Relationship between vemurafenib plasma concentrations and survival outcomes in patients with advanced melanoma
Purpose To validate a plasma vemurafenib steady-state trough concentration (Css,min) threshold that predicts survival outcomes of patients with BrafV600 mutated melanoma.
A pooled analysis of individual patient data from two advanced melanoma trials involving vemurafenib ± cobimetinib therapy was performed. Day 23 was chosen as the landmark time when steady-state concentration reached. Optimal Css,min threshold was determined via assessment of discriminative performance and model fitting. Association between vemurafenib Css,min and survival was modelled using Cox proportional hazards regression.
Vemurafenib plasma concentration data were available for 402 patients who were on stable dose for the first 3 weeks. When compared to a previously described plasma vemurafenib Css,min threshold of 42 mg/L, we identified that a cutoff concentration of 50 mg/L by day 23 was strongly associated with progression-free survival and overall survival. The association remained statistically significant after adjusting for important clinical confounding variables. Sub-group analysis showed that while the addition of cobimetinib resulted in a lower day 23 plasma vemurafenib Css,min, the threshold was still associated with overall survival and not in the monotherapy cohort.
A plasma vemurafenib Css,min threshold of 50 mg/L is strongly associated with survival outcomes in patients with advanced melanoma. This new threshold needs to be validated prospectively in future studies.
KeywordsVemurafenib Melanoma Threshold concentration Survival
Several kinase inhibitors are available for the management of both haematological and solid cancers. While a fixed dosage regimen is routinely used, these drugs are characterized by a highly variable pharmacokinetics (PK) and hence exposure. However, relationships between exposure, response and toxicity are increasingly described indicating the potential for plasma concentration guided dosing strategies . Vemurafenib, a serine–threonine kinase inhibitor, is approved as monotherapy for the treatment of both BRAFV600 mutated advanced melanoma and BRAFV600 mutated Erdheim–Chester disease by the US Food and Drug Administration. It is also approved in combination with cobimetinib, a mitogen-activated protein kinase (MEK) inhibitor, for the treatment of BRAFV600 mutated advanced melanoma. The current recommended starting dose of vemurafenib is 960 mg twice a day given orally with dose modifications allowed for significant adverse events.
The clinical PK of vemurafenib is well established, with oral absorption showing high inter-patient variability (coefficient of variation 101%) . Doses of 960 mg twice daily are associated with a median time to maximum drug concentration of 4–5 h and mean (± SD) maximum plasma steady-state concentrations (Cmax) of 62 ± 23 mg/L. The elimination half-life of vemurafenib varies over time, changing from 25 h after a single dose to 57 h over multiple doses, with steady state being achieved by 15–22 days . Once at steady state, vemurafenib exhibits linear PK with the area under the concentration curve (AUC) over 8 h being 392 ± 126 mg h/L and the apparent oral clearance 31 L/day (coefficient of variation 32%) . While the bioavailability of vemurafenib is unknown, food has been shown to have a significant effect on vemurafenib plasma concentrations with a 5-fold increase in AUC and a 2.5-fold increase in Cmax.
Due to the high inter-patient and intra-patient variability reported for plasma vemurafenib concentrations, previous studies have evaluated exposure–response and exposure–toxicity relationships using a single plasma concentration threshold. Kramkimel et al. analysed 159 samples from 39 patients and reported that plasma vemurafenib concentrations below 40.4 mg/L at day 15 were associated with significantly shorter progression-free survival (PFS) and a lower incidence of grade ≥ 2 rash . Goldwirt et al. (148 samples from 48 patients) identified a threshold of 42 mg/L during the 1st year of therapy that differentiated responders and non-responders . A similar proportion of responders and non-responders was reported by another group using a threshold of 42 mg/L (data from 23 patients), and both studies did not find an exposure–toxicity relationship [6, 7]. Moreover, Funck-Brentano et al. suggested that a higher threshold may be required if a lack of early response is noted .
The current study aimed to validate the proposed plasma vemurafenib steady-state trough concentration (Css,min) threshold as a predictor of PFS and overall survival (OS) in patients with advanced melanoma.
