3.1 Characteristics of the subjects

The study enrolled 219 NSCLC patients receiving immunotherapy, comprising 191 males (87.21%) and 28 females (12.79%), and a median age of 63 years (IQR: 57–69 years). History of smoking was noted in 60.27% of these patients. Tumor staging included 65 patients (29.68%) at stage III and 154 patients (70.32%) at stage IV. Pathologically, there were 91 cases (41.55%) of squamous cell carcinoma and 128 cases of non-squamous non-small-cell lung cancer (non-sq-NSCLC), which comprised 109 cases (49.77%) of adenocarcinoma and 19 cases (8.68%) of other types of NSCLC. Regarding treatment, 14 patients received monotherapy with ICIs, while 205 patients underwent combination therapy. Regarding comorbidities, chronic obstructive pulmonary disease (COPD) was the most frequent, affecting 47 patients (21.46%), followed by diabetes and hypertension, each affecting 26 patients (11.87%), and coronary heart disease (CAD) in 14 patients (6.39%). There were 12 cases (5.48%) with EGFR mutations, 3 cases (1.37%) with either EML4-ALK (2 cases) or ROS-1 fusions (1 case), and 21 other mutations (9.59%) which included 6 cases of KRAS G12 C, 11 HER2 mutations, 2 MET exon 14 skipping mutations, and one each of BRAF V600 mutation and RET fusion. A total of 172 patients (78.54%) received first-line immunotherapy, while the rest were treated with second-line or more advanced immunotherapy. Within this cohort, before starting immunotherapy, 11 patients (5.02%) had received targeted therapies (EGFR-TKI or ALK-TKI), 17 patients (7.76%) had received chemotherapy alone, and 19 patients (8.68%) had been treated with chemotherapy combined with anti-angiogenic therapy. According to the NRS2002 scores before treatment, patients were divided into high nutritional risk (NRS2002 ≥ 3) and low nutritional risk (NRS2002 < 3) groups. Using X-tile for cutoff value calculation, the optimal cutoff values were determined to be 45.5 for PNI, 40.0 g/L for ALB, and 4.92 for NLR, as shown in Fig. 1. A comparative analysis of general clinical features between the two groups revealed statistical differences in age, PS, COPD, hypertension, target therapy, and line of ICIs treatment, optimal efficacy, BMI, PNI, ALB, and NLR (P < 0.05), Table 1.

Fig. 1

Determining the optimal cutoff values for PNI, ALB, and NLR using X-tile. A Prognostic Nutritional Index (PNI) determination; B Albumin (ALB) determination; C Neutrophil-to-Lymphocyte Ratio (NLR) determination

Table 1 Clinical characteristics of NSCLC patients by different NRS2002 Score Groups3.2 Survival and prognostic analysis after ICIs therapy

As of May 31, 2024, the median follow-up duration for the cohort was 29 months (IQR: 25.96–32.04). The cohort experienced 140 cases of disease progression and 93 all-cause deaths. Best response assessment across all patients showed 1 case (0.46%) achieving CR, 98 cases (44.75%) achieving PR, 95 cases (43.38%) classified as SD, and 25 cases (11.42%) presenting PD, with an objective response rate of 45.21%. Survival analyses were performed based on clinical characteristics, where univariate analysis identified tumor stage IV, anti-angiogenic drugs, later-line therapy, ECOG-PS, irAEs, NRS2002 scores, albumin, PNI, and NLR as factors related to OS. Multivariate analysis further demonstrated that, besides the number of treatment lines and irAEs, only NLR (HR = 2.83; 95% CI 1.68–4.76; P < 0.001) and NRS2002 (HR = 2.80; 95% CI 1.71–4.60; P < 0.001) were significantly correlated with OS (Table 2). Multivariate analysis concerning PFS showed that NLR (HR = 2.47; 95% CI 1.56–3.90; P < 0.001) and NRS2002 (HR = 1.55; 95% CI 1.02–2.37; P = 0.041) maintained significant effects on PFS (Table 3).

Table 2 Univariate and multivariate analyses of OSTable 3 Univariate and multivariate analyses of PFS3.3 Survival analysis of combined NRS2002 and NLR as prognostic indicators

