Journal of Cancer Research and Clinical Oncology

, Volume 139, Issue 11, pp 1917–1926

Small molecule targeted therapies for the second-line treatment for metastatic renal cell carcinoma: a systematic review and indirect comparison of safety and efficacy

Authors

    • University of Ioannina
  • Susanne Schmitz
    • Trinity College
  • Reuben J. Broom
    • Auckland City Hospital
Original Paper

DOI: 10.1007/s00432-013-1510-5

Cite this article as:
Dranitsaris, G., Schmitz, S. & Broom, R.J. J Cancer Res Clin Oncol (2013) 139: 1917. doi:10.1007/s00432-013-1510-5

Abstract

Background

Patients with metastatic renal cell carcinoma (mRCC) and a good performance status typically receive an anti-vascular endothelial growth factor receptor (VEGFR) TKI (sunitinib or pazopanib) as initial therapy. Upon disease progression or intolerance, there are four orally administered agents approved in the second-line setting (including cytokine-refractory). However, head-to-head comparative trial data are limited. In this study, an indirect statistical comparison of safety and efficacy was undertaken between axitinib, sorafenib, pazopanib and everolimus in second-line therapy mRCC.

Methods

A systematic review of major databases was conducted from January 2005 to June 2013 for randomized controlled trials (RCTs) evaluating at least one of the four agents in second-line mRCC. Bayesian mixed treatment comparison models were fitted to assess relative effectiveness on multiple endpoints such as objective response rates, dose-limiting grade III/IV toxicities, treatment discontinuations and progression-free survival (PFS).

Results

Four RCTs met the inclusion criteria. All four agents seem able to induce tumor shrinkage and to provide patients with a clinically meaningful PFS benefit. Axitinib was superior to pazopanib [hazard ratio (HR) 0.64; 95 % credible interval (95 % Crl) 0.42–0.96] and sorafenib (HR 0.70; 95 % Crl 0.57–0.87) in terms of PFS. However, axitinib was associated with an elevated risk of fatigue and to a lesser extent stomatitis.

Conclusions

Keeping in mind the caveats associated with cross-trial statistical comparisons, axitinib provides superior PFS relative to pazopanib and sorafenib. Everolimus, an mammalian target of rapamycin inhibitor, is mechanistically distinct from the other agents and remains a useful option for patient’s post-anti-VEGFR TKI failure.

Keywords

SorafenibPazopanibAxitinibEverolimusIndirect comparisonRenal cell carcinoma

Introduction

Kidney cancer is the eighth most common neoplasm affecting men and women with over 15,000 deaths annually in the USA and Canada alone (Surveillance Epidemiology and End Results (SEER) 2013; National Cancer Institute of Canada 2012). Renal cell carcinoma (RCC) is the most common form of kidney cancer, accounting for approximately 90 % of all cases Nelson et al. (2007). Approximately 70–80 % of patients have localized disease when first diagnosed, with surgery being an effective intervention in many cases (Surveillance Epidemiology and End Results 2013; Nelson et al. 2007). For patient with metastatic disease at diagnosis or for those with a metastatic recurrence, clinically effective options were few prior to 2005. Chemotherapy and hormonal therapy are largely ineffective, and high dose interleukin-2 was reserved for good performance status patients with favorable prognostic factors (Motzer et al. 1996; Bukowski 1997). As a result, interferon alfa (IFN) was the most commonly used treatment and became the control arm in randomized trials evaluating the new generation of small molecule targeted therapies (Motzer et al. 2002).

Sunitinib was the first drug to demonstrate a progression-free survival (PFS) benefit over IFN in previously untreated metastatic RCC patients (median 11 vs. 5 months; p < 0.01) (Motzer et al. 2007). However, this drug is associated with clinically significant toxicity such as diarrhea and vascular irritation (Motzer et al. 2007; Grünwald et al. 2010). Sorafenib, an oral multikinase inhibitor, was evaluated against IFN in the first-line setting and against placebo in patients who were previously treated with cytokines (Escudier et al. 2007, 2009a, b). Sorafenib did not demonstrate a significant improvement in PFS relative to IFN in patients who were previously untreated (Escudier et al. 2009). In second-line patients, there was a significant improvement in PFS over placebo (5.5 vs. 2.8 months; p < 0.01), but there was no significant survival benefit (median 17.8 vs. 15.2 months; p = 0.146) (Motzer et al. 2008; Sternberg et al. 2010).

