Real‑world evidence for first‑line immuno‑oncology combinations in metastatic renal cell carcinoma: an evidence synthesis through May 2026

Introduction
Over the past five to eight years, immune checkpoint inhibitor (ICI)–based combinations have displaced vascular endothelial growth factor tyrosine kinase inhibitor (VEGF‑TKI) monotherapy as first‑line standard of care in metastatic renal cell carcinoma (mRCC). Randomized controlled trials (RCTs) established efficacy; real‑world evidence (RWE) tests external validity across broader, unselected populations, different health systems, and pragmatic care settings, and captures outcomes such as real‑world progression, time to next treatment, and tolerability under routine management. This narrative review synthesizes RWE on first‑line ICI‑based combinations (IO‑IO and IO‑TKI) through May 20, 2026, with global scope and explicit attention to the United States, Europe, and China.

Definitions: IO‑IO refers primarily to nivolumab plus ipilimumab; IO‑TKI refers to combinations such as pembrolizumab plus axitinib, nivolumab plus cabozantinib, pembrolizumab plus lenvatinib, and avelumab plus axitinib.

1) Evidence inventory and study design

The contemporary RWE base comprises:

• Retrospective electronic health record (EHR) cohorts and registries: US community oncology (The US Oncology Network) single‑arm effectiveness studies of first‑line nivolumab–ipilimumab; International Metastatic RCC Database Consortium (IMDC) registry updates; single‑center European series (Royal Marsden) including special populations; and national registry analyses (Norway) that contextualize ICI exposure at the population level 10711125.
• Comparative effectiveness analyses: a multinational chart review directly comparing two IO‑TKIs (pembrolizumab–axitinib vs nivolumab–cabozantinib; ARON‑1) and a US Flatiron Health EHR analysis comparing IO‑IO vs pooled IO‑TKIs using propensity score weighting 14.
• Pragmatic/phase IV prospective real‑world cohorts: the ongoing non‑interventional CaboCombo study of first‑line nivolumab–cabozantinib (design and endpoints reported; results pending) 22.
• Country‑ and regimen‑specific series: Japanese multicenter experience in older adults comparing IO‑IO and IO‑TKI; a Japan multicenter avelumab–axitinib cohort (J‑DART); and a 3‑year follow‑up Japanese nivolumab–ipilimumab cohort 91718.

Network meta‑analyses of RCTs provide indirect regimen‑level context (OS/PFS/response ranking) but are not RWE; they are cited here to contextualize observed real‑world comparative signals 62830.

China: Regulatory adoption of toripalimab plus axitinib as a domestically developed first-line IO-TKI regimen (reported in 2025) signals rapid uptake, but the retrieved materials highlight limited published China‑specific RWE from registries/EHRs to date. A Chinese multicenter real‑world analysis suggested PD‑1 inhibitor plus TKI may outperform TKI monotherapy in metastatic non‑clear cell RCC (mnccRCC), though details are limited in the retrieved summaries 151316.

2) Population and setting

Real‑world cohorts include broader, sicker populations than RCTs:

• US community oncology (first‑line nivolumab–ipilimumab, 2018–2019): median age 63; 39.6% IMDC poor risk; 19.8–24.5% ECOG ≥2; 6.4% with brain metastases—features largely excluded from CheckMate 214 103.
• IMDC registry update (first‑line nivolumab–ipilimumab): n=1,145 across all IMDC risks; 86% clear‑cell histology; 14% sarcomatoid features; median follow‑up ~20 months 11.
• ARON‑1 multinational IO‑TKI comparison: n=760 across 55 centers/16 countries; 78% IMDC intermediate/poor; 85% clear‑cell; nivolumab–cabozantinib recipients were somewhat enriched for poor‑prognosis features (more poor risk; more bone/liver metastases) 1.
• Flatiron Health comparative analysis: 1,438 IMDC intermediate/poor‑risk clear‑cell mRCC patients receiving first‑line IO‑IO or IO‑TKI across US community and academic practices 4.
• Special populations: a UK cohort documented outcomes in rare variant histologies and patients with brain metastases; a Japanese multicenter study focused on older adults (≥75 years) receiving IO‑IO or IO‑TKI; J‑DART enrolled a higher‑risk, older Japanese cohort for avelumab–axitinib 12917.
• Norway population registry (RECON3): population‑wide treatment and survival trends, including the association of ICI exposure in later lines, though first‑line IO‑combination adoption was limited during the observation window 5.

China: The retrieved materials emphasize a maturing IO‑TKI era with regulatory approval, but few PubMed‑indexed, first‑line IO‑combination RWE publications from China were identified; this is an important evidence gap 1513.

