Renal cell carcinoma results in an estimated 54,000 new cases and 13,000 deaths each year in the United States 1. Approximately 30% of patients present with metastatic disease (mRCC), often associated with poor prognosis and subsequent 5-year survival rate 23 Until recently, treatment was limited to cytokine therapies (interleukin-2 [IL-2] and interferon-alfa [IFN-α ]) that produce modest response rates (< 20%) and substantial toxicities, although occasional complete responses have been reported 4. Findings from randomized trials have shown that treatment with cytokine therapy results in an overall median survival of 13 months (range 6–28) 5.

As a greater understanding of the molecular mechanisms involved in the pathogenesis of metastatic renal cell carcinoma (mRCC) develops, more advanced treatment options are emerging 46. Clear-cell carcinoma histology in particular, is associated with the von Hippel-Lindau (VHL) tumour suppressor gene. Loss of this gene function results in an over expression of several hypoxia-responsive proteins, including the vascular endothelial growth factor (VEGF) and platelet derived growth factor (PDGF), capable of promoting tumour growth and angiogenesis 7. Newer antiangiogenesis-targeting agents are targeting these factors and have recently provided more promising treatment options. New VEGF targeted regimens (bevacizumab, sorafenib, and sunitinib) and temsirolimus (CCI-779), an inhibitor of mammalian target of rapamycin (mTOR) kinase, in particular, have now been evaluated, compared to interferon as a first-line treatment6, although they have not been directly evaluated in head-to-head trials. Sunitinib and sorafenib have shown positive PFS endpoints when compared to IFN-α or placebo, in phase III trials 89. Similarly, phase III trials of bevacizumab plus interferon have shown superiority in PFS, response, and tolerability compared with interferon alone 4. Temsirolimus has demonstrated positive PFS in patients with a poor prognosis 10.

In a commentary in The Lancet, Motzer and Basch (2007) highlight that important information on the clinical setting, nature of the treatment program, and survival outcomes is necessary to inform the management of patients as well as future trial designs 6. Other considerations such as cost, patient experience, and adverse events also have an important place in the decision-making process and have complicated the delivery of these new interventional drugs in some settings 11. Thus, to determine optimal treatment in a rapidly advancing era of targeted therapy 6, we conducted a meta-analyses of all randomized controlled trials examining either bevacizumab, sorafenib, sunitinib or temsirolimus for the treatment of mRCC. In the absence of head-to-head evaluations, we applied an adjusted indirect comparison analysis 12.

Our primary literature search identified 150 full-length articles. There was excellent agreement between reviewers on narrowing this to 16 full-text articles review (φ = 0.87). Of these 11, were excluded as they were not original studies, were not randomized controlled trials, did not examine at least one of the selected drug therapies (i.e., bevacizumab, sunitinib) or in one case, only examined outcomes in treatment responders. The remaining 5 reports were included. While numerous abstracts from relevant conferences and annual meetings were reviewed, only 2 were included 1819 as they reported original studies. One of these 18 was subsequently published in full during the review stage of this manuscript and checked for accuracy 20 Additionally, 9 abstracts examined reported on the preliminary findings of 5 included studies 212223242526272829 and were subsequently excluded. Figure 1 displays the flow-diagram of the 7 included studies totalling 3,957 patients.

In total, 2 full-length studies 430 and 1 abstract 18 examined bevacizumab; 1 full-length study 31 and 1 abstract 19 examined sorafenib; 1 full-length report 8 presented follow-up data examining sunitinib for the treatment of mRCC; and 1full-length report examined temsirolimus 10. Interferon-α was the most common comparator in 5 included studies 41819810 while placebos were used in 2 studies 239 Study and patient population characteristics for each included trial are displayed in Additional File 1 [See Additional File 1].

While all included studies were randomized controlled trials, among the 5 full-length articles, only 1 detailed the sequence generation process 4 Only this same study adequately described allocation concealment. Descriptions of who was blinded were available in 3 studies 4831. All 5 full-length reports employed an intent-to-treat design in their statistical analyses 48303110. Finally, all 5 full length reports described adequate safety monitoring, with all five employing the use of a data and safety monitoring board (DSMB) 48303110. Four trials employed their stopping rules 893019. Treatment outcomes by study are displayed in Additional File 2 [See Additional File 2].

Our study should be of interest to patients and clinicians alike. Until only recently, mRCC patients had few treatment options beyond cytokine therapies. Our study demonstrates consistent PFS benefits with the new targeted therapies. Insights into the histology of RCC appear to now result in considerably improved PFS and response rates. Our study represents a first step at determining the relative effectiveness of each new intervention.

