Relationship between apparent systemic clearance of vemurafenib and toxicity in patients with melanoma
Ganessan Kichenadasse, PhD; Jim Henry Hughes, PhD; Alia Fahmy; Andrew Rowland, PhD; Michael J Sorich, PhD; Ashley H Hopkins, PhD
Abstract
Vemurafenib, a BRAF inhibitor, is commonly used in combination of cobimetinib for the treatment of melanoma. In the current study, we evaluated the relationship between vemurafenib exposure, as measured by the estimated apparent clearance (CLB) at steady state and any grade ≥ 3 toxicity, grade ≥ 3 skin rash or toxicity requiring dose modification using pooled data from 3 prospective clinical trials involving 898 patients. A total of 69% had any grade ≥ 3 toxicity; grade ≥ 3 skin rash in 15% and 47% had a dose reduction/interruption or cessation. The median vemurafenib CLB was 1.35 L/h (interquartile range 1.15 – 1.65 L/h). Lower vemurafenib CLB was significantly associated with an increased risk of grade ≥ 3 toxicity (HR 0.62, P <0.001), grade ≥ 3 rash (HR 0.29, P <0.001), and adverse events requiring vemurafenib dose reduction/interruption or cessation (HR 0.5, P <0.001). When the patients were divided into three groups based on the vemurafenib CLB thresholds, those with low CLB (< 1.22 L/h) had significantly increased incidence of any grade ≥ 3 toxicity or skin rash or dose adjustment, interruption or cessation at 12 months and at day 28 when compared to those with medium (≥ 1.22 and < 1.55 L/h) or high (> 1.55 L/h) vemurafenib CLB. In conclusion, the estimated CLB of vemurafenib is associated with severe toxicities and dose adjustment or cessation, suggesting that an early estimation of vemurafenib exposure may be useful in identifying patients at risk of experiencing toxicity.
Keywords – 4-6 Melanoma; vemurafenib; clearance; toxicity
Introduction
Vemurafenib, a serine-threonine kinase inhibitor that targets wild type and mutated forms of BRAF such as BRAFV600E (BRAFi), and other kinases (CRAF, ARAF, SRMS, Ack1, MAP4K5 and FGR), is currently approved for the treatment of BRAF mutated advanced and early stage melanoma and Erdheim-Chester disease 1. Vemurafenib is also commonly used in combination with a MEK inhibitor (MEKi), cobimetinib, due to synergistic activity with improved efficacy. However, most patients receiving BRAFi +/- MEKi therapies experience some form of toxicities. Among the toxicities that occur from vemurafenib, skin related adverse events manifesting as various types of skin rash, photosensitivity, keratoacanthoma and cutaneous squamous cell carcinoma are the most common 2. Other common toxicities include pyrexia, arthralgia, fatigue, diarrhoea and nausea.
Given the known high inter-individual heterogeneity in response and occurrence of toxicities, prior studies have explored various predictive factors that may be associated with these outcomes from vemurafenib. While female sex has been found to be associated with increased toxicity incidence, there are limited number of other predictive factors 2,3. Though a threshold plasma concentration for vemurafenib that predicted efficacy has been explored previously, it is unclear if it is associated with toxicities 3-6. Moreover, multiple serial concentration measurements are required to predict the occurrence of toxicity and some toxicities can occur prior to steady-state concentration being achieved. Hence, early markers are required to predict toxicities prior to their occurrence. Other early pharmacokinetic (PK) measures such as the area under the curve (AUC) and apparent systemic (body) clearance (ClB) which can be estimated even with a single dose prior to the development of toxicity are potentially useful. In the current study, we hypothesised that apparent clearance (ClB) may be associated with severe toxicities. The main aim of the study was to determine if the apparent oral ClB of vemurafenib is associated with grade 3 or more toxicities in patients with advanced melanoma.
