Biomarker analyses from the phase III randomized CLEAR trial: lenvatinib plus pembrolizumab versus sunitinib in advanced renal cell carcinoma

Abstract

Background: In CLEAR, lenvatinib + pembrolizumab (L + P) significantly improved efficacy versus sunitinib in first-line treatment of patients with advanced renal cell carcinoma (aRCC). We report results from CLEAR biomarker analyses.

Patients and methods: Programmed death-ligand 1 (PD-L1) immunohistochemistry (IHC) and next-generation sequencing assays (whole exome sequencing/RNA sequencing) were carried out on archival tumor specimens. For IHC-derived/RNA sequencing analyses, a continuous analysis was carried out adjusting by Karnofsky performance status (KPS) score for: PD-L1 combined positive score (CPS) versus best overall response (BOR)/progression-free survival (PFS); and each gene signature score [T-cell inflamed gene expression profile (Tcell_infGEP)/non-Tcell_infGEP signatures including proliferation and angiogenesis] versus BOR/PFS. Association between mutation status of RCC driver genes and PFS were analyzed for genes for which ≥20 patients per arm had oncogenic alterations. Association of molecular subtypes with outcome was evaluated with baseline KPS adjustments. The set of biomarkers evaluated and statistical significance criteria for PD-L1 CPS, gene signature scores, and molecular subtypes were prespecified.

Results: Within-arm analyses using continuous values showed no association between PD-L1 levels and BOR/PFS for either treatment. PFS hazard ratios between arms were similar regardless of the mutant or wild-type subgroups of RCC driver genes (VHL, PBRM1, SETD2, BAP1, KDM5C). No associations between PFS and gene signature scores were observed for L + P. With sunitinib, high proliferation and MYC signature scores showed shorter PFS; high angiogenesis and microvessel density signature scores showed longer PFS. Six new molecular subtypes were defined. Tumors of patients with favorable/intermediate risk were enriched in angiogenesis and angiogenesis/stromal clusters; those with poor risk were enriched in proliferative and unclassified (low-Tcell_infGEP/low-angiogenesis/low-proliferation) clusters. No association between molecular subtypes and PFS for L + P/sunitinib was observed (after adjustment for KPS and gene signatures that were individually associated with PFS).

Conclusions: Improvements in objective response rate and PFS for L + P versus sunitinib in aRCC were observed consistently across a range of biomarker subgroups defined using RCC driver mutations, PD-L1, gene expression signatures, and molecular subtypes.

Keywords: CLEAR; RCC; biomarkers; lenvatinib; pembrolizumab.

Figure 1.. Baseline mutation rates and ORR by mutational status of RCC driver genes.

A) Mutational landscape in the lenvatinib plus pembrolizumab and sunitinib arms; B) ORR in mutant and wildtype subgroups for select mutations (≥20 mutants per arm). Oncogenic mutations were annotated by OncoKB platform. Please see further details regarding this analysis under “Methods” of the Supplementary Appendix. CI, confidence interval; KPS, Karnofsky performance status; L+P, lenvatinib plus pembrolizumab; Mut, mutant; ORR, objective response rate; RCC; renal cell carcinoma; S, sunitinib; WT, wildtype.

Figure 2.. Between-arm PFS analysis by mutational status of RCC driver genes.

A) Kaplan-Meier curves of PFS between arms for mutant and wildtype subgroups of RCC driver genes, and B) further PFS analysis of RCC driver gene mutation status between arms. *Interpretation is limited by small sample size. For Figure 2B, evaluations were by Cox regression model (adjusted by KPS) between arms in Mut and WT subgroups, separately. CI, confidence interval; HR, hazard ratio; KPS, Karnofsky performance status; L+P, lenvatinib plus pembrolizumab; Mut, mutant; PFS, progression-free survival; RCC, renal cell carcinoma; S, sunitinib; Strat, stratification; WT, wildtype.

Figure 3.. Analyses of gene signature scores.

A) Within-arm association analysis of continuous gene signature scores by BOR, and B) between-arm association analysis of gene signature scores by PFS for signature-high and -low subgroups separately; C) Kaplan-Meier curves of PFS between arms for signature-high and -low subgroups separately. Cutoff values (1st tertile of Tcell_infGEP, median of non-Tcell_infGEP signatures) were determined based on the combined three arms. High/non-low: >= cutoff; low: <cutoff. For figure 3A, P-values in the boxplots were for within-arm testing and were calculated using logistic regression adjusted by KPS, one-sided and nominal for Tcell_infGEP in the L+P arm, two-sided and nominal for Tcell_infGEP in the S arm, and two-sided and BH FDR adjusted for non-Tcell_infGEP signatures. Angio36, angiogenic & growth factor pathway; BOR, best overall response; BH, Benjamini Hochberg; CI, confidence interval; EMT, epithelial-mesenchymal transition; FDR, false discovery rate; HR, hazard ratio; KPS, Karnofsky performance status; L+P, lenvatinib plus pembrolizumab; m/gMDSC, monocytic/granulocytic myeloid-derived suppressor cells; MVD, microvessel density; PFS, progression-free survival; S, sunitinib; Tcell_infGEP, T-cell inflamed gene expression profile.

Figure 4.. Distribution and analyses of molecular subtypes.

A) Distribution of molecular subtypes across MSKCC risk categories and PD-L1 subgroups (CPS≥1 as PD-L1+ and CPS<1 as PD-L1-). Analysis of B) ORR and C) PFS by molecular subtypes in the lenvatinib plus pembrolizumab and sunitinib arms. The “unclassified” cluster was defined as low- Tcell_infGEP/low-angiogenesis/low-proliferation (Supplementary Figure S2). *Associations between molecular subtypes and PFS were tested by a 5-degree of freedom likelihood ratio test (while adjusting for KPS). An association between PFS and the S arm (P=0.0009) was further evaluated by adding signatures (associated significantly with KPS-adjusted PFS; including proliferation, MYC, angiogenesis and microvessel density) as covariates to test the independent value of the molecular subtypes. Molecular subtypes were associated with PFS in the S arm but not after adjustment for KPS and gene signatures that were shown to individually associate with PFS in the S arm (Pr[>Chisq]=0.1721). CPS, combined positive score; IHC, immunohistochemistry; KPS, Karnofsky performance status; L+P, lenvatinib plus pembrolizumab; MSKCC, Memorial Sloan Kettering Cancer Center; NE, not estimable; ORR, objective response rate; PD-L1, programmed death ligand 1; PFS, progression-free survival; S, sunitinib; Tcell_infGEP, T-cell inflamed gene expression profile.