Oncology · Medical Literature
Bayesian Analysis to Refine East Asian Subgroup Estimates in the CLEAR Trial (Lenvatinib Plus Pembrolizumab vs. Sunitinib) for Advanced Renal Cell Carcinoma.
Abstract
This study aimed to refine estimates of lenvatinib plus pembrolizumab (Len/Pembro) efficacy in East Asian patients with advanced renal cell carcinoma (aRCC) in the CLEAR trial using Bayesian borrowing from Western Europe and North America (WENA), and to assess safety while preserving region-specific toxicity. Published hazard ratios (HRs) and 95% confidence intervals (CIs) for progression-free survival (PFS) and overall survival (OS) in WENA and East Asian subgroups were reanalyzed using Bayesian normal models (priors: neutral to strong borrowing). Posterior HRs, 95% credible intervals (CrIs), and benefit probabilities were derived. Adverse events (AEs) in East Asian (n=75) versus non-East Asian (n=277) patients were assessed with independent Beta-Binomial models with Jeffreys priors to estimate posterior incidences and probabilities of excess toxicity. The observed East Asian PFS HR was 0.38 (95% CI=0.23-0.62), similar to the WENA HR of 0.49 (95% CI=0.37-0.64). Bayesian borrowing yielded posterior PFS HRs of 0.40-0.46 with narrower CrIs and >99% probability of benefit. For OS, East Asian data alone were imprecise (HR= 0.71, 95% CI=0.30-1.71), but borrowing from WENA (HR=0.78, 95% CI=0.57-1.05) gave posterior HRs 0.75-0.77; under strong borrowing 0.77 (95% CrI=0.58-1.03) with 96% probability. East Asians showed excess toxicity for any-grade and grade ≥3 proteinuria and hand-foot syndrome, with signals for hypothyroidism, dysphonia, and stomatitis. Bayesian information borrowing demonstrates a clear progression-free survival benefit and supports a credible overall survival advantage for first-line lenvatinib plus pembrolizumab in East Asian patients with advanced renal cell carcinoma. Independent safety modeling identified a distinct regional pattern of higher VEGFR-related toxicities, emphasizing the need for proactive toxicity management in this population.
Clinical relevance
Can Bayesian borrowing from the Western Europe and North America (WENA) subgroup of the CLEAR trial yield more precise estimates of lenvatinib plus pembrolizumab efficacy and region-specific toxicity for East Asian patients with advanced renal cell carcinoma? __SNAPSHOT__ {"studyDesign":"Post-hoc Bayesian reanalysis of published subgroup data from a phase III randomized controlled trial (CLEAR; NCT02811861); not a new prospective study","setting":"Multicenter, international; reanalysis of CLEAR trial subgroup data (East Asia [Japan and Republic of Korea] vs. Western Europe and North America); specific institution types not reported in abstract","sampleSize":"East Asian subgroup: n=75 (lenvatinib plus pembrolizumab arm); non-East Asian comparator for safety modeling: n=277; WENA subgroup size for efficacy borrowing: not reported in abstract","population":"Treatment-naïve adults with advanced clear-cell renal cell carcinoma (aRCC); East Asian patients (Japan and Republic of Korea) enrolled in the CLEAR trial","intervention":"Lenvatinib 20 mg orally once daily plus pembrolizumab 200 mg IV every 3 weeks (Len/Pembro)","comparator":"Sunitinib 50 mg orally once daily (4 weeks on/2 weeks off); WENA subgroup used as external prior for Bayesian efficacy borrowing","primaryOutcome":"Posterior progression-free survival (PFS) hazard ratio (HR) and overall survival (OS) HR derived from Bayesian normal models with neutral-to-strong borrowing priors; posterior adverse event (AE) incidence probabilities derived from independent Beta-Binomial models with Jeffreys priors","followUp":"Inherited from the CLEAR trial follow-up; specific data cutoff used for this reanalysis not reported in abstract","fundingSource":"Not reported in abstract"}
Key findings
For progression-free survival (PFS), the observed East Asian subgroup HR was 0.38 (95% CI: 0.23–0.62), directionally consistent with the WENA HR of 0.49 (95% CI: 0.37–0.64). Bayesian borrowing from the WENA subgroup yielded posterior PFS HRs of 0.40–0.46 with narrower credible intervals (CrIs) and a posterior probability of PFS benefit exceeding 99%. For overall survival (OS), the East Asian subgroup alone produced an imprecise point estimate (HR: 0.71; 95% CI: 0.30–1.71), with the wide CI spanning 1.0 and precluding definitive conclusions. Bayesian borrowing from the WENA OS data (HR: 0.78; 95% CI reported as truncated in the submitted abstract—full CI not available from provided text) refined this estimate, though the exact posterior OS HR and its CrI are not fully reported in the abstract provided. NNT cannot be calculated from available data (absolute event rates by arm not reported). For safety, East Asian patients (n=75) were modeled separately from non-East Asian patients (n=277) using independent Beta-Binomial models with Jeffreys priors to assess posterior AE incidences and probabilities of excess toxicity; specific posterior AE estimates and excess toxicity probabilities are truncated in the provided abstract and thus not fully reportable here. Prior publications of the CLEAR East Asian subgroup (Rha et al., Int J Cancer 2023) document that hand-foot syndrome occurred in 66.7% of East Asian patients on Len/Pembro versus 28.7% in the global population, and dose reductions due to TKI-associated adverse events were more frequent in East Asian patients than in the global population—context consistent with the safety modeling rationale of this Bayesian reanalysis.
