The study retrospectively collected data of rectal cancer patients with no prior history of CVD, who were initially diagnosed from January 1, 2010 to December 31, 2020, in the First Affiliated Hospital of Gannan Medical University. The composite primary outcome was the incidence of one or more CVDs after treatment for rectal cancer patients during inpatient hospital visits. The cardiovascular complications of rectal cancer treatment can be divided into nine categories (19): 1) Coronary artery disease; 2) Arrhythmia; 3) Congestive heart failure; 4) Valvular disease; 5) Pulmonary hypertension; 6) Thrombotic disease; 7) Peripheral vascular disease; 8) Stroke; 9) Pericardial complication. According to the tenth International Classification of Diseases (ICD-10 codes), patients with a first-time diagnosis of rectal cancer (C19-C20) and with CRP measurement data were eligible. The patients were excluded according to the following criteria: 1) Missing CRP data; 2) Occurrence of an acute infection; 3) Having other serious and fatal diseases; 4) Those who did not confirm the diagnosis, were discharged, transferred, and uncooperative with the treatment. The flow chart of the study population recruitment is shown in Figure 1 .

From January 1, 2015 to December 31, 2020, data on patients with CVD was collected from the First Affiliated Hospital of Gannan Medical University. A total of 175 patients with CVD were identified through 1:1 matching on gender, age, and nine categories of CVD of rectal cancer patients with CVD.

The following indicators were extracted from the electronic medical record: age, gender, CRP, chemotherapy, neutrophils, lymphocytes, white blood cell (WBC) count, eosinophils, basophils, carcinoembryonic antigen (CEA), hyperglycemia, blood lipids, an infection history, and surgical therapy. In the laboratory, serum CRP is determined by means of transmission turbidimetry. For rectal cancer patients treated with chemotherapy, the records of their serum CRP levels were extracted more than two weeks after the chemotherapy. For the indicators of blood analytes, the patients were categorized as normal and abnormal according to laboratory results (including neutrophils, lymphocytes, WBCs, eosinophils, basophils, and blood lipids). Hyperglycemia was defined as a whole blood glucose level greater than 150mg/dl or 8.5mmol/l without diabetes (20); Dyslipidemia was defined as one of the following conditions according to the 2016 Chinese Adult Dyslipidemia Prevention Guidelines (21): total cholesterol ≥ 6.2mmol/L; triglyceride ≥ 2.3mmol/L; low-density lipoprotein cholesterol  ≥4.1mmol/L; and high-density lipoprotein cholesterol < 1.0mmol/L. The data extracted in this study were only relevant medical records obtained from patients and did not involve personally identifiable information, which met the ethical review requirements.

CRP, a continuous variable, was logarithmically transformed. Then the intersection of the two normal curves of the two groups (with and without CVD in rectal cancer patients) was used to determine the cut-off value of CRP (3.3 mg/L). Categorical variables were expressed as frequencies (percentages) and were compared for intergroup differences using the χ² test. Normally distributed continuous variables between two groups were compared using a t-test. Otherwise, the Kruskal-Wallis test was applied. The multivariable logistic regression model was used to identify the independent effects of CRP on CVD in patients with rectal cancer using the significant variables (p<0.05) in the univariable logistic regression analysis. The matching of the propensity score was performed in a 1:1 ratio to adjust for confounding factors and further determine the independence of CRP. The discrimination capability of the prediction model was assessed by examining the change in the area under the receiver operator characteristic curve (ROC AUC) and net reclassification improvement (NRI) when CRP was included in multivariable models. Multiple logistic regression models were built to assess the association between the risk of CRP and coronary artery disease. Subgroup analysis based on gender was used to estimate a specific association between CRP and CVD. In all regressions, odds ratios (ORs) and their 95% confidence intervals (95% CIs) were calculated. All analyses were performed using the R 4.2.2 software. The ggplot2 package was used to make the violin and box plots. Statistical significance was considered as a two-sided p < 0.05.