Once sunitinib was approved for the first-line treatment for metastatic RCC, trials evaluating the next generation agents focused primarily on the second-line setting. Everolimus is an oral mammalian target of rapamycin (mTOR) inhibitor that was tested under a randomized phase III setting. Patients who had progressed on sunitinib or sorafenib were randomized in a two-to-one ratio to everolimus or placebo (Motzer et al. 2008). Patients in the active therapy arm had a median PFS of 4.9 months compared to 1.9 months in the placebo control (p < 0.001). However, there was no difference in overall survival (OS) or quality of life between groups. The major grade III/IV side effects of everolimus were anemia (9 %), hyperglycemia (12 %) and stomatitis (3 %) (Motzer et al. 2008).

The fourth small molecule to be approved for the treatment of good performance status RCC patients was pazopanib. Clinical activity in RCC was demonstrated through a randomized, double-blind, placebo-controlled trial in 435 subjects with locally metastatic and/or metastatic RCC who had received no prior therapy or one prior cytokine-based systemic treatment (Sternberg et al. 2010). Patients were stratified by prior cytokine exposure and randomized 2–1 to receive pazopanib 800 mg once daily or placebo. The primary objective was PFS; the secondary endpoints included OS and overall response rate (RR). The median PFS for the overall population was 9.2 months in the pazopanib arm versus 4.2 months in the placebo group (p < 0.001). The median PFS for the treatment-naïve subgroup was 11.1 versus 2.8 months and in the cytokine pretreated subgroup 7.4 versus 4.2 months (p < 0.001). Pazopanib was well tolerated, and the major grade III/IV toxicities were reported at frequencies of <5 % (Sternberg et al. 2010). The main distinguishing adverse event with pazopanib compared to others in class is a higher likelihood of liver function test derangement. OS was not statistically different between the pazopanib arm and the placebo arm [hazard ratio (HR) 0.91, stratified log-rank p value 0.22].

The latest agent to gain regulatory approval worldwide in the second-line setting was axitinib. A randomized trial provided head-to-head evidence comparing axitinib with sorafenib. A total of 723 patients who had previously progressed on first-line therapy consisting of sunitinib, bevacizumab in combination with IFN or cytokines were randomized 1–1 to receive axitinib at a dose of 5 mg twice daily or to receive sorafenib at the clinically approved dose (400 mg twice daily) (Rini et al. 2011). Patients in the axitinib arm had a median PFS of 6.7 months compared to 4.7 months with sorafenib (HR 0.66, p < 0.001). In addition, twice as many patients in the sorafenib arm discontinued therapy because of toxicity (8 vs. 4 %). Patients randomized to axitinib experienced a significant prolongation in time to clinical deterioration compared to the control group (HR 0.83, p = 0.14) (Rini et al. 2011).

There are now five small molecule oral-targeted therapies approved for the first- and second-line treatment for metastatic RCC. Medical decision making is guided by patient and disease-related factors as well as prior therapies. Patient factors that are critical in the selection of optimal therapy include performance status, the presence of comorbidities, prognostic factors and the ability of the patient to tolerate therapy (Escudier et al. 2012a, b). Important disease-related factors that guide therapy include histology and tumor burden (Escudier et al. 2012a, b). Sunitinib and pazopanib are often recommended as first-line therapy in patients with a good performance status (Escudier et al. 2012). However, a decision-making challenge manifests itself in patients who have progressed on sunitinib or cytokines. Practice guidelines state that everolimus and axitinib are reasonable treatments in patients who progressed on sunitinib (Escudier et al. 2012). However, in the patients previously treated with cytokines, axitinib, sorafenib and pazopanib are all recommended therapies (Escudier et al. 2012a, b).

There are no randomized trials comparing the safety and efficacy of everolimus to axitinib in patients who are refractory to sunitinib. In the absence of a randomized trial, statistical methods can be used to indirectly evaluate two or more drugs. The advantage of using indirect statistical techniques to conduct comparative effectiveness evaluations is their utilization of the best available evidence to provide answers to questions that have not been addressed through a randomized trial. In this study, Bayesian mixed treatment comparison (MTC) models were developed to perform an indirect comparison on the safety and efficacy between the second- line small molecule targeted therapies that have been approved for patients with metastatic RCC.

Methods

Literature review and meta-analysis of randomized trials

A computer literature search of PubMed, Embase, the Cochrane Database and Google Scholar was conducted from January 2005 to June 2013 for published randomized trials evaluating sorafenib, axitinib, everolimus and pazopanib for the second-line treatment for metastatic RCC. Search terms consisted of “{advanced RCC} OR {metastatic RCC} OR {sorafenib} OR {axitinib} OR {everolimus} OR {pazopanib} AND {randomized clinical trial} AND {second line}.” Combination therapy trials were not considered because this remains experimental.

