This retrospective case-control study was approved by the Institutional Review Board (IRB no. CE21406B) of Taichung Veterans General Hospital (TCVGH; Taichung, Taiwan); all data were retrospectively collected from the health information system of TCVGH under this approved protocol The inclusion criteria for the study population were: i) Patients with early-stage breast cancer treated with breast reconstruction immediately after mastectomy between 1 January 2011 and 31 December 2018, and ii) completion of the chemotherapy treatment course in an adjuvant or neoadjuvant setting; breast cancer diagnosed at clinical stages I and IIIB was considered early-stage. Patients who did not meet the inclusion criteria were excluded. Data of 139 women with breast cancer met the inclusion criteria. The clinical characteristics and treatment characteristics of the study population were extracted from the patients' medical records or the cancer registry database.

A post hoc power analysis for two independent groups with dichotomous outcomes was used to evaluate the statistical power of the population in the current study. It was assumed that the probability of type I error (α) was 0.05, the population size of the current study cohort was 136; the sample size of neoadjuvant cases was 64, with a mortality rate of 23.4%, and the sample size of adjuvant controls was 75, with a mortality rate of 1.3%. The case-control sample size ratio was 1.172 and the absolute difference between the two groups was 0.221. With the power estimation formula, the estimated critical Z value for the assumed α was ~2.122 (Φ), equal to the power of 0.983, indicating that the assumed sample size may lead to results with 99.5% statistical power.

For the matched case and control cohorts, there was an assumed sample size of 37 for both the neoadjuvant cases and adjuvant controls, with mortality rates of 14 and 0%, respectively; the case-control sample size ratio was 1 and an absolute between-group difference was 0.14. With the power estimation formula, Φ was ~0.416, equal to the power of 0.661, indicating the assumed sample size may lead to results with 66.1% statistical power. In other words, the findings of the matched cohort may only provide conservative inference.

Several clinical characteristics were considered, including age, clinical stage and molecular subtype at diagnosis. The molecular subtypes of all patients were classified according to the results of standard histological examinations, including microscopic analysis with standard immunohistochemistry (IHC) for estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER2), which was performed on tissues from the primary tumor excised during mastectomy. Pathohistological testing was conducted in the Department of Pathology and Laboratory Medicine at TCVGH. Cases with + (mild), ++ (moderate) or +++ (strong) IHC results (according to the intensity of staining) for ER and PR were defined as ER⁺ and PR⁺ cases, respectively, whereas the remaining patients were defined as ER⁻ and PR⁻. Moreover, cases with +++ IHC result for HER2 were defined as HER2⁺ cases, whereas those with - or + IHC results for HER2 were defined as HER2⁻cases. In cases with ++ IHC results for HER2, fluorescence in situ hybridization (FISH) was performed using VENTANA HER2 Dual ISH DNA Probe Cocktail assay to analyze the HER2-neu gene amplification status. Subsequently, cases with ++ IHC and + FISH results for HER2 were defined as HER2⁺, whereas the remaining cases were defined as HER2⁻. In accordance with the classification system for molecular subtypes presented in the St. Gallen International Expert Consensus on the Primary Therapy of Early Breast Cancer 2013 (21), breast cancer was classified as luminal A (ER⁺/PR⁺ and HER2⁻), luminal B (ER⁺/PR⁺ and HER2⁺), HER2-enriched (ER⁻/PR⁻ and HER2⁺) or triple-negative (ER⁻/PR⁻ and HER2⁻).

Overall survival (OS) was considered a primary endpoint; patients who succumbed during the within-study follow-up duration were considered left-censored, whereas the remaining patients were considered right-censored. The follow-up interval was between the date of initial diagnosis and the date of censoring.

The treatment characteristics of the study population were classified according to the treatment type: Chemotherapy, radiotherapy, hormone therapy or targeted therapy (e.g. anti-HER2 therapy). In all patients, the treatment selection was mainly dependent on the individual clinical indications (i.e., age, tumor size, lymph node invasion and molecular subtype at diagnosis), as suggested by the NCCN guidelines for breast cancer treatment (1). The chemotherapy drugs used in the current settings included an anthracycline (epirubicin) with an alkylating agent (cyclophosphamide) followed by a taxane (paclitaxel or docetaxel). An antineoplastic agent (carboplatin) was used for patients with triple-negative breast cancer (TNBC) or HER2-enriched breast cancer. Most patients with T3, T4 or lymph node metastasis received intensity-modulated radiotherapy, whereas a few patients were treated with external beam radiation therapy. The hormone therapy agents used in premenopausal and postmenopausal patients were tamoxifen and an aromatase inhibitor (letrozole or anastrozole), respectively. Dual-blockade anti-HER2 agents (i.e. pertuzumab and trastuzumab) were used for HER2-enriched patients. Here, pertuzumab was administered at 840 mg in the first cycle and then at 420 mg in the following cycles, whereas trastuzumab was administered at 8 mg/kg in the first cycle and then at 6 mg/kg in the following cycles. All HER2-enriched patients received complete dual-blockade or single-blockade therapy for one year.

