A Real-World Efficacy and Safety of KEYNOTE-522 Regimen in Patients With Early Triple-Negative Breast Cancer

Abstract

Purpose

Based on the KEYNOTE-522 study, neoadjuvant pembrolizumab plus chemotherapy has become the standard treatment for early-stage triple-negative breast cancer (TNBC). This study evaluated the real-world efficacy, safety, and predictors of pathologic complete response (pCR) in Korean patients.

Methods

We conducted a retrospective cohort study of 174 patients with early-stage TNBC who received the KEYNOTE-522 regimen (neoadjuvant pembrolizumab plus paclitaxel and carboplatin, followed by doxorubicin and cyclophosphamide) at a tertiary cancer center between August 2022 and July 2024. We assessed the primary endpoints, including pCR rate and event-free survival (EFS). We performed univariable and multivariable logistic regression analyses to identify independent predictors of pCR.

Results

The median patient age was 50 years (range, 24–74 years). The clinical stages were II and III in 79.3% and 20.1% of patients, respectively, and 10.9% had clinical N3 disease. The overall pCR rate was 62.1%, and the N3 subgroup had a pCR rate of 47.4%. On multivariable analysis, high baseline Ki-67 expression (≥ median, 75%) was significantly associated with pCR (odds ratio, 2.84; 95% confidence interval, 1.45 to 5.66; p = 0.002). At a median follow-up of 18.4 months, the 12-month EFS rate was 97.4%, with significantly superior outcomes observed in patients who achieved pCR compared with those who did not achieve pCR (100% vs. 93.1%, p = 0.007). The treatment completion rate was 92.0%, and immune-related adverse events occurred in 13.8% of patients.

Conclusion

In this real-world analysis of one of the largest Asian cohorts of patients with early-stage TNBC treated with neoadjuvant pembrolizumab, the KEYNOTE-522 regimen demonstrated substantial efficacy and manageable toxicity, consistent with the original trial findings.

INTRODUCTION

Triple-negative breast cancer (TNBC) is characterized by higher recurrence and mortality rates than other breast cancer subtypes [1]. Perioperative chemotherapy plays a central role in the management of early-stage TNBC [2,3], and neoadjuvant chemotherapy is widely used as the standard treatment for patients with stage T1cN0 or higher disease. Neoadjuvant chemotherapy enables tumor downsizing while facilitating assessment of early tumor response based on pathologic responses [4,5]. The achievement of a pathologic complete response (pCR) has been consistently associated with improved long-term outcomes among patients with early-stage TNBC treated with neoadjuvant chemotherapy, making pCR a vital endpoint for evaluating treatment efficacy [6,7].

The unique tumor microenvironment of TNBC, characterized by higher tumor-infiltrating lymphocytes (TILs) and increased expression of programmed death ligand 1 compared with other breast cancer subtypes, provides a strong biological rationale for integration of immunotherapy [8,9,10]. Recently, the KEYNOTE-522 trial demonstrated that adding pembrolizumab to neoadjuvant chemotherapy significantly improved pCR rates, event-free survival (EFS), and overall survival in patients with stage II–III TNBC, establishing this immunochemotherapy combination as a new standard of care [2,11]. However, Asian patients comprised only a minority (18%) of the KEYNOTE-522 study population, in which 136 patients were randomized and treated with neoadjuvant pembrolizumab [12]. Moreover, real-world evidence remains limited regarding the implementation of this intensive immunochemotherapy regimen in this population. This is crucial because neoadjuvant pembrolizumab plus chemotherapy showed a trend toward greater benefits in the Asian subgroup of the KEYNOTE-522 trial population in terms of EFS and pCR rates [12]. Furthermore, the Korean subgroup exhibited a trend toward an even more pronounced benefit [13]. Although these findings are exploratory, the magnitude of the benefits observed in Asian patients raises the need to assess the real-world effectiveness in this population.

In this study, we analyzed the real-world outcomes of patients with early-stage TNBC who received neoadjuvant pembrolizumab plus chemotherapy at a tertiary referral center in the Republic of Korea. We aimed to assess the efficacy and safety of this regimen in routine clinical practice, identify the factors associated with treatment response, and provide insights into optimal treatment strategies for Asian populations in real-world settings.

METHODS

Study design and population

We retrospectively identified and included patients with early-stage TNBC who received neoadjuvant treatment with pembrolizumab plus chemotherapy (paclitaxel and carboplatin, followed by doxorubicin and cyclophosphamide) at Asan Medical Center, a tertiary referral center in Seoul, Republic of Korea, between August 2022 and July 2024 [11]. The treatment indications followed the inclusion criteria of the KEYNOTE-522 trial, except that patients with N3 disease were included. Patients were excluded from the analysis if they had concurrent malignant disease, multiple breast primaries, or evidence of a coexisting non-TNBC subtype. All data were extracted from the electronic medical records.

This retrospective study was approved by the Institutional Review Board (IRB) of Asan Medical Center (IRB No. 2024-1535). All the research procedures adhered to the principles outlined in the Declaration of Helsinki and followed the ethical guidelines established by our institutional research committee. Given the retrospective design of this study, the IRB waived the need for informed consent.

