Role of plasma angiogenesis factors in the efficacy of first‐line chemotherapy combined with biologics in RAS wild‐type metastatic colorectal cancer: Results from the GI‐SCREEN CRC‐Ukit study

Abstract Background Several biomarkers have been established for metastatic colorectal cancer (mCRC). We investigated whether plasma angiogenesis factors could predict the efficacy of biologics combined with chemotherapy in first‐line (1L) treatment in patients with RAS wild‐type mCRC and the dynamics of plasma angiogenesis factors at progression during 1L treatment. Methods In this multicenter prospective observational study, serial plasma samples were prospectively collected at pretreatment and progression stages; 17 plasma angiogenesis factors were analyzed using the multiplex assay with Luminex® technology. Interactions between the pretreatment measurements and treatment groups on progression‐free survival (PFS) and overall survival (OS) in patients with RAS wild‐type were assessed using the propensity‐score weighted Cox proportional hazards model. Results From February 2018 to September 2020, 202 patients were enrolled in the 1L cohort; 133 patients had RAS wild‐type (chemotherapy plus bevacizumab [BEV group, n = 33] and plus anti‐epidermal growth factor receptor monoclonal antibodies [aEGFR group, n = 100]). A trend of strong interaction on PFS was observed for interleukin‐8 (IL‐8) (p = 0.0752) and soluble vascular cell adhesion molecule‐1 (sVCAM‐1) (p = 0.0156). Regarding OS, IL‐8 (p = 0.0283), soluble vascular endothelial growth factor‐receptor‐1 (sVEGFR‐1) (p = 0.0777) and sVCAM‐1 (p = 0.0011) tended to differentiate the treatment effect. In 112 patients, plasma samples were evaluable for dynamic analysis (57 and 55 from the BEV and aEGFR groups, respectively). In the BEV group, six factors significantly increased during progression, whereas two decreased. In the aEGFR group, three factors significantly increased, and six decreased. Conclusion Pretreatment plasma IL‐8 and sVCAM‐1 levels could be predictive biomarkers to distinguish BEV and anti‐EGFR mAbs when combined with chemotherapy in the 1L treatment of RAS wild‐type mCRC. Several plasma angiogenesis factors showed significant change at progression in 1L chemotherapy plus biologics for RAS wild‐type mCRC, which are potential biomarkers for selecting an optimal angiogenesis inhibitor in second‐line treatment.


| INTRODUCTION
Colorectal cancer (CRC) is the second leading cause of cancer-related deaths worldwide, and metastatic or recurrent CRC has a poor prognosis. 1][4][5][6] It has been reported that plasma vascular endothelial growth factor D (VEGF-D) and VEGF-A are potential predictive biomarkers for antiangiogenic inhibitors in 1L and second-line (2L) treatments.A plasma biomarker analysis of the CALGB80405 trial 7 on 1L chemotherapy plus BEV versus chemotherapy plus cetuximab (CET) showed that patients with low serum VEGF-D levels had significantly longer overall survival (OS) from chemotherapy plus BEV than CET (hazard ratio [HR] 0.62; 95% confidence interval [CI] 0.41-0.92).Additionally, the 1L MAX trial, which was a Phase III trial comparing capecitabine, capecitabine plus BEV (CB), and CB plus mitomycin (CBM), showed that patients with low VEGF-D expression had significantly longer OS when treated with the BEV combined regimen (CB plus CBM vs. C) (HR 0.35; 95% CI: 0.13-0.90). 8n the 2L RAISE trial, 9 which was a Phase III trial comparing fluorouracil, leucovorin, and irinotecan (FOLFIRI) plus ramucirumab (RAM) with FOLFIRI plus placebo, patients receiving RAM in the high VEGF-D group (≥ 115 pg/ mL) had a significantly longer survival than those receiving the placebo (HR 0.73; 95% CI: 0.60-0.89,p = 0.0022), while those with low VEGF-D (<115 pg/mL) had shorter survival (HR 1.32; 95% CI: 1.02-1.70;p = 0.0344).Additionally, a significant interaction was observed between VEGF-D levels (cut off, 115 pg/mL) and treatment arms (p = 0.0005).The WJOG6210G trial, which was a randomized Phase II trial comparing FOLFIRI plus panitumumab (PANI) and FOLFIRI plus BEV for patients with RAS wild-type BEV refractory, demonstrated that FOLF-IRI plus BEV had a better OS than FOLFIRI plus PANI in patients with low serum VEGF-D (HR 0.82; 95% CI: 0.35-2.01) 10 and those with high serum VEGF-A (HR 0.67; 95% CI 0.48-1.69). 11However, the role of plasma angiogenesisrelated mediators as predictors of anti-EGFR mAbs and angiogenesis inhibitors remains unknown.
Therefore, we aimed to investigate whether plasma angiogenesis factors could predict the efficacy of biologics combined with chemotherapy in 1L treatment in patients with RAS wild-type mCRC and the dynamics of plasma angiogenesis factors at progression during 1L.

