Combining VEGFR tyrosine kinase inhibitors and PD-1/PD-L1 inhibitors versus VEGFR tyrosine kinase inhibitors monotherapy in renal cell carcinoma: a target trial emulation

Technical Summary: Real-World Effectiveness of VEGFR-TKI and PD-1/PD-L1 Combinations in Renal Cell Carcinoma

Problem Statement
While pivotal randomized clinical trials (RCTs) such as KEYNOTE-426, JAVELIN Renal 101, CheckMate 9ER, and CLEAR have established the efficacy of combining VEGFR tyrosine kinase inhibitors (TKIs) with PD-1/PD-L1 immune checkpoint inhibitors (ICIs) as a first-line standard of care for renal cell carcinoma (RCC), these trials excluded patients with common comorbidities and strict performance status requirements. Consequently, the real-world generalizability of these survival benefits remains uncertain. Traditional observational studies often suffer from selection bias, confounding by indication, and immortal time bias, which can obscure causal inference. This study aims to determine whether adding PD-1/PD-L1 inhibitors to VEGFR-TKI therapy improves overall survival (OS) in a real-world RCC cohort, utilizing a rigorous analytic framework to emulate a target trial.

Methodology
The study employed a Target Trial Emulation (TTE) framework using retrospective electronic health record (EHR) data from the University of Florida Health Integrated Data Repository (IDR), covering the period from September 2009 to June 2023.

• Cohort Construction: From an initial pool of 107,783 patients, 387 adult patients (≥18 years) with confirmed RCC and at least one VEGFR-TKI prescription were identified. Patients were categorized into an experimental group (EXP: VEGFR-TKI + PD-1/PD-L1 inhibitor, n=123) and a control group (CTRL: VEGFR-TKI monotherapy, n=264).
• Exposure Definition: The index date was the first VEGFR-TKI prescription. The EXP group included patients receiving any combination of the specified drug classes; the CTRL group included those with no documented PD-1/PD-L1 exposure.
• Bias Reduction: To emulate randomization, the study utilized variable-ratio propensity score matching (PSM) (up to 2:1) across 13 baseline covariates, including demographics (age, sex, race) and nine comorbidities derived from the NCI Comorbidity Index (e.g., CKD, liver disease, myocardial infarction). A caliper of 0.2 standard deviations was applied.
• Statistical Analysis:
  • Survival Estimation: Weighted Kaplan-Meier (KM) curves were generated for the matched cohort (n=319).
  • Hazard Modeling: A cluster-robust Cox proportional hazards model was used to estimate hazard ratios (HR), accounting for within-subclass correlation.
  • Non-Proportional Hazards Handling: Recognizing that the proportional hazards (PH) assumption was violated (Schoenfeld residuals test, P = 9.7 × 10⁻⁴), the primary effect measure was defined as the Restricted Mean Survival Time (RMST) at a truncation time ($\tau$) of 24 months. This approach provides a model-free, clinically interpretable estimate of absolute survival gain without assuming constant hazards.

Key Results

• Covariate Balance: Prior to matching, significant imbalances existed between groups, particularly in chronic kidney disease (SMD=0.316), liver disease (SMD=0.227), and myocardial infarction (SMD=0.226). Post-matching, all 13 covariates achieved balance with standardized mean differences (SMD) < 0.10.
• Survival Outcomes:
  • 12-Month OS: The combination arm showed a significantly higher survival rate (81.8%; 95% CI, 74.7–89.6%) compared to monotherapy (68.1%; 95% CI, 61.1–76.0%), an absolute difference of 13.7 percentage points.
  • 24-Month OS: The survival advantage narrowed but remained numerically higher for the combination arm (61.1% vs. 56.7%).
  • Hazard Ratio: The Cox HR was 0.718 (95% CI, 0.484–1.064; P = 0.0986), which did not reach statistical significance, likely due to the violation of the PH assumption and the time-varying nature of the treatment effect.
  • RMST Analysis: The RMST difference at 24 months was 2.79 months (95% CI, 0.93–4.65; P = 0.0033) in favor of the combination therapy. The RMST ratio was 1.17 (P = 0.0037), indicating a 17% increase in mean survival time within the 24-month window.
• Pattern of Effect: The survival curves exhibited early separation followed by gradual convergence, consistent with the rapid anti-angiogenic effects of TKIs followed by the delayed, durable response of immunotherapy.

Key Contributions

1. Real-World Generalizability: This study provides evidence that the survival benefits of VEGFR-TKI and PD-1/PD-L1 combinations observed in controlled RCTs are maintained in a real-world population with higher comorbidity burdens and heterogeneity.
2. Methodological Rigor: By applying a TTE framework with variable-ratio PSM and cluster-robust variance estimation, the study mitigates common biases in observational data.
3. Appropriate Effect Measures: The paper demonstrates the critical importance of using RMST over Cox HR when proportional hazards are violated. While the HR was non-significant, the RMST revealed a statistically significant and clinically meaningful absolute survival gain, offering a more accurate summary of the treatment effect in this context.
4. Robustness to Comorbidities: The results suggest that the combination therapy remains effective even in patients with significant comorbidities (e.g., CKD, liver disease) who might have been excluded from pivotal trials, countering potential "healthy user" bias.

Significance and Claims
The authors conclude that adding PD-1/PD-L1 inhibitors to VEGFR-TKI therapy is associated with a statistically significant and clinically meaningful gain in restricted mean overall survival (2.79 months over 24 months) in a real-world RCC cohort. The study supports the generalizability of combination therapy regimens beyond the strict eligibility criteria of pivotal trials. Furthermore, it underscores the necessity of selecting appropriate statistical measures (such as RMST) when evaluating ICI-based therapies, where non-proportional hazards are common. The findings validate the use of target trial emulation to generate high-quality real-world evidence that complements randomized trial data, particularly for treatment decisions in complex, comorbid patient populations.