A network meta‐analysis of efficacy and safety of first‐line and second‐line therapies for the management of metastatic renal cell carcinoma

Metastatic renal cell carcinoma (mRCC) is the most common type of kidney cancers. Disease‐specific survival for mRCC has been significantly improved with the introduction of new targeted agents since 2005. However, there is a lack of head‐to‐head clinical trials comparing the efficacy between therapies. This study compared indirectly progression‐free survival (PFS) and overall survival (OS) among first‐line and second‐line therapies in patients with mRCC using network meta‐analysis (NMA).


| WHAT IS K NOWN AND OBJEC TIVE
In the United States, kidney cancer is the eighth and tenth most prevalent form of cancer in men and women, respectively. An estimated 73 820 incident cases and 14 770 deaths from kidney cancer were reported in 2019. 1 Renal cell carcinoma (RCC) is the most common type of kidney cancer and accounts for approximately 90% of the diagnosed kidney cancer cases. 2 There are several types of RCC, including clear cell RCC, papillary RCC, chromophobe RCC, collecting duct carcinoma and other rare types of RCC. Of the different types of RCC, clear cell RCC is the most common accounting for a prevalence of approximately 70%. 2,3 The major treatment options for RCC include surgery, radiation therapy, immunotherapy, targeted therapy and chemotherapy.
Approximately 70%-80% of patients with RCC have localized disease at the time of diagnosis, and surgery (eg nephrectomy) is an effective intervention and a potential cure for those with early-stage RCC. 4,5 However, nephrectomy alone is not sufficient for patients experiencing recurrence after nephrectomy or having metastatic disease at the time of diagnosis, and few options were available for those patients before 2005. 4,6 Traditional treatments, including radiation therapy, chemotherapy, and hormone therapy, are less used in patient with metastatic RCC (mRCC), whereas the most commonly used treatments have been targeted therapies or immunotherapy. 6,7 Prior to 2005, immunotherapy was a dominant treatment option in patients with mRCC. Starting in the late 1980s, cytokines were considered the main treatment options for mRCC. 8 The cytokines used most often to treat mRCC were interleukin-2 (IL-2) and interferon α (IFN-α). 9 Since 2005, the use of cytokines have gradually declined due to the advent of the newer targeted treatment agents for the treatment of mRCC, which significantly improved disease-specific survival for mRCC. 10 By June 2020, the following targeted agents have been approved for the treatment of mRCC by the Food and Drug Administration (FDA): sorafenib (Nexavar ® ), sunitinib (Sutent ® ), temsirolimus (Torisel ® ), everolimus (Afinitor ® ), bevacizumab (Avastin ® ), pazopanib (Votrient ® ), axitinib (Inlyta ® ), cabozantinib (Cabometyx ™ ) and lenvatinib (Lenvima ® ). Tivozanib (Fotivda ® ) has been approved by the European Medicines Agency (EMA). These agents can be classified into following groups by their mechanism: small-molecule multiple tyrosine kinases inhibitor (TKI) including sorafenib, sunitinib, pazopanib, axitinib, cabozantinib and lenvatinib work by directly inhibiting multiple receptors such as the vascular endothelial growth factor receptor (VEGFR), platelet-derived growth factor receptor (PDGFR) or Fms-like tyrosine kinase receptor-3 and thus blocking angiogenesis, and tivozanib is the VEGFR TKI, which inhibits VEGFR; temsirolimus and everolimus work by inhibiting mammalian target of rapamycin (mTOR); and bevacizumab works as monoclonal antibodies that bind directly with vascular endothelial growth factor (VEGF) and prevent the engagement with its receptor. 11,12 In addition, the FDA has approved a biologic agent, nivolumab (Opdivo ® ), which is a PD-1 (programmed cell death 1) immune checkpoint inhibitor, for the treatment of mRCC in 2015. Ipilimumab (Yervoy ® ), which is a monoclonal antibody that targets cytotoxic T-lymphocyte antigen-4, was approved as the combination therapy with nivolumab in April 2018. 13 Additional PD-1 immune checkpoint inhibitor such as pembrolizumab (Keytruda ® ) and avelumab (Bavencio ® ) were approved with axitinib in combination therapies in 2019. 14,15 Appendix S1 summarizes the targeted and biologic agents approved for the treatment of mRCC.
Several randomized controlled trials (RCTs) have demonstrated the efficacy and safety of each of these targeted therapies in patients with mRCC, but head-to-head comparisons are lacking. Previous indirect network meta-analysis (NMA) studies have compared the efficacy and safety of some targeted therapies for the treatment of mRCC to address this gap in RCTs. [16][17][18] However, there are no prospective clinical trials directly comparing the efficacy and safety AE (HR = 0.54, 95% CrI = 0.40-0.71; HR = 0.98, 95% CrI = 0.42-1.97, respectively).
What is new and conclusion: With respect to PFS and OS improvement, cabozantinib, avelumab + axitinib and pembrolizumab + axitinib are likely to be the preferred options for the first-line therapy and cabozantinib and axitinib for the second-line therapy in the management of mRCC. Regarding safety, avelumab + axitinib and temsirolimus were considered preferred treatment options in first-line and second-line therapies. More future research is needed to establish subgroup analyses, allowing evaluation of the impact of some of the differences in patient characteristics, including treatment effect modifiers.