Methods
Ethics approval was exempted by the local ethics committee for the current study. Individual patient level data from three clinical trials that evaluated vemurafenib with or without cobimetinib in patients with advanced melanoma was obtained through sharing from www.clinicalstudydatarequest.com under Roche’s transparency policy. The three clinical (NCT01689519) 7. BRIM-2 was a single arm phase II vemurafenib monotherapy study that included patients with previously treated BRAF-mutant melanoma. BRIM-3 was a randomised phase III trial comparing vemurafenib monotherapy with dacarbazine. coBRIM was a randomised phase III study in which patients with previously untreated BRAF-mutated melanoma received vemurafenib/ placebo or vemurafenib with cobimetinib. Vemurafenib was given at a starting dose of 960 mg twice daily as oral administration. All the three trials enrolled participants with similar key eligibility criteria. Patients were at least 18 years old, diagnosed with unresectable stage IIIC or stage IV melanoma that tested positive for a BRAFV600 mutation (either previously treated (BRIM-2) or untreated (BRIM-3 and coBRIM), had an Eastern Cooperative Oncology Group performance status (ECOG PS) of 01, and adequate organ function (haematological, hepatic, renal, and cardiac).
Predictors and outcomes
The main outcomes of interest were any grade 3 or more (grade ≥ 3) toxicities as well as grade ≥ 3 rash as reported by the investigators in the respective trials using National Cancer Institute – common toxicity criteria v4.0 (NCI-CTCv4.0). Secondary outcome measures included trial protocol mandated or clinically necessitated adverse event driven dose reductions /treatment interruptions/cessation due to toxicity of vemurafenib. These outcomes may have occurred any time during trial participation. A sensitivity analysis was also conducted for early onset toxicities or dose reduction/interruption/cessation (i.e. within first 28 days of vemurafenib initiation). Data on baseline clinical and tumor related characteristics, treatment details and toxicities were extracted at an individual patient level within a secure platform provided by the data sharing service.
Apparent oral ClB was estimated using all available dosing and plasma concentration data in BRIM-2, BRIM-3, and coBRIM trials. Plasma samples for concentration were collected at various time points according to the trial protocol as described in supplementary data. Vemurafenib plasma concentrations collected in BRIM-2 and BRIM-3 have previously been described using a one-compartment population pharmacokinetic model with first-order absorption, first-order elimination and time-varying bioavailability 8. The open-source PKADVAN package for R (version 3.4) 9 was used to implement the model, providing model predictions for empirical Bayesian estimation of vemurafenib pharmacokinetic parameters, including apparent oral ClB at steady-state. The empirical Bayesian estimates were determined for each individual using an optimisation algorithm designed to maximise the likelihood of estimates considering both the observed data and the prior model (see supplementary material for equation).
Statistical analysis
Associations between vemurafenib apparent oral ClB at steady state and any type of grade ≥ 3 toxicity, grade ≥ 3 rash, and dose reduction or treatment interruption were modelled using Cox proportional hazards regression and are reported as hazard ratios (HR) with 95% confidence intervals (95% CI). All analyses were stratified by study and treatment arm. Analyses accounting for confounding by age, sex, and ECOG status were also conducted.
Potential non-linear associations were evaluated using cubic splines with three knots with visual checks. The pooled data were split into terciles to obtain thresholds of ClB. Patients were then divided into three groups according to vemurafenib apparent ClB thresholds (low: for the lowest tercile, medium: middle tercile, and high: for the highest tercile). Goodness of fit were evaluated using Akaike information criterion (AIC) and concordance statistic (cstatistic). The Kaplan-Meier method was used to estimate the cumulative probability at 12 months and 28-days after vemurafenib initiation for any grade ≥ 3 toxicities, grade ≥ 3 skin rash and dose reduction or treatment interruption over time for each of the three CLB groups and reported as percentages and 95%CI. The effect of the addition of cobimetinib to vemurafenib on the three outcomes when compared to monotherapy with vemurafenib was also evaluated as sub-group analysis. The influence of cobimetinib on any grade ≥ 3 toxicity was also evaluated with a Cox proportional hazards model of the treatment arms of the coBRIM clinical trial. All analyses were conducted in R (version 4.3). Statistical tests were two-sided, with a P value <0.05 considered statistically significant.
