Cytoreductive nephrectomy (CN) is performed in select patients with metastatic renal cell carcinoma (mRCC) to reduce tumor burden before or alongside systemic therapy, but identifying which patients will benefit remains challenging.
Systemic inflammatory response (SIR) biomarkers derived from routine blood tests have emerged as candidate prognostic markers because inflammation drives tumor progression and treatment resistance in mRCC.
Three specific SIR biomarkers were evaluated: the albumin-to-globulin ratio (AGR), the de Ritis ratio (DRR), and the systemic immune-inflammation index (SII), all measurable from standard preoperative labs.
This study uses machine learning to identify the optimal SIR biomarker panel for predicting cancer-specific survival (CSS) after cytoreductive nephrectomy in a large multicenter international cohort.
The study analyzed 613 mRCC patients from a multicenter international database who underwent cytoreductive nephrectomy, providing a large and diverse real-world population.
Optimal cutoff values for each SIR biomarker were established: AGR cutoff 1.43, DRR cutoff 1.2, and SII cutoff 710, based on receiver operating characteristic analysis for CSS prediction.
LASSO (Least Absolute Shrinkage and Selection Operator) regression with 10-fold cross-validation was used as the primary variable selection method to identify the most predictive combination of SIR biomarkers and clinical variables.
The cohort was divided into training and testing sets to enable unbiased performance assessment, with multivariable Cox regression used to derive hazard ratios for selected predictors.
Low AGR was associated with worse CSS in both the training cohort (HR 1.40) and the testing cohort (HR 1.78), making it the most consistently predictive individual SIR biomarker.
High SII (HR 1.51) and high DRR (HR 1.41) were associated with worse CSS in the testing cohort, though their associations were not as robust across both cohorts as AGR.
The combined SIR biomarker panel achieved a C-index of 64.4% in the testing cohort, indicating moderate discriminative ability for ranking patients by CSS risk.
A C-index of 64.4% represents a modest improvement over chance but falls short of what would be required for confident individual-level clinical decision-making.
Decision curve analysis showed that the SIR biomarker model provided only marginally improved net benefit over a standard model without the inflammatory markers across most clinically relevant probability thresholds.
This finding challenges the assumption that easily measurable inflammatory markers meaningfully improve prognostication in the specific context of cytoreductive nephrectomy.
The modest C-index and limited net benefit improvement suggest that SIR biomarkers may capture some prognostic signal but are insufficient as standalone risk stratification tools for CN patient selection.
The results highlight the difficulty of improving upon standard clinical models in mRCC, where tumor burden, performance status, and treatment history already account for much of the prognostic variance.
The retrospective and multicenter design introduces heterogeneity in CN timing, systemic therapy protocols, and patient selection practices that may confound the SIR biomarker associations.
The study period spans the pre-immunotherapy era, and SIR biomarker performance in the context of contemporary ICI-based regimens combined with CN requires separate evaluation.
Missing data for some SIR biomarker variables in a subset of patients may have introduced selection bias, as patients with complete labs may systematically differ from those without.
Future studies should evaluate whether SIR biomarkers add value specifically in combination with immunotherapy response prediction models in mRCC patients being considered for CN.
This large multicenter study concludes that while individual SIR biomarkers (AGR, SII, DRR) show statistically significant associations with CSS after CN, their collective clinical utility is limited.
The SIR biomarker panel failed to meaningfully add discriminative or decision-relevant benefit over the standard clinical model, suggesting they should not change CN decision-making in isolation.
These findings do not negate the biological relevance of systemic inflammation in mRCC but underscore the challenge of translating population-level associations into individual patient benefit.
More complex multi-modal approaches incorporating imaging features, genomic data, and immunotherapy biomarkers alongside SIR indices may be needed to achieve clinically meaningful prognostication in mRCC patients considered for CN.