Materials and methods
Individual patient data from the previously published BRIM-3 (NCT01006980) and coBRIM (NCT01689519) clinical trials were accessed through Roche’s data sharing policy. BRIM-3 was a monotherapy trial that compared vemurafenib and dacarbazine, while coBRIM was a combination therapy trial of vemurafenib/cobimetinib vs vemurafenib monotherapy for the treatment of advanced BRAFV600 mutated melanoma.
The primary outcome assessed for the current study was PFS, while the secondary outcomes were OS and best overall response (BOR). Patients who fulfilled all the following criteria were included in the primary analysis: had at least one Css,min of vemurafenib available by day 23 of cycle 1 (D23); had no dose changes for at least 14 days prior to the sample collection; and had not progressed or died before D23. Vemurafenib concentrations were considered to be at steady state after 14 days at a consistent dose; a day 23 landmark was utilised (as opposed to day 15) to account for different sampling times between clinical trials. Sensitivity analysis was also performed with no dose changes for a minimum of 7 days prior to sample collection.
Associations between plasma vemurafenib Css,min and PFS/OS were modelled using Cox proportional hazards regression and reported as hazard ratios (HR) with 95% confidence intervals (95% CI). Statistical tests were two-sided and a P value less than 0.05 was considered statistically significant. Clinically relevant confounders, including, age, gender, ECOG performance status (PS), stage of melanoma, BRAF V600 mutation type, LDH concentration and sites of metastases were accounted for in adjusted analyses. BOR included patients who achieved a complete or partial response using RECIST 1.1. Potential non-linear associations were evaluated using restricted cubic splines with 3–5 knots and subsequent visual checks; an optimal Css,min threshold was determined via assessment of discriminative performance (concordance statistic − c-statistic), model fit [Akaike information criterion (AIC)] and consistency of the PFS/OS association. c-statistic was used instead of the commonly used receiver operating curve (ROC) to define thresholds of concentrations. Survival curves were estimated using the Kaplan–Meier method. All analyses were conducted in R (version 3.4.3) using the survival package . Ethics approval was waived by the Southern Adelaide Clinical health Research Ethics Committee.
A total of 830 patients (583 on vemurafenib monotherapy and 247 on the vemurafenib/cobimetinib combination) from the two clinical trials were available for selection. After exclusion of 428 patients (62 for lack of plasma concentration data, 352 for dose adjustments 15 days prior to the cutoff, and 14 for disease progression, death or loss of follow-up prior to day 23), 402 were available for further analysis. A summary of patient characteristics is described in the supplementary material (Table S1). The median follow-up was 25.3 months, median PFS was 7.2 months and the median OS was 14.7 months. The median plasma vemurafenib Css,min was 54.4 mg/L (interquartile range 42.5–69.7 mg/L).
There was no significant association between plasma vemurafenib Css,min, either at 42 mg/L or 50 mg/L, and BOR (odds ratio 1.17, 95% CI 0.71–1.94; P = 0.53 and 1.38, 95% CI 0.880–2.15; P = 0.16, respectively) (Figure S1).
Next, we evaluated the effect of cobimetinib on vemurafenib and outcomes as a part of sub-group analysis (Fig. 2). The median day 23 vemurafenib (Cmin, ss) was lower in the combination cohort when compared to vemurafenib alone (52 vs 57 mg/L, P = 0.008), Table S1. The threshold of 50 mg/L was associated with OS (HR 0.7, 95% CI 0.52–0.95, P = 0.02), but not for PFS or BOR only with combined cobimetinib and vemurafenib (Table S5). There was no association between the threshold of 50 mg/L and PFS, OS or BOR in the monotherapy cohorts.
This analysis of prospectively collected data from 402 patients with advanced melanoma demonstrated that a plasma vemurafenib Css,min threshold of ≥ 50 mg/L was associated with improved survival outcomes. Clinically relevant confounding factors were systematically evaluated and adjusted in our analyses, thereby improving the validity of the association identified between plasma vemurafenib Css,min and outcomes.