3) Comparators and follow‑up

Comparators vary by study:

• Single‑arm effectiveness: first‑line nivolumab–ipilimumab in US practice and in IMDC registry; Royal Marsden single‑center series 1071112.
• Head‑to‑head real‑world comparison of two IO‑TKIs: pembrolizumab–axitinib vs nivolumab–cabozantinib (ARON‑1) 1.
• Propensity score‑weighted comparison: IO‑IO vs pooled IO‑TKIs (pembrolizumab–axitinib, nivolumab–cabozantinib, pembrolizumab–lenvatinib, avelumab–axitinib) in Flatiron Health 4.

Follow‑up and censoring:

• Median follow‑up ranged from ~9.7 months (earlier US Oncology report) to ~22.4 months (extended follow‑up) for US community nivolumab–ipilimumab; ~20 months in the IMDC update; ~46.7 months in the IMDC complete‑response (CR) durability analysis; ~15–18 months in ARON‑1; ~21 months overall (25.9 vs 17.0 months across arms) in Flatiron; ~10.4 months in J‑DART; ~34.4 months in the Japanese 3‑year nivolumab–ipilimumab series; ~20–24 months in the Royal Marsden cohort 71011814171812. Most studies applied standard administrative censoring at last visit/study end; completeness of follow‑up and missingness were variably reported.

4) Key endpoints and effectiveness signals

Overall survival (OS), (real‑world) progression‑free survival [(rw)PFS], time to next treatment (TTNT), time to treatment discontinuation (TTD), and response metrics were emphasized; where specified, real‑world endpoints relied on clinician‑documented progression or treatment patterns.

• US community nivolumab–ipilimumab (2018–2019): median OS 38.4 months (12‑ and 24‑month OS 76.5% and 59.5%); median rwPFS 11.1 months; real‑world response rate (rwRR) 42.9% (mostly partial responses); median time to treatment discontinuation ~5.8 months. Clinically significant treatment‑related adverse events (TRAEs) occurred in 26.2%; any TRAE in 47.6%—lower than in trials, likely reflecting under‑documentation in routine practice 10. An earlier analysis in the same network reported median PFS 17.1 months and 12‑month OS 75.4% with similar ORR (43.2%) over shorter follow‑up 7.
• IMDC registry (first‑line nivolumab–ipilimumab): risk‑stratified median OS 47.8 months (favorable), 51.1 months (intermediate), and 18.3 months (poor); median TTD 6.5, 5.7, and 3.6 months; and median TTNT 24.3, 11.8, and 8.2 months, respectively. Immune‑related AEs were noted in ~48% of evaluable patients. Conditional survival analyses provided pragmatic counseling milestones (e.g., survival probabilities contingent on being alive or remaining on therapy at 6 or 12 months) 11.
• Comparative IO‑TKI vs IO‑IO (Flatiron, propensity‑weighted): IO‑TKIs had significantly longer rwTTNT than nivolumab–ipilimumab (13.1 vs 7.8 months; HR 0.78; p<0.001), but real‑world OS was similar (25.0 vs 28.8 months; HR 1.01; p=0.91). Individual IO‑TKI regimen median rwOS ranged from ~17 months (pembrolizumab–lenvatinib) to ~25 months (avelumab–axitinib and pembrolizumab–axitinib), though heterogeneity and unequal follow‑up limit regimen‑level inferences 4.
• Comparative IO‑TKI vs IO‑TKI (ARON‑1): nivolumab–cabozantinib showed longer PFS than pembrolizumab–axitinib (27.6 vs 16.2 months; p=0.003), with no significant OS difference (OS not reached vs 55.7 months; p=0.51). Overall response rate (ORR) favored nivolumab–cabozantinib (59% vs 49%), with higher complete response in favorable risk (27% vs 7%). The nivolumab–cabozantinib cohort had worse baseline prognostic features; the nonrandomized design and lack of propensity adjustment preclude causal attribution 1.
• Durability of complete response (IMDC, ESMO 2025): CRs were uncommon overall (~3.1%), highest with IO‑IO (~6.3%), then IO‑VEGF (~3.9%) and VEGF monotherapy (~1.5%). Among complete responders, 4‑year OS exceeded 90% across regimens (92.9–100%), and 4‑year TTNT approached 81–86%, with numerically highest TTNT after IO‑IO (85.7% vs 80.9% with IO‑VEGF). Achieving CR confers excellent but not guaranteed long‑term disease control 8.
• Older adults (Japan): In patients ≥75 years, IO‑IO and IO‑TKI produced broadly similar PFS/OS to younger cohorts overall, but within older adults, IO‑TKI was associated with significantly lower ORR (55% vs 81%) and markedly higher discontinuation due to AEs (60% vs 32%) despite similar initial dose and relative dose intensity—suggesting IO‑IO may be better tolerated in this age group. Among younger adults, IO‑TKI had PFS advantage over IO‑IO (HR 2.37; p=0.02) 9.
• Avelumab–axitinib (Japan, J‑DART): ORR 48.8% (CR 7.0%); median PFS 15.3 months; median TTD 15.2 months; frequent dose modifications and interruptions; ~10% required corticosteroids for immune‑related AEs 17.
• Japanese nivolumab–ipilimumab (≥3‑year follow‑up): median OS 49.0 months; median PFS 9.0 months; ORR 48% (CR 14%); grade ≥3 toxicity 45%; 50% treatment interruptions and 45% corticosteroid use—highlighting both durability and real‑world toxicity 18.
• Royal Marsden (UK) single center: nivolumab–ipilimumab achieved median OS 31.9 months and PFS 9.0 months in clear‑cell RCC; in rare variants, OS 25.9 months and ORR 38.2%; in brain metastases, OS 14.5 months and ORR 27.8%. Grade 3–4 immune‑related AEs (irAEs) occurred in 28.4% and were associated with longer OS on landmark analysis; higher steroid doses were also associated with longer OS, consistent with irAEs as a pharmacodynamic correlate of response 12.