K E Y W O R D S
network meta-analysis, overall survival, progression-free survival, renal cell carcinoma, safety, targeted therapy of all possible options of targeted and biologic agents by including nivolumab, cabozantinib, ipilimumab, pembrolizumab and avelumab.
Furthermore, there is no evidence in the literature regarding the efficacy and safety in both the first-line treatment and the second-line treatment for mRCC within a study. Therefore, the primary objective of this study was to compare overall survival (OS) and progression-free survival (PFS) of targeted and biologic agents approved by the FDA or EMA assessing both the first-line and the second-line treatments for mRCC. In addition, a secondary objective was to compare discontinuation rates due to toxicity in patients with mRCC.

| Study selection and evaluation
The study selection process was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guideline. Two independent reviewers (JH and CP) performed a first-stage screening of the titles and the abstracts identified in the initial search against the inclusion and exclusion criteria. Based on this initial screening, selected full-text articles were obtained and reviewed in the second stage of screening.
Studies that met the following inclusion criteria and clear description of the patients' characteristics were included in the NMA:  In order to assess whether it was feasible to perform a valid NMA to compare interventions of interest, a feasibility assessment was conducted before performing the NMA. The NMA feasibility algorithm was adopted from Cope et al (2014). 19 The first step included network identification to ensure treatments of interest are connected. Second, an assessment of similarity/transitivity was performed to identify differences in treatment definitions, outcome characteristics, and study and patient characteristics. Third, a heterogeneity of observed treatment effects was assessed using the consistency test. The consistency for closed loop was tested if there was a discrepancy between direct and indirect comparisons. 20 The risk of bias in each included trial was assessed using the Cochrane risk-of-bias assessment tool to determine the quality of studies by two independent reviewers (SP and JH). Seven specific domains were assessed: selection bias including random sequence generation and allocation concealment, blinding of participants and personnel, blinding of outcome assessment, completeness of outcome data, selective outcome reporting and other sources of bias. 21 Each study was assigned a 'low, 'high' or 'unclear/unknown' risk of bias for each item.