Treatment

In the neoadjuvant phase, patients received pembrolizumab 200 mg intravenously every 3 weeks (cycles 1–8) in combination with chemotherapy consisting of paclitaxel 80 mg/m² weekly plus carboplatin (area under the curve five every 3 weeks) for cycles 1–4, followed by doxorubicin 60 mg/m² every 3 weeks and cyclophosphamide 600 mg/m² every 3 weeks for cycles 5–8. In the adjuvant phase, the treatment was administered at the discretion of the treating physicians.

Assessments and endpoints

Breast cancer was classified as TNBC when tumor cells showed < 1% nuclear staining for estrogen receptor and progesterone receptor by immunohistochemistry and showed no overexpression of human epidermal growth factor receptor 2 (HER2), defined as immunohistochemistry (IHC) 0 or IHC 1+ or 2+ without gene amplification, according to the American Society of Clinical Oncology/College of American Pathologists guidelines [14,15]. TILs were evaluated by expert pathologists according to the recommendations of the International TILs Working Group [16]. Cutoff values of 30% and 50% for TILs were applied in this study based on previously published studies that reported their prognostic significance [17]. The pCR status (ypT0/Tis ypN0) and residual cancer burden (RCB) class were assessed by a dedicated breast pathologist using established criteria, including primary tumor bed dimensions, overall cellularity, proportion of in situ carcinoma, number of positive lymph nodes, and diameter of the largest metastatic lesion [18]. Patients’ menopausal status was retrieved from electronic medical records based on the treating physician’s assessment.

Before initiating neoadjuvant systemic therapy, patients underwent mammography; breast ultrasound; breast magnetic resonance imaging (MRI); computed tomography of the chest, abdomen, and pelvis; and bone scanning. After cycles four and eight, mammography, breast ultrasound, and breast MRI were performed. The time from the start of neoadjuvant systemic therapy to the first occurrence of disease progression that rendered the patient ineligible for definitive surgery, invasive disease recurrence (local, regional, or distant), second primary invasive cancer, or death from any cause was defined as EFS [19]. Adverse events following chemotherapy were graded according to the Common Terminology Criteria for Adverse Events, version 5.0 [20]. Immune-related adverse events (irAEs) were assessed throughout the neoadjuvant and adjuvant treatment phases.

Statistical analysis

For comparisons of baseline characteristics between groups, continuous variables were analyzed using Student’s t-test or the Mann–Whitney U test, as appropriate, and categorical variables were compared using the chi-square test or Fisher’s exact test, depending on the expected cell counts. The Kaplan–Meier method was used to estimate EFS, and between-group comparisons were performed using the log-rank test. Clinicopathologic factors associated with pCR were evaluated using univariable and multivariable logistic regression analyses, with the results reported as odds ratios (ORs) and their corresponding 95% confidence intervals (CIs). Firth’s penalized logistic regression was used for the univariable and multivariable models. Statistical significance was defined as a two-sided p-value of < 0.05. All statistical analyses were performed using R software (version 4.4.1; R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

Patient characteristics

A total of 181 patients with early-stage TNBC who received neoadjuvant pembrolizumab plus chemotherapy during the study period were identified. Among these, six were excluded owing to loss to follow-up during neoadjuvant chemotherapy as they sought further care at other hospitals closer to their residences. One additional patient with concurrent hormone-positive breast cancer was excluded. Thus, 174 patients were included in the final analysis.

The median age at diagnosis was 50 years (range, 24–74 years). All patients were females. The clinical stages at diagnosis were I, II, and III in one (0.6%), 138 (79.3%), and 35 (20.1%) patients, respectively. Among the 149 (85.6%) patients tested for germline BRCA mutations, seven (4.0%) and two (1.1%) harbored germline BRCA1 and BRCA2 mutations, respectively. The median TIL level was 15% (interquartile range [IQR], 5 to 35), and the median Ki-67 index was 75% (IQR, 54 to 85) (Table 1).

Table 1. Baseline characteristics.