Sequencing and post‑progression context: In line with the CONTACT‑03 trial, a US EHR study showed that cabozantinib plus PD‑1/L1 inhibitor after prior ICI performed similarly to cabozantinib monotherapy for rwTTNT and rwOS, arguing against routine ICI “re‑challenge” after first‑line ICI progression 20. Although outside first‑line scope, these data inform expectations for OS in RWE (influenced by subsequent lines) and support the observation that IO‑IO vs IO‑TKI OS differences may be attenuated by downstream therapy 4.

5) Critical appraisal of validity

• Selection bias and confounding by indication: Treatment selection (IO‑IO vs specific IO‑TKI) may reflect unmeasured factors (tumor burden, sarcomatoid features, comorbidities, patient preference), biasing comparative effectiveness. Propensity score methods (Flatiron) balanced measured covariates (standardized mean difference <0.1 after reweighting) but cannot eliminate residual confounding; pooling heterogeneous IO‑TKIs risks ecological bias 4.
• Channeling and era effects: Differential drug availability and timing (e.g., ARON‑1’s 80% pembrolizumab–axitinib vs 20% nivolumab–cabozantinib with shorter follow‑up in the latter) complicate comparisons. Notably, real‑world PFS for nivolumab–cabozantinib in ARON‑1 exceeded trial values, suggesting selection or outcome ascertainment differences 1.
• Immortal time and survivorship biases: Population registries assessing ICI exposure across multiple lines (e.g., Norway) are vulnerable to immortal time bias; sensitivity analyses help but cannot fully remove it 5.
• Endpoint ascertainment: Real‑world PFS and response often rely on clinician assessment and treatment discontinuation patterns rather than adjudicated RECIST, risking misclassification. AE documentation is consistently lower than in RCTs (e.g., any‑grade TRAE ~48% in US practice vs >90% in CheckMate 214), likely reflecting under‑reporting 10.
• Follow‑up maturity and censoring: Shorter, unequal follow‑up across arms in comparative studies (e.g., Flatiron; ARON‑1) undermines OS comparisons, while TTNT/PFS may be more sensitive to regimen‑specific disease control but remain subject to management heterogeneity 41.

Overall, RWE is practice‑informing but not determinative for regimen superiority. The strongest comparative signals—IO‑TKIs yielding longer TTNT/PFS than IO‑IO with similar OS—are consistent across Flatiron and ARON‑1 but should be interpreted with caution given observational design.

6) Generalizability and evidence strength

• United States: Community oncology data confirm that nivolumab–ipilimumab delivers trial‑consistent ORR and (rw)PFS in sicker, more heterogeneous patients, supporting broad external validity. Flatiron analyses, drawing from geographically diverse practices, enhance generalizability but retain unmeasured confounding 104.
• Europe: Single‑center experience (Royal Marsden) demonstrates activity in rare variants and brain metastases; multinational ARON‑1 suggests regimen‑level PFS differences across health systems but is limited by nonrandom allocation 121.
• Asia: Japanese multicenter studies provide age‑stratified insights, suggesting IO‑IO may be more tolerable than IO‑TKI in older adults without compromising outcomes—data of immediate practical relevance in aging populations 91718. The retrieved materials highlight limited China‑specific, PubMed‑indexed RWE for first‑line IO‑combinations to date despite regulatory approval and rapid clinical adoption, underscoring a critical evidence gap 1513.
• Alignment with RCTs: RWE largely confirms trial signals—IO‑TKIs deliver superior progression control relative to dual ICI, while dual ICI achieves the highest probability of deep responses (CR) in some analyses. Network meta‑analyses of RCTs rank nivolumab–cabozantinib highest for OS probability and pembrolizumab–lenvatinib for PFS, with IO‑IO achieving the highest CR probability in intermediate/poor‑risk disease—providing context for the observed real‑world TTNT/PFS advantage of IO‑TKIs and the CR durability observed after IO‑IO 6. These indirect comparisons, however, cannot substitute for head‑to‑head RWE/RCTs.