Eligibility criteria regarding the validity of trial design and analysis were used to identify potential studies. To be eligible, studies must have used a randomized design with at least 50 patients enrolled into each group. Patients must have been adults 18 years of age or older, diagnosed with metastatic RCC and with the primary population being good to intermediate risk according to the prognostic criteria developed by Motzer et al. (2004). Unpublished randomized trials presented in abstract form at professional meetings were not eligible unless access to full study reports was available. Care was taken to avoid inclusion of duplicate publications.

Once trials meeting the inclusion criteria were identified, the following data were extracted: sample size, year of publication, line of therapy being evaluated, drug regimen, dosage, definition of primary and secondary endpoints, trial duration, median duration of therapy, if patient cross-over was allowed, number of withdrawals due to adverse drug reactions and all relevant clinical and safety outcomes. Once all of the data were collected, it was entered into a database for the subsequent statistical analysis.

Study outcomes for indirect comparison

The primary efficacy outcomes were PFS evaluated by independent assessment and objective tumor RR. Response rate was defined as complete (CR) or partial response based on the Response Evaluation Criteria in Solid Tumors (RECIST) criteria (Therasse et al. 2000). There are a series of drug-induced grade III/IV dose-limiting toxicities (DLT) that are unique to the small molecule targeted therapies and can compromise patient quality of life (Wong et al. 2012). Therefore, the following clinically relevant DLT were evaluated: grade III/IV diarrhea, fatigue, rash, hand–foot skin reaction and stomatitis. To compare the entire toxicity profile of all the second-line agents, the final parameter that was indirectly compared was adverse events leading to the permanent discontinuation of treatment.

Indirect statistical comparisons between drugs

Mixed treatment comparison modeling is a generalization of a meta-analytic technique, which was first introduced by Lu and Ades (2004) and is now a standard tool in comparative effectiveness analysis. The outcome of a Bayesian analysis is a posterior distribution for each parameter of interest, which can be summarized by the mean and a 95 % credible interval (95 % Crl) that captures the uncertainty. Such models allow for the inclusion of all available evidence and simultaneously performs indirect comparisons between treatments, where no direct evidence is available.

A Bayesian MTC model was fitted for each of the efficacy and toxicity outcomes using WinBUGs and R. Relative HR were estimated for PFS assuming an exponential survival model. The model estimates odds ratios (ORs) for tumor response and treatment discontinuation; relative rates are estimated for toxicities. As suggested by Cai et al. (2010), a Poisson likelihood was assumed in the case of toxicities to overcome the issue of small event rates. Figure 1 illustrates the network diagram for second-line small molecules approved for the second-line treatment for metastatic RCC.
https://static-content.springer.com/image/art%3A10.1007%2Fs00432-013-1510-5/MediaObjects/432_2013_1510_Fig1_HTML.gif
Fig. 1

Network diagram for RCC analysis

Results

A total of 548 citations were identified and reviewed. A total of 4 randomized trials meeting the inclusion criteria were appropriate for the statistical pooling exercise. Reasons for study rejection included duplicate publications or review articles, early phase studies and studies evaluating non-clinical outcomes (Fig. 2). Of the four trials, one compared sorafenib to placebo in patients who were treated with cytokines in the first-line setting (Table 1) (Escudier et al. 2007, 2009). The second trial compared everolimus to placebo following exposure to first-line sunitinib, sorafenib or other therapies (Motzer et al. 2008; Lu and Ades 2004). In contrast, the trial by Sternberg et al. stratified patients by prior cytokine exposure (yes vs. no) and randomized them to receive pazopanib or placebo. As a result, only data reported in patients with prior cytokine exposure were used in the indirect comparison of PFS. However, toxicity data from the entire treated population were used in the indirect comparison of safety (Table 2). The final trial identified in the systematic review was a head-to-head comparison of axitinib and sorafenib following treatment with standard first-line interventions (Rini et al. 2011).
https://static-content.springer.com/image/art%3A10.1007%2Fs00432-013-1510-5/MediaObjects/432_2013_1510_Fig2_HTML.gif
Fig. 2

Consort diagram of study selection

Table 1

Randomized trials evaluating oral therapies in second-line metastatic RCC

Citation

Line

Study arms

Num Pat

Good to inter riska (%)