To gain a better understanding of the mechanism that determines CRP expression levels based on the genotype, we retrieved gene expression profile datasets from the Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo/). The R limma package was then used to identify differentially expressed genes, and ClusterProfiler of the R package was employed for Gene Ontology (GO) analysis. Moreover, the Kyoto Encyclopedia of Genes and Genomes (KEGG) Pathway gene annotation was obtained from the Database for Annotation, Visualization, and Integrated Discovery (DAVID) tool (https://david.ncifcrf.gov/home.jsp). All functions with a p-value of less than 0.05 were considered statistically significant.

GSE32384 (22) had 24 liver tissue samples from colorectal liver metastases (CRLM), including 19 samples with oxaliplatin-based chemotherapy and 5 samples without any chemotherapy. We first assessed the difference in CRP levels between the chemotherapy and the non-chemotherapy groups. Subsequently, we conducted DEGs analysis between two groups (FDR <0.1). Functional enrichment analysis (GO biological process) was conducted on the identified DEGs to discover biological functions involved in the CRP gene.

GSE20060 (23) had 385 samples (including replicates of the same sample) from human aortic endothelial cells treated with or without oxidized phospholipids. After averaging the expressed values of the same samples, a total of 96 control and 96 treatment samples were obtained. Then DEGs were calculated between the two groups (FDR <0.01) and the CRP-related genes from DEGs were identified. GO and KEGG pathway analyses were used for enrichment analysis.

We also identified hub genes by selecting common DEGs that were significant in both datasets, thus providing an additional list of gene functions.

A total of 2,985 patients with rectal cancer were diagnosed from 2010 to 2020, of whom 827 met the mentioned exclusion criteria and were selected for our study, including 175 with CVD and 652 without CVD following treatment. The clinical characteristics of the two groups were represented in Table 1 . Afterwards, we analyzed CRP levels among nine categories of CVD in rectal cancer patients with CVD and in patients with CVD, respectively. The nine categories of CVD among these patients were shown in Table 2 . The results showed that there were no differences in CRP levels among the nine categories in the two populations ( Figures 2A, B ), indicating a sensible combination of these categories for CVD.

To recognize whether CVD in rectal cancer patients was distinct from that in CVD patients, we compared the CRP levels between the two groups. The results showed that the CRP levels were significantly higher in rectal cancer patients with CVD than those in patients with CVD ( Figure 2C ). This finding indicated that rectal cancer patients with CVD should not be confused with CVD patients.

To further identify the relationship between CRP and CVD in patients with rectal cancer, we conducted a univariate logistic regression analysis. The results found that increased CRP was significantly associated with CVD in rectal cancer patients (OR: 2.767, 95% CI 1.646-4.592, p < 0.001). Additionally, the other five variables were significant factors in the univariate analysis, as shown in Table 3 . To control for the potential influence of these variables on CRP, we performed multivariable logistic regression using the significant factors in the univariable regression. The results showed that CRP remained a risk factor for CVD in patients with rectal cancer (OR: 2.411, 95% CI: 1.378-4.165, p=0.002). Furthermore, 1:1 propensity score matching with a caliper value of 0.1 was performed to verify the stability of the above findings. The propensity score matched 332 pairs of rectal cancer patients with and without CVD. The results consistently demonstrated that CRP was considered as an independent risk factor for CVD in patients with rectal cancer ( Table 4 ).

To evaluate the value of CRP in the diagnosis of CVD in rectal cancer, ROC AUC and NRI were described. When adding CRP to multivariable models including age, CEA, infection history, and chemotherapy, the ROC AUC showed an increase of 0.17 compared to the model without CRP ( Figure 2D , Table 5 ). The addition of CRP to risk factor model also improved NRI by 6.9% (95% CI 5–14%). These results demonstrated that CRP improved the accuracy of diagnosing CVD in patients with rectal cancer ( Table 5 ).