Previous studies reported a lower concentration threshold, 40.4 or more than 42 mg/L, than that identified in the current study [4, 6, 7]. However, these studies included fewer patients and did not address confounding variables potentially affecting survival outcomes, and the association was not explored for both PFS and OS [4, 5, 6, 7]. By contrast, the present study used large high-quality data and the vemurafenib Css,min threshold (≥ 50 mg/L) was demonstrated as significantly associated with both OS and PFS. The association between vemurafenib Css,min threshold (≥ 50 mg/L) and OS was seen for the combined population of two trials and for the combination therapy sub-group, while a non-significant trend was seen in the monotherapy group.
While it is common to use ROC to define thresholds of concentrations, in our study, we have used c-statistic, which is equivalent to ROC. Both ROC and c-statistic are used for discrimination of a model generated to predict outcomes. While ROC is to visualize, c-statistic quantifies the discriminative ability of a model. The higher the c-statistic is, the better the model prediction for outcome of interest will be . We found that vemurafenib Css,min threshold of ≥ 50 mg/L had the highest c-statistic among various cutoffs modelled to predict outcomes.
A population PK analysis for vemurafenib in a cohort of advanced melanoma patients was performed by Roche as part of the submission for regulatory review. The dataset comprised 5515 plasma samples from 459 patients including participants from BRIM3 trial [2, 10]. In contrast to our study where the plasma vemurafenib Css,min was the PK parameter of interest, the relationship between mean AUC0–8 h on day 15 and response was explored. While an increase in tumour response with increasing exposure was noted, there was no clear exposure–response (PFS or OS) relationship at a dose of 960 mg bd. Similarly, another population PK model using 147 plasma samples from 26 patients with non-melanoma diseases with BRAFV600 mutations reported overlapping mean plasma vemurafenib concentrations across BOR categories . It appears that no significant relationship between vemurafenib exposure and tumour response has been consistently described across all studies.
In our study, we did not find any association between the plasma vemurafenib Css,min threshold of ≥ 50 mg/L and BOR. It is unclear why there was an association with survival outcomes in the absence of association with tumour response. Similarly, the reason behind day 23 plasma vemurafenib concentration being associated with OS is poorly understood. It is unclear how achieving optimal vemurafenib therapeutic concentrations may influence subsequent post-trial treatment and thereby, OS. However, the improvement in PFS likely contributed by improved depth of response may influence survival outcomes .
It was previously established that vemurafenib exposure was not altered by the addition of cobimetinib . In the current study, we noted a significantly lower median day 23 plasma concentration of vemurafenib when combined with cobimetinib than when given as monotherapy. The mechanisms behind this potential drug–drug interaction are unclear. Moreover, the association between the threshold of 50 mg/L and survival was seen only in the combination and not seen in the monotherapy cohort. The small size of these two cohorts may have precluded meaningful evaluation of association between concentration threshold and survival outcomes.
There are several limitations in our study. The current study was an exploratory, post hoc analysis. Hence, the new proposed threshold requires validation in prospective clinical trials with an adequate number of patients and plasma samples. A large proportion of patients were excluded from the current analysis due to dose adjustments within 15 days prior to the D23 cutoff. Further, the current study evaluated plasma concentrations within the 1st month of starting vemurafenib. Although it is unclear if the threshold remains valid beyond this time, the association between D23 plasma vemurafenib Css,min values and OS suggests that the relationship may persist beyond the 1st month. Moreover, the exposure–toxicity relationship was not evaluated to determine if there is an upper limit for survival benefit with acceptable toxicities.
Another limitation is the use of a single threshold trough plasma concentration (Css,min) rather than area under the curve (AUC) approach. While the latter is increasingly considered as a better pharmacokinetic parameter for relationship with response/toxicity, the lack of access to the data to calculate vemurafenib AUC precluded its assessment in the current study. In addition, the measurement of trough concentration may reduce the barriers for clinical translation of dose individualisation for kinase inhibitors .