Med age

Resp rate (%)

Med PFS (mo)

Med OS (mo)

Escudier (2007, 2009)

2nd post-cytokines

So versus P

451, 452

100, 99

58, 59

9.8 versus 1.8

5.5 versus 2.8b

17.8 versus 15.2c

Motzer (2008, 2012)

Post SU, So or cytokines

E versus P

272, 138

85, 85

61, 60

1 versus 0

4.9 versus 1.9d

NR versus 8.8e

Sternberg (2010)

1st and 2nd post-cytokines

Pa versus P

290, 145

94, 93

59, 60

30.3 versus 3.4

Overall: 9.2 versus 4.2

Overall: NR

32 versus 4

1st line: 11.1 versus 2.8f

1st line: NR

29 versus 3

2nd line: 7.4 versus 4.2f

2nd line: NR

Rini (2011)

Post SU, B+IFN, or cytokines

A versus So

361, 362

65, 64

61, 61

19 versus 9

6.7 versus 4.7g

NR

PFS progression-free survival, OS overall survival, Su sunitinib, INF interferon alfa, P placebo, So sorafenib, E everolimus, Pa pazopanib, A axitinib, B bevacizumab, T temsirolimus, NR not reached

aAccording to the Memorial Sloan-Kettering Cancer Center risk criteria (Motzer et al. 2004)

bHR 0.44; 95 % CI 0.35–0.55; p < 0.001

cHR 0.88; p = 0.146

dHR 0.30; 95 % CI 0.22–0.40, p < 0.001

eHR 0.83; 95 % CI, 0.50 to 1.37, p = 0.23

fFirst-line HR 0.40, 95 % CI 0.27–0.60, p < 0.001. Second-line HR 0.54, 95 % CI, 0.35–0.84, p < 0.001

gHR 0.66, p < 0.001

Table 2

Dose-limiting grade III/IV toxicities reported in the randomized trials of small molecules in metastatic RCC

Citation

Study arms

Diarrhea (%)

Fatigue (%)

Rash (%)

HFS (%)

Stomatitis (%)

Drug D/C (%)

Escudier (2007, 2009)

So versus P

2, 1

5, 4

1, <1

6, 0

0, 0a

10, 8

Motzer (2008, 2012)

E versus P

1, 0

3, <1

<1, 0

0, 0a

3, 0

10, 4

Sternberg (2010)

Pa versus P

3.8, <1

2.4, 2.8

0, 0a

0, 0a

0, 0a

14, 3

Rini (2011)

A versus So

11, 7

11, 5

<1, 4

5, 16

1, <1

6, 9

P placebo, So sorafenib, E everolimus, Pa pazopanib, A axitinib, HFS hand–foot skin reaction, D/C permanent drug discontinuation, NR not reported

aThese events were not reported in the particular trial. Therefore, it was assumed that they did not occur

Patients enrolled into each trial were comparable with respect to median age, gender and the enrollment of patients with a good performance status (primarily ECOG PS of 0 or 1) (Escudier et al. 2007; Motzer et al. 2008; Sternberg et al. 2010; Rini et al. 2011). However, trial heterogeneity was noted in prognostic factors [as assessed by the Memorial Sloan-Kettering Cancer Center risk criteria (Motzer et al. 2004)] and prior first-line therapies (Table 1). This heterogeneity has to be considered when evaluating the results of the indirect comparison of PFS and to a lesser extent DLT.

Indirect comparison of efficacy and safety between drugs

The first clinical outcome evaluated in the indirect analysis was tumor RR, and it was expressed as an OR. An OR greater than one indicates improved tumor response. A Crl around the point estimate is reported as a measure of uncertainty. A 95 % Crl above 1.0 gives a 95 % probability of improved tumor response. The findings revealed that pazopanib, axitinib and sorafenib were superior to placebo in terms of tumor response (Table 3). In addition, axitinib was also superior to sorafenib. Patients treated with axitinib were twice as likely to achieve a tumor response compared to similar patients receiving sorafenib (OR 2.3; 95 % Crl 1.45–3.73). None of the other interdrug comparisons indicated significant differences between the second-line small molecule targeted therapies (Table 3).
Table 3

Summary of indirect statistical comparisons of clinical outcomes using a Bayesian mixed treatment comparison model

Comparison

Outcome (95 % Crl)

Significant improvementc

OR for tumor responsea

 Everolimus versus placebo

3.12 (0.30–56.0)

