Correlation Between Bevacizumab Exposure and Survival in Patients with Metastatic Colorectal Cancer

Abstract

Background

Bevacizumab treatment is subject to large interpatient variability in efficacy, which may partly be explained by differences in complex bevacizumab pharmacokinetic characteristics that influence bevacizumab exposure. Exposure–response relationships have been identified for other monoclonal antibodies. We aimed to identify possible exposure–survival relationships in bevacizumab‐treated patients with metastatic colorectal cancer (mCRC).

Materials and Methods

Patients with mCRC who started first‐line bevacizumab‐based chemotherapy between July 2012 and July 2014, and from whom serial blood samples and survival were prospectively collected, were included. Follow‐up was carried out until July 2018. Total bevacizumab trough concentrations were measured from cycle 2 to cycle 30 of treatment. The receiver operating characteristic (ROC) curve analysis and Cox analysis were used to identify the relationship between concentrations and overall survival (OS). In addition, OS was compared between different trough concentration groups.

Results

One hundred fifty‐seven blood samples from 46 patients were evaluable for analyses. ROC analysis showed a clear separation in survival based on trough levels (area under the curve = 0.739, p = .009). Cox regression also showed a strong positive correlation between trough levels and survival (p = .0004). Three distinct groups of exposure were identified: low (median trough concentration [C_tm] ≤41.9 mg/L); medium (C_tm 43–87.2 mg/L) with median OS of 12.8 and 36 months, respectively (p = .0003); and high (C_tm ≥7.9 mg/L), where the majority of patients were still alive 60 months after the initiation of treatment.

Conclusion

This study shows that survival was proportional to the magnitude of exposure in patients with mCRC. Further clinical research should focus on clarifying these exposure–outcome relationships in order to optimize dosing.

Implications for Practice

Bevacizumab‐based chemotherapy is standard first‐line treatment in metastatic colorectal cancer. Moreover, bevacizumab presents complicated pharmacokinetics, and in many cases, clinical outcomes can be highly variable, with some patients responding remarkably well and others not. This study's results show that patients who experienced longer overall survival also had significantly higher exposure to bevacizumab. Therefore, bevacizumab trough concentrations could be used both as a predictive biomarker and as a tool for treatment monitoring and optimization. Finally, the development of validated, rapid, and sensitive assays for bevacizumab concentration measurements in combination with these results may lead to a therapeutic drug monitoring–guided approach in bevacizumab treatment with better clinical outcomes.

Short abstract

Bevacizumab concentrations as a predictive biomarker for clinical outcomes could represent an advance in therapeutics. This study investigated the relationship between bevacizumab exposure and overall survival in metastatic colorectal cancer.

Introduction

Colorectal cancer (CRC) is the second most commonly diagnosed cancer in women and third most common in men 1. It is estimated that approximately 20% of newly diagnosed patients have distant metastatic disease at the time of presentation 2. Major advances in the treatment of metastatic colorectal cancer (mCRC) have been made in the last 15 years. When fluorouracil was the only available active agent, overall survival (OS) was approximately 1 year. However, the average median survival duration is now 3 years, and 5‐year survival rates are as high as 20%. These improvements have been mainly driven by the availability of new active agents, which include conventional cytotoxic agents such as irinotecan and oxaliplatin, and biologic agents targeting angiogenesis and the epidermal growth factor receptor (EGFR) 3.

