Network Meta-analysis of Progression-Free Survival and Overall Survival in First-Line Treatment of BRAF Mutation-Positive Metastatic Melanoma

The treatment of metastatic melanoma has been transformed in recent years by the introduction of immune checkpoints inhibitors and therapies targeted on the MAP kinase (MAPK) pathway. Immune checkpoint inhibitors block proteins on cytotoxic T-cells or cancer cells which inhibit T cell signaling, such as the cytotoxic lymphocyte-associated protein 4 (CTLA-4) or the programmed death-1 (PD-1) protein, thus reactivating an immune system response against tumor cells. Ipilimumab, an anti-CTLA-4 antibody, was the first immune checkpoint inhibitor to be approved in melanoma. In controlled trials, ipilimumab has demonstrated consistent activity against melanoma in pretreated and chemotherapy-naïve patients [1, 2, 3]. More recently, two novel anti-PD-1 immune checkpoint inhibitors, pembrolizumab and nivolumab, have been approved based on randomized controlled trials demonstrating improved outcomes versus ipilimumab alone [4, 5, 6].

Dysregulation of BRAF signaling in the MAPK pathway is a key driver of metastatic melanoma [7]. Two inhibitors of the BRAF protein, vemurafenib and dabrafenib, have been approved for treatment of patients with BRAF V600 mutation-positive unresectable stage IIIC or stage IV (advanced or metastatic) melanoma. These approvals were based on the results of the BRIM-3 (ClinicalTrials.gov identifier: NCT01006980) and BREAK-3 (ClinicalTrials.gov identifier: NCT01227889) randomized controlled trials, respectively, which demonstrated improved outcomes versus dacarbazine (DTIC) in front-line treatment of BRAF V600 mutation-positive unresectable or metastatic melanoma [8, 9, 10, 11, 12, 13, 14]. Trametinib is a highly potent and selective small-molecule inhibitor of the MEK kinase that is downstream of and activated by BRAF in the MAPK pathway. Trametinib has been shown to be more effective than chemotherapy in treatment-naïve or previously treated patients with BRAF V600 mutation-positive unresectable or metastatic melanoma [15]. These results, as well as improved understanding of the mechanisms of resistance to targeted therapies, have led to the exploration of combination approaches to targeted therapy. The efficacy and safety of dabrafenib in combination with trametinib has been evaluated in two phase 3 trials, the COMBI-d (ClinicalTrials.gov identifier: NCT01584648) and COMBI-v (ClinicalTrials.gov identifier: NCT01597908) trials, as well as the phase 2 BRF113220 (ClinicalTrials.gov identifier: NCT01072175) trial [16, 17, 18, 19, 20, 21]. Results of these trials have demonstrated that dabrafenib plus trametinib improves progression-free survival (PFS) and overall survival (OS) compared with dabrafenib monotherapy or vemurafenib monotherapy treatment-naïve patients with BRAF V600 mutation-positive unresectable or metastatic melanoma.

Economic evaluations of novel treatments such as dabrafenib plus trametinib require estimates of the relative treatment effects versus other available therapies on PFS and OS. The objective of this study was to conduct a network meta-analysis of HRs for PFS and OS of dabrafenib plus trametinib versus other first-line treatments of BRAF mutation-positive advanced or metastatic melanoma for use in an economic evaluation of dabrafenib plus trametinib in this indication for submission to Canadian reimbursement authorities.

This study was a network meta-analysis of HRs for PFS and OS reported in randomized controlled trials of first-line treatments for patients with BRAF mutation-positive advanced or metastatic melanoma. Comparators were limited to first-line treatments for patients with metastatic melanoma that were approved by Health Canada as of February 2015 and included dabrafenib plus trametinib, dabrafenib, vemurafenib, trametinib, ipilimumab, and DTIC. Nivolumab, pembrolizumab, and cobimetinib plus vemurafenib were not included in the analysis as these treatments were not yet approved of that date.

