ASCO 2017: The Dynamic Landscape of Renal Cell Carcinoma Biomarkers: Can We Predict Prognosis, Treatment Response, and Outcome?

Chicago, IL ( Dr. Brugarolas provided a succinct summary of the key points, highlights and potential limitations, of the these three excellent ASCO 2017 abstracts (Abstract 5422 DJ George et al, Abstract 5423 MH Voss et al, Abstract 5424 MI Carlo et al). He started with Abstract 5424 by Carlo et al, in which the authors assessed the prevalence of cancer susceptibility germline mutations in patients with advanced renal cell carcinoma (RCC).

Using the MSK IMPACT next-generation sequencing prospective study, they identified 226 patients over a 1 year period. Specifically, Dr. Brugarolas notes that the population has a relative high proportion of young (<45 years old patients), patients with bilateral disease, and non-clear cell histology – which may not represent a usual practice. The authors identified 38 germline mutations in 38 patients, and CHEK2 was the most common (24%). Other common mutations included BAP2, APC, while there was a smaller proportion of DNA damage repair gene mutations. Importantly, germline mutations were not significant predictor of presentation. 

The authors next assessed how many of these mutations would have been missed by ACMG (American College of Medical Genetics) criteria for genetics referral. Current guidelines recommend referral for early onset or aggressive clear cell RCC, any non-clear cell histology, or any clinical history consistent with genetic syndromes. About 62% of clear cell RCC patients with germline mutations would have been missed. 

His main input regarding their conclusions for this abstract are as follows:
1. Their population may not be representative of the general population presenting with advanced RCC
2. He agrees that the ACMG criteria needs to be updated with regards to non-clear cell RCC and inclusion of unclassified RCC.
3. He feels the lack of correlation with clinical criteria is likely due to sample size
4. CHEK2 mutation was associated with an increased risk of RCC and loss of heterozygosity (LOH) in the tumor
5. BAP2 mutations seen in both clear cell and non-clear cell patients – which differs from prior studies. Two of this cohort’s patients had both ccRCC and nccRCC – but only one nccRCC was clearly independent

He then moved on to abstract 5423 by Voss et al. In this nice study, the authors look back at the COMPARZ trial, which compared pazopanib to sunitinib in the setting of advanced or metastatic RCC. They utilized that patient data and available tissue to correlate PBRM1 and BAP1 mutation rates with clinical outcomes, regardless of treatment arm. RNA and DNA data was available for 352 of the patients. Of these patients, 15% had BAP1 mutation and 44% had PBRM1 mutation. Dr. Brugarolas notes that there is some relationship between these genes that may be confounding some of the results from the study; he requested secondary analysis after abstract submission to clarify.

The original study found that PBRM1 mutation was associated with improved PFS and OS, but the groups were unbalanced for BAP1 mutations. A PBRM1 mutant tumor is less likely (OR 0.3) to harbor a BAP1 mutation than wild type. They subsequently noted reduced PFS and OS with BAP1 mutations. 

A prior study in Nature Genetics 2012 (Pena-Lops et al) identified 4 molecular subtypes based on different combinations of BAP1 and PBRM1 mutation, associated with different clinical outcomes. In order to compare, the authors, on Dr. Brugarolas’ request, did the secondary analysis. Similar to the 2012 study, 4 distinct clinical outcomes were noted. PBRM1-BAP1- patients had significantly worse OS and PFS. Subsequent analysis linking angiogenic signature demonstrated that high angiogenic signature was associated with improved PFS and OS. 

His main input regarding their conclusions for this abstract are as follows:
1) Acquired mutations in PBRM1 and BAP1 are common mutations in advanced RCC
2) PBRM1 and BAP1 mutations are not independent and should be considered together
3) Loss of PBRM1 appears to enhance to the proangiogenic microenvironment – he feels this is provocative, but there are significant limitations to the dataset.
4) It may be more interesting to look at treatment response prediction based on PBRM1/BAP1 mutation

Last, he discussed abstract 4522 by George et al, in which the authors looked at the association of HGF levels (during treatment) to clinical response in patients treated in the ALLIANCE CALGB 90206 study. In that study, patients were treated with interferon alpha (IFNA) +/- bevacizumab (BEV) for advanced RCC. They found that baseline HGF levels were predictive of overall survival (OS) – high HGF levels predicting worse OS. Importantly, the absolute HGF level at 4 weeks and a decrease in HGF levels at 4 weeks were prognostic in multivariable analysis. 

Dr. Brugarolas didn’t have much to add to the authors’ conclusion. They extend the work of others by demonstrating that HGF change during therapy has prognostic value. HGF levels that stay low on therapy have the best OS, and may be treated with VEGF-selected therapy. HGF levels that start high and stay high on therapy potentially identify patients that may need HGF-MET axis targeted therapy. 

Presented By: James Brugarolas, MD, PhD, The University of Texas Southwestern Medical Center, Dallas, TX

Written By: Thenappan Chandrasekar, MD, Clinical Fellow, University of Toronto 
Twitter: @tchandra_uromd 

at the 2017 ASCO Annual Meeting - June 2 - 6, 2017 - Chicago, Illinois, USA