Bevacizumab, a recombinant humanized immunoglobulin G (IgG)1 monoclonal antibody, is the first approved antiangiogenetic agent for the treatment of mCRC 4. The mechanism of action of bevacizumab includes binding to circulating vascular endothelial growth factor A (VEGF‐A) and blocking of VEGF‐A binding to its receptors (VEGFR‐1 and VEGFR‐2) on the surface of endothelial cells, which results in the inhibition of tumor angiogenesis, growth, and metastases 5. In the absence of validated predictive factors for treatment benefit, the default position in the management of mCRC had been to add bevacizumab to the chemotherapy backbone chosen for the individual patient, particularly for patients with RAS mutated tumors, for whom an anti‐EGFR agent is contraindicated 6. Benefits from bevacizumab are often questioned and clinical outcomes are highly variable 7. Like most monoclonal antibodies, bevacizumab exhibits more complex and variable pharmacokinetic and pharmacodynamics characteristics than small molecules. Variability was associated with several factors such as tumor burden, binding to their molecular targets, and others 8, 9, 10, 11. This variability could cause significant differences in exposure to bevacizumab, which subsequently may affect clinical response, as already shown with other anticancer monoclonal antibodies. For example, patients with breast cancer treated with therapeutic drug monitoring (TDM)‐1 12, those with mCRC treated with cetuximab 13, patients with melanoma treated with ipilimumab 14, those with non‐small cell lung cancer treated with nivolumab 15, and patients with soft tissue sarcoma treated with olaratumab 16.

Bevacizumab concentrations as a predictive biomarker for clinical outcomes would represent a major advance in therapeutics. Therefore, the aim of the present study was to investigate the relationship between bevacizumab exposure and OS in mCRC.

Materials and Methods

Study Design and Patients

This was a prospective, real‐world study, conducted at the Department of Oncology, University Hospital of Patras, Greece. Forty‐six patients with mCRC who were treated with first‐line bevacizumab‐containing treatment between July 2012 and July 2014 were included in the study. Serial blood samples from cycle 2 up to cycle 30 and survival data were prospectively collected. Follow‐up was carried out until July 2018. Patients were 18 years of age or older, had Eastern Cooperative Oncology Group performance status 0–2 and histologically confirmed mCRC. All patients received standard of care first‐line treatment with bevacizumab in combination with oxaliplatin or irinotecan and fluoropyrimidine‐based chemotherapy. Bevacizumab was administered as an intravenous infusion at a dose of 5 mg/kg once every 2 weeks in combination with 5‐fluorouracil/leucovorin/irinotecan or oxaliplatin (BEV‐FOLFIRI or BEV‐FOLFOX, respectively) or at a dose of 7.5 mg/kg once every 3 weeks in combination with capecitabine/irinotecan or oxaliplatin (BEV‐CapIRI or BEV‐CapOX, respectively) in 3‐week cycles. Treatment was initially administered for six (BEV‐FOLFIRI, BEV‐FOLFOX) or four (BEV‐CapIRI, BEV‐CapOX) cycles; patients who responded to treatment continued with bevacizumab‐based maintenance treatment. Radiographic evaluation was performed every 8–12 weeks or when clinically indicated. The response was evaluated according to RECIST criteria version 1.1. The treating physicians decided maintenance treatment. The primary endpoint of the study was OS.

The study was conducted in accordance with the Declaration of Helsinki and the International Conference on Harmonisation Good Clinical Practice. Approval was obtained by the hospital's ethics committee. Prior to study enrollment, all patients provided signed informed consent.

Bevacizumab Serum Measurements

Trough (predose) concentrations of bevacizumab were measured in serum from cycle 2 up to cycle 30. This period was sufficient to account for the minimum time needed to investigate the response of each patient to the therapy while maintaining an acceptable number of follow‐up observations. A previously published and validated enzyme‐linked immunosorbent assay (ELISA) method by Panoilia et al. was used, where the detection limit was 0.033 mg/L and the range of linearity was between 5 and 75 mg/L with precision 5.6 % 8. Measurements were performed as follows: Microtiter Nunc Maxisorp 96‐well plates were coated with recombinant human VEGF165 (R&D Systems Europe, Abingdon, UK) at a concentration of 0.15 mg/L in carbonate–bicarbonate buffer (1 M, pH 9.6) overnight at 4°C (100 μL per well). After washing four times with phosphate‐buffered saline (PBS) containing 0.05 % Tween 20, the wells were blocked with PBS containing 1% bovine serum albumin (BSA; 200 μL per well) and were incubated for 2 hours at room temperature. Afterward, plates were washed and 100 μL of 1:100 diluted standards and samples in 1% PBS–BSA were added and were incubated for 1 hour at 37°C in an incubator shaker. Then, the plates were washed again, and 100 μL of peroxidase‐conjugated goat antihuman IgG specific for Fc fragment (AbD Serotec, A Bio‐Rad Company, Oxford, UK) diluted in 1% PBS–BSA was added to each well. After 1‐hour incubation at room temperature followed by washing, 100 μL OPD (Sigma‐Aldrich, Darmstadt, Germany) was added and the reaction could develop at room temperature in the dark. The color reaction was stopped with the addition of sulfuric acid (2 M, 50 μL per well). Optical density wasmeasured at 450 nm with a correction at 650 nm using an ELISA plate reader (ThermoMax, Molecular Devices, San Diego, CA, USA). Duplicate readings for 1:100 diluted standards and samples were performed. OriginPro 8.0 software (Origin‐Lab Corporation, Northampton, MA, USA) was used to determine the best‐fit line of the standard curve with regression analysis. The concentrations read from the standard curve were multiplied by the dilution factor.