Trials included in the analysis were identified from two systematic literature reviews. Details regarding the reviews have been published previously [22, 23]. Both reviews followed the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) guidelines [24]. The first review was undertaken to evaluate efficacy, safety, and tolerability of dabrafenib and trametinib monotherapy versus other first-line treatments for unresectable advanced or metastatic melanoma [22]. The second review was undertaken to evaluate the efficacy and safety of the dabrafenib plus trametinib therapy compared to other first-line or second-line treatments for patients with unresectable or metastatic melanoma [23]. For both reviews, Embase^®, MEDLINE^® (including MEDLINE^® In-Process and other non-indexed citations), Cochrane Central Trials Register, and key conferences (i.e., American Society of Clinical Oncology, European Society for Medical Oncology/European Cancer Organization, International Congress of the Society for Melanoma Research) were searched. Additionally, data reported in public assessment reports from the European Medicines Agency and Food and Drug Administration websites were searched, as were Health Technology Assessment (HTA) guidance documents published in English from selected HTA authorities (e.g., National Institute for Health and Care Excellence). The first review included randomized and non-randomized studies of dabrafenib, trametinib, DTIC, ipilimumab, vemurafenib, fotemustine, and temozolomide published through October 2012. The second review included randomized studies of dabrafenib plus trametinib, dabrafenib, trametinib, vemurafenib-cobimetinib combination, vemurafenib, ipilimumab, nivolumab, and pembrolizumab published through October 2015. For both reviews, screening and extraction was conducted by two independent reviewers. The first review identified 123 studies which met all inclusion criteria, of which 24 were randomized controlled trials in treatment-naïve patients [22]. The second review identified 12 randomized controlled trials of one or more of the treatment of interest in patients with unresectable or metastatic melanoma [23].

Combining the results of these two reviews, and focusing on trials of previously untreated patients receiving any one of the comparators approved in Canada at the time of the analysis, and which were required to estimate HRs for PFS and OS for the comparators versus dabrafenib plus trametinib, yielded seven studies, including the COMBI-d and BRF113220 (dabrafenib–trametinib versus dabrafenib), COMBI-v (dabrafenib–trametinib versus dabrafenib), BREAK-3 (dabrafenib versus DTIC), BRIM-3(vemurafenib versus DTIC), METRIC (ClinicalTrials.gov identifier: NCT01245062, trametinib versus chemotherapy), and CA184-024 (ClinicalTrials.gov identifier: NCT00324155, ipilimumab plus DTIC vs DTIC). HRs for PFS and OS were based on analyses of the most recent data cuts available from each trial at the time the evaluation was completed [9, 13, 15, 16, 18, 20, 25]. HRs for PFS and OS from the METRIC trial were based on analyses of the first-line subgroup. HRs for OS from the BRIM-3, BREAK-3, METRIC, and BRF113220 trials were based on Rank Preserving Structural Failure Time (RPSFT) analyses to adjust for crossover from control to active therapy [10, 14, 26, 27]. HRs for PFS and OS for ipilimumab were based on results of the CA184-024 trial [25]. Although this trial included both BRAF mutation-positive and BRAF wild-type patients, it was the only study available at the time of the analysis that reported both PFS and OS for ipilimumab as first-line treatment of metastatic melanoma. Also, although the approved dosage for ipilimumab is 3 mg/kg q3w as monotherapy, this trial compared ipilimumab 10 mg/kg plus DTIC versus DTIC. In the economic evaluation which this analysis informed, results for ipilimumab 10 mg/kg plus DTIC were used to estimate outcomes for patients receiving the approved dosage.

HRs were estimated by Bayesian network meta-analysis using WinBUGS 1.4.3© (MRC Biostatistics Unit, Cambridge, UK) with input data taking the form of the log transformed HRs from each trial and the corresponding standard errors [28]. For each comparison, the log HRs for PFS and OS were estimated alternatively using multivariate network meta-analysis (in which treatment effects on PFS and OS were estimated simultaneously) and using traditional or univariate network meta-analysis (in which the analyses of PFS and OS were conducted separately). A multivariate network meta-analysis uses the correlation of the HRs for PFS and OS in the network to inform parameter estimation [29]. The use of this approach is reasonable given the well-established correlation between treatment effects on PFS and OS in metastatic melanoma [30].

Given the small number of trials with which to estimate random effects (in the analysis assuming no class effect for BRAF inhibitors, there was only one comparison for which there was more than one trial to estimate the random effects), a fixed effects model was employed. Model parameters were estimated using the WinBUGS software package (version 1.43) (Medical Research Council (MRC) Biostatistics Unit (BSU), Cambridge, UK). WinBUGS code used in the analysis are provided in the supplementary material.

