram, as reflected by goodness of fit, was measured using a calibration curve. A DFS nomogram was also established in the same manner.

Statistical analysis
All statistical analyses were performed using the SPSS 21.0 and R software (version 3.1.1). The chisquare test was employed to analyze the differences in categorical indexes between LVI+ and LVIpatients. Multivariable logistic regression was utilized to identify the independent variables affecting LVI+. Cox regression analysis was adopted to identify the potent prognostic factors in patients with CRC. ROC analysis was performed to assess the predictive ability of the survival nomogram. The Kaplan-Meier method and log-rank test were used to estimate the prognostic significance of LVI and the nomogram. Two-sided P values of less than 0.05 were considered significant.

Features of the LVI+ and LVI-groups
LVI tumors were identified in 319 (21.64%) of the 1,474 CRC patients with CRC in the Wuhan Union cohort. As listed in Table 1, CRC patients with LVI+ exhibited more advanced T stage (P<0.001), N stage (P<0.001), and TNM stage (P<0.001) than those with LVI-. Among 549 CRC patients in TCGA cohort, 193 patients (35.15%) showed LVI+. Similar to the Wuhan Union cohort, CRC patients with LVI+ showed more advanced T stage (P=0.003), N stage (P<0.001), and TNM stage (P<0.001) than those with LVI-. Hence, the presence of LVI was closely correlated with aggressive tumor behavior in patients with CRC.

Factors that independently affected the incidence of LVI
To determine which clinical variables could independently affect the incidence of LVI, univariable and multivariable logistical regression analyses were performed in the Wuhan Union cohort ( Table 2). The multivariable logistical regression analysis showed that T4 stage (P=0.017), N3 stage (P<0.001), stage IV (P<0.001), and absence of radiotherapy (P=0.015) were independent variables affecting the incidence of LVI in patients with CRC.

Establishment of the survival nomogram
A Cox regression model was employed to explore the influence of LVI and other covariates on OS in patients with CRC from Wuhan Union cohort. As shown in Table 3, LVI (P<0.001), T4 stage (P=0.031), stage IV (P=0.016), absence of adjuvant chemotherapy (P<0.001), and carcinoembryonic antigen (CEA) (P<0.001) were all independent risk factors for unfavorable OS in patients with CRC. As shown in Figure 2A, the OS nomogram included five risk factors that may increase the probability of having a poor OS. The evaluative indexes, such as predictive performance measured using the AUC (1 year, 3 years, and 5 years: 0.82, 0.786, and 0.736, respectively), ( Figure 3A) and calibration curves showed good agreement ( Figure 4A-C). With regard to the DFS (Table 3), multivariable Cox analysis identified four critical variables that were significantly correlated with DFS, including LVI (P<0.001), T stage (P=0.021), TNM stage (P<0.001), and CEA (P=0.005). We also adopted the four critical indexes in order to develop a DFS nomogram for CRC patients ( Figure  2B). The 1-year, 3-year, and 5-year AUC values of td-ROC were 0.876, 0.823, and 0.817, respectively ( Figure 3C). In addition, the calibration curves showed that the 1-year, 3-year, and 5-year OS rates between the predicted DFS nomogram and actual observed values exhibited good concordance ( Figure  4D-F).

Validation of the survival nomogram in TCGA cohort
First, we used TCGA cohort to externally verify the discrimination and calibration of the OS nomogram. As shown in Figure 3B