The effect of exposure to bevacizumab on clinical outcome was defined by using the median trough concentration (C_tm) at steady state, occurring after cycle 2–4, during treatment. Trough concentration during a dosing interval reflects the lowest target saturation within this dosing interval. It should be sufficiently high to achieve desired pharmacodynamic effect (maximal VEGF inhibition) and clinical outcomes. C_tm was also selected, as it summarizes the serum exposure retrospectively and captures any missing doses or dose delays during treatment.

Statistical Analysis

All categorical data including therapy were tabulated and presented as frequencies and counts. The type I error rate or significance level was set to 0.05. Area under the receiver operating characteristic curve analysis was performed to evaluate our ability to classify the outcome (survival) depending on the bevacizumab levels. For the time‐to‐event analyses, we used the Kaplan‐Meier estimate, and all lost‐to‐follow‐up cases were censored up to the most recent available time point. The log‐rank test was used for the comparison of the OS distribution for different patient groups, whereas the effects of age, sex, and bevacizumab levels were analyzed using Cox regression. In terms of the sample size calculation, there were feasibility limitations in recruitment, and therefore, the investigators calculated the minimum effect size that this study could capture for a fixed number of patients. At a power level of 80% and alpha set as 0.05, for 46 patients with 60 months of follow‐up and 2 years of recruitment, a minimum of 1.5 years of difference in the median of OS between the compared groups can be statistically detected. Alternatively, this is approximately equal to a 0.34 hazard ratio θ under a Cox regression model, meaning that this study would able to identify predictors with large effect size. The Kaplan‐Meier graphs were generated using R version 3.5.2 for Windows (R Foundation for Statistical Computing, Vienna, Austria). The analyses and power calculations were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC).

Results

A total of 46 consecutive patients with mCRC were enrolled in the study. Overall, patients had a mean age of 64.5 years (range 31–86) and were predominantly male (28/46; 61%). The most commonly used initial treatment was BEV‐FOLFOX (46%) and the most common maintenance treatment was bevacizumab monotherapy (30%). The mean number of metastatic sites was 2 (range 1–5; Table 1).

Table 1.

Baseline patient characteristics and therapeutic regimens used

Characteristic	Total no. of patients (n = 46)	Frequency, %
Sex
Female	18	39
Male	28	61
Age, years
< 55	14	30
55–65	12	26
> 65	20	44
Metastatic site
1	15	33
2	12	26
> 2	19	41
Initial treatment
Bevacizumab‐mFOLFOX6	22	48
Bevacizumab‐FOLFIRI	13	28
Bevacizumab‐CapOX	2	4
Bevacizumab‐CapIRI	9	20
Maintenance treatment
Bevacizumab‐mFOLFOX6	4	9
Bevacizumab‐FOLFIRI	5	10
Bevacizumab‐CapIRI	9	20
Bevacizumab‐De Gramont	11	24
Bevacizumab‐Capecitabine	3	7
Bevacizumab monotherapy	14	30
Response to treatment
Complete response	1	2
Partial response	16	35
Stable disease	25	54
Disease progression	4	9