| Study design and participants
The SCRUM-Japan GI-SCREEN CRC-Ukit trial was a multicenter prospective observational study investigating the association between plasma angiogenesis-related factors and clinical outcomes in mCRC.Patients who met the following criteria were eligible for this trial: pathologically or histologically confirmed mCRC and RAS mutational status and age ≥20 years.In the 1L cohort, patients who had not been administered chemotherapy for metastatic disease were scheduled to receive chemotherapy plus either BEV or anti-EGFR mAbs.In the 2L cohort, patients immediately after completion of chemotherapy plus either BEV or anti-EGFR mAbs for the 1L treatment for mCRC were scheduled to receive FOLFIRI plus RAM or FOLFIRI plus aflibercept (AFL) or chemotherapy plus BEV.Longitudinal measurements of plasma angiogenesis-related factors were analyzed using the multiplex assay with Luminex® technology.Here, progression, whereas two decreased.In the aEGFR group, three factors significantly increased, and six decreased.

Conclusion:
Pretreatment plasma IL-8 and sVCAM-1 levels could be predictive biomarkers to distinguish BEV and anti-EGFR mAbs when combined with chemotherapy in the 1L treatment of RAS wild-type mCRC.Several plasma angiogenesis factors showed significant change at progression in 1L chemotherapy plus biologics for RAS wild-type mCRC, which are potential biomarkers for selecting an optimal angiogenesis inhibitor in second-line treatment.

K E Y W O R D S
angiogenesis, biomarkers, chemotherapy, colorectal cancer we investigated whether plasma angiogenesis-related mediators could predict the efficacy of biologics combined with chemotherapy as a 1L treatment in patients with RAS wild-type mCRC and the dynamics of plasma angiogenesisrelated mediators during progression in 1L treatment.
A comparison of the efficacy parameters, such as response rate (RR), progression-free survival (PFS), and OS of each biologic (1L chemotherapy plus BEV [BEV group] or 1L chemotherapy plus anti-EGFR mAbs [aEGFR group]) depending on plasma angiogenesis-related mediators were performed in patients with RAS wild-type mCRC (efficacy analysis set).Additionally, analysis of dynamics in plasma angiogenesis-related mediators was performed in patients with paired plasma samples (pretreatment and disease progression), regardless of RAS mutational status (dynamic analysis set).The analyses of the 2L cohort will be reported elsewhere in the future.
The Institutional Review Board at all participating centers reviewed and approved this study.This study was registered in the University Hospital Medical Information Network Clinical Trials Registry (UMIN000028616).This study was conducted following the Declaration of Helsinki.Written informed consent was obtained from all patients enrolled in this trial.

| Data collection and assessments
The following clinical characteristics of the eligible patients were collected: age, sex, presence of primary tumor, RAS/ BRAF mutational status, and microsatellite instability/mismatch repair (MSI/MMR) status.Regarding efficacies, antitumor effects of 1L treatment, PFS and OS, and RR were assessed.Tumor response assessment was conducted by treating physicians based on Response Evaluation Criteria in Solid Tumors ver1.1. 12Furthermore, PFS was defined as the time from the initiation of 1L chemotherapy to disease progression or death from any cause, whichever occurred first.OS was defined as the time from the initiation of 1L chemotherapy to death from any cause.Patients who were lost to follow-up were censored at the date of the last response evaluation for PFS and the date of the last contact for OS.
First, angiogenesis factors from plasma samples were bound to their respective specific antibodies using Lu-minex® beads.Next, biotinylated detection antibodies were conjugated to antigen-antibody complex and labeled with streptavidin-phycoerythrin.Subsequently, beads were analyzed using Luminex® 200, and fluorescent signals were calculated as concentrations from the standard curve for each substance.Finally, these angiogenesis factors were measured using Milliplex® assay kits (Merck). 13,14

| Statistical analyses
A propensity-score weighted Cox regression model was applied to screen out baseline plasma angiogenesis-related mediators that would strongly interact with the treatment group (either BEV or aEGFR group) on PFS or OS.Continuous plasma angiogenesis variables were dichotomized according to their median values.The propensity score was estimated using a multivariable logistic regression model, including tumor location, an indicator of whether the number of combined cytotoxic agents was ≥3, their interaction term, sex, and age.Age was modeled as a natural cubic spline.In the propensity-score weighted Cox regression model, the inverse of the estimated propensity score was used as a weight, and a robust sandwich estimator was used to calculate the standard error.The significance level for testing interaction terms was set at an alpha of 10%.
For baseline patient characteristics, categorical and continuous variables were compared using Fisher's exact and Wilcoxon rank-sum tests, respectively.Additionally, survival curves were estimated and compared using the Kaplan-Meier method and log-rank test, respectively.

