Clinicopathologic characteristics

Based on the inclusion and exclusion criteria, a total of 601 patients with HR+/HER2− breast cancer who did not achieve pCR following neoadjuvant chemotherapy were included in this study. The baseline characteristics of the patients are summarized in Table 1. The median age of the cohort was 45 years, and 434 patients (72.2%) were premenopausal. Prior to neoadjuvant chemotherapy, the majority of patients were clinically staged as IIA to IIIA (88.4%). After neoadjuvant chemotherapy, 60.5% of patients had a postoperative clinical stage of I to IIB. A total of 210 patients (35.4%) underwent breast-conserving surgery, and 495 patients (82.4%) received radiotherapy. In terms of adjuvant endocrine therapy, 226 patients (37.6%) received ovarian function suppression (OFS) via surgery or medication, 330 patients (56.2%) were treated with aromatase inhibitors (AI), and 257 patients (43.8%) received tamoxifen (TAM). The median follow-up time for the cohort was 47 months, with a minimum follow-up of 7 months and a maximum of 111 months. The 5-year DFS rate was 77.8%, and the 5-year OS rate was 87.5%.

Table 1 Clinicopathological characteristics of the study cohortPrognostic prediction of IHC4 score

Based on the predefined cut-off values for the IHC4 variables, patients were stratified into low, intermediate, and high-risk groups using the IHC4 score. A higher IHC4 risk group was significantly associated with poorer prognosis, as demonstrated by a marked reduction in both DFS and OS. The Kaplan–Meier curve (Fig. 2A) shows that the IHC4 score effectively identified patients in the intermediate-risk group (hazard ratio [HR] = 5.16, 95% confidence interval [CI]: 1.24–21.4, p = 0.024) and the high-risk group (HR = 14.1, 95% CI: 3.46–57.4, p < 0.001) with significantly shorter DFS compared to the low-risk group. Furthermore, the Kaplan–Meier curve (Fig. 2E) illustrates that the IHC4 score was able to distinguish high-risk patients (HR = 14.9, 95% CI: 2.06–108.3, p = 0.008) with significantly reduced OS compared to the low-risk subgroup; intermediate-risk patients also showed numerically shorter OS (HR = 5.10, 95% CI: 0.68–38.2, p = 0.113), though without statistical significance. After adjusting for BMI, age, tumor size, lymph node status, menopausal status and treatments, multivariate analysis confirmed that the IHC4 score was significantly associated with worse DFS and OS (Table 2, 3), indicating its strong prognostic value.

Fig. 2

Kaplan–Meier survival curves for non-pCR HR+/HER2− breast cancer patients following NAC. Panels A-D present DFS stratified by IHC4, RCB, MP, and CPS-EG scores, respectively. Panel E–H depict OS stratified by IHC4, RCB, MP, and CPS-EG scores, respectively

Table 2 Univariate and multivariate Cox regression analyses for DFSTable 3 Univariate and multivariate Cox regression analyses for OSPrognostic prediction of RCB score

Patients were also classified into RCB I, RCB II, and RCB III groups using the RCB score. Higher RCB risk groups were significantly associated with worse prognosis, with notable reductions in both DFS and OS. As shown in the Kaplan–Meier curve (Fig. 2B), the RCB score identified patients in the RCB III group (HR = 3.55, 95% CI: 1.11–11.12, p = 0.033) with significantly shorter DFS compared to the RCB I subgroup. Although RCB II patients had numerically shorter DFS (HR = 1.38, 95% CI: 0.42–4.54, p = 0.59) compared to the RCB I group, the difference was not statistically significant. The Kaplan–Meier curve (Fig. 2F) further shows that the RCB score did not reveal significant differences in OS between the RCB II group (HR = 0.86, 95% CI: 0.20–3.75, p = 0.84) and the RCB III group (HR = 2.54, 95% CI: 0.61–10.5, p = 0.20) compared to the low-risk subgroup.

Prognostic prediction of MP score

Patients were also stratified into MP1, MP2, MP3, and MP4 groups based on the MP score. Higher MP risk groups did not show a statistically significant association with worse prognosis, as demonstrated by the lack of significant differences in DFS and OS. The Kaplan–Meier curve (Fig. 2C) shows that patients in the MP1, MP2, and MP3 risk groups had numerically shorter DFS compared to the MP4 group, but no statistically significant differences were observed (HR = 1.42, 95% CI: 0.74–2.74, p = 0.30; HR = 1.01, 95% CI: 0.56–1.80, p = 0.99; HR = 0.72, 95% CI: 0.39–1.31, p = 0.29). Additionally, as illustrated in the Kaplan–Meier curve (Fig. 2G), OS was shorter in the MP1, MP2, and MP3 risk groups compared to the MP4 group, but again, the differences were not statistically significant (HR = 1.76, 95% CI: 0.73–4.24, p = 0.21; HR = 0.97, 95% CI: 0.42–2.21, p = 0.94; HR = 0.82, 95% CI: 0.36–1.87, p = 0.64).