| Statistical analysis
Network meta-analysis is a methodology for synthesizing evidence to inform relative treatment effects from indirect and direct comparisons. In contrast to conventional meta-analyses, NMA allows simultaneous comparisons and pooling of multiple treatments. 22 This study utilized the generalized linear modelling framework using a Bayesian Markov Chain Monte Carlo (MCMC) approach to conduct the NMA. Results of the NMA represented the posterior distributions of the model parameters, including point estimates of HR and 95% credible intervals (CrI), indicating the range of true underlying effects with 95% probability. 23 In addition, the surface under the cumulative ranking curve (SUCRA), and rank probability for all treatments, and the probability that one treatment is better than a specific comparator were also calculated. SUCRA is a proportion, interpreted as the probability that the intervention in an outcome would be ranked first, which equals 100% when the treatment is certain to be the best and 0% when it would be the worst. 24 Model specification recommended by NICE was employed for each end point. 25 For the PFS and OS, because the different follow-up time by trials may cause bias when estimates were calculated, HR and its SE was used instead of PFS or OS time. We assumed a normal distribution for the continuous relative measure (ie HR) of treatment effect of arm k relative to arm 1 in trial i, y ik , with variance The case where the y ik are log-hazard ratios was applied, with variance based on the normal theory approximation.
For the safety end points of the discontinuation due to AE, HR was generated adjusting for follow-up times. 25 The different length of follow-up time in each trial was adjusted by the underlying as- where r ik is the number of events in arm k of trial i, with follow-up time f i , the treatment effects δ ik is the log-hazard ratios form of study-specific treatment effects, θ ik is linear predictor of complementary log-log (cloglog) form of probability of events, and μ ik is trial-specific baselines.
This model allowed to generate the probability of a discontinuation due to AE within follow-up time. The pooled relative estimates were presented as HR for each pairwise comparisons.
The sampling was performed using OpenBUGS version 3.2.3 (Free Software Foundation, Inc., Boston, MA, USA), which is a standalone software application for the Bayesian analysis using the MCMC method. 26 Non-informative prior distributions were used. Point estimates of the HR for each pair of treatments and 95% CrIs were calculated from 200 000 simulated draws from the posterior distribution after a burn-in of 100 000 iterations with three chains. The code for OpenBUGS is shown in Appendix S2. StataIC 14 (StataCorp LLC, College Station, TX, USA) was used to create network plots for each endpoint and meta-analysis for treatment effects in a pairwise comparison (metan command) to assess the heterogeneity of treatment effects. 20 To check the assumption of consistency, the ifplot command was used to identify all closed loops in the network to estimate the respective inconsistency factors and their uncertainties. This provided the inconsistency plot, presenting the estimated inconsistency factor that is truncated at 0. 27

| First-line therapy
For the first-line therapy, most interventions were statistically superior to placebo in increasing PFS of patients with mRCC.
Cabozantinib had 91% of SUCRA and the probability of being first rank was 0.42, which was the highest among all treatments available followed by avelumab + axitinib and pembrolizumab + axitinib (HR = 0.27, SUCRA = 90%) ( Table 2). For OS, pembrolizumab + axitinib had a high likelihood of being the preferred treatment for OS     (14) 229 (61)

TA B L E 1 Continued
In the second-line therapy, the advantage of cabozantinib and axitinib was particularly superior in improving OS and PFS, respectively. Especially, axitinib which also showed lowest risk of discon-

| WHAT IS NE W AND CON CLUS I ON
This study provides the updated insight stakeholders on the efficacy and safety of a range of approved treatments for RCC. The results of NMA showed that cabozantinib and combination therapy of avelumab + axitinib, pembrolizumab + axitinib and/or nivolumab + ipilimumab are likely to be the preferred options for the first-line therapy based on PFS and OS, whereas the regimen of cabozantinib and axitinib was superior to other treatments for the second-line therapy.
Regarding safety, avelumab + axitinib, temsirolimus and axitinib were considered preferred treatment options in the first-line and second-line therapies.

CO N FLI C T O F I NTE R E S T
All authors have no conflicts of interest to disclose.

AUTH O R S ' CO NTR I B UTI O N S
JHH: developed the study design, conducted statistical analyses, drafted the initial manuscript and approved the final manuscript as