Practice implications
This Bayesian reanalysis established that the directional PFS benefit of lenvatinib plus pembrolizumab over sunitinib observed in East Asian CLEAR participants (HR 0.38) is numerically at least as large as in the WENA subgroup (HR 0.49), and that Bayesian borrowing substantially narrows the credible intervals around the East Asian PFS estimate, yielding a posterior probability of PFS benefit >99%. This supports the applicability of the Len/Pembro combination as first-line therapy in East Asian patients with aRCC. However, OS benefit in East Asian patients alone was statistically highly imprecise (HR 0.71; 95% CI 0.30–1.71), and the degree to which Bayesian borrowing from WENA OS data meaningfully resolves this uncertainty cannot be fully assessed from the truncated abstract. The safety modeling approach, preserving region-specific toxicity by using independent Beta-Binomial models rather than borrowing AE rates across regions, is methodologically appropriate given the known higher incidence of TKI-related toxicities (particularly hand-foot syndrome and proteinuria) in Asian populations. Importantly, this is a post-hoc statistical reanalysis of published summary-level data—not individual patient data—and generates no new clinical outcomes. Findings apply to treatment-naïve East Asian patients with advanced clear-cell RCC. Non-East Asian patients outside WENA, patients with non-clear-cell histology, patients with prior systemic therapy, and those with CNS metastases were not included in the source trial. Only the abstract was available for this review; full model specifications, prior sensitivity analyses, complete posterior AE estimates, and OS credible intervals were not assessable. __LIMITATIONS__ 1. Post-hoc reanalysis of published summary-level subgroup data, not individual patient data (IPD); this limits adjustment for confounders and restricts granularity of the Bayesian models. 2. The East Asian subgroup (n=75 in the Len/Pembro arm) is small, making the original frequentist OS estimate highly imprecise (HR 0.71; 95% CI 0.30–1.71); Bayesian borrowing formally improves precision but introduces assumptions about exchangeability between East Asian and WENA populations that may not fully hold. 3. The abstract is truncated; complete OS posterior CrIs, full AE posterior incidences, and the sensitivity of results to prior strength (neutral vs. strong borrowing) cannot be fully evaluated. 4. The CLEAR trial was open-label, introducing performance and ascertainment bias risk. 5. Subgroup analyses were not pre-specified primary endpoints of the CLEAR trial; Bayesian reanalysis of post-hoc subgroup data compounds multiplicity concerns. 6. Subsequent anticancer therapy use differed substantially between arms in the East Asian subset (PD-1/PD-L1 inhibitor use: 9.3% in Len/Pembro arm vs. 43.1% in sunitinib arm), which confounds the OS comparison in ways not specifically addressed in this reanalysis. 7. Funding source is not reported, which limits assessment of potential conflicts of interest. 8. Generalizability beyond Japan and the Republic of Korea (the East Asian subgroup composition) to other East Asian populations is uncertain. __BOTTOM_LINE_CATEGORY__ confirmatory __BOTTOM_LINE__ Bayesian borrowing from the WENA subgroup of the CLEAR trial substantially narrows the credible intervals around the East Asian PFS estimate for lenvatinib plus pembrolizumab versus sunitinib, yielding a posterior probability of PFS benefit exceeding 99% and supporting the use of this regimen in treatment-naïve East Asian patients with advanced clear-cell RCC. OS benefit in East Asian patients alone remains imprecisely estimated, and clinicians should be aware of the higher TKI-related toxicity burden—particularly hand-foot syndrome—in this population, which warrants proactive dose modification planning. This analysis is confirmatory and precision-enhancing in nature; it does not alter current practice recommendations but strengthens the statistical confidence in subgroup-specific efficacy estimates.