Two sets of analyses were performed. In one, it was assumed that the effects of dabrafenib and vemurafenib on PFS and OS may be different (“no class-effect”), and separate HRs were estimated for dabrafenib and vemurafenib. In the other analysis, it was assumed that the effects of dabrafenib and vemurafenib on PFS and OS are the same (“class effect”). In this analysis, HRs for other comparators versus BRAF inhibitor monotherapy were calculated. This analysis permitted the combining of data on PFS and OS from the COMBI-d, COMBI-v, and BRF113220 trials in order to potentially improve the precision of the comparisons. Both analyses were conducted alternatively including and excluding the phase 2 BRF113220 trial. Multivariate network meta-analyses (in which the HRs for PFS and OS were estimated simultaneously to account for the correlation of these outcomes), were conducted both including and excluding trametinib and ipilimumab plus DTIC from the network, in order to account for the possibility that the correlation between treatment effects on OS and PFS differ by drug class.

The heterogeneity of effect sizes reported for all direct comparisons with multiple sources of direct evidence was evaluated using Cochrane’s Q and the I ² statistics. Consistency of direct and indirect comparisons within closed loops defined by the evidence network was confirmed by plotting the confidence intervals (CIs) of the difference in log HRs between direct and indirect comparisons.

This article does not contain any new studies with human or animal subjects performed by any of the authors.

Evidence network diagrams for PFS and OS for the analysis assuming no class-effect for BRAF inhibitors and the analysis assuming class-effect for BRAF inhibitor are shown in Fig. 1a, b, respectively. In the latter analysis, HRs for PFS and OS for the comparison of dabrafenib plus trametinib versus BRAF inhibitor monotherapy from COMB-v, COMBI-d and BRF113220 are pooled, as are the HRs for BRAF inhibitor monotherapy versus DTIC from BREAK-3 and BRIM-3.

Open image in new window

Results of the network meta-analysis on HRs for PFS and OS for the analysis assuming no class-effect for BRAF inhibitors are shown in Table 3 and in Fig. 2. All HRs (95% CrI) for PFS significantly favored dabrafenib plus trametinib: 0.23 (0.18–0.29) versus DTIC, 0.32 (0.24–0.42) versus ipilimumab plus DTIC, 0.52 (0.32–0.83) versus trametinib, 0.57 (0.48–0.69) versus vemurafenib, and 0.59 (0.50–0.71) versus dabrafenib. All HRs (95% CrI) for OS also significantly favored dabrafenib plus trametinib: 0.41 (0.29–0.56) versus DTIC, 0.52 (0.38–0.71) versus ipilimumab plus DTIC, 0.68 (0.47–0.95) versus trametinib, 0.69 (0.57–0.84) versus vemurafenib, and 0.72 (0.60–0.85) versus dabrafenib. Results were similarly in favor of dabrafenib plus trametinib when the Phase II BRF113220 trial was excluded from the network. Excluding ipilimumab plus DTIC and trametinib from the network had little impact on the HRs for the other therapies. When HRs for PFS and OS were estimated separately using univariate network meta-analysis (i.e., without accounting for the correlation of treatment effects on PFS and OS), the beneficial effects on OS of dabrafenib plus trametinib versus ipilimumab plus DTIC and dabrafenib plus trametinib versus trametinib were reduced compared with estimates from the multivariate analysis in which the HRs for PFS and OS were estimated simultaneously (compared with trametinib, HR = 0.68 for multivariate analysis versus 0.98 for univariate analysis; compared with ipilimumab plus DTIC, HR = 0.52 for multivariate analysis versus 0.60 for univariate analysis).