A vemurafenib steady-state trough plasma concentration (Css,min) threshold of 50 mg/L is strongly associated survival outcomes for patients with advanced melanoma. This new threshold needs to be validated prospectively in future studies.
The current study was undertaken with the financial support of Cancer Council South Australia’s Beat Cancer Project on behalf of its donors and the State Government through the Department of Health (Grant ID: 1159924 and 1127220). A.R. is supported by a Beat Cancer Mid-Career Research Fellowship from Cancer Council SA. A.M.H is a researcher funded by a Postdoctoral Fellowship from the National Breast Cancer Foundation, Australia (PF-17-007).
Compliance with ethical standards
Conflict of interest
MJS and AR report investigator-initiated project grants from Pfizer, outside the scope of the submitted work. The authors have no other conflicts of interest to disclose.
- 3.FDA (2017) Clinical pharmacology and biopharmaceutics review. https://www.accessdata.fda.gov/drugsatfda_docs/nda/2017/202429Orig1s016ClinPharmR.pdf. Accessed 27 Jul 2019
- 4.Kramkimel N, Thomas-Schoemann A, Sakji L, Golmard J, Noe G, Regnier-Rosencher E, Chapuis N, Maubec E, Vidal M, Avril M, Goldwasser F, Mortier L, Dupin N, Blanchet B (2016) Vemurafenib pharmacokinetics and its correlation with efficacy and safety in outpatients with advanced BRAF-mutated melanoma. Target Oncol 11(1):59–69. https://doi.org/10.1007/s11523-015-0375-8 CrossRefPubMedGoogle Scholar
- 5.Goldwirt L, Chami I, Feugeas JP, Pages C, Brunet-Possenti F, Allayous C, Baroudjian B, Madelaine I, Sauvageon H, Mourah S, Lebbe C (2016) Reply to ‘Plasma vemurafenib concentrations in advanced BRAFV600mut melanoma patients: impact on tumour response and tolerance’ by Funck-Brentano et al. Ann Oncol 27(2):363–364. https://doi.org/10.1093/annonc/mdv538 CrossRefPubMedGoogle Scholar
- 6.Funck-Brentano E, Alvarez JC, Longvert C, Abe E, Beauchet A, Funck-Brentano C, Saiag P (2015) Plasma vemurafenib concentrations in advanced BRAFV600mut melanoma patients: impact on tumour response and tolerance. Ann Oncol 26(7):1470–1475. https://doi.org/10.1093/annonc/mdv189 CrossRefPubMedGoogle Scholar
- 7.Funck-Brentano E, Alvarez JC, Longvert C, Abe E, Beauchet A, Saiag P, Funck-Brentano C (2016) 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 27(2):364–365. https://doi.org/10.1093/annonc/mdv549 CrossRefPubMedGoogle Scholar
- 8.Team RC (2017) R: A language and environment for statistical computing, vol 2019. R Foundation for Statistical Computing, ViennaGoogle Scholar
- 10.(TGA) TGA (2012) Australian Public Assessment Report for Vemurafenib. Commonwealth of Australia 2012. https://www.tga.gov.au/sites/default/files/auspar-vemurafenib-121217.pdf. Accessed 27 Jul 2019
- 11.Lewis KD, Larkin J, Ribas A, Flaherty KT, McArthur GA, Ascierto PA, Dreno B, Yan Y, Wongchenko M, McKenna E, Zhu Q, Mun Y, Hauschild A (2019) 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 121(7):522–528. https://doi.org/10.1038/s41416-019-0546-y CrossRefPubMedGoogle Scholar
- 12.Ribas A, Gonzalez R, Pavlick A, Hamid O, Gajewski TF, Daud A, Flaherty L, Logan T, Chmielowski B, Lewis K, Kee D, Boasberg P, Yin M, Chan I, Musib L, Choong N, Puzanov I, McArthur GA (2014) Combination of vemurafenib and cobimetinib in patients with advanced BRAF(V600)-mutated melanoma: a phase 1b study. Lancet Oncol 15(9):954–965. https://doi.org/10.1016/S1470-2045(14)70301-8 CrossRefPubMedGoogle Scholar