Significantly different immunological score in lung adenocarcinoma and squamous cell carcinoma and a proposal for a new immune staging system

Figures & data

Figure 1. Construction of the IS_NSCLC by the LASSO model. LASSO coefficient profiles of 20 significant immune features with the top 20 smallest P-values in LUAD training cohort (a) and LUSC training cohort (d). For tuning parameter selection in the LASSO model, by 10-time cross-validation via 1-SE criteria, λ = 0.194 was chosen in LUAD training cohort (b), and λ = 0.283 was chosen in LUSC training cohort (e). Representative multiplex fluorescent immunohistochemistry images show the final features enrolled in IS_LUAD (c) and in IS_LUSC (f). Bar, 100 µm. IS, immunological score; NSCLC, non-small cell lung cancer; LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma; CT, the core of tumor; IM, invasive margin.

Figure 2. IS_NSCLC performance in time-dependent ROC curves and Kaplan–Meier survival analyses in different cohorts. (a) LUAD training cohort. (b) LUAD validation cohort. (c) LUSC training cohort. (d) LUSC validation cohort. AUCs at 1, 3, and 5 y were used to assess prognostic accuracy. P-values were calculated by the log-rank test. DFS, disease-free survival; OS, overall survival; ROC, receiver operator characteristic; AUC, area under the curve.

Table 1. Multivariable Cox regression analyses of the IS_NSCLC, clinicopathological characteristics, and survival.

LUAD cohort			LUSC cohort
Variable	HR (95% CI)	P	Variable	HR (95% CI)	P
Disease-free survival			Disease-free survival
Training Cohort (n= 102)			Training Cohort (n= 50)
ISLUAD	2.766 (2.154, 3.550)	<0.001	ISLUSC	2.866 (1.898, 4.329)	<0.001
T stage (T3+ T4 vs. T1+ T2)	2.328 (1.139, 4.759)	0.021	N stage (N1+ N2+ N3 vs. N0)	2.947 (1.234, 7.039)	0.015
Validation Cohort (n= 102)			Validation Cohort (n= 50)
ISLUAD	1.952 (1.501, 2.540)	<0.001	ISLUSC	3.638 (2.190, 6.044)	<0.001
TNM stage (III vs. I+ II)	1.353 (1.059, 1.728)	0.016
Tumor volume (≥10 cm^3 vs. <10 cm^3)	1.910 (1.172, 3.112)	0.009
Overall survival			Overall survival
Training Cohort (n= 102)			Training Cohort (n= 50)
ISLUAD	2.153 (1.642, 2.824)	<0.001	ISLUSC	2.575 (1.728, 3.836)	<0.001
Validation Cohort (n= 102)			Validation Cohort (n= 50)
ISLUAD	1.930 (1.402, 2.656)	<0.001	ISLUSC	2.502 (1.644, 3.805)	<0.001
T stage (T3+ T4 vs. T1+ T2)	2.930 (1.503, 5.711)	0.002

IS, immunological score; NSCLC, non-small cell lung cancer; LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma.

Figure 3. Kaplan–Meier survival analysis stratified by the TNM stages. Kaplan–Meier survival analysis of DFS and OS for all the NSCLC patients according to the IS_NSCLC stratified by the TNM stages. P-values were calculated by the log-rank test. (a) Stage I LUAD. (b) Stage I LUSC. (c) Stage II LUAD. (d) Stage II LUSC. (e) Stage III LUAD. (f) Stage III LUSC.

Figure 4. From TNM to TN-I. DFS curves of LUAD/LUSC patients according to IS_NSCLC (a, b) and TNM stage (c, d). Patients were grouped into three groups (e, f) according to the combination of TNM stage (t, n) and IS_NSCLC (i) (T-N-I risk score) by survival (3-y DFS, <20%, 20%-70%, and >70%; 5-y OS, <30%, 30%-70%, and >70%), resulting into a TN-immunological score (TN-I) staging table (g), adding significant prognostic impact to each TNM stage. ROC curves show the comparisons of the area under the curve of TN-I stage, IS_NSCLC, and the IS_NSCLC classifier (high vs. low) with TNM stage and the IS proposed by SITC in LUAD cohort (h) and LUSC cohort (I). SITC, the Society for Immunotherapy of Cancer.

Figure 5. Nomograms for predicting the 3-y recurrence and 5-y mortality rate after surgery in NSCLC patients. Nomograms (a) are showed at the top. Receiver operator characteristic curves (b) exhibit the prediction accuracy of each nomogram model. Model performance is shown by calibration plots (c) depicting the agreement between predicted and observed probabilities of each model. Decision curve analysis (d) shows the net clinical benefit of the nomograms.