Based on the fact that coronary artery disease (CAD) accounts for the highest proportion of CVD mortality (estimated 1 in 4 deaths) (24, 25), we conducted an investigation into the association between CRP and the risk of CAD. As would be expected, the results showed that CRP was associated with a higher risk of CAD in model 1 (OR: 4.217, 95% CI 1.885-9.434, p< 0.001), and in model 2 (OR: 3.531, 95% CI 1.533-8.134, p=0.005) when compared to those non-CAD ( Table 6 ).

In view of systemic differences between men and women (26, 27), subgroup analyses by gender were performed to determine a specific association of CRP with CVD. The results showed that CRP was a predictor of CVD risk in men, whereas no correlation between CRP and CVD was observed in women ( Figures 3A, B ). What’s more, an interaction between CRP and gender was found in the entire population (p for interaction = 0.05).

To evaluate the influence of chemotherapy on CRP levels, we compared CRP levels of patients with rectal cancer before and after chemotherapy. There were 536 rectal cancer patients previously treated with chemotherapy, of whom the CRP levels of 145 were examined. We found a significant increase in CRP levels after chemotherapy compared to before chemotherapy (p<0.001) ( Figure 3C ). Considering the impact of age on tolerance in patients following chemotherapy, we also investigated the levels of CRP between groups of those aged ≥ 51 years and ≤ 50 years after chemotherapy treatment. Results showed that CRP levels were higher in patients aged over 51 years old compared to those aged under 50 years old, suggesting that elderly patients were more vulnerable to chemotherapy ( Figure 3D ).

The liver is not only the main release site for CRP but also the main pathway for chemotherapeutic metabolism. Accordingly, we selected the gene expression profile dataset of colorectal cancer liver metastases receiving oxaliplatin-based chemotherapy (GSE32384). To evaluate the impact of chemotherapy on CRP expression levels, we compared the differences in CRP levels between 19 samples with chemotherapy and five samples without any chemotherapy. The results showed that the CRP levels increased in the chemotherapy group in contrast to the group without any chemotherapy (p=0.005), indicating that chemotherapy can precipitate an increase in CRP ( Figure 3E ). Then, 96 eligible DEGs were identified by differential gene analysis between the two groups. To gain a better understanding of the role of differentially expressed genes in chemotherapy, we performed a GO analysis. The functional gene set analysis identified opsonization, low-density lipoprotein particle binding, regulation of tube diameter, regulation of blood vessel diameter vascular process in circulatory system, regulation of interleukin-8 production, etc. ( Figure 4A ).

Oxidized phospholipids induce inflammatory responses via oxidative stress (28), which provides a possibility to explore the impact of CRP on vascular endothelium. We selected gene expression profile dataset of the vascular endothelial cells treated with oxidized phospholipids (GSE20060) simulating an inflammatory environment. To explore the possible biological functions and mechanisms of CRP in the vascular endothelium, we conducted DEG analyses, as well as GO and KEGG analyses.

In the GSE20060 dataset, a comparison between the oxidized and non-oxidized phospholipids treatments identified 2,696 DEGs. A Spearman correlation was conducted to analyze the CRP-related genes from the 2696 DEGs, which identified 140 genes. According to p < 0.05, the GO analysis of 140 CRP-related genes identified significantly enriched terms in biological processes (BP), cellular components (CC), and molecular functions (MF), as shown in ( Figure 4B ). Similarly, these CRP-related genes were significantly enriched in KEGG pathways, including neuroactive ligand-receptor interaction, calcium signaling pathway, Cytokine-cytokine receptor interaction, and so on ( Figure 4C ).

We identified 30 common DEGs of the two datasets that were likely to be of great significance. To provide a broader understanding of the biological connections between chemotherapy and CVD, the existing functions of these genes were summarized in Supplementary Table 1 . These current studies reported that 15 genes were associated with CVD, among which CCL1, LDLR, and TENT5A were identified as hub genes that had susceptibility effects for atherosclerosis (29–31). Our findings suggested that these hub genes were likely to be associated with genetic risk factors for atherosclerosis after chemotherapy.