This study included 113 nonagenarian Caucasian patients living in Italy, of whom only one patient was male. Of these patients, 43 (38.1%) underwent TS, 34 (30.1%) underwent CS, and 36 (31.9%) did not undergo surgery (and represent the NS group).

The median age of the entire cohort was 93 years (range 90–99). The most common comorbidities were hypertension (66.4%), osteoporosis (29.2%), diabetes (24.8%), and atrial fibrillation (23.0%) [Table 1]. Approximately half of the patients presented with a good clinical condition, as indicated by Karnofsky scores of 80–100 (54.9%), ECOG scores of 0–1 (54.9%), and AN-CCI scores of 0–1 (48.6%). Cognitive function was normal in 54.9% of patients, and 74.3% exhibited normal humoral status according to the GDS.

Five patients presented with bilateral tumors. A palpable mass was the clinical presentation in 115/118 cases (97.5%), with 28.8% also showing visible findings and 22.0% showing palpable axillary nodes. Most tumors were invasive carcinomas of no special type (78.8%) and were categorized as Luminal A (84.7%). Neither ductal carcinoma in situ nor triple-negative BC cases were recorded. The global median tumor size was 27 mm, with significant differences among the groups (28 mm in the TS group, 21 mm in the AS group, and 31 mm in the NS group; p < 0.001). At presentation, half of the cases (50.0%) were classified as T2 neoplasia, while T4 accounted for 18.6% of cases. Lymph node metastases were detected in 22.0% of cases (Table 1).

Primary systemic therapy (PST), always hormone-based, was administered to 48/113 patients (42.5%), of whom 12/48 (25.0%) underwent subsequent surgery, while 36 (31.9% of the entire cohort) did not (Table 2).

HT as a definitive treatment was delivered to 31.9% of patients, with a median duration of 24 months, predominantly using aromatase inhibitors (AIs; 86.1%). Among these patients, 30.6% had to switch type of HT due to adverse effects, most commonly arthralgia. Partial clinical response was documented in 66.7% of patients, while two patients achieved complete clinical response; disease progression was recorded in 22.2% of patients, while two patients showed stable disease. Neoadjuvant HT was administered to 15.6% of the surgical patients, mainly using AIs (91.7%), for a median duration of 19 months. This approach was chosen for patients who were initially non-surgical candidates due to transient comorbidities, those who needed a reduction in BC dimensions to become operable, or for patients who initially refused surgery. Of these patients, 66.7% showed partial response, whereas 33.3% experienced disease progression. The following surgical treatment was TS and CS for 8 and 4 of these patients, respectively.

Chemotherapy was not administered to any patients, neither in the neoadjuvant setting nor as part of an exclusive medical treatment regimen. This decision was mainly due to the frailty of this specific age group of patients. After multidisciplinary discussion, it was determined that the risks outweighed the potential benefits, even in cases of human epidermal growth factor receptor 2-positive tumors, where trastuzumab would have posed a risk of severe cardiotoxicity, especially in patients already suffering from heart disease.

Seventy-seven patients (68.1% of the entire cohort, corresponding to 79 tumors) underwent surgery, of whom 65 (84.4%) underwent upfront surgery. Mastectomy was performed in 39.2% of tumors, while 60.8% underwent WLE. Mastectomy was the most common procedure in the TS group (27/44, 61.4%), while WLE was preferred in the CS group (31/35, 88.6%; p < 0.001). Axillary surgery was only performed in the TS group, with SLNB performed in 30/44 cN0 cases (68.2%) and ALND in all 12 cN+ patients. In two cases, sentinel lymph nodes were positive and ALND was consequently performed.

The overall positive and close margin rates were 11.4% and 13.9%, respectively, and no significant differences were observed between the two surgical groups (p = 0.151). Re-operation for margin clearance was performed in three TS selected cases with positive margins, while no close margin was considered for additional surgery.

The overall rate of patients with at least one postoperative complication was 51.9%, with breast seroma being the most frequent (22.8% of all procedures), occurring after 41.9% of mastectomies and 10.4% of WLEs (p < 0.001). The TS group experienced a significantly higher postoperative complication rate, including minor complications, than the CS group (74.4% vs. 23.5%; p < 0.001), with a reported axillary lymphocele rate of 31.8%, mainly after ALND (71.4% vs. 13.3% after SLNB; p < 0.01). Mild arm lymphedema was recorded after ALND in 18.2% of TS procedures. All postoperative complications were conservatively managed and no surgical re-intervention was required. The 30-day postoperative mortality rate was zero.

Adjuvant HT was administered to 72.7% of surgical patients, with a median duration of 34 months; in 7.1% of cases, adjuvant HT was suspended due to adverse effects. AIs were used in 91.1% of patients. None of the patients received adjuvant chemotherapy. Adjuvant radiotherapy was recommended for 21/77 surgical patients (27.3%), but only 9.1% adhered to the treatment.

Among all surgical patients, the RR was 10.4%, whereas the NS group experienced an overall (local and/or distant) disease progression rate of 22.2% (p = 0.049). Recurrence management consisted of additional surgery in the TS group only, performed in three of five recorded recurrences (60.0%). Specifically, this included two WLEs in patients initially staged IIB who subsequently developed ipsilateral BC recurrence, and one ALND in a stage IIIB patient with axillary recurrence.

HT was the treatment used for 80% of TS group recurrences and 33.3% of CS group recurrences. Additionally, 50% of NS patients experiencing disease progression underwent drug switching during the course of HT. Supportive care as the sole recurrence treatment was provided to 66.7% and 37.5% of CS and NS patients, respectively.

Over a median estimated follow-up of 77 months (real median follow-up of 42 months and a loss to follow-up rate of 7.4%), 54.9% (62/113) of the cohort died, with only 12.9% of these deaths caused by BC. BC-related mortality was significantly higher in the NS group than in the TS and CS groups (25% vs. 0% and 7.1%, respectively; p = 0.01). The leading causes of non-BC-related deaths were heart failure (43.5%), respiratory failure (19.4%), and cancer at other sites (12.9%).

The median global OS was 52 months, with median DFS and median PFS not reached in any of the three groups (Fig. 1). Surgical patients experienced a significantly higher OS than NS patients (p = 0.04) (Fig. 2a). When considering non-BC-related deaths as a competing risk, surgical patients showed a significantly lower BC-related mortality rate (p = 0.002), with no notable difference in mortality from other causes between the groups (p = 0.8) (Fig. 2b). No significant differences in OS or DFS were observed between the TS and CS groups (p = 0.6 and p = 0.8, respectively) (Fig. 3). A specific survival analysis revealed no significant difference in OS (medians of 66 and 51 months, respectively) and DFS between clinically node-negative (cN0) patients at diagnosis undergoing TS or CS (p = 0.7 and p = 0.8, respectively) (Fig. 4).

a Overall survival in the whole population; b disease-free survival in surgical patients; and c progression-free survival in non-surgical patients

a Overall survival and b cumulative incidence of death in surgical versus non-surgical patients. mo months, BC breast cancer

a Overall survival and b disease-free survival in all surgical patients: traditional surgery versus current-standard surgery

a Overall survival and b disease-free survival in cN0 surgical patients: traditional surgery versus current-standard surgery

A Cox Model analysis, including all the surgical patients, revealed that the Charlson score at diagnosis was the only significant predictor of poorer OS (HR 1.24, 95% CI 1.03–1.48; p = 0.022). Conversely, age, tumor stage, and, most notably, the type of surgery did not significantly affect the OS (Table 3).