Clinical Need Sintilimab, a PD-1 inhibitor widely used in China for NSCLC, produces significant survival benefits but also immune-related adverse events (irAEs) in a subset of patients. Identifying at-risk patients before treatment would allow more personalized monitoring and dose adjustment.
The LIPI Index The Lung Immune Prognostic Index (LIPI) is a simple inflammatory score derived from routine blood tests: the derived neutrophil-to-lymphocyte ratio (dNLR) and lactate dehydrogenase (LDH). Both reflect systemic inflammation and immune activation status at baseline.
Dual Prediction Framework This study developed two separate nomograms - one predicting overall survival (OS) and one predicting irAEs - using LIPI scores alongside clinical variables, creating a comprehensive prognostic and safety prediction toolkit for sintilimab-treated patients.
Multicenter Validation The study enrolled 1,197 NSCLC patients from three tertiary centers; after propensity score matching (PSM), 178 training, 89 internal validation, and 89 external validation patients were analyzed with balanced baseline characteristics.
dNLR Calculation The derived neutrophil-to-lymphocyte ratio (dNLR) is calculated as: neutrophil count divided by (white blood cell count minus neutrophil count). A dNLR greater than 3 earns 1 point; dNLR of 3 or less earns 0 points.
LDH Threshold Lactate dehydrogenase (LDH) above 245 IU/L earns 1 point; at or below that threshold earns 0 points. Elevated LDH reflects high tumor burden and increased cellular turnover - both associated with worse immunotherapy outcomes.
Three-Tier Scoring Total LIPI scores of 0, 1, and 2 define good, intermediate, and poor prognosis groups respectively. This three-tier classification is more informative than a binary cutoff, capturing gradient relationships between inflammatory burden and outcome.
Clinical Accessibility Because LIPI requires only a complete blood count and basic metabolic panel - tests available at every clinic - this score is universally applicable without specialized assays, making it practical for routine clinical integration in any healthcare setting.
Independent OS Predictors Multivariate Cox regression identified four independent risk factors for OS: stage IV disease (HR=1.725), second-line or later treatment (HR=1.302), intermediate LIPI score (HR=1.736), poor LIPI score (HR=1.568), and albumin level above 35 g/L (HR=1.802).
AUC Performance The OS nomogram achieved time-dependent AUCs of 0.826/0.770 (1-year/2-year) in training, 0.850/0.834 in internal validation, and 0.837/0.805 in external validation - strong and consistent discrimination across all three cohorts.
Concordance Index C-index values of 0.778 (training), 0.793 (internal), and 0.790 (external) confirmed excellent calibration - the model's predicted survival probabilities closely match observed patient outcomes across all cohorts.
Albumin Level as Predictor Albumin level was the strongest individual predictor (HR=1.802), reflecting that nutritional status and systemic inflammation are powerful modulators of immunotherapy efficacy - a clinically intuitive finding given albumin's role as a negative acute-phase reactant.
Independent irAE Predictors Cox multivariable analysis identified age under 60 (HR=1.821), female sex (HR=2.258), ECOG performance status of 2 or above (HR=1.258), and intermediate/poor LIPI scores (HR=1.822/1.593) as independent predictors of irAE occurrence.
irAE AUC Performance The irAE nomogram achieved time-dependent AUCs of 0.754-0.835 for 1-2 year predictions, with C-indices of 0.805 (training), 0.825 (internal validation), and 0.775 (external validation) - strong performance for a difficult-to-predict outcome.
LIPI as irAE Predictor The finding that intermediate and poor LIPI scores predict higher irAE risk is biologically plausible: elevated baseline inflammatory markers suggest a pre-activated immune state more prone to further dysregulation upon PD-1 blockade.
Female Sex and Young Age The association of female sex and younger age with irAEs is consistent with known immunological sex differences - women mount more robust immune responses but are also more susceptible to autoimmune phenomena, a relevant background for understanding irAE risk.
Why PSM Was Needed Before matching, the training set (n=538), internal validation (n=268), and external validation (n=391) cohorts showed significant baseline differences in age, sex, pathological type, clinical stage, LIPI score, and albumin levels (all p<0.05).
PSM Methodology Propensity score matching was performed at a 2:1:1 ratio using nearest-neighbor matching without replacement, with a caliper width of 0.1, implemented through the pm3 package in R. This reduced each cohort to smaller but balanced samples.
Post-Matching Balance After PSM, all clinical characteristics were well-balanced across the three cohorts (all p>0.05), allowing valid comparisons of nomogram performance that are not confounded by baseline differences.
Reduced Sample Size Trade-Off PSM reduced the total usable sample from 1,197 to 356 patients - a substantial reduction. This trade-off between confounding control and statistical power is a limitation acknowledged by the authors, favoring validation of results in larger unmatched cohorts.
Risk-Stratified Monitoring Patients with poor LIPI scores could be flagged for more intensive irAE monitoring (more frequent clinical visits, laboratory tests, and imaging) and prophylactic corticosteroid protocols, potentially preventing irAE-related treatment discontinuation.
Treatment Line Selection The strong impact of treatment line (first vs. second or later) on OS supports using the OS nomogram to counsel patients about realistic survival expectations when sintilimab is used later in their treatment course.
Extending to Other Immunotherapy Agents While this study focused on sintilimab, the LIPI framework and nomogram methodology are likely applicable to other PD-1/PD-L1 inhibitors (nivolumab, pembrolizumab, atezolizumab) - future validation studies should confirm this generalizability.
Integration with PD-L1 Testing Currently, LIPI provides predictive information independent of PD-L1 TPS or TMB. Future multi-variable models combining LIPI with these molecular biomarkers could further improve OS and irAE prediction, creating comprehensive pre-treatment risk profiles.