Bevacizumab‐mFOLFOX6 (bevacizumab 5 mg/kg, oxaliplatin 85 mg/m², folinic acid 400 mg/m², fluorouracil 400 mg/m² bolus, fluorouracil 2,400 mg/m² over 46 hours every 2 weeks), Bevacizumab‐FOLFIRI (bevacizumab 5 mg/kg, irinotecan 180 mg/m², folinic acid 400 mg/m², fluorouracil 400 mg/m² bolus, fluorouracil 2,400 mg/m² over 46 hours every 2 weeks), Bevacizumab‐CapOX (bevacizumab 7.5 mg/kg on d1, oxaliplatin 130 mg/m² on d1, capecitabine 1,000 mg/m² per 12 hours d1–14 every 3 weeks), Bevacizumab‐CapIRI (bevacizumab 7.5 mg/kg on d1, irenotecan 250 mg/m² on d1, capecitabine 1,000 mg/m² per 12 hours d1–14 every 3 weeks), Bevacizumab‐De Gramont (bevacizumab 5 mg/kg, folinic acid 200 mg/m² on d1,2, fluorouracil 400 mg/m² bolus on d1,2, fluorouracil 2,400 mg/m² over 22 hours on d1,2 every 2 weeks), Bevacizumab‐Capecitabine (bevacizumab 7.5 mg/kg on d1, capecitabine 1,000 mg/m² per 12 hours d1–14 every 3 weeks), Bevacizumab monotherapy (5 mg/kg every 2 weeks or 7.5 mg/kg every 3 weeks).

After a median follow‐up of 60 months, 88.1% of patients progressed and 58.0% died. For all patients, median PFS and OS were 9 and 18 months, respectively. In terms of response, complete response was achieved by one patient, partial response by 35% of patients, stable disease by 54% of patients, and progressive disease by 9% of patients, all at the first imaging evaluation.

In total, 157 trough concentration samples were analyzed. It was found that the median trough concentration at steady state was 74.6 mg/L, ranging from 10.9 to 195.5 mg/L.

The receiver operating characteristic curve analysis using the bevacizumab trough levels as the predictor and OS as the classification produced an area under the curve = 0.739 with p = .009. This suggests a distinct separation between the patients who did and did not survive in the studied cohort, depending on their bevacizumab trough levels.

This finding was further studied using Cox proportional hazards modeling with the survival status and time as the outcome and bevacizumab trough levels as the predictor (independent variable). The results of the Cox regression showed a strong positive correlation between bevacizumab C_tm and survival (p = .0004), verifying the previous findings and suggesting the continuous nature of the effect.

To quantify the results in a meaningful manner and demonstrate potentially translational aspects of these findings, patients were separated into three exposure groups depending on bevacizumab trough levels, namely, high with C_tm ≥87.9 mg/L, medium with C_tm between 43 and 87.2 mg/L, and low with C_tm ≤41.9 mg/L. Patients’ characteristics did not differ significantly between the three exposure groups (Table 2). It was found that median OS in patients of the low exposure group was 12.8 months, whereas that in the patients of the medium exposure group was 36 months (p = .0003). In the high exposure group, OS could not be determined, as more than 50% of patients were still alive after the 60 months follow‐up period (Fig. 1 and Table 1). Sex, age, and KRAS status were not associated with OS, as any combination did not add to the Cox model.

Table 2.

Baseline characteristics and OS in the three different groups of bevacizumab exposure

Characteristic	Ctm ≤41.9 mg/L	Ctm 43–87.2 mg/L	Ctm ≥87.9 mg/L	p value
Median age, years	63.3	63.4	61.3	.89
Sex (males)	73%	60%	50%	.56
KRAS (mutant)	57%	55%	33%	.77
Median OS (range), months	12.8 (7.7–17.9)	36 (23.4–48.5)	N/A	.0003

Median OS was not reached for the high exposure group.

Bolded p value is statistically significant.

Abbreviations: C_tm, median bevacizumab trough concentration at steady state; N/A, not applicable; OS, overall survival.

