1. Clinical characteristics
This study included 100 patients diagnosed with TNBC who underwent NAC combined with surgery. The case-screening flowchart was shown in Fig. 1. The median age of participants was 52. There were 8, 37, and 55 patients with tumor stages T1, T2, and T3, respectively. A total of 82 patients had lymph node metastasis at diagnosis, whereas the other 18 did not. Among the tumor immunohistochemistry results, 45 patients were HER-2 negative, 35 were 1+, and 20 were 2 + FISH-. A total of 76 patients had a high Ki-67 level (> 30), and 24 had a low level (≤ 30). The clinical and pathological characteristics of the patients with TNBC were presented in Table 1.
Table 1
Clinical feature |
Age (years) | Median | 52 |
| Rang | 23–75 |
Tumor size | T1 | 8 |
| T2 | 37 |
| T3 | 55 |
HER-2 | 0 | 45 |
| 1 | 35 |
| 2 (FISH-) | 20 |
N | have | 82 |
| none | 18 |
Ki-67 | ≤ 30 | 24 |
| > 30 | 76 |
2. The relationship between inflammation indicators and the efficacy of NAC
IL-6 (p = 0.0398), PLT (p = 0.0157), PLR (p = 0.0143), and SII (p = 0.0108) were significantly lower in the pCR group than in the non-pCR group, whereas Ki-67 (p = 0.0137) in the pCR group was distinctly higher than that in the non-pCR group, as displayed in Fig. 2.
3. Prognostic relevance of inflammation indicators and clinical features
Indicators were divided into two groups based on their median values. The high-level IL-6 (p = 0.0157), PLR (p = 0.0001), CRP (p = 0.0298), and SII (p = 0.0011) levels were significantly correlated with worse DFS compared to the low-level group. Additionally, the prognosis of the group with high lymphocyte (p = 0.0267) and Ki-67 (p = 0.0298) levels was better than that of the low-level group, as showed in Fig. 3.
4. Development of a prognostic model
The present study suggests that IL-6, SII, PLR, and Ki-67 can serve as predictive indicators of NAC efficacy and postoperative DFS. Univariate Cox regression and least absolute shrinkage and selection operator (LASSO) regression were used to construct an optimization model utilizing these four indicators. The risk model formula based on prognostic parameters is as follows:
Risk score = (IL-6 value * 0.05196) + (SII value * 0.00029) + (Ki-67 value * -0.01468) + (PLR value * 0.00109).
The patients were divided into low- and high-risk groups based on the median risk score. Survival curves were analyzed using the Kaplan-Meier method. Patients in the low-risk score group exhibited a significantly better DFS than those in the high-risk score group (Fig. 4A; p = 0.0007; HR, 3.630; 95% CI: 1.787–7.375). All participants were randomly assigned to either the test or training group using R package to validate the accuracy of the prognostic model. Prognostic analysis was conducted for both groups of patients. Regardless of whether or not the test was performed (Fig. 4B; p = 0.0262; HR, 2.936; 95% CI: 1.137–7.749) or training (Fig. 4C; p = 0.0111; HR, 3.228; 95% CI: 1.169–8.912) group, patients with a low-risk score consistently demonstrated better DFS.
5. Cox regression analysis of the prognostic model
Univariate regression analysis revealed that a high-risk score (HR, 3.932; 95% CI: 1.676–9.224; p = 0.002) and small tumor size (HR, 2.408; 95% CI: 1.061–5.463; p = 0.036) were correlated with worse DFS, as listed in Table 2. Multifactor analysis identified a high-risk score (HR, 4.597; 95% CI: 1.782–11.858; p = 0.002) and small tumor size (HR, 2.485; 95% CI: 1.084–5.696; p = 0.032) as independent risk factors (Table 2).
Table 2
COX regression analysis of prognostic model
univariate analyses | multivariate analyses |
| P value | Hazard ratio | | P value | Hazard ratio |
Risk scores | 0.002 | 3.932 (1.676–9.224) | Risk scores | 0.002 | 4.597 (1.782–11.858) |
Tumor size | 0.036 | 2.408 (1.061–5.463) | Tumor size | 0.032 | 2.458 (1.084–5.696) |
Age (years) | 0.588 | 0.815 (0.389–1.710) | Age (years) | - | - |
HER2 | 0.633 | 1.198 (0.571–2.511) | HER2 | - | - |
6. Evaluation of prognostic model accuracy
The accuracy of the prognostic model was further validated using the ROC curve, as displayed in Fig. 5.
7. Establishment and validation of the nomogram
Four indicators were utilized along with the Miller-Payne (MP) scores[17] to develop a nomogram for predicting the one-, two-, and three-year DFS. Calibration curves were generated to illustrate acceptable consistency between the actual and nomogram-predicted survival rates. (Fig. 6)