Prognostic prediction of CPS-EG score

Using the predefined cut-off values for the CPS-EG variables, patients were categorized into low and high-risk groups based on the CPS-EG score. A higher CPS-EG risk group was significantly correlated with poorer prognosis, as indicated by a marked reduction in DFS and OS. The Kaplan–Meier curve (Fig. 2D) demonstrates that the CPS-EG score effectively identified high-risk patients (HR = 4.33, 95% CI: 2.79–6.33, p < 0.001) with significantly shorter DFS compared to the low-risk subgroup. Likewise, the Kaplan–Meier curve (Fig. 2H) shows that the CPS-EG score also identified high-risk patients (HR = 5.00, 95% CI: 2.98–8.40, p < 0.001) with significantly reduced OS compared to the low-risk subgroup. After adjusting for BMI, age, menopausal status and treatments, multivariate analysis confirmed that the CPS-EG score was significantly associated with worse DFS and OS (Table 2, 3).

Comparison of prognostic stratification performance

In summary, apart from the MP score, which did not achieve statistical significance in prognostic stratification, the IHC4, RCB, and CPS-EG scoring systems were all able to stratify prognosis in HR (+)/HER2 (−) breast cancer patients who did not achieve pCR following neoadjuvant chemotherapy. Subsequently, we compared the prognostic stratification performance of these three scoring systems using AUC and the concordance index (C-index). We calculated the time-dependent ROC curves for DFS and OS at the 1- to 5-year time points for each of the three scores, plotted the AUC curves over time, and analyzed the temporal changes in AUC from 1 to 5 years. The graphical results showed that, for both DFS and OS, the AUC of the IHC4 and CPS-EG scores was consistently higher than that of the RCB score during the 1- to 5-year period (Fig. 3B and E).

Fig. 3

Kaplan–Meier analysis and time-dependent AUC curves and C-index comparisons for DFS and OS in HR+/HER2− breast cancer patients post-NAC, stratified by the COMBINE score and compared with IHC4, RCB, and CPS-EG scores. Panel A: Kaplan–Meier curve for DFS based on COMBINE score stratification. Panel B: Time-dependent AUC curves for DFS over 1–5 years for each scoring system. Panel C: C-index comparisons for DFS across scoring systems. Panel D: Kaplan–Meier curve for OS based on COMBINE score stratification. Panel E: Time-dependent AUC curves for OS over 1–5 years for each scoring system. Panel F: C-index comparisons for OS across scoring systems

Furthermore, both the IHC4 and CPS-EG scores demonstrated superior prognostic performance compared to the RCB score, with statistically significant differences in the concordance index for DFS (0.69 vs. 0.63, p < 0.01; 0.69 vs. 0.63, p < 0.01) and OS (0.70 vs. 0.64, p < 0.01; 0.73 vs. 0.64, p < 0.01). However, when comparing the prognostic performance between the IHC4 and CPS-EG scores, there was no statistically significant difference in the concordance index for DFS (0.69 vs. 0.69, p = 0.86; 0.70 vs. 0.73, p = 0.37) or OS (0.70 vs. 0.73, p = 0.37) (Fig. 3C and F).

IHC4 provides additional prognostic insight in CPS-EG scoring

To assess whether the IHC4 score provides additional prognostic information beyond the CPS-EG score, we re-stratified patients within the CPS-EG low-risk group using the IHC4 score. Among the 429 patients classified as low-risk by CPS-EG, 145 (33.8%) were reclassified as high-risk, 218 (50.8%) as intermediate-risk, and 66 (15.4%) as low-risk. Increased risk levels were significantly associated with worse DFS (high-risk vs. low-risk, HR = 5.80, p = 0.006; intermediate-risk vs. low-risk, HR = 3.16, p = 0.101; high-risk vs. intermediate-risk, HR = 1.91, p = 0.031) and OS (high-risk vs. low-risk, HR = 6.17, p = 0.045; intermediate-risk vs. low-risk, HR = 2.89, p = 0.29; high-risk vs. intermediate-risk, HR = 2.19, p = 0.067) (Fig. 4A and C).

