Targeted therapies for previously treated advanced or metastatic renal cell carcinoma: systematic review and network meta-analysis

Objective To compare the effectiveness and safety of treatments for advanced or metastatic renal cell carcinoma (amRCC) after treatment with vascular endothelial growth factor (VEGF)-targeted treatment. Design Systematic review and network meta-analysis of randomised controlled trials (RCTs) and comparative observational studies. MEDLINE, EMBASE and Cochrane Library were searched up to January 2018. Participants People with amRCC requiring treatment after VEGF-targeted treatment. Interventions Axitinib, cabozantinib, everolimus, lenvatinib with everolimus, nivolumab, sorafenib and best supportive care (BSC). Outcomes Primary outcomes were overall survival (OS) and progression-free survival (PFS); secondary outcomes were objective response rate (ORR), adverse events, and health-related quality of life (HRQoL). Results Twelve studies were included (n=5144): five RCTs and seven observational studies. Lenvatinib with everolimus significantly increased OS and PFS over everolimus (HR 0.61, 95% Credible Interval [95%CrI]: 0.36 to 0.96 and 0.47, 95%CrI: 0.26 to 0.77, respectively) as did cabozantinib (HR 0.66, 95%CrI: 0.53 to 0.82 and 0.51, 95%CrI: 0.41 to 0.63, respectively). This remained the case when observational evidence was included. Nivolumab also significantly improved OS versus everolimus (HR 0.74, 95%CrI: 0.57 to 0.93). OS sensitivity analysis, including observational studies, indicates everolimus being more effective than axitinib and sorafenib. However, inconsistency was identified in the OS sensitivity analysis. PFS sensitivity analysis suggests axitinib is more effective than everolimus, which may be more effective than sorafenib. The results for ORR supported the OS and PFS analyses. Nivolumab is associated with fewer grade 3 or grade 4 adverse events than lenvatinib with everolimus or cabozantinib. HRQoL could not be analysed due to differences in tools used. Conclusions Lenvatinib with everolimus, cabozantinib and nivolumab are effective in prolonging the survival for people with amRCC subsequent to VEGF-targeted treatment, but there is considerable uncertainty about how they compare to each other and how much better they are than axitinib and sorafenib. PROSPERO registration number CRD42017071540.

6. The conclusion of the abstract should also be softened based on the limitations. 7. It is also important to note different mechanisms of action of the therapies evaluated, because this may also direct treatment decisions. Grouping all the anti-VEGF agents as TKIs, does not distinguish potential other targets, especially with cabozantinib. (bkgnd section) 8. Similarly, Nivolumab should not be characterized as simply a monoclonal antibody (bkgnd line 41) but should be characterized by its target, PD-1, and therefore its effect is through inhibition of the target. 9. The point of 7 and 8 is that these differences may also lead to differences in patient selection for these treatments based on biomarkers.

REVIEWER
Daniel Spratt University of Michigan USA REVIEW RETURNED 04-Sep-2018

GENERAL COMMENTS
This is a very well performed network meta-analysis investigating optimal targeted therapies in RCC treated patients. The bayesian approach allows for probabilities to be generated which are highly informative and a welcomed method to inform what an optimal treatment may be without conducting a dedicated RCT (for which there are too many combination now to conduct RCTs for every comparison). This is a rare time for which I do not have any substantive comments and I believe the authors have performed a rigorous study, quantified limitations and potential biases, and stated appropriate conclusions supported by the data.

GENERAL COMMENTS
Thank you for the opportunity to review this interesting paper that provides a comprehensive review on an important health related issue. However, there are some concerns that should be addressed in order to clarify the aims and methods of this paper.

INTRODUCTION
First and second paragraphs should be joined and organized from general information to more specific one.
The fifth paragraph should be limited to the information in other research papers, all the information regarding the treatment options should be stated in the previous paragraph.
The objective in the last paragraph does not match with the one in objective section and with the first paragraph in the discussion section. Please, unify in order to provide a clear aim to the reader.
There are complete paragraphs without references. This should me amended.

METHODS/RESULTS
There are secondary outcomes included from which there is no information in the introduction section. If these outcomes are important to understand the picture, then they and their role in the research should be stated in the introduction section.
Search strategy is not update. Please, update it, nine months are too much for a review.
The used of statistical methods should be referenced. (i.e., random or fixed effects models; heterogeneity interpretation,…) Each supplementary table should be referenced in the appropriated place of the test.
Non-RCT are defined as: An experimental study in which people are allocated to different interventions using methods that are not random. It should be clarified why cohort studies and restrospective chart reviews are included. Additionally, I doubt on how including these designs could fit in a network meta-analysis.