We sought out all randomized controlled trials of targeted drugs for mRCC. In the end, five papers and two abstracts met our inclusion criteria. Analysis of the populations from those studies revealed there to be three distinct groups of patients: treatment naïve patients of favorable to intermediate risk, refractory patients of favorable to intermediate risk, and refractory patients of intermediate to poor risk. As we hoped to measure the overall treatment effect of targeted therapy, we included both the treatment naïve and refractory patients of favorable to intermediate risk. While the refractory group were likely to be enrolled later after their time of diagnosis, and hence at risk for shorter PFS by definition, we felt that PFS was still a reasonable outcome measure based on their favorable (and similar) prognostic risk. In a pooled analysis this would likely bias the results in a negative direction that would make a statistically positive result more likely to be true. For patients with a poorer prognosis, PFS (our a priori primary endpoint) is likely shorter whether or not previous treatment has been tried. Interestingly, as can be seen in Additional Files 1 and 2 [See additional Files 1 and 2], median duration of treatment and hazard ratios for PFS and investigator assessed response rate were comparable between the poor-risk patients and favorable-risk ones.

Establishing prognosis of mRCC is difficult as there are a variety of patient, tumour, and biochemical factors that have been shown to influence prognosis 333435. Some of these include, among other variables, prior nephrectomy, tissue pathology, tumour stage (nodes and number of metastatic sites), serum calcium, alkaline phosphatase, and hemoglobin values, and scores of performance such as Eastern Co-operative Oncology Group (ECOG) performance status or Karnofsky score. We accepted individual study authors' assessment of prognosis into favorable-risk, intermediate, and poor-risk patients. While some criteria varied between papers, measures of ECOG or Karnofsky scores were also used by authors and we accepted risk stratification based upon them. From Additional Files 1 and 2 [See Additional Files 1 and 2], however, the PFS, and clinician determined response to therapy was similar between the other groups included in our statistical models. While they are not perfectly matched groups from a prognostic point of view, we felt they are similar enough for inclusion into our analysis. Further limitations of included trials based on more specific prognostic factors would have made any analysis impossible given the variety of different prognostic scores available for metastatic renal cell carcinoma.

There are several important strengths and limitations to consider when interpreting our analysis. Strengths include our extensive searching and successful contact with authors for clarifications and in one case, access to individual patient data. Our data analysis approach used a strategy that permits strong inferences about the relative effectiveness of interventions in the absence of head-to-head direct evidence. While head-to-head trials remain the highest level of evidence of therapeutic effectiveness, the method we employed has demonstrated comparatively consistent findings between this indirect method and head-to-head trials 3637. Limitations of our review include the absence of head-to-head trials, the potential that other trials exist that we were unable to identify, and the small number of included studies, thus negating the possibility of exploring possible sub-group analyses and exploring heterogeneity among study populations. Indeed, such analyses can only be undertaken when there are a large number of included studies.

Our method of analysis employed the adjusted indirect comparisons meta-analysis method. This method may be unfamiliar to some readers. Strengths of this method are that it borrows strength from each trial and maintains the benefits of randomization. Clinical, as well as statistical, rationale is required for combining treatments. Head-to-head (direct comparisons) trials remain the highest level of evidence of therapeutic effectiveness and in our review the only head-to-head trials compared these new therapies versus cytokine therapy, IFN-α. While our comparative data is hypothesis generating, it highlights the need for appropriate new clinical trials that examine the direct relative effectiveness of these interventions and also examine the combined effectiveness of these interventions.

There has been a recent public outcry over access to these mRCC interventions in the UK. The UK National Institute for Clinical Excellence (NICE) has issued provisional guidance not to recommend any of the drugs on the National Health Service due to their cost 11. Our review finds that the clinical benefits of these interventions are clear. However, NICE evaluated the cost of the drugs per quality adjusted life year (QALY) and found all four treatments exceeding the Institute's £30,000 threshold (bevacizumab £171,301, sorafenib tosylate £102,498, sunitinib £71,462, temsirolimus £94,385). This troubling finding indicates two important public health challenges: 1) drugs need to be available at reduced costs; and 2) thresholds for cost-effectiveness need to be adjusted according to relative clinical efficacy 38. In May 2009, NICE will issue final guidance that will take into account new supplementary criteria to be applied in drugs used at 'End of Life', and because of this it is expected that one or more of the reviewed drugs will be recommended. It is difficult to envision our own clinical decision-making, if cost alone were the deciding factor.

The findings of our study raise intriguing questions about the future of treatment options for mRCC. Sunitinib, bevacizumab, and temsirolimus all offer improved PFS compared to IFN-α. The current availability of these active agents allows new treatment options and strategies to extend PFS, but in some settings are too expensive for public availability. The optimal treatment strategy in mRCC is likely to be defined in future randomized trials that should examine combination therapies of these active agents.

Over the past 5 years Edward Mills has consulted to Pfizer Ltd., Mac Reviews Inc., World Health Organization, GSK, and Boehringher Ingelheim: all on methodological issues. Chris O'Regan has been employed by Pfizer Ltd. Lehana Thabane consults with GSK on statistical issues. Beth Rachlis has no conflict of interest to declare. Dan Perri has no conflict of interest to declare.

EM, CO, BR, and DP conceived the study. EM, CO, BR, LT, and DP acquired the data. EM, CO, BR, LT, and DP analyzed the data. EM, CO, BR, LT, and DP wrote and approved the final manuscript.