Table 3

Estimated HRs (treatment versus comparator) for PFS and OS from network meta-analysis for all comparisons

Outcome/treatment	Comparator
	Dabrafenib + Trametinib	Vemurafenib	Dabrafenib	Trametinib	Ipilimumab + DTIC	DTIC
PFS
Dabrafenib + trametinib		0.56 (0.46, 0.69)	0.55 (0.43, 0.69)
Vemurafenib	0.57 (0.48, 0.69)					0.38 (0.32, 0.46)
Dabrafenib	0.59 (0.50, 0.71)	1.03 (0.82, 1.30)				0.37 (0.23, 0.57)
Trametinib	0.52 (0.32, 0.83)	0.91 (0.58, 1.42)	0.88 (0.54, 1.43)			0.44 (0.28, 0.69)
Ipilimumab + DTIC	0.32 (0.24, 0.42)	0.55 (0.43, 0.70)	0.53 (0.39, 0.73)	0.61 (0.38, 0.95)		0.76 (0.63, 0.93)
DTIC	0.23 (0.18, 0.29)	0.40 (0.34, 0.47)	0.39 (0.30, 0.50)	0.44 (0.29, 0.67)	0.73 (0.61, 0.87)
OS
Dabrafenib + trametinib		0.69 (0.53, 0.89)	0.69 (0.48, 0.99)
Vemurafenib	0.69 (0.57, 0.84)					0.64 (0.47, 0.88)
Dabrafenib	0.72 (0.60, 0.85)	1.03 (0.84, 1.28)				0.55 (0.21, 1.43)
Trametinib	0.68 (0.47, 0.95)	0.98 (0.70, 1.38)	0.95 (0.66, 1.35)			0.44 (0.20, 1.00)
Ipilimumab + DTIC	0.52 (0.38, 0.71)	0.75 (0.58, 0.97)	0.72 (0.54, 0.97)	0.77 (0.53, 1.07)		0.72 (0.59, 0.87)
DTIC	0.41 (0.29, 0.56)	0.59 (0.46, 0.74)	0.57 (0.42, 0.76)	0.60 (0.41, 0.83)	0.78 (0.67, 0.90)

Values above the diagonal represent direct evidence, whereas those below the diagonal represent direct and indirect evidence

HR hazard ratio, PFS progression-free survival, OS overall survival, DTIC dacarbazine

Open image in new window

Results of the network meta-analysis on HRs for PFS and OS for the analysis assuming a class-effect for BRAF inhibitors are shown in Fig. 3. These results were generally similar to those for the analysis in which separate HRs were estimated for dabrafenib and vemurafenib monotherapy. When all trials and treatments were included, the HRs for PFS and OS for dabrafenib plus trametinib were significantly less than 1.0 for all comparisons. The HRs and 95% CrI for dabrafenib plus trametinib versus trametinib, ipilimumab plus DTIC, and DTIC were similar to those obtained in the analysis in which no class-effect was assumed.

Open image in new window

For all comparisons of OS based on multiple sources of direct evidence, the I ² statistic was estimated to be 0%, indicating no identifiable heterogeneity. For PFS, significant heterogeneity of treatment effects was observed only for the PFS HR for the comparison of dabrafenib plus trametinib versus dabrafenib based on COMBI-d and BRF113220 (Q = 3.873, p = 0.049, I ² = 74.2%). When assuming a class effect, COMBI-V is included in the comparison, heterogeneity remains elevated, but drops below the threshold for significance (Q = 4.338, p = 0.114, I ² = 53.9%). No heterogeneity was identified in the comparison of BRAF monotherapies and DTIC based on BREAK-3 and BRIM-3 (Q = 0.011, p = 0.915, I ² = 0.0%). There was no evidence of inconsistency of direct and indirect estimates of treatment effects within the closed loop formed by the BREAK-3, BRIM-3, COMBI-v, and COMBI-d studies, as the CIs for the differences of log HRs between direct and indirect effects spanned zero for both PFS and OS.

This study was a network meta-analysis of HRs for PFS and OS of dabrafenib plus trametinib versus other first-line treatments of BRAF mutation-positive metastatic melanoma that was conducted to inform an economic evaluation of dabrafenib plus trametinib as first-line treatment of metastatic melanoma in Canada. Results of this analysis suggest that dabrafenib plus trametinib yields improved PFS and OS compared to other first-line treatment for BRAF mutation-positive metastatic melanoma. The results for dabrafenib-trametinib were more favorable compared with trametinib and ipilimumab plus DTIC when HRs for PFS and OS were estimated using multivariate network meta-analysis to account for the correlation of treatment effects on PFS and OS than when they were estimated separately using traditional univariate network meta-analysis. Results were generally similar when it was assumed there was a class effect on PFS and OS for BRAF inhibitor monotherapy compared with analyses in which no such assumption was made. Results also were similar when the phase 2 BRF113220 trial was included or excluded.