Figure 1.

Kaplan‐Meier curves of overall survival according to median trough bevacizumab levels between cycles 2 and 30.

Discussion

Today, bevacizumab has become standard treatment in patients with mCRC. However, clinical outcomes can be highly variable, with some patients responding remarkably well and others not. Thus, because of this heterogeneous response, the real clinical impact of bevacizumab remains unclear. Pharmacokinetic parameters are among the most important factors, which influence drug action and clinical response. Bevacizumab presents very variable and complicated pharmacokinetics. Therefore, differences in the pharmacokinetic parameters could explain the interindividual variability observed in patients. The mean half‐life is approximately 20 days; however, large individual differences were noted ranging between 11 and 50 days, a variability that is anticipated to affect response to treatment 4.

In this study, we aimed to assess the relationship between bevacizumab exposure, measured as C_tm, and overall survival in patients with mCRC receiving first‐line bevacizumab in combination with oxaliplatin or irinotecan and fluoropyrimidine‐based chemotherapy. We demonstrated that patients who experienced longer OS also had significantly higher exposure to bevacizumab. Three distinct groups of patients were identified: the low exposure group (C_tm ≤41.9 mg/L) with median OS of 12.8 months, the medium exposure group (C_tm 43–87.2 mg/L) with median OS of 36 months (p = .0003), and the high exposure group (C_tm ≥87.9 mg/L), where the majority of patients were still alive 60 months after the initiation of treatment.

It has been previously reported that lower trough bevacizumab concentrations are associated with poorer clinical outcomes 17, 18. However, our study for the first time has identified three distinct groups of subpopulations within patients with mCRC, where bevacizumab exposure is directly and positively related to OS. These findings could be interpreted and used as prognostic markers for survival but more importantly could be used as guide for TDM in order to optimize dosing during the treatment; such an optimization will lead to better clinical outcomes and improved survival. TDM is a standard‐of‐care and cost‐effective practice for several classes of antibiotics, immunosuppressants, antiepileptics, human immunodeficiency virus agents, and many other drugs in clinical practice 19. Moreover, the TDM‐guided approach has been proved to improve safety and efficacy of both classic cytotoxic drugs and targeted therapies with small‐molecule tyrosine kinase inhibitors 20, 21, 22, 23, 24, 25, 26. Similarly, there are some studies that support the TDM approach for the treatment with monoclonal antibodies such as rituximab and cetuximab 27, 28, 29. In the case of bevacizumab in mCRC, the TDM approach has the practical advantage that trough concentrations provide the most relevant clinical information. Therefore, trough concentration measurement prior to the next dose could guide clinical decisions and potential dose adjustments in order to improve survival. In addition, implementation of such an approach seems feasible as the progress in the field for bevacizumab includes the development of validated, robust, rapid, and sensitive assays for bevacizumab levels measurements, which could be suitable for use in clinical routine as part of personalized treatment 17, 30.

Conclusion

A significant correlation between exposure to bevacizumab and survival was observed. Further studies for clarification of the dose, exposure, and clinical outcomes relationship in patients treated with bevacizumab are warranted. Considering the fundamental role of bevacizumab in the therapeutics of mCRC, the highly variable patient responses, and the wide range in clearance rates, the measurement of blood bevacizumab levels can be used to achieve long‐term responses and benefit from bevacizumab‐based therapies.

Author Contributions

Conception/design: Apostolos Papachristos, Gregory Sivolapenko

Provision of study material or patients: Apostolos Papachristos, Haralabos Kalofonos, Gregory Sivolapenko

Collection and/or assembly of data: Apostolos Papachristos, Polychronis Kemos

Data analysis and interpretation: Apostolos Papachristos, Polychronis Kemos, Haralabos Kalofonos, Gregory Sivolapenko

Manuscript writing: Apostolos Papachristos, Polychronis Kemos, Gregory Sivolapenko

Final approval of manuscript: Apostolos Papachristos, Polychronis Kemos, Haralabos Kalofonos, Gregory Sivolapenko

Disclosures

The authors indicated no financial relationships.