DISCUSSION
The authors' study definition change across the paper, please unify it (network, comprehensive, …) Second and third paragraph are related to strengths and limitations, these should be stated at the end of the discussion section.
I miss more discussion on the results across the discussion section. Please, discuss all the outcomes of interest comparing them with previous findings.

VERSION 1 -AUTHOR RESPONSE
Reviewer 1: Janice Dutcher 1. No response required 2. We appreciate the reviewer's suggestion regarding type and amount of prior therapy, which was a subgroup analysis in the original health technology assessment. Too few studies were available to conduct reliable subgroup analyses, so they were not included in the protocol for this update. Inclusion criteria for type and amount of prior therapies in each included study is included in Table 1, and we have added the point to the discussion. 3. We have amended the opening of the discussion and abstract conclusions to better account for the uncertainties and limitations listed. 4-6. The first two paragraphs of the discussion have been reorganised and reworded to explain the uncertainty in the treatment hierarchy, which has also been amended in the abstract conclusions. 7-9. A sentence has been added to the background about targets and mechanism of action.
Reviewer 2: Daniel Spratt No changes suggested Reviewer 3: Celia Álvarez Bueno The first two paragraphs have been merged and reorganised as requested.
We appreciate the reviewer's suggestion to changes of the fifth paragraph, but note that the treatment options are stated in the previous paragraph and that paragraph five is focused on other research papers available. The wording of objectives has been unified in the background, objectives and discussion. We have added additional references to the background. The final sentence of the introduction has been amended to add in the relevance of all outcomes. We appreciate the reviewer's comment about the date of the search but updating at this late stage would further delay the results being published. We are aware of emerging evidence in the field and are relatively certain that any studies published since the last search would be observational, so only the sensitivity analyses would be affected. Statistical methods, including those requested by the reviewer (random or fixed effects models; heterogeneity interpretation) are described in the Data synthesis section. Table numbers from the supplementary file have been added to the text. We accept the possible confusion with the definition of non-RCT, which we intend to mean comparative evidence not from RCTs. We have amended the wording from 'non-RCTs' to observational studies throughout the manuscript. Only comparative observational studies were eligible, meaning the between-group data could be included in NMA in the same way as for the RCTs.
We have unified the way the design is described to systematic review and network meta-analysis, in line with the title. The discussion was written in accordance with the structure suggested by BMJ Open so we have not reorganised as suggested by the reviewer.
We have added a paragraph to compare results for secondary outcomes with previous reviews.

GENERAL COMMENTS
1. I am still concerned regarding the clinical applicability of these data, based on how the original studies were designed, and thus the data provided for a meta-analysis. we have subsequently learned much about RCC heterogeneity, and patients who particularly would respond to MTOR inhibitors and Anti-PD1 inhibitors -and these are subgroups, that might not be appropriate for secondary anti-VEGF. Additionally, the design of many of the "registration" trials for new anti-VEGF TKIs purposely chose a weak, or obsolete comparator, thus biasing the result.
4.Again, patient and clinical characteristics may inform the type of treatment to utilize -based again on the heterogeneity that we are discovering.
5. Discussion -I would suggest NOT providing a hierarchy of treatment choice -again, based on issues discussed in #4, not would I do so for toxicity -again, this depends. Some of lack of tox of Nivo is lack of response, yet if get immune toxicity -may predict for response 7. Again, would not make strong recommendations based on comment above, but also -only 1 Nivo trial, indirect comparisons, and the Levatinib/Ev trial is phase II.
8. Pg 25 -should not recommend sorafenib -old and not as good as any of the other anti-VEGF drugs. just cheap.

REVIEWER
Celia Álvarez-Bueno Universidad de Castilla-La Mancha Health and Social Research Center, Spain REVIEW RETURNED 29-Nov-2018

GENERAL COMMENTS
The reviewer completed the checklist but made no further comments.

VERSION 2 -AUTHOR RESPONSE
1. I am still concerned regarding the clinical applicability of these data, based on how the original studies were designed, and thus the data provided for a meta-analysis. We have subsequently learned much about RCC heterogeneity, and patients who particularly would respond to MTOR inhibitors and Anti-PD1 inhibitors -and these are subgroups, that might not be appropriate for secondary anti-VEGF. Additionally, the design of many of the "registration" trials for new anti-VEGF TKIs purposely chose a weak, or obsolete comparator, thus biasing the result. Response: The systematic review was conducted in a methodologically robust manner, where the issued raised by the reviewer were assessed and discussed, where necessary. The trials used in the network for the network meta-analysis were considered sufficiently robust with any potential biases explicitly stated (e.g. in the subsequent analysis including observational studies). In terms of the comparators within the trials, some have limited use in clinical practice. However, this will not have an effect on the results of the network meta-analysis. It is highlighted in the review that sequencing of treatments may be very important but that there aren't data available at the moment to explore this. No change made.