A targeted search identified several network meta-analyses of treatments for metastatic melanoma [22, 31, 32, 33, 34]. Two of these articles focused on comparisons of chemotherapies and interferon [32, 33]. One study focused on comparisons of ipilimumab versus immunotherapies, chemotherapies, and biochemotherapies, but did not consider BRAF or MEK inhibitors [34]. Srivastava and colleagues conducted a systematic review and indirect treatment comparison of dabrafenib and trametinib as monotherapy versus other treatments in previously untreated metastatic melanoma patients; this study did not include BRAF-MEK combination therapy, however [22]. Only one study was identified that included dabrafenib plus trametinib [31]. This study by Mai and colleagues compared BRAF-MEK inhibition with BRAF inhibitor monotherapy, MEK inhibitor monotherapy, BRAF inhibitor plus chemotherapy, and MEK inhibitor plus chemotherapy [31]. The study did not consider ipilimumab. The authors reported that combined BRAF–MEK inhibition significantly prolonged PFS compared with BRAF inhibition alone (HR = 0.58, 95% CI 0.51–0.67, p < 0.0001) and MEK inhibition alone (HR = 0.29, 95% CI 0.22–0.37, p < 0.0001). Similar results were obtained for OS for BRAF–MEK inhibition versus BRAF inhibition, (HR = 0.67, 95% CI 0.56–0.81, p < 0.0001) and for BRAF–MEK inhibition versus MEK inhibition (HR = 0.48, 95% CI 0.36–0.65, p < 0.0001). These results are qualitatively similar to those reported herein.

Limitations of this study should be noted. Perhaps most importantly, this analysis did not consider new immunotherapies such as nivolumab and pembrolizumab, or the combination of vemurafenib plus the MEK inhibitor, cobimetinib, as these therapies were not approved in Canada at the time this analysis was conducted and therefore were not required in the economic evaluation for which this network meta-analysis was performed. Future network meta-analyses of therapies for metastatic melanoma should include these novel treatments if feasible. It should be noted, however, that the robustness of any such analyses will depend on the availability of efficacy data on both PFS and OS in patients with BRAF mutations positive disease.

Differences in patients, study design, and duration of follow-up may have affected treatment effects thus violating the similarity assumption and confounding the comparisons. For example, the COMBI-d and CA184-024 trials were double-blind, whereas the other trials in the network were open label. The CA184-024 trial included both BRAF mutation-positive and wild-type patients. These factors may modify treatment effects and bias the comparisons. Given the small number of studies it was not possible to assess the impact of these differences on results.

This study used HRs as the measure of treatment effect. Use of HRs requires an assumption of proportional hazards which may not hold across all studies. Ouwens et al. [35] have described an approach for conducting a network meta-analysis of parametric survival distributions that may address the limitations of conducting the analysis based on HRs. However, this approach requires additional assumptions regarding the similarity of underlying parametric survival distributions across trials that may not hold. Nevertheless, this is a potentially important area for future research.

The approved dosage for ipilimumab in North America and Europe is 3 mg/kg once every three weeks. However, the systematic literature reviews did not identify any controlled trials of ipilimumab monotherapy versus DTIC at this dosage. Accordingly, the HRs for PFS and OS for ipilimumab 10 mg/kg plus DTIC versus DTIC from the CA184-024 trial were used in the evidence network. The use of the results of this analysis to estimate the relative efficacy of ipilimumab monotherapy 3 mg/kg versus other comparators requires the assumptions (1) that there is no difference in treatment effect on PFS and OS for the 10 and 3 mg/kg doses of ipilimumab and (2) that the addition of DTIC to ipilimumab has no impact on treatment effects on PFS or OS. Data on the impact of the addition of DTIC to ipilimumab on outcomes are limited and inconclusive. In the MDX010-08 trial, an open label phase 2 study of ipilimumab 3 mg/kg plus DTIC versus ipilimumab 3 mg/kg in 76 previously treated patients with metastatic melanoma [36], the addition of DTIC to ipilimumab was associated with nominally improved response, PFS, and OS. These results were not conclusive, however, given the small size of the trial. Median PFS was 99 and 85 days in the ipilimumab plus DTIC and ipilimumab groups, respectively. Median OS was 14.3 versus 11.4 months for the ipilimumab plus DTIC and ipilimumab groups, respectively.