Results
The pooled cohort of patients included a total of 962 patients who initiated vemurafenib treatment (715 on vemurafenib monotherapy, and 247 on vemurafenib plus cobimetinib). Plasma concentration and exposure datasets were combined to create a longitudinal pharmacokinetic dataset consisting of both dose and plasma concentration. Sixty-four patients had no valid plasma concentration or dosing data and were excluded from the analysis. A total of 5,654 plasma concentrations including 1,422 from cycle1 day15 timepoint were available from the remaining patients. As such, the final pooled dataset consisted of 898 patients (653 on vemurafenib monotherapy, and 245 on vemurafenib plus cobimetinib). A summary of patient characteristics is described in the Table 1. Median follow-up for the pooled cohort was 26 months.
Outcomes
Table 2 and supplementary table S1 present a summary of three main study outcomes documented in BRIM-2, BRIM-3, and coBRIM. Within the available data, any grade ≥ 3 toxicity was experienced by 69% (N = 619) of patients, and grade ≥ 3 skin rash occurred in 15% (N =139). 47% had dose reduction or interruption while 12% had a permanent cessation of vemurafenib. The rate of any grade ≥ 3 toxicity was higher in patients receiving vemurafenib with cobimetinib compared to those who received vemurafenib alone (HR 1.33, [95%CI 1.07 – 1.65], P = 0.009). A significant proportion of adverse events occurred in the first 28 days of treatment; 32% (N =289) of all patients experienced any grade ≥ 3 toxicity and 13% (116) grade ≥ 3 rash. A dose reduction, interruption or cessation of vemurafenib treatment occurred in 59% (N = 528) of all patients within the available data, while 36% (N = 326) of patients had a dose reduction, interruption or cessation by day 28.
Association between vemurafenib clearance (ClB)and outcomes
The median estimated apparent oral clearance (ClB) of vemurafenib was 1.35 L/h (interquartile range (IQR) 1.15 – 1.65 L/h) for the whole cohort with significant difference between male and female (1.42 L/h [IQR 1.19-1.76] vs 1.27 L/h [IQR 1.04-1.53], P < 0.001). The association between vemurafenib CLB and toxicity outcomes were best described by a linear relationship (supplementary table S2). Lower vemurafenib CLB was significantly associated with an increased risk of grade ≥ 3 toxicity (HR 0.62, [95%CI 0.52 – 0.74], P <0.001), grade ≥ 3 rash (HR 0.29, [95%CI 0.18 – 0.46], P <0.001), and adverse events requiring vemurafenib dose reduction, interruption or cessation (HR 0.50, [95%CI 0.41 – 0.63], P <0.001). The association between vemurafenib CLB and the three outcomes remained statistically significant after additionally adjusting for age, sex and ECOG PS (supplementary table S3). The association between vemurafenib CLB and toxicity outcomes were also consistent between patient receiving vemurafenib monotherapy or vemurafenib plus cobimetinib (supplementary table S4).
Patients were then divided into three groups according to vemurafenib CLB terciles (low: < 1.22 L/h, medium: ≥ 1.22 and < 1.55 L/h, and high: ≥ 1.55 L/h). Figure 1 presents the cumulative incidence of grade ≥ 3 toxicity, grade ≥ 3 rash, and adverse events requiring vemurafenib dose reduction, interruption or cessation according to vemurafenib CLB groups. For the low CLB group, the cumulative probability of grade ≥ 3 toxicity, grade ≥ 3 rash and adverse events requiring vemurafenib dose reduction or interruption at 12 months were 78%, 23% and 71% respectively. For the medium CLB group, 67% had grade ≥ 3 toxicity, 14% had grade ≥ 3 rash and 63% had adverse events requiring vemurafenib dose alterations. On the other hand, 66% experienced grade ≥ 3 toxicity, 10% had grade ≥ 3 rash and 53% experienced adverse events requiring vemurafenib dose alterations among the high CLB group. The cumulative probability of grade ≥ 3 toxicity, grade ≥ 3 rash, and adverse events requiring vemurafenib dose reduction or interruption at 28 days range 43%, 21% and 47% for the low CLB group, 37%, 12% and 37% for the medium CLB group and 24%, 7% and 25% for the high CLB group, respectively. Supplementary figures SF1-3 present cumulative incidence estimates of grade ≥ 3 toxicity, grade ≥ 3 rash, and adverse events requiring vemurafenib dose reduction, interruption or cessation according to vemurafenib CLB, subset by patient receiving vemurafenib monotherapy or vemurafenib plus cobimetinib.