| Patients
From February 2018 to September 2020, 504 patients with mCRC were enrolled at 27 institutions in Japan.Among the 501 eligible patients, 202 (40%) were enrolled as a 1L treatment cohort (Figure 1).In the efficacy analysis set, after excluding 68 patients with RAS mutations and one whose estimated propensity score was too small, 133 were included in the RAS wild-type 1L cohort (BEV group: n = 33, aEGFR group: n = 100 [reference]).
Baseline characteristics of the RAS wild-type 1L cohort are presented in Table 1.Several characteristics, including tumor location (p < 0.001), BRAF V600E mutational status (p < 0.001), MSI status (p = 0.002), and combined cytotoxic agents (p = 0.002), were imbalanced between the two groups.At the data cutoff date as of September 30, 2021, the median follow-up time was significantly longer in the BEV group (35.7 months) than in the aEGFR group (28.8 months) (HR 0.436; 95% CI: 0.251-0.757,p = 0.002).
The propensity-score weighted Cox model for PFS is presented in Table 2. Significant interactions were found in IL-8 (below median; aEGFR group 13.9 months vs.

Note:
For interferon-γ and IL-6, more than half of the patients were below the lower detection limit (2.00 pg/mL), and no analysis was performed.
Abbreviations: aEGFR, anti-epidermal growth factor receptor; BEV, bevacizumab; CI, confidence interval; HGF, hepatocyte growth factor; HR, hazard ratio; IL, interleukin; NE, not-estimable; OPN, osteopontin; PlGF,   T A B L E 3 Overall survival by each baseline angiogenesis factor in the efficacy analysis set.T A B L E 3 (Continued)

| RR baseline plasma angiogenesis-related mediators
The RR in the efficacy analysis set was 74.2% and 48.4% in the aEGFR and BEV groups, respectively, significantly superior to that in the aEGFR group (odds ratio [OR] 0.326; 95% CI: 0.141-0.753,p = 0.014) (Table S1).

| Dynamics of plasma angiogenesis-related mediators in pretreatment and at progression
In the dynamics analysis set, after excluding 90 patients due to the ongoing treatment, discontinuation without progression, and lack of sample collection at progression, 112 were analyzed using the dynamics of plasma angiogenesis-related mediators (BEV group: n = 57, aEGFR group: n = 55) (Figure 1).
Baseline plasma angiogenesis-related mediators in the efficacy analysis set are presented in Table S3.IL-6, IL-8, OPN, and TIMP-1 were significantly higher in the aEGFR group than in the BEV group.

| DISCUSSION
We demonstrated that baseline plasma IL-8 and sVCAM-1 levels could be biomarkers that distinguish between BEV and anti-EGFR mAbs combined with cytotoxic agents in 1L treatment for patients with RAS wild-type T A B L E 4 Dynamics of plasma angiogenesis-related factors at pretreatment and disease progression in the dynamic analysis set.mCRC.Furthermore, in the analysis of dynamics for angiogenesis-mediators during pre-and post-1L treatment, several angiogenesis-related mediators were significantly changed during progression.These mediators may become biomarkers for selecting an optimal angiogenesis inhibitor in 2L treatment for patients with RAS wild-type mCRC.

