A total of 1859 patients with invasive breast cancer underwent radical surgery between 2006 and 2021. Among them, 1633 (87.8%) patients were identified as having invasive breast carcin“oma of no special type, whereas 226 (12.2%) were classified as having a special type. After excluding 105 (5.6%) patients with “favorable histologies”, the remaining 121 patients (6.6%) with special-type breast cancer were included in the final analysis. (Fig. 1)

Study flowchart. NCCN: the National Comprehensive Cancer Network

The clinicopathological characteristics of the 121 patients are presented in Table 1. Median age was 59.8 years (IQR 49.9–69.1), and median follow up period was 77.8 months (IQR 37.2–114.3). The most prevalent histological type was invasive lobular carcinoma (39 cases, 32.2%), followed by invasive micropapillary carcinoma (35 cases, 28.9%) and carcinoma with apocrine differentiation (15 cases, 12.4%) (Fig. 1). The T and N stages most frequently observed were T1 in 54 (44.6%) patients, and N0 in 68 (56.2%) patients. The observed subtypes were as follows: 91 cases (75.2%) were HR + HER2-, 13 (10.7%) were HR+/-HER2+, and 17 (14.1%) were HR-HER2-. Adjuvant chemotherapy and radiotherapy was administered in 73 (60.3%) and 65 (53.7%) patients, respectively. Table 1 shows the patient backgrounds for each histological type. A subsequent evaluation of previous reports and our own data, employing a T-test, demonstrated a comparable distribution pattern for SUVmax (Table S1). Notably, the prevalence of HER2-positive mucinous carcinoma was higher, because luminal-type and low-grade mucinous carcinomas were excluded owing to their favorable prognosis. Additionally, triple-negative carcinomas were more prevalent among carcinomas with apocrine differentiation, adenomyoepithelial carcinomas and papillary carcinomas compared to other histologies (Table S2).

Overall, 66 patients (54.5%) had SUVmax ≤ 3.0 and 55 patients (45.5%) had SUVmax > 3.0. Additionally, SUVmax > 3.0 was significantly associated with higher T stage, N stage, higher nuclear grade, and adjuvant chemotherapy than SUVmax ≤ 3.0 in Fisher’s exact test (P = 0.012, P = 0.006, P < 0.005, P < 0.005, respectively). (Table 2) Among all patients, recurrence occurred in 16 (13.2%) while nine (7.4%) died. The 5-year RFS was 94.6% (SE ± 3.02%) for SUVmax ≤ 3.0 and 82.9% (SE ± 5.63%) SUVmax > 3.0 (P = 0.005). The 5-year OS was 96.7% (SE ± 2.32%) for SUVmax ≤ 3.0 and 89.2% (SE ± 4.61%) SUVmax > 3.0 (P = 0.167) (Fig. 2a, b). The prognosis for RFS was statistically significantly more favorable when the SUVmax was ≤ 3.0.

Relapse free survival curve and Overall survival of patients according to SUVmax values. a Relapse free survival and b Overall survival of all patients. RFS relapse free survival, OS overall survival, SUVmax maximum standardized uptake value. 1 Log-rank test, *significant P-value < 0.05

Table 3 presents the 5-year RFS and OS rates according to the SUVmax for each histological type. Over the five-year period, no cases of recurrence or mortality due to invasive lobular carcinoma, mucinous carcinoma, tubular carcinoma, encapsulated papillary carcinoma, solid papillary carcinoma, or papillary carcinoma were observed. However, three cases of recurrence were observed in patients with invasive lobular carcinoma after a ten-year period. In the context of micropapillary carcinomas, the 5-year RFS rate was 93.8% for SUVmax ≤ 3.0 and 80.1% for SUVmax > 3.0, and the 5-year OS was 94.1% for SUVmax ≤ 3.0 and 93.3% SUVmax > 3.0. In the case of apocrine carcinomas, the 5-year RFS was 100% for SUVmax ≤ 3.0 and 80.0% for SUVmax > 3.0. No cases of mortality were observed (Table 3). Subsequent evaluation of prognostic differences based on the SUVmax was conducted for each molecular subtype (Table 4). In the HR + HER2-subtype, the 5-year RFS was 95.2% (SE ± 3.29%) for SUVmax ≤ 3.0 and 87.4% (SE ± 6.07%) for SUVmax > 3.0 (P = 0.020), indicating a statistically significantly better prognosis with SUVmax ≤ 3.0. Whereas no statistically significant differences were observed between the HR+/- HER2 + and HR-HER2- subtypes. Furthermore, no substantial variations in 5-year OS were identified among the subtypes.

The univariate analysis identified nodal metastasis (HR 5.19, 95% CI 1.48–18.2, P = 0.010) and SUVmax (HR 1.11, 95% CI 1.05–1.18, P < 0.005) as factors potentially associated with RFS (Table 5). Subsequently, in the multivariate analysis model included clinically important subtypes and the presence or absence of chemotherapy, in addition to SUVmax and N stage, which showed statistical significance in the univariate analysis. SUVmax (HR 1.13, 95% CI 1.05–1.23, P < 0.005), nodal metastasis (HR 2.62, 95% CI 1.57–4.35, P < 0.005) and HR-HER- (HR 1.13, 95% CI 1.87–1.23, P < 0.005) were significantly associated with RFS (Table 5). In the multivariable Cox regression analysis, the interaction term between SUVmax and nodal status was not statistically significant (HR = 0.85, 95% CI: 0.05–13.3, P = 0.909), indicating that the effect of SUVmax does not significantly differ depending on the presence of nodal status. In addition, time-dependent AUCs at 1, 3, and 5 years were 0.74, 0.69, and 0.71, respectively. The multivariate Cox model including SUVmax and nodal status, Harrell’s concordance index (C-index) was calculated. The C-index for relapse-free survival prediction was 0.728, indicating acceptable discriminative ability. The VIF values for N factor, Subtype, SUVmax, and CT factor were 1.28, 1.36, 1.06, and 1.22, respectively, all of which are well below the conventional threshold of 5. Internal validation using 1000 bootstrap resamples was performed to assess the stability of the Cox proportional hazards model. The results indicated a 95% confidence interval of 0.171–1.443. In the context of overall survival, the univariate analysis identified SUVmax (HR 1.14, 95% CI 1.05–1.23 P < 0.005) as factors potentially associated with OS. In the multivariate analysis model used the same factors of RFS analysis. SUVmax (HR 1.11, 95% CI 1.00–1.24, P = 0.040) demonstrated a significant association with OS (Table 6).