Discussion
In this large study, we have identified that apparent oral clearance CLB of vemurafenib is associated with any grade ≥ 3 toxicity, grade ≥ 3 rash, and adverse events requiring dose reduction, interruption or cessation. The results of this study highlight a significant relationship between increased vemurafenib exposure and toxicity. This provides a strategy to prospectively identify patients at greatest risk of toxicity based on an estimated CLB for vemurafenib. Patients with a low vemurafenib estimated CLB (below 1.22 L/h) were more likely to experience any grade ≥ 3 toxicity and grade ≥ 3 rash and receive a dose reduction or interruption in treatment compared to those with a medium or high estimated CLB. Clinically relevant confounders were adjusted for, including treatment with cobimetinib, improving the validity of observed associations between estimated vemurafenib CLB and toxicity.
The relationship between vemurafenib exposure and toxicity has been considered previously, however, has remained poorly defined. Kramkimel et al. identified (n = 159 samples from 39 patients) reported an association between increased vemurafenib exposure at day 15 and grade ≥ 2 skin rash 3. The present analysis similarly identified a relationship between increased vemurafenib exposure at day 28 and increased incidence of grade ≥ 3 skin rash. In contrast to the current study, Kramkimel et al. did not find a relationship between increased vemurafenib exposure and incidence of other grade ≥ 3 toxicity 3. This difference is likely due to the small sample size in contrast to the large cohort with more than 800 patients and improved power in the current study.
In clinical practice, clearance as a pharmacokinetic parameter that determines the dose of drug is rarely measured directly. However, it can be estimated using the dose administered and area under the curve (AUC) through measuring plasma drug concentration at steady state. The previously established threshold for efficacy and the results from the current study highlight the potential for therapeutic drug monitoring for vemurafenib to improve efficacy and reduce toxicities3,4,6. While it is common to measure CLB after a steady state is reached or when toxicity has occurred, predicting an apparent CLB prior to starting therapy or prior to the occurrence of toxicities is not commonly performed. Novel strategies are required to to estimate apparent CLB prior to the occurrence of toxicities which may allow the prescribers to individualise the dose for a particular patient. Prevention and early detection of toxicities during drug therapies especially for cancer may prevent early cessation of a potentially effective agent. Hence, measurement of plasma concentration of vemurafenib to improve survival/response and to estimate clearance that is associated with toxicity prediction appears to be a useful strategy. Further research should involve prospective studies to explore the potential for implementation of therapeutic drug monitoring for vemurafenib.
It is well established that vemurafenib is metabolised by hepatic cytochrome P4503A (CYP3A) and UDP glucuronosyltransferase 1 (UGT1A1), and in vitro evidence suggests it is also a substrate for the ATP-binding cassette transporter, subfamily B member 1 (ABCB1, or P-glycoprotein) 1,8. A 40% increase in the AUC of a single dose of vemurafenib was observed following concurrent administration with rifampicin (a strong CYP3A inducer) 8. Food has also been shown to increase the Cmax and AUC of vemurafenib by 2.5-fold and 5-fold, respectively 1. Given that vemurafenib is a substrate of several drug metabolising enzymes and transporters, endogenous biomarkers of vemurafenib clearance may offer an alternative approach to identifying patients at risk of experiencing adverse events that vemurafenib is a substrate of several drug metabolising enzymes and transporters. Plasma-derived extracellular nanovesicles are a novel source of biomarker and have been used to characterise the activity of the drug metabolising enzymes and transporter involved in the metabolism of vemurafenib 10,11. Endogenous biomarkers (and extracellular nanovesicles in particular) could be used to identify patients likely to experience toxicities prior to starting vemurafenib who should be initiated on a reduced dose.
