Patient characteristics and breast cancer subtypes
The patient characteristics are demonstrated in Table 1. The median age at diagnosis was 49.9 years. Among the 485 HER2-negative patients, 306 (63.1%) were HR+ and 179 (36.9%) were HR-. The molecular subtypes of luminal A, luminal B, HER2, basal-like, and normal types were noted in 153 (31.5%), 75 (15.5%), 35 (7.2%), 179 (36.9%), and 43 (8.9%) cases, respectively. In assessments based on tumor size, 279 (57.5%) cases were categorized as T0-2, while evaluations of nodal status showed that 151 cases (31.1%) had no lymph node metastasis. Seventeen cases (3.5%) included no information on histological grade status. After neoadjuvant chemotherapy treatment, 93 (19.2%) patients achieved pCR, while 373 (76.9%) patients showed residual disease. As for the prognostic outcomes, 106 (21.9%) patients showed no relapse in five years after primary diagnosis, whereas 379 (78.1%) patients experienced recurrence within five years.
The expression levels of each IGS classified by molecular subtype are shown in Figure 1. Each model tended to show differences in expression across subtypes (Kruskal–Wallis P value < 0.001). As expected, significantly higher expression levels were found in more aggressive subtypes, such as HER2 or basal-like subtypes.
Scatter Plot Matrix For Each Igs
We then assessed the strength of the correlation between the models by using scatter plot matrix-calculated correlation coefficients determined with Pearson's rank correlation test. As shown in Figure 2, all pairs of IGSs showed high correlations with each other: Ascierto (r: 0.65-0.83), Schmidt (r: 0.58-1), Bianchini (r: 0.56-1), TILsGS (r: 0.56-0.66), and IRSN-23 (r: 0.6-0.67).
Correlations Between Distant Relapse-free Survival And Igss
In the univariate Cox proportional analysis of 5-year DRFS, in all cancer patients (n = 485), 3 of the 5 IGSs showed significant prognostic value (Table 2), including Ascierto: low vs. high (HR 0.577, 95% CI 0.353-0.944, P = 0.028); TILsGS: low vs. high (HR 0.281, 95% CI 0.164-0.482, P < 0.001); IRSN-23: low vs. high (HR 0.420, 95% CI 0.254-0.693, P < 0.001). In HR+ patients (n = 306), two out of five IGS had considerable importance (Table 2), including TILsGS: low vs. high (HR 0.358, 95% CI 0.166-0.775, P = 0.009); IRSN-23: low vs. high (HR 0.319, 95% CI 0.141-0.724, P = 0.006). However, in HR- patients (n = 179), there was no significant difference between the groups (Table 2). Kaplan–Meier curves for survival are shown in Supplementary figure1.
In the multivariate Cox proportional analysis of 5-year DRFS (Table 2), in all cancer patients (n = 485), only IRSN-23 had significant prognostic value: low vs. high (HR 0.459, 95% CI 0.224-0.941, P = 0.033). In HR+ patients (n = 306), two of the five IGSs had considerable importance, including TILsGS: low vs. high (HR 0.402, 95% CI 0.181-0.892, P = 0.025) and IRSN-23: low vs. high (HR 0.394, 95% CI 0.166-0.932, P = 0.034). However, in HR- patients, no IGSs were statistically significant.
Correlations Between Pcr And Igss
In the univariate logistic regression analysis of the prediction of chemotherapy response, for all cancer patients (n = 466), 4 of the 5 IGSs showed significant predictive values (Table 3), including Ascierto: intermediate vs. high (OR 0.382, 95% CI 0.215-0.661, P < 0.001), low vs. high (OR 0.363, 95% CI 0.203-0.631, P < 0.001); Schmidt: intermediate vs. high (OR 0.460, 95% CI 0.267-0.809, P = 0.007), low vs. high (OR 0.400, 95% CI 0.249-0.762, P = 0.004); Bianchini: intermediate vs. high (OR 0.459, 95% CI 0.263-0.790, P = 0.005), low vs. high (OR 0.430, 95% CI 0.242-0.750, P = 0.003); TILsGS: intermediate vs. high (OR 0.484, 95% CI 0.284-0.814, P = 0.007), and low vs. high (OR 0.212, 95% CI 0.108-0.394, P < 0.001). In HR+ patients (n = 306), 4 of the 5 IGSs showed considerable importance (Table 3), including Ascierto: intermediate vs. high (OR 0.258, 95% CI 0.100-0.615, P = 0.003), low vs. high (OR 0.268, 95% CI 0.112-0.609, P = 0.002); Schmidt: intermediate vs. high (OR 0.397, 95% CI 0.151-0.459, P = 0.034), low vs. high (OR 0.345, 95% CI 0.143-0.797, P = 0.014); Bianchini: intermediate vs. high (OR 0.400, 95% CI 0.168-0.926, P = 0.034), low vs. high (OR 0.340, 95% CI 0.139-0.798, P = 0.014); TILsGS: intermediate vs. high (OR 0.392, 95% CI 0.163-0.915, P = 0.032), low vs. high (OR 0.253, 95% CI 0.103-0.595, P = 0.002). However, there was no statistically significant difference in HR- patients (Table 3). Boxplots for the predictive results of pCR and RD are shown in Supplementary figure 2.
Also, among the multivariate logistic regression analysis of the prediction to chemotherapy response (Table 3), in HR+ patients, 4 of the 5 IGSs (Ascierto, Schmidt, Bianchini, and TILsGS ) showed considerable importance, including Ascierto: intermediate vs. high (OR 0.217, 95% CI 0.077-0.559, P = 0.002), low vs. high (OR 0.274, 95% CI 0.106-0.670, P = 0.005); Schmidt: intermediate vs. high (OR 0.357, 95% CI 0.137-0.880, P = 0.028), low vs. high (OR 0.309, 95% CI 0.118-0.771, P = 0.013); Bianchini: intermediate vs. high (OR 0.313, 95% CI 0.120-0.783, P = 0.015), low vs. high (OR 0.276, 95% CI 0.102-0.712, P = 0.009); TILsGS: low vs. high (OR 0.418, 95% CI 0.159-1.075, P = 0.019). However, no IGS showed statistical significance in all cancer patients and in HR- patients.