Inconclusive

 Everolimus versus axitinib

0.24 (0.02–5.0)

Inconclusive

 Everolimus versus pazopanib

0.28 (0.02–6.8)

Inconclusive

 Everolimus versus sorafenib

0.56 (0.05–11.0)

Inconclusive

 Pazopanib versus placebo

11.3 (3.2–50)

Yes, pazopanib better

 Pazopanib versus axitinib

0.87 (0.18–5.0)

Inconclusive

 Pazopanib versus sorafenib

2.0 (0.46–10.7)

Inconclusive

 Axitinib versus placebo

12.9 (5.5–32.3)

Yes, axitinib better

 Axitinib versus sorafenib

2.3 (1.45–3.73)

Yes, axitinib better

 Sorafenib versus placebo

5.6 (2.76–12.3)

Yes, sorafenib better

HR for PFSb

 Everolimus versus placebo

0.48 (0.36–0.64)

Yes, everolimus better

 Everolimus versus axitinib

1.32 (0.88–2.0)

Inconclusive

 Everolimus versus pazopanib

0.84 (0.56–1.28)

Inconclusive

 Everolimus versus sorafenib

0.93 (0.66–1.32)

Inconclusive

 Pazopanib versus placebo

0.56 (0.42–0.75)

Yes, pazopanib better

 Pazopanib versus axitinib

1.57 (1.05–2.36)

No, axitinib better

 Pazopanib versus sorafenib

1.10 (0.78–1.56)

Inconclusive

 Axitinib versus placebo

0.36 (0.27–0.48)

Yes, axitinib better

 Axitinib versus sorafenib

0.70 (0.57–0.87)

Yes, axitinib better

 Sorafenib versus placebo

0.51 (0.42–0.62)

Yes, sorafenib better

HR hazard ratio, PFS progression-free survival, Crl credible interval, OR odds ratio

aAn OR of greater than 1.0 suggests increased benefit with the former agent relative to the latter. A significant benefit is suggested if the 95 % Crl is above and does not cross 1.0

bA HR of less than 1.0 suggests increased benefit with the former agent relative to the latter. A significant benefit is suggested if the 95 % Crl is below and does not cross 1.0

cAt the 5 % level of significance

The second clinical endpoint evaluated in the network meta-analysis was the hazard ratio (HR) for PFS. A hazard in this case indicates the instantaneous risk of disease progression. A HR of less than one therefore indicates an improved PFS of one agent evaluated against a comparator. A 95 % Crl below one allows for 95 % certainty for an improvement in PFS.

All four small molecules were superior to placebo with respect to PFS (Table 3). The indirect comparison suggested a similar PFS with everolimus, pazopanib and sorafenib (Fig. 3). In contrast, axitinib was associated with a superior PFS when compared to pazopanib and sorafenib (Fig. 3). No statistically significant difference was found between axitinib and everolimus. Given the comparable clinical outcomes with at least three of the four drugs in the second-line setting, medical decision making should be guided by which agents were used in the first-line setting, patient comorbidities and the risk of grade III/IV DLT.
https://static-content.springer.com/image/art%3A10.1007%2Fs00432-013-1510-5/MediaObjects/432_2013_1510_Fig3_HTML.gif
Fig. 3

Forest plot for progression-free survival in the second-line setting: Plot shows hazard ratios (HRs) of second-line treatments compared to placebo and each other. A HR of 1.0 indicates no treatment difference

Everolimus and pazopanib were both associated with a higher risk of treatment discontinuations due to adverse events relative to placebo (Table 4). Indirect estimates indicated that patients being treated with pazopanib or everolimus as an alternative to axitinib have an increased risk of discontinuation caused by adverse events (Fig. 4). The data also suggest a reduced risk of treatment discontinuation with sorafenib compared to pazopanib. However, the OR suggesting a reduced risk of sorafenib compared to everolimus did not reach statistical significance (Fig. 4).
Table 4

Summary of indirect statistical comparisons of toxicity outcomes using a Bayesian mixed treatment comparison model

Comparison

Outcome (95 % Crl)

Significant incremental riskb

OR for drug discontinuationsa

 Everolimus versus placebo

3.29 (1.29 to 10)

Yes, with everolimus

 Everolimus versus axitinib

4.0 (1.2 to 14.5)

Yes, with everolimus

 Everolimus versus pazopanib

0.66 (0.16 to 2.71)

Inconclusive

 Everolimus versus sorafenib

2.57 (0.91 to 8.26)

Inconclusive

 Pazopanib versus placebo

5.0 (2.02 to 14.9)