In sum, the RWE base is supportive and practice‑informing. It extends trial conclusions to underrepresented populations (ECOG ≥2, brain metastases, rare variants, older adults) and helps tailor regimen choice: IO‑TKIs for superior progression control and longer TTNT, IO‑IO for tolerability in selected older adults and potentially greater CR durability in a small subset. Definitive regimen superiority remains unproven in RWE due to residual confounding and follow‑up imbalances.

Summary table: Key RWE on first‑line IO‑based combinations in mRCC

Abbreviations: AE, adverse event; CR, complete response; EHR, electronic health record; IMDC, International Metastatic RCC Database Consortium; int, intermediate; ICI, immune checkpoint inhibitor; IO‑IO, dual immunotherapy; IO‑TKI, immunotherapy plus tyrosine kinase inhibitor; mRCC, metastatic renal cell carcinoma; ORR, objective response rate; OS, overall survival; PFS, progression‑free survival; QoL, quality of life; rwOS, real‑world OS; rwPFS, real‑world PFS; rwTTNT, real‑world time to next therapy; TTD, time to treatment discontinuation.

Final synthesis and clinical takeaways

• Effectiveness in routine practice: Across geographies and settings, nivolumab–ipilimumab delivers trial‑consistent response rates and PFS in broader, sicker populations, with median OS approximating 32–49 months depending on risk and cohort maturity 1071811. IO‑TKIs achieve strong progression control and high response rates in real‑world cohorts (avelumab–axitinib PFS ~15 months; pembrolizumab–axitinib and nivolumab–cabozantinib robust in comparative studies) 1714.
• Comparative signals: The most robust comparative RWE shows IO‑TKIs prolong time to next therapy (and PFS) over IO‑IO, while overall survival is similar—consistent with RCT trends and likely reflecting effective post‑progression options and cross‑over in practice 41. These data are practice‑informing but not causal due to residual confounding and follow‑up imbalances.
• Depth and durability of response: Real‑world CRs remain uncommon but durable across regimens; IO‑IO shows the highest CR proportion and numerically longest TTNT among complete responders, supporting IO‑IO when deep, durable remissions are a priority 8.
• Special populations and tolerability: In older adults (≥75 years), IO‑IO appears more tolerable than IO‑TKI (lower discontinuation due to AEs) without obvious loss of efficacy, informing regimen selection in geriatric practice 9. Nivolumab–ipilimumab demonstrates activity in rare variants and brain metastases, populations underrepresented in trials 12.
• Safety in RWE: Toxicity is substantial and may be under‑documented in EHRs; grade ≥3 immune‑related toxicities are common in some cohorts, and dose modifications/interruptions are frequent with IO‑TKIs—yet do not necessarily compromise outcomes 101817. Associations between higher‑grade irAEs and better survival, observed in single‑center data, suggest irAEs as a pharmacodynamic marker, meriting prospective validation 12.
• China: Clinical adoption is accelerating (toripalimab–axitinib approval), but PubMed‑indexed national RWE remains sparse in the retrieved materials; country‑level registries/EHR studies are needed to quantify effectiveness, safety, and access in Chinese practice. Early real‑world signals in non‑clear‑cell RCC favor PD‑1+TKI over TKI alone but require more granular data and longer follow‑up 151316.

Bottom line: RWE across the US, Europe, and Asia confirms that first‑line IO‑based combinations deliver meaningful real‑world benefit in unselected mRCC populations and broadly corroborate pivotal trial signals. Comparative RWE supports IO‑TKIs when maximizing disease control and time to next therapy is paramount, while IO‑IO remains a strong option—especially for intermediate/poor‑risk patients seeking durable responses and in older adults where tolerability may favor dual ICI. Given observational limitations, regimen choice should be individualized, incorporating patient age, comorbidities, risk group, goals of care (depth vs continuity of control), and local expertise, while the field prioritizes prospective pragmatic registries and geographically diverse RWE to close remaining evidence gaps.