The overall study population that underwent immediate breast reconstruction after mastectomy was considered an unmatched cohort, comprising 75 patients who received adjuvant chemotherapy and 64 control patients who received neoadjuvant chemotherapy. The treatment responses after neoadjuvant chemotherapy were reported. Furthermore, age group, clinical stage and molecular subtype at diagnosis were used as matching criteria to generate a well-balanced case-control subgroup matched at a 1:1 ratio, to further reduce the bias and clarify the feasibility role of neoadjuvant and adjuvant chemotherapy in the study population. In this study, 37 neoadjuvant cases were well matched with 37 adjuvant controls, which was used as a matched cohort (n=74).

The clinical characteristics, treatment characteristics and OS rates of the study population are presented the mean ± SD or as frequencies and percentages. All analyses were performed using R (version 4.1.2; R core team, 2021). The distribution difference of baseline characteristics between the cases and controls was estimated using an independent two-sample Student's t-test, χ² test or Fisher's exact test. The survival rates and the correspond 95% confidence interval (CI) of different subgroups were computed using Kaplan-Meier estimators, and OS was compared between the subgroups using the log-rank test. A post-hoc pairwise log-rank test were performed for multiple comparison correction. All P-values were two-sided, and P<0.05 was considered to indicate a statistically significant difference.

Table I summarizes the clinicopathological characteristics, treatment and OS of the unmatched cohort. The mean age at diagnosis of the unmatched cohort was 43.0±8.6 years; 31 (22.3%) of the patients were aged ≥50 and 108 (77.7%) were <50 years. The mean ages at diagnosis of the adjuvant controls and neoadjuvant cases were 44.1±9.2 and 41.8±7.8 years, respectively, and the differences between the adjuvant controls and neoadjuvant cases were non-significant (P=0.117). Similar ratios were obtained for the age group distribution between the adjuvant controls and neoadjuvant cases. However, the clinical stages at diagnosis differed significantly between the adjuvant controls and neoadjuvant cases. For example, significantly more neoadjuvant cases were at clinical stages II and III compare with the adjuvant controls (79.7 vs. 45.3% and 15.6 vs. 1.3%, respectively; both P<0.001). Moreover, more neoadjuvant cases received radiotherapy compared with the adjuvant controls (53.1 vs. 32.0%; P=0.012). According to the clinical tumor staging, the tumor sizes ranged between 2 and 5 cm; most patients were at stages T1 (n=51; 36.7%) or T2 (n=76; 54.7%), followed by those at T3 (n=8; 5.8%). One (0.7%) patient in the adjuvant group was at TisN1; the remaining 3 (2.2%) patients were neoadjuvant cases at stage T4 at diagnosis. However, their tumor size decreased after neoadjuvant chemotherapy to a pathological tumor stage of T3, T4b or T1a. In total, 19 (13.7%) HER2-enriched breast cancer, 87 (62.6%) luminal A breast cancer, 26 (18.7%) luminal B breast cancer and 7 (5.0%) TNBC cases were identified in the unmatched cohort; the adjuvant controls and neoadjuvant cases demonstrated a similar molecular subtype distribution (P=0.453). The proportions of neoadjuvant cases and adjuvant controls receiving hormone and targeted therapy demonstrated a non-significant difference. In addition, 10 (15.6%) of the 64 neoadjuvant cases reached pathologic complete response (pCR) after neoadjuvant chemotherapy. However, 16 (11.5%) patients succumbed during follow-up; this proportion was higher among neoadjuvant cases compared with adjuvant controls (23.4 vs. 1.3%; P<0.001).

Fig. 1 presents the Kaplan-Meier plots for OS of the different subgroups in the unmatched cohort. The OS rates (95% CI) of the neoadjuvant cases and adjuvant controls were 50 (28.5-87.7) and 98.5% (95.5-100), respectively (Fig. 1A), with the neoadjuvant cases exhibiting significantly poorer OS compared with the adjuvant controls (P<0.001). Among the neoadjuvant cases, the OS rate was higher in patients who achieved pCR compared with those who did not [100 vs. 42.9% (95% CI, 21.4-86.1); P=0.072; Fig. 1B].

Fig. 1C and D present a comparison of the OS according to different molecular subtypes among the adjuvant controls and neoadjuvant cases, respectively. The difference in OS between the adjuvant controls with different molecular subtypes was significant (P<0.001; Fig. 1C). The survival curves of individuals with HER2-enriched, luminal A and luminal B breast cancer overlapped because they had similar survival outcomes; their OS rates were all 100%, whereas the OS rate in individuals with TNBC was 66.7% (95% CI, 30–100). However, the post-hoc pairwise comparison results revealed that individuals with TNBC had significantly poorer OS only when compared with those with luminal A breast cancer (P<0.001). By contrast, the differences in the neoadjuvant cases with different molecular subtypes were not significant (P=0.029; Fig. 1D). The discrepancy between neoadjuvant cases with TNBC and those with other subtypes was large in the post-hoc pairwise comparison: The differences between neoadjuvant cases with TNBC (40.0% OS; 95% CI, 13.7-100) and those with HER2-enriched (83.3% OS; 95% CI, 58.3-100; P=0.560), luminal A (44.0% OS; 95% CI, 21.0-92.3; P=0.300) and luminal B (84.4% OS; 95% CI, 66.6-100; P=0.300) breast cancer were not significant. Taken together, both the neoadjuvant cases and adjuvant controls with TNBC demonstrated poor OS, whereas neoadjuvant therapy led to no OS benefits in the patients with luminal A breast cancer (blue lines in Fig. 1C and D; the adjuvant controls achieved improved OS outcome compared to neoadjuvant cases). Moreover, the patients with HER2-enriched and luminal B breast cancer in both the neoadjuvant and adjuvant cohorts demonstrated similar OS improvement.