There are several limitations of this study. This study was a post hoc analysis in which vemurafenib CLB was estimated using trough plasma concentration (Css,min) data which may have been collected after toxicity occurrence. The results from this study should be considered as hypothesis generating and further validation studies should be conducted to confirm these findings. As such, prospective clinical trials with a sufficient number of patients and plasma samples obtained prior to day 28 are required to define a vemurafenib Css,min threshold and an optimal sampling time associated with toxicity. Moreover, we have not evaluated the time gap between the measurement of plasma concentration of vemurafenib and the occurrence of toxicity outcomes. If there is sufficient lead-time, it would be useful to assess vemurafenib clearance earlier than the occurrence of toxicities so as to adjust the dose and prevent significant toxicities.
Conclusions
The estimated apparent clearance of vemurafenib is associated with severe toxicities and dose adjustment or cessation. The results of the PLX4032 current study suggest that an early estimation of vemurafenib exposure may be useful in identifying patients at risk of experiencing toxicity.
References
1. Zhang W, Heinzmann D, Grippo JF. Clinical Pharmacokinetics of Vemurafenib. Clin Pharmacokinet. 2017;56(9):1033-1043.
2. Hopkins AM, Rathod AD, Rowland A, Kichenadasse G, Sorich MJ. Risk factors for severe rash with use of vemurafenib alone or in combination with cobimetinib for advanced melanoma: pooled analysis of clinical trials. BMC Cancer. 2020;20(1):157.
3. Kramkimel N, Thomas-Schoemann A, Sakji L, et al. Vemurafenib pharmacokinetics and its correlation with efficacy and safety in outpatients with advanced BRAF-mutated melanoma. Target Oncol. 2016;11(1):59-69.
4. Funck-Brentano E, Alvarez JC, Longvert C, et al. Plasma vemurafenib concentrations in advanced BRAFV600mut melanoma patients: impact on tumour response and tolerance. Ann Oncol. 2015;26(7):1470-1475.
5. Funck-Brentano E, Alvarez JC, Longvert C, et al. Reply to the letter to the editor ‘Plasma vemurafenib concentrations in advanced BRAFV600mut melanoma patients: impact on tumor response and tolerance’ by Funck-Brentano et al. Ann Oncol. 2016;27(2):364-365.
6. Kichenadasse G, Hughes JH, Miners JO, et al. Relationship between vemurafenib plasma concentrations and survival outcomes in patients with advanced melanoma. Cancer Chemother Pharmacol. 2020;85(3):615-620.
7. Lewis KD, Larkin J, Ribas A, et al. Impact of depth of response on survival in patients treated with cobimetinib +/- vemurafenib: pooled analysis of BRIM-2, BRIM-3, BRIM-7 and coBRIM. Br J Cancer. 2019;121(7):522-528.
8. FDA. Clinical pharmacology and biopharmaceutics review. https://www.accessdata.fda.gov/drugsatfda_docs/nda/2017/202429Orig1s016ClinPharmR.pdf. Published 2017. Accessed 24 December, 2020.
9. Abulhelwa A, Foster D, Upton R. PKADVAN. https://github.com/abuhelwa/PKADVAN_Rpackage. Published 2017. Accessed 24th December, 2020.
10. Rodrigues D, Rowland A. From Endogenous Compounds as Biomarkers to Plasma-Derived Nanovesicles as Liquid Biopsy; Has the Golden Age of Translational PharmacokineticsAbsorption, Distribution, Metabolism, Excretion-Drug-Drug Interaction Science Finally Arrived? Clin Pharmacol Ther. 2019;105(6):1407-1420.
11. Rowland A, Ruanglertboon W, van Dyk M, et al. Plasma extracellular nanovesicle (exosome)derived biomarkers for drug metabolism pathways: a novel approach to characterize variability in drug exposure. Br J Clin Pharmacol. 2019;85(1):216-226.