Given the significance of NRS2002 score and NLR in survival analysis of NSCLC patients receiving immunotherapy, this study developed a novel scoring system to further delineate risk stratification. The method involved assigning 1 point for NRS2002 scores ≥ 3 and NLR > 4.92, with all other conditions assigned 0 points. The scores were summed, and patients were categorized into three groups based on total score: 0 points, 1 point, and 2 points, to assess survival prognosis differences among risk groups. Survival analysis across the 0, 1, and 2-point groups revealed that median OS was not reached for the 0-point group, while it was 16 months (95% CI 13.04–18.96) and 8 months (95% CI 2.61–13.39) for the 1-point and 2-point groups, respectively (P < 0.001). The median PFS was 16 months (95% CI 11.41–20.59), 8.5 months (95% CI 5.47–11.53), and 2.5 months (95% CI 0.99–4.02) for the 0, 1, and 2 point groups, respectively (P < 0.001). Kaplan–Meier analysis demonstrated a progressive decrease in both OS and PFS with increasing NRS-NLR scores. As shown in Fig. 2A and B, the 0-point group had the best survival probability, followed by the 1-point group, with the 2-point group having the poorest prognosis (P < 0.0001). Additionally, subgroup analysis stratified by tumor stage revealed significant differences in OS and PFS across all subgroups (P < 0.0001). Further stratified analysis by tumor histology demonstrated a consistent trend, where OS and PFS progressively decreased with increasing NRS-NLR scores, regardless of tumor stage. Higher NRS-NLR scores were consistently associated with lower survival rates and poorer prognoses.

Fig. 2

Kaplan–Meier survival curves of patients in different NRS-NLR Score Groups

3.4 Predictive value of combined NRS-NLR score

The prognostic significance of the NRS-NLR composite score was evaluated in comparison to NRS2002, PD-L1, NRS-NLR, the NRS combined index (NRS2002 + ALB + PNI), the NLR combined index (NLR + ALB + PNI), and NLR alone. Time-dependent ROC curve analysis revealed that at 12 months, the NRS-NLR score showed the highest AUC value (AUC = 0.81; 95% CI 0.72–0.90), compared to an AUC of 0.73 (95% CI 0.63–0.82) for the NRS2002 score, 0.63 (95% CI 0.50–0.75) for PD-L1, 0.70 (95% CI 0.60–0.78) for NLR, 0.71 (95% CI 0.60–0.81) for the NRS combined index, and 0.72 (95% CI 0.60–0.83) for the NLR combined index (Fig. 3A). Although the predictive power of all scoring systems decreased over time, the NRS-NLR score remained relatively robust, with AUCs of 0.73 (95% CI 0.66–0.80) and 0.70 (95% CI 0.64–0.76) at 24 and 36 months, respectively. The NRS2002 and NLR scores showed comparable performance in predicting long-term survival, with AUCs of 0.65 (95% CI 0.59–0.72) and 0.64 (95% CI 0.57–0.70) at 24 months, and 0.63 (95% CI 0.57–0.68) and 0.63 (95% CI 0.56–0.68) at 36 months. The predictive abilities of the NRS combined index and NLR combined index were similar, with AUCs of 0.66 (95% CI 0.58–0.74) and 0.66 (95% CI 0.58–0.74) at 24 months, and 0.66 (95% CI 0.57–0.74) and 0.66 (95% CI 0.59–0.73) at 36 months. The discriminative ability of PD-L1 remained relatively unchanged, with AUC values of 0.62 (95% CI 0.53–0.70) and 0.56 (95% CI 0.49–0.64) at 24 and 36 months, respectively (Fig. 3B and C).

Fig. 3

Time-dependent ROC curves for NRS-NLR, NRS2002, NLR, PD-L1, NRS2002 + ALB + PNI and NLR + ALB + PNI

3.5 Construction and predictive performance of a nomogram model based on NRS-NLR

To visually illustrate the significance of the NRS-NLR score in predicting the efficacy of treatment in NSCLC patients, we incorporated factors with significant influence from survival analysis to construct a nomogram (Fig. 4A). This model incorporates key variables such as cancer staging, line of therapy, PNI, ALB, irAEs, and the combined NRS-NLR score. Each variable's contribution to survival outcomes is quantified through regression coefficients from our statistical analysis, and corresponding points are assigned to each.

Fig. 4

Nomogram and its predictive performance evaluation. A Nomogram combining NRS-NLR with key covariates (stage, line of therapy, PNI, ALB, irAEs); B Time-dependent ROC analysis; C Calibration curves; D Decision curve analysis (DCA) 

The model indicated that the NRS-NLR score plays a critical role in prognostic prediction, where higher scores are associated with an elevated risk of mortality, in line with survival analysis findings. Moreover, the nomogram model built on NRS-NLR exhibited strong predictive performance for the prognosis of NSCLC patients treated with ICIs, as depicted in Fig. 4B, with AUC values of 0.84 (95% CI 0.77–0.91), 0.85 (95% CI 0.79–0.91), and 0.78 (95% CI 0.69–0.88) at 12, 24, and 36 months, respectively. The model's accuracy and clinical utility were assessed through calibration curves and decision curve analysis (DCA), with the results presented in Fig. 4C and D. The calibration curves indicated a good concordance between predicted survival probabilities and actual observed outcomes, particularly in predicting survival outcomes for patients at moderate to high risk. The DCA curves further validated that this scoring system provides significant positive net benefits in clinical decision-making.