In kidney cancer, the Genitourinary Cancer Symposium unveiled data from the phase III KEYNOTE-426 trial. This study compares axitinib with pembrolizumab to sunitinib in patients with treatment-naïve metastatic renal cell carcinoma (mRCC). We have a bit of a spoiler for the data already - a press release over 3 months ago noted that KEYNOTE-426 met it's the endpoints of both progression-free survival (PFS) and overall survival (OS). The press release overshadowed a near simultaneous presentation of JAVELIN-101, the phase III study comparing axitinib with avelumab to sunitinib. JAVELIN-101 showed an improvement in PFS, but had yet to show an improvement in OS at the time of a preliminary analysis at ESMO 2018.
I thought this blog post could center on how we should interpret the results of KEYNOTE-426 in the context of current studies. A press release alone is insufficient to guide clinical practice - here are the three key questions I'll be asking about the data:
1. What are the hazard ratios for PFS and OS? Is there a good separation between the Kaplan-Meier (KM) curves? These are obvious, but there are some tall trees that axitinib/pembrolizumab is up against. The hazard ratio for OS benefit with nivolumab/ipilimumab versus sunitinib in CheckMate-214 was 0.63 for poor- and intermediate-risk patients, with pretty clear and wide separation of the KM curves.
2. What was the rate of crossover to immunotherapy in the control arm in KEYNOTE-426? An OS survival benefit is great, but if patients on the control arm rarely received immunotherapy, one could make the argument that a sequence of sunitinib followed by immunotherapy (e.g., nivolumab monotherapy) could have produced identical results. In the Immotion151 study, comparing sunitinib to bevacizumab with atezolizumab, only 20% of patients crossed over from sunitinib onto nivolumab. As time goes on, I would anticipate that these numbers would increase. However, there is a huge geographical dependence here - study sites in resource-strapped areas of the world may have limited access to active second-line therapies. This could inflate the OS signal beyond what would be anticipated if study sites had appropriate access to salvage treatment.
3. What proportion of patients stays on the combination? In the context of studies like KEYNOTE-426, I would surmise that there are four biological subtypes of patients. First, there is a subset of patients who need both therapies (e.g., patients with inflamed cancers that also have some degree of VEGF-dependence). There are then likely a second and third group of patients who either have VEGF-dependent or inflamed tumors, who would benefit from either targeted therapy or immunotherapy, respectively. In these patients, combined strategies (e.g., axitinib/pembrolizumab) may be excessive - only one type of therapy may be necessary. Finally, there is a fourth group of patients that represents a huge area of need - those that respond to neither type of treatment. These patients, who demonstrate primary progressive disease, may be candidates for novel strategies such as CAR-T or vaccine therapies currently in development. Coming back to KEYNOTE-426, I'd love to see what the level of attrition is for doublet therapy. If a considerable number of long-term responders ultimately require just one of the two therapies, it suggests that we may have to go back to the drawing board and identify predictors of response to monotherapy.
I'll close by commenting on the unique situation we have in mRCC. Specifically, we now have a total of five pivotal studies (CABOSUN, JAVELIN-101, Immotion-151, CheckMate-214 and [shortly] KEYNOTE-426) which employ a sunitinib control arm with the antiquated dosing regimen of 50 mg 4 weeks on and 2 weeks off. With all the caveats of cross-trial comparisons, we have a unique opportunity to try and pool these datasets and come up with relative estimates of therapeutic efficacy. Not sure if all the key stakeholders would be willing to play ball, but this could offer huge insights for investigators.
Written by: Sumanta Kumar Pal, MD