Fig. 2 presents a comparison of the OS between the adjuvant controls and neoadjuvant cases in the unmatched cohort according to their treatment characteristics. In both the adjuvant and neoadjuvant settings, the combined use of chemotherapy either with radiotherapy [95.7% (95% CI, 87.7-100) vs. 54.4%, (95% CI, 33.4-88.7), respectively; P=0.058; Fig. 2A] or with targeted therapy [91.7% (95% CI, 77.3-100) vs. 40.7% (95% CI, 10.0-88.7), respectively; P=0.314; Fig. 2C] led to different OS outcomes which were not significant. The neoadjuvant cases demonstrated significantly poorer OS than did the adjuvant controls in the subgroups without radiotherapy (100 vs. 0%, respectively; P=0.001; Fig. 2B) and without therapy [91.7%, (95% CI, 77.3–100%) vs. 40.7% (95% CI, 29.1–100%); P<0.001; Fig. 2D].

The patients who received hormone therapy combined with neoadjuvant chemotherapy demonstrated significantly poorer OS than those who received hormone therapy combined with adjuvant chemotherapy [100 vs. 52.4%, respectively; 95% CI, 27–100%, P<0.001; Fig. 2E]. By contrast, no significant difference was identified between the patients who did not receive hormone therapy in the adjuvant control [90.9% (95% CI, 75.4–100%)] and neoadjuvant case [47.3% (95% CI, 20.1–100%) groups (P=0.171; Fig. 2F)].

To further understand the survival differences between the matched neoadjuvant cases and adjuvant controls, an age-matched, clinical stage-matched and molecular subtype-matched cohort was generated comprising 74 patients (n=37 patients/group). Table II presents the distribution of the clinicopathological characteristics, treatment characteristics and OS rates in the matched cohort. The mean age at diagnosis in the matched cohort was 42.0±8.4 years; 60 (81.1%) patients in this cohort were <50 years old. Because one adjuvant control at clinical stage III did not match with any of the neoadjuvant cases based on the aforementioned criteria, the matched cohort did not include patients at clinical stage III. The matched cohort contained 6 (8.1%) clinical stage I and 68 (91.9%) clinical stage II cases. The distribution of molecular subtypes was similar to that of the unmatched cohort; thus, the treatment characteristics also demonstrated a similar distribution between the matched neoadjuvant cases and adjuvant controls. In total, 5 (13.5%) of the 37 matched cases achieved pCR after neoadjuvant chemotherapy. However, 5 (13.5%) patients died during the follow-up duration; all of these patients were matched neoadjuvant cases (P=0.054) not in receipt of pCR.

Fig. 3 presents the Kaplan-Meier plots for the OS of the different subgroups in the matched cohort. The neoadjuvant cases demonstrated significantly poorer OS compared with the adjuvant controls [63.3% (95% CI, 37.5-100) vs. 100%, respectively; P=0.044; Fig. 3A]. However, in terms of treatment characteristics and molecular subtypes, the neoadjuvant cases and adjuvant controls demonstrated differences which were not significant (Fig. 3B-D). In addition, similar survival curves were noted in the population with poor OS, as shown in Fig. 3A (blue line, matched neoadjuvant cases), B (red line, non-pCR matched neoadjuvant cases) and D (blue line, luminal A matched neoadjuvant cases). Mortality events were also only noted in the neoadjuvant cases with luminal A breast cancer [OS rate=71.8% (95% CI, 49.3–100%); Fig. 3D, blue line].

As shown in Fig. 4, the matched neoadjuvant cases and adjuvant controls demonstrated similar survival outcomes for those without radiotherapy (P=1.000; Fig. 4B) and hormone therapy (P=0.480; Fig. 4F). While the matched adjuvant controls with radiotherapy (P=0.171; Fig. 4A) and hormone therapy (P=0.074; Fig. 4E) showed improved survival outcomes compared with matched neoadjuvant cases. However, the matched neoadjuvant cases who did not receive targeted therapy [OS rate=60.4%; 95% CI, 34.8-100] demonstrated significantly poorer OS than the matched adjuvant controls who did not receive targeted therapy (OS rate=100%; 95% CI; P=0.044; Fig. 4D). However, the combined use of targeted therapy and chemotherapy led to the same favorable OS in both the matched adjuvant controls and neoadjuvant cases (P=1.000; Fig. 4C).