Discussion
In this study, we initially investigated the oncological significance of LVI in CRC patients who underwent surgical resection and found that LVI was a histologic index of advanced stage in CRC. T4 stage, N3 stage, stage IV, and absence of radiotherapy were independent risk factors for LVI, which were never reported in the previous studies. Then, we further probed the prognostic value of LVI both in the Wuhan Union cohort and TCGA cohort, and the results were almost consistent. Finally, we constructed a survival nomogram based on the LVI status for the risk stratification of CRC patients who underwent surgical treatment in the Wuhan Union cohort. We also validated the survival nomogram using the clinical data from TCGA cohort. Indeed, this multicenter retrospective analysis systemically illustrated the oncological impact of LVI and was the first to establish a survival nomogram incorporating the LVI status for the risk stratification of patients with CRC.
The incidence of LVI was not unified in the published studies. Kim et al. [15] demonstrated an LVI incidence rate of 8.5% in patients with stage I CRC. A clinical study based on the Swedish colorectal cancer registry [11] reported an LVI incidence rate of 15% in patients with stage II CRC. A recent study from Germany showed an LVI detection rate of 22.0% in patients with stage II CRC. A clinical study conducted in 3,707 stage I-III CRC patients from South Korea reported an LVI incidence of 39.7%, which was highest among the rates reported in previous studies related to CRC. Another retrospective study [12] revealed that the overall detection rate of LVI was 12.3% among patients with CRC. A previous study [7] examining the National Cancer Data Base reported an LVI detection rate of 26.3%. In our study, the LVI detection rates were 21.64% in the Wuhan Union cohort and 35.15% in TCGA cohort, which was in line with the results of other studies related to LVI. More importantly, no study has determined the independent risk factors for LVI. The multivariable logistical regression proved that T4 stage, N stage, stage IV, and absence of radiotherapy were closely correlated with the occurrence of LVI, which can serve as a reference for predicting the risk of LVI in CRC patients who underwent surgery.    Although several studies have already reported the association between the presence of LVI and CRC, most of them investigated this association in patients with a specific stage, such as stage I [15], stage II [16] [11], and stage I-III [17]. Unfortunately, no study has evaluated the correlation of LVI with oncological and prognostic implications in CRC patients who underwent surgical resection. Our study included CRC patients in stage I-IV treated with surgical resection, and our results were more representative than those of previous studies. More importantly, we appropriately utilized TCGA database to validate the results of the Wuhan Union cohort and obtained similar results, implicating that our survival nomogram can be applicable to different populations. From this standpoint, our study could be viewed as a multicenter study with sufficient CRC patients. The College of American Pathologists recommended the assessment of CRC patients for presence of LVI due to its highly important clinical significance [18]. However, the clinical significance of LVI was somewhat underestimated by some studies. A great number of survival nomograms were created for the risk stratification of CRC patients with unfavorable outcomes [19][20][21][22][23][24][25]. However, none of them included the significant features of LVI. In our study, a multivariable Cox regression was used to identify the clinical indexes that were significantly correlated with survival in CRC patients. Unsurprisingly, LVI positivity was a potent prognostic marker for both poor OS and DFS, and this finding was similar to the survival analysis results of other clinical studies. Encouragingly, the survival nomogram incorporating LVI obtained more excellent predictive performance in the prediction of survival than the TNM stage. Furthermore, when validated in TCGA cohort, the survival nomogram still possessed an extremely good predictive performance.
Although this is a multicenter research with relatively large sample size, two limitations still exist. One of the limitations was the retrospective nature of the study, and several relevant variables, such as tumor budding and microsatellite instability, were not included in our analysis. Another limitation was that the survival nomogram for CRC patients was not validated in the perspective cohort. Hence, well-designed perspective studies investigating the oncological and prognostic significance of LVI in CRC patients are warranted in the future.

Conclusion
Our data showed that LVI may serve as a significant indicator for aggressive tumor behavior and is remarkably correlated with worse prognosis in CRC patients. Moreover, the survival nomogram containing LVI obtained an extremely good predictive ability for predicting OS and DFS in CRC patients, indicating that the survival nomogram can be utilized as a useful prognostic system for individual estimation of prognosis.

Ethics Committee Approval and Patient Consent
The study plan was checked and approved prior to the beginning of the research by the ethics Committee of Wuhan Union Hospital (No. 2018-S377). The informed consent was obtained from the all the participants.

Data Sharing Statement
All data in our study are available from the corresponding author upon reasonable request.