Regarding the assumption that the efficacy of ipilimumab 10 mg/kg is equivalent to 3 mg/kg, the CA184-004 trial compared ipilimumab 3 versus 10 mg/kg in a mixed population of 101 melanoma patients with and without prior treatment. There was no evidence of improved PFS or OS with 10 versus 3 mg/kg in this trial [37, 38]. The CA184-022 trial compared 0.3, 3, and 10 mg/kg in 214 previously treated patients with melanoma [39, 40]. This study demonstrated a statistically significant dose–response relationship on best overall response rates. Median OS and 1- and 2-year survival rates were nominally improved with ipilimumab 10 mg/kg compared with 3 and 0.3 mg/kg. These results are suggestive that adding DTIC to ipilimumab may provide some benefit and that 10 mg/kg may improve outcomes versus 3 mg/kg. Accordingly, the use of the CA184-024 trial to estimate the relative effectiveness of ipilimumab 3 mg/kg versus DTIC may bias estimates of the relative effectiveness in favor of ipilimumab and versus BRAF and MEK inhibitors.

Trials of BRAF and MEK inhibitors focused on patients with BRAF mutations, whereas the CA184-024 trial focused on a mix of patients with BRAF mutations and wild-type (WT) disease. Known BRAF mutation status was not required for inclusion in the CA184-024 trial and data on PFS and OS for subgroups of patients defined on BRAF mutation status have not been reported. Evidence suggests that BRAF-mutant melanoma is biologically distinct from BRAF WT disease [41, 42, 43, 44]. Also, numerous studies have shown that BRAF-mutant melanoma is associated with more aggressive disease and worse patient outcomes [41, 42, 43, 44, 45, 46, 47]. Because data on treatment effects for ipilimumab versus DTIC by BRAF mutation status are unavailable, it was not possible to assess potential effect modification from this factor.

To the best of our knowledge, this is the first study to use both univariate and multivariate network meta-analysis to estimate HRs for PFS and OS. Multivariate network meta-analysis of HRs for PFS and OS accounts for the well-established correlation of treatment effects on PFS and OS. In the COMBI-d, COMBI-v, BREAK-3, and BRIM-3 trials, treatment effects on PFS were greater than treatment effects on OS. In the METRIC trial, the estimated treatment effect of trametinib on OS was equal to that on PFS (HR = 0.44 for both). In the CA184-024 trial, the estimated treatment effect of ipilimumab plus DTIC on OS was greater than that on PFS (HR = 0.76 for PFS and HR = 0.72 for OS). As a consequence, the multivariate network meta-analysis attenuated the estimated treatment effects on OS for trametinib and ipilimumab plus DTIC. For example, the HR for OS for trametinib versus DTIC observed in METRIC was 0.44 (95% CI 0.20–1.00), whereas that obtained from the multivariate network meta-analysis was 0.60 (95% CI 0.41–0.83). Similarly, the HR for OS for ipilimumab plus DTIC versus DTIC observed in CA184-024 was 0.72 (95% CI 0.59–0.87), whereas that obtained from the multivariate network meta-analysis was 0.78 (95% CI 0.67–0.90). The results of these analyses may be biased if there are unique attributes of trametinib and ipilimumab that would result in differential relative effects on OS versus PFS compared with the other therapies. For trametinib, we know of no reason to believe that this would be the case; rather, it is more likely a consequence of the small number of subjects in the first-line subgroup of the METRIC trial and the imprecision in the RPSFT method for adjustment for crossover. For ipilimumab, however, it is possible that the immune-related effects may be delayed relative to the cytotoxic effects of BRAF and MEK inhibitors, which would explain a relatively large effect on OS versus that on PFS. The analysis based on the multivariate network meta-analysis may have biased the results in favor of the combination and against ipilimumab plus DTIC therefore should be interpreted cautiously.

In conclusion, this network meta-analysis demonstrates improved PFS and OS with dabrafenib plus trametinib versus dabrafenib, trametinib, vemurafenib, ipilimumab, and DTIC as first-line therapy for patients with BRAF mutation-positive metastatic melanoma. Future research should be conducted which includes other novel treatments, if feasible, and based on network meta-analysis of survival distributions rather than HRs to account for non-proportionality of hazards.