Yes, with pazopanib

 Pazopanib versus axitinib

6.07 (1.88 to 22.4)

Yes, with pazopanib

 Pazopanib versus sorafenib

3.90 (1.41 to 12.6)

Yes, with pazopanib

 Axitinib versus placebo

0.82 (0.40 to 1.70)

Inconclusive

 Axitinib versus sorafenib

0.64 (0.36 to 1.12)

Inconclusive

 Sorafenib versus placebo

1.28 (0.81 to 2.04)

Inconclusive

RR for grade III/IV toxicitiesa

 Diarrhea

  Everolimus versus placebo

4.1 (0.46 to 66.3)

Inconclusive

  Everolimus versus axitinib

0.91 (0.07 to 18.2)

Inconclusive

  Everolimus versus pazopanib

0.95 (0.05 to 22.4)

Inconclusive

  Everolimus versus sorafenib

1.33 (0.11 to 25.3)

Inconclusive

  Pazopanib versus placebo

4.3 (0.96 to 29.1)

Inconclusive

  Pazopanib versus axitinib

0.95 (0.14 to 8.68)

Inconclusive

  Pazopanib versus sorafenib

1.40 (0.21 to 12.0)

Inconclusive

  Axitinib versus placebo

4.55 (1.49 to 15.3)

Yes, with axitinib

  Axitinib versus sorafenib

1.47 (0.89 to 2.45)

Inconclusive

  Sorafenib versus placebo

3.10 (1.14 to 9.41)

Yes, with sorafenib

 Fatigue

  Everolimus versus placebo

3.18 (0.65 to 22.0)

Inconclusive

  Everolimus versus axitinib

1.03 (0.17 to 8.31)

Inconclusive

  Everolimus versus pazopanib

3.77 (0.49 to 35.6)

Inconclusive

  Everolimus versus sorafenib

2.4 (0.42 to 18.2)

Inconclusive

  Pazopanib versus placebo

0.84 (0.26 to 2.98)

Inconclusive

  Pazopanib versus axitinib

0.27 (0.06 to 1.25)

Inconclusive

  Pazopanib versus sorafenib

0.63 (0.16 to 2.57)

Inconclusive

  Axitinib versus placebo

3.08 (1.33 to 7.31)

Yes, with axitinib

  Axitinib versus sorafenib

2.30 (1.34 to 4.10)

Yes, with axitinib

  Sorafenib versus placebo

1.34 (0.71 to 2.57)

Inconclusive

 Hand–foot skin reaction

  Everolimus versus placebo

0.11 (0.00 to 7.4)

Inconclusive

  Everolimus versus axitinib

0.01 (0.001 to 1.2)

Inconclusive

  Everolimus versus pazopanib

1.04 (0.001 to >100)

Inconclusive

  Everolimus versus sorafenib

0.003 (0.001 to 0.37)

Yes, with sorafenib

  Pazopanib versus placebo

0.11 (0.001 to 7.2)

Inconclusive

  Pazopanib versus axitinib

0.01 (0.001 to 1.16)

Inconclusive

  Pazopanib versus sorafenib

0.003 (0.00 to 0.35)

Yes, with sorafenib

  Axitinib versus placebo

9.01 (1.74 to 70.9)

Yes, with axitinib

  Axitinib versus sorafenib

0.31 (0.18 to 0.52)

Yes, with sorafenib

  Sorafenib versus placebo

29.0 (6.1 to >100)

Yes, with sorafenib

 Rash

  Everolimus versus placebo

2.1 (0.16 to 40.3)

Inconclusive

  Everolimus versus axitinib

7.2 (0.22 to >100)

Inconclusive

  Everolimus versus pazopanib

18.4 (0.13 to >100)

Inconclusive

  Everolimus versus sorafenib

0.70 (0.03 to 21.1)

Inconclusive

  Pazopanib versus placebo

0.11 (0.001 to 6.7)

Inconclusive

  Pazopanib versus axitinib

0.39 (0.001 to 46)

Inconclusive

  Pazopanib versus sorafenib

  

  Axitinib versus placebo

0.29 (0.03 to 2.85)

Inconclusive

  Axitinib versus sorafenib

0.10 (0.02 to 0.41)

Yes, with sorafenib

  Sorafenib versus placebo

2.95 (0.56 to 20.1)

Inconclusive

 Stomatitis

  Everolimus versus placebo

8.8 (1.2 to >100)