Fig. 4

Kaplan–Meier curves for DFS and OS in non-pCR HR+/HER2− breast cancer patients following NAC, further stratified by IHC4 score after initial classification by the CPS-EG score. Panels A and C illustrate DFS and OS for CPS-EG low-risk patients (n = 429), respectively. Panels B and D represent DFS and OS for CPS-EG high-risk patients (n = 172), respectively, stratified by IHC4 score

Similarly, in the 172 patients classified as high-risk by CPS-EG, 106 (61.6%) were categorized as high-risk, 59 (32.6%) as intermediate-risk, and 10 (5.8%) as low-risk. Higher risk levels were associated with worse DFS (high-risk vs. low-risk, HR = 7.12, p = 0.003; intermediate-risk vs. low-risk, HR = 2.45, p = 0.22; high-risk vs. intermediate-risk, HR = 1.94, p = 0.03) and worse OS (high-risk vs. low-risk, HR = 5.64, p = 0.065; intermediate-risk vs. low-risk, HR = 2.31, p = 0.243; high-risk vs. intermediate-risk, HR = 3.28, p = 0.004) (Fig. 4B and D).

Calculation of the COMBINE score

We found that only the IHC4 and CPS-EG scores showed statistical significance in the univariate and multivariable Cox regression analyses for DFS and OS (see Tables 2 and 3). The calculated β values were 0.809 for DFS and 0.855 for OS in the IHC4 score, and 1.228 for DFS and 1.322 for OS in the CPS-EG score, all of which were approximately equal to 1(Supplemental Table 1).

Based on these findings, we evaluated two scoring methodologies:

1.

β-Derived Composite Score: This approach used the exact β coefficients from the regression analyses to calculate the COMBINE score.

2.

Simplified Scoring System: This approach assigned fixed scores (0, 1, or 2) to predefined risk categories in the IHC4 and CPS-EG systems.

Performance comparisons between these methods showed comparable prognostic utility, as evidenced by overlapping time-dependent AUC curves and similar c-index values (Supplementary Fig. 1). Given its simplicity and clinical applicability, we adopted the simplified scoring approach for the final COMBINE score.

In the simplified system, high-risk patients in the IHC4 system were assigned 2 points, intermediate-risk patients 1 point, and low-risk patients 0 points. In the CPS-EG system, high-risk patients received 1 point, while low-risk patients received 0 points. The total COMBINE score, ranging from 0 to 3, is calculated by summing these points. A score of 3 denotes high risk, 2 indicates moderate-to-high risk, 1 represents low-to-moderate risk, and 0 signifies low risk. This scoring system offers refined stratification to guide personalized clinical decision-making.

Development and prognostic value of the COMBINE score

As previously noted, the IHC4 score can provide supplementary prognostic information to the CPS-EG score. Based on this, we combined the IHC4 and CPS-EG scores to create a composite scoring system (detailed scoring criteria are described in the Methods section). We then stratified all patients according to the COMBINE score. The results showed a clear correlation between higher risk groups and poorer prognosis across different risk stratifications. This was reflected in significantly worse DFS (high-risk vs. low-risk, HR = 21.3, p < 0.001; intermediate-high-risk vs. low-risk, HR = 7.48, p = 0.001; intermediate-low-risk vs. low-risk, HR = 3.03, p = 0.011; high-risk vs. intermediate-low-risk, HR = 7.45, p < 0.001; intermediate-high-risk vs. intermediate-low-risk, HR = 2.78, p < 0.001; high-risk vs. intermediate-high-risk, HR = 3.09, p < 0.001) (Fig. 3A) and OS (high-risk vs. low-risk, HR = 22.7, p < 0.001; intermediate-high-risk vs. low-risk, HR = 7.81,p = 0.001; intermediate-low-risk vs. low-risk, HR = 2.78, p = 0.031; high-risk vs. intermediate-low-risk, HR = 8.73, p < 0.001; intermediate-high-risk vs. intermediate-low-risk, HR = 2.83, p = 0.03; high-risk vs. intermediate-high-risk, HR = 3.08, p < 0.001) (Fig. 3D).

Notably, patients with a COMBINE score of 0 had a 5-year DFS of 96.5% and a 5-year OS of 100%. In contrast, patients with a COMBINE score of 3 had a 5-year DFS of only 55.1% and a 5-year OS of 63.4%. To compare the prognostic performance of the COMBINE score with CPS-EG and IHC4 in stratifying DFS and OS, we plotted time-dependent AUC curves over a 1–5 year period. The results showed that, for both DFS and OS, the AUC for the COMBINE score was consistently higher than that of both the IHC4 and CPS-EG scores during the 1–5 year period (all p < 0.05, Supplemental Tables 2, 3) (Fig. 3B and E).

Furthermore, the C-index for DFS based on the COMBINE score was significantly higher than that of CPS-EG and IHC4 (0.76 vs. 0.69, p < 0.001; 0.76 vs. 0.69, p < 0.001). Similarly, the C-index for OS based on the COMBINE score was significantly greater than that of IHC4 and CPS-EG (0.78vs. 0.70, p < 0.001; 0.78 vs. 0.73, p < 0.001) (Fig. 3C and F).