Median (pg
A pro-inflammatory cytokine, IL-8, alternatively known as CXCL8, 15 plays a significant role in tumor growth, angiogenesis (formation of new blood vessels), and immune cell recruitment. 16BEV is a monoclonal antibody that targets VEGF, a key signaling molecule involved in angiogenesis. 17t inhibits VEGF, thereby reducing the formation of new blood vessels in tumors.IL-8 promotes angiogenesis by stimulating endothelial cell migration and proliferation. 16n the context of anti-angiogenic inhibitors, high levels of IL-8 may contribute to resistance to the drug's antiangiogenic effects.Tumors with elevated IL-8 expression may have increased vascularization and are more likely to continue growing despite anti-angiogenic inhibitor.In our study, IL-8 showed a significant interaction with PFS and OS, and low IL-8 demonstrated favorable PFS and OS in the BEV group.9][20][21] Correlations between IL-8 and angiogenesis inhibitors have been reported in other cancers.From hepatocellular carcinoma and renal cell carcinoma analyses, higher baseline plasma levels of IL-8 were correlated with worse outcomes in treatment with angiogenesis inhibitors, such as sunitinib or BEV.3][24] However, a few reports exist on the significance of baseline IL-8 in BEV combination chemotherapy for mCRC; thus, further investigation is needed.
VCAM-1 is a Type-1 transmembrane glycoprotein essential for cell adhesion needed for metastasis and is expressed on endothelial cells in response to VEGF stimulation or inflammation. 25VCAM-1 is soluble and may function as an attractant for endothelial cells. 268][29][30] In our analysis, sVCAM-1 showed a significant interaction with PFS and OS, and high sVCAM-1 demonstrated favorable PFS and OS in the BEV group.In a previous study, a baseline high level of sVCAM-1 was significantly correlated with shorter OS in patients with mCRC receiving capecitabine, oxaliplatin, and BEV, 21 of which results were inconsistent with our findings.However, Liu et al. reported that higher baseline levels of VCAM-1 appeared to predict PFS (interaction p = 0.022) and OS (interaction p = 0.012) benefits from chemotherapy plus regorafenib over that plus placebo, 31 which supports our findings of sVCAM-1 as the potential predictive factor for BEV.
In our analysis, VEGF-A was decreased and VEGF-D was significantly increased at progression in the aEGFR group.In contrast, VEGF-A was decreased, and both PlGF and VEGF-D were significantly increased at progression in the BEV group.The dynamics in VEGF-A, VEGF-D, and PlGF after BEV administration, regardless of the primary organ, have been reported, [32][33][34] and we found consistent results in our analysis.Although angiogenesis inhibitors, such as BEV, RAM, and AFL, are key drugs used in 2L, no clear selection criteria are available.Given the results of our analysis, using the RAM for the 1L patients with anti-EGFR mAbs refractory and using either RAM or AFL in the 1L BEV refractory patients might be theoretically effective options because each drug inhibits VEGF-D and PlGF.Therefore, we plan to analyze the relationship between the dynamics of plasma angiogenesis-related factors and clinical outcome using a 2L cohort.We aim to present this aspect in another study.
VEGF-A and VEGF-D values may be potential predictive biomarkers for BEV. 7,8,10,11In our study, the aEGFR group tended to have better PFS and OS outcomes in patients with high VEGF-A levels (HR 0.504 and 0.574, respectively).In contrast, the BEV treatment tended to result in better survival outcomes in patients with low VEGF-D levels (HR 0.554/0.597)without significant interaction.However, the role of VEGF-A and VEGF-D as predictors for anti-EGFR mAbs and BEV remains controversial.Therefore, it warrants further investigation.
This study had some limitations.First, this was an observational study with a small number of patients with RAS wild-type.Second, because the treatment regimen was decided by the physician's choice, some significant background differences were observed, mainly for the primary tumor location, BRAF status, MSI status, and combined cytotoxic agents (doublet/mono or triplet), which reflect the present standard-of-care in the world.Therefore, in the future, we plan to conduct validation trials and basic research using cell lines to verify the results of this study.
In conclusion, pretreatment plasma IL-8 and sVCAM-1 levels could be predictive biomarkers to distinguish between BEV and aEGFR when combined with chemotherapy in the 1L treatment for RAS wild-type mCRC; thus, further investigation is warranted.Additionally, several plasma angiogenesis factors significantly changed during progression, implying that they might become biomarkers for selecting an optimal angiogenesis inhibitor in 2L treatment.

F I G U R E 1
Flow diagram of this study.anti-EGFR mAbs, anti-epidermal growth factor receptor monoclonal antibodies; BEV, bevacizumab.

F I G U R E 3
Progression-free survival (PFS) and overall survival (OS) by sVCAM-1.Kaplan-Meier curves of the PFS in patients with below (A) and above (B) the median sVCAM-1.Kaplan-Meier curve of the OS in patients with below (C) and above (D) the median sVCAM-1.p-values # are calculated using the log-rank test, two-sided.HR* estimated by unstratified Cox model.aEGFR, anti-epidermal growth factor receptor; BEV, bevacizumab; CI, confidence interval; HR, hazard ratio; NE, not estimable; sVCAM-1, soluble vascular cell adhesion molecule-1.