Yes, with everolimus

  Everolimus versus axitinib

7.4 (0.09 to >100)

Inconclusive

  Everolimus versus pazopanib

77.8 (0.68 to >100)

Inconclusive

  Everolimus versus sorafenib

60 (0.97 to >100)

Inconclusive

  Pazopanib versus placebo

0.11 (0.001 to 7.4)

Inconclusive

  Pazopanib versus axitinib

0.10 (0.001 to 34.3)

Inconclusive

  Pazopanib versus sorafenib

0.77 (0.001 to >100)

Inconclusive

  Axitinib versus placebo

1.19 (0.02 to 58.2)

Inconclusive

  Axitinib versus sorafenib

8.1 (1.1 to >100)

Yes, with axitinib

  Sorafenib versus placebo

0.15 (0.001 to 4.79)

Inconclusive

OR odds ratio, RR relative risk, Crl credible interval

aA RR or OR of greater than 1.0 suggests increased risk. A significant incremental risk is suggested if the 95 % Crl is above and does not contain 1.0

bAt the 5 % level of significance

https://static-content.springer.com/image/art%3A10.1007%2Fs00432-013-1510-5/MediaObjects/432_2013_1510_Fig4_HTML.gif
Fig. 4

Forest plot for drug discontinuations due to adverse events: Plot shows odds ratios (ORs) of second-line treatments relative to placebo and each other. An OR of 1.0 indicates no difference between treatments

Relevant dose-limiting grade III/IV toxicities from the perspective of the patient are diarrhea, fatigue, hand–foot skin reaction, rash and stomatitis. Point estimates of relative rates indicate an increased risk of grade III/IV diarrhea relative to placebo for axitinib and sorafenib. However, there were no significant incremental risk differences between drugs with respect to diarrhea (Table 4). The indirect analysis also identified some toxicities that are unique to each drug. Fatigue was more prevalent with axitinib. In contrast, hand–foot skin reaction was the most problematic toxicity for sorafenib. Sorafenib was associated with a significant risk in hand–foot skin reaction relative to the three other agents. While no significant differences were found, grade III/IV rash was more prevalent with sorafenib and everolimus (Table 4). The final side effect evaluated was stomatitis. Stomatitis appeared to be a relevant toxicity with everolimus and to a lesser extent axitinib.

Discussion

Sunitinib is a recommended first-line agent for patients of all prognostic groups, but especially in younger patients with aggressive disease but with a good performance status (Escudier et al. 2012a, b). Upon progression, sequential therapy is recommended, particularly in patients with an adequate performance status (Escudier et al. 2012; Calvo et al. 2012). There are currently four small molecule targeted therapies approved for second-line therapy in metastatic RCC. Sorafenib, pazopanib and axitinib target the vascular endothelial growth factor receptors (VEGFR) VEGFR2 and VEGFR3, as well as the platelet-derived growth factor receptor-beta (PDGFRβ) and cKIT, with varying affinity to these targets among the three agents (Roskoski 2007; Sonpavde et al. 2008). In contrast, everolimus is an inhibitor of mTOR, which is a key component of the intracellular signaling pathway regulating cell growth and proliferation, metabolism and angiogenesis (Thomas et al. 2006).

Each of these agents has demonstrated activity in the second-line setting, primarily against placebo. There have been few head-to-head comparisons to establish the optimal treatment paradigm. The only trial comparing two such agents determined that axitinib was associated with a statistically significant improvement in PFS over sorafenib, but the benefit failed to result in a prolongation in OS (HR 0.97, p = 0.37) (Rini et al. 2011; Motzer et al. 2013). The inability to detect a statistically significant survival benefit, as with all of the small molecule targeted agents in RCC, is likely related to trial mandated patient cross-over into the alternative arm upon progression and the impact of subsequent therapies on the duration of survival post-progression (Broglio and Berry 2009).

Given the lack of comparative data from randomized trials, indirect statistical techniques were used to compare the efficacy and safety between everolimus, pazopanib, sorafenib and axitinib in the second-line setting. Even though several groups have published indirect comparisons in this disease site, none of the studies included an assessment of tumor RRs, DLT and permanent treatment discontinuations (Mills et al. 2009; Di Lorenzo et al. 2011; Larkin et al. 2013).

The findings of the current study suggest that axitinib and pazopanib have a higher likelihood of inducing an objective tumor response. Therefore, these two agents should be considered if tumor shrinkage is an immediate and desired clinical objective. Among the four agents, axitinib also appeared to provide the greatest benefit in terms of PFS, although in patients who had received prior anti-VEGFR therapy, the PFS was similar to that for everolimus. Notwithstanding, an incremental PFS benefit has not been correlated with an improvement in OS in metastatic RCC (Escudier et al. 2012). A gain in PFS can therefore not be directly translated into a survival benefit, particularly with the evolving availability of agents in the third-line setting (Motzer et al. 2013).

All metastatic RCC patients will eventually experience disease progression. Since all four agents are active in the second-line setting, drug tolerability should be of primary importance in guiding therapy because it has been reported that approximately one-third of patients are lost with each subsequent line of therapy and only about 50 % of patients receive two or more lines of therapy (Escudier et al. 2012; Levy et al. 2013).

The indirect statistical analysis revealed that the risk of treatment discontinuation was highest with everolimus and pazopanib. In contrast, axitinib and sorafenib had somewhat lower rates of drug toxicities leading to discontinuation. Grade III/IV diarrhea is common with all four agents, with higher rates relative to placebo. Each agent also appears to possess a predominant dose-limiting toxicity. Fatigue was identified as a clinically relevant toxicity with axitinib and everolimus, with little difference in risk between them [RR 1.03 (95 % Crl 0.17–8.33)]. Hand–foot skin reaction and grade III/IV rash were the most problematic events for sorafenib. Prevalence of hand–foot skin reaction was also higher with axitinib, but significantly less than with sorafenib. Risk of grade III/IV stomatitis was elevated with everolimus. Lastly, pazopanib appeared to be well tolerated and having the lowest risk of fatigue, hand–foot skin reaction, rash and stomatitis relative to the other three small molecules. However, it did have the highest rate of drug discontinuation, likely due to elevations in liver function tests.

When choosing between an anti-VEGFR therapy (such as axitinib) and an mTOR inhibitor (such as everolimus) in the second-line setting, taking patient factors into consideration is paramount. Patients at risk for vascular toxicities, such as arteriopathy, may be more safely treated with everolimus, which does not work directly on the vasculature. Conversely, patients with diabetes or lung disease may be more safely treated with an anti-VEGFR agent that does not carry the same risk of hyperglycemia nor pneumonitis. Furthermore, from the pivotal trials, substantially more poor risk patients were included in the everolimus study, so this may be a more appropriate choice for patients with moderately to poorly impaired performance status.

There are several limitations in this study that need to be acknowledged. All meta-analyses are affected by the quality of the studies analyzed. For that reason, we limited our analysis to prospective randomized trials with sufficient sample size and with disease progression being measured by independent review. However, the axitinib trial was not double-blind. Regardless of the data source, our analysis was indirect and does not replace a well-designed non-inferiority trial. There were only four randomized trials suitable for indirect analysis. Hence, it must be acknowledged that this may have limited our statistical power to find differences. Furthermore, certain toxicities were rare, which made an estimation of treatment risk difficult. Poisson models were fitted to overcome this issue; however, uncertainty remained high in some of the comparisons. The absence of a statistically significant difference in the estimates cannot be interpreted as an actual lack of difference between treatments. The analysis was not a non-inferiority study because a prespecified “minimally clinically important difference” in efficacy between two drugs has not been established by regulators. As suggested by the clinical characteristics of patients enrolled into the four studies, heterogeneity between study arms was also evident. OS was not assessed in our network analysis because trial mandated treatment cross-overs contaminated this outcome. This is illustrated by the fact that not none of the four agents were able to demonstrate an OS benefit against the control arm of their respective trials (Escudier et al. 2007; Motzer et al. 2008, 2013; Sternberg et al. 2010). Fortunately, there are ongoing randomized trials focusing on sequential therapy (Escudier et al. 2012). Lastly, the failure to find significant differences in clinical and safety endpoints between two drugs through indirect methods does not confirm comparable safety and efficacy.

Conclusions

The findings of this indirect analysis of prospective randomized trials suggest that everolimus, sorafenib, axitinib and pazopanib are all able to induce tumor shrinkage and provide patients with a clinically meaningful PFS benefit. Keeping in mind caveats associated with cross-trial statistical comparisons, our findings also suggest a superior PFS benefit associated with axitinib relative to pazopanib and sorafenib. However, this comes at cost of a higher risk of fatigue and to a lesser extent stomatitis. Given its distinct mechanism of action and differing toxicity profile, everolimus is an effective option after an anti-VEGFR progression.

Conflict of interest

None.

Copyright information

© Springer-Verlag Berlin Heidelberg 2013