Characteristics		All patients (n = 174)
Age at diagnosis (median, range)		50 (24–74)
	< 50 yr	87 (50.0)
	≥ 50 yr	87 (50.0)
Menopausal status
	Premenopausal	97 (55.7)
	Postmenopausal	77 (44.3)
Germline BRCA status
	No mutation	140 (80.5)
	gBRCA1	7 (4.0)
	gBRCA2	2 (1.1)
	Unknown (not tested)	25 (14.4)
Baseline histopathology
	Ductal	170 (97.7)
	Lobular	0 (0)
	Others	4 (2.3)
Baseline histologic grade
	G1–2	37 (21.3)
	G3	137 (78.7)
Baseline nuclear grade
	G1–2	36 (20.7)
	G3	138 (79.3)
Baseline ER Allred score
	0	164 (94.3)
	1	0 (0)
	2	10 (5.7)
Baseline PR Allred score
	0	173 (99.4)
	1	0 (0)
	2	1 (0.6)
Baseline HER2 IHC
	Null	12 (6.9)
	Ultralow	107 (61.5)
	Low (1+ or 2+ and SISH-negative)	55 (31.6)
Baseline Ki67 (median, Q1–Q3)		75 (54–85)
	< 20	5 (2.9)
	20–49	37 (21.3)
	50–74	40 (23.0)
	(median)	92 (52.9)
Baseline TIL (median, Q1–Q3)		15 (5–35)
	< 30	119 (68.4)
	30–49	26 (14.9)
	≥ 50	29 (16.7)
cT stage
	cT1	8 (4.6)
	cT2	134 (77.0)
	cT3	24 (13.8)
	cT4	8 (4.6)
cN stage
	cN0	98 (56.3)
	cN1	48 (27.6)
	cN2	9 (5.2)
	cN3	19 (10.9)
Clinical stage at diagnosis
	I	1 (0.6)
	II	138 (79.3)
	III	35 (20.1)

Values are presented as number (%).

ER = estrogen receptor; PR = progesterone receptor; HER2 = human epidermal growth factor receptor 2; IHC = immunohistochemistry; SISH = silver-enhanced in situ hybridization; TIL = tumor-infiltrating lymphocyte.

Pathologic response to neoadjuvant treatment and characteristics of residual tumors

Among the 174 included patients, 108 (62.1%) achieved pCR, whereas 66 (37.9%) had residual disease in the surgical specimen. Among those with residual disease, 23 (13.2%) were classified as RCB-I, 33 (19.0%) as RCB-II, and 10 (5.7%) as RCB-III (Table 2).

Table 2. Pathologic and clinical outcomes after neoadjuvant therapy.

Clinical and pathologic response		All patients (n = 174)
Residual cancer burden
	pCR	108 (62.1)
	I	23 (13.2)
	II	33 (19.0)
	III	10 (5.7)
Miller-Payne grade
	1	3 (1.7)
	2	5 (2.9)
	3	24 (13.8)
	4	23 (13.2)
	5	118 (67.8)
Lymphovascular invasion
	No	159 (91.4)
	Yes	14 (8.0)
ypT stage
	ypT0	119 (68.4)
	ypT1	31 (17.8)
	ypT2	18 (10.3)
	ypT3	6 (3.4)
	ypT4	0 (0)
ypN stage
	ypN0	147 (84.5)
	ypN1	17 (9.8)
	ypN2	7 (4.0)
	ypN3	3 (1.7)
ypStage
	0	109 (62.6)
	I	28 (16.1)
	II	25 (14.4)
	III	12 (6.9)

Values are presented as number (%).

pCR = pathologic complete response.

Among the 66 patients with residual disease, postoperative residual tumors exhibited a median decrease in Ki-67 of −28.2% (IQR, −58.2% to −0.1%), and TILs of 0% (IQR, −10.0% to 0.0%). The decrease in the Ki-67 in the post-treatment samples was numerically larger in the lower RCB subgroup; however, the inter-group differences were not statistically significant (Kruskal–Wallis p = 0.537). The median change in TIL did not differ according to RCB (Kruskal–Wallis p = 0.877). Residual disease demonstrated hormone receptor positivity in six patients (9.1%), and the distribution of HER2 status was similar to that of the baseline samples (Supplementary Figure 1). Supplementary Table 1 presents the pathologic and clinical outcomes of patients with residual disease following neoadjuvant therapy.

Subgroup of patients with N3 disease

Among the 19 patients (10.9%) with clinical N3 disease, the clinical staging at diagnosis was N3a, N3b, and N3c in two (10.5%), eleven (57.9%), and six (31.6%), respectively. The pCR rate in this group was 47.4% (n = 9). Among the 10 patients with residual disease (52.6%), the RCB classes were RCB-I, RCB-II, and RCB-III in two (10.5%), five (26.3%), and three (15.8%), respectively. Residual disease involved the breast in one patient (10.0%), the axilla in three (30.0%), and both in six (60.0%).

Predictors of pCR

The baseline characteristics did not differ significantly between patients who achieved pCR and those who did not, except that patients who achieved pCR were younger and exhibited higher baseline Ki-67 and TIL values (Supplementary Table 2).

Univariable and multivariable logistic regression analyses were conducted to assess the association between the clinicopathologic characteristics and pCR (Table 3). In univariable analysis, high Ki-67 (≥ median, 75%), high TIL (≥ 50%), and nuclear grade 3 were associated with higher odds of pCR. The clinical stage and germline BRCA status were not associated with pCR. In multivariable analysis, only a high baseline Ki-67 (≥ 75%) remained significant (OR, 2.84; 95% CI, 1.45 to 5.66; p = 0.002).

Table 3. Univariable and multivariable analyses for pathologic complete response.

Variable		Univariable OR (95% CI)	Univariable p-value	Multivariable OR (95% CI)	Multivariable p-value
Age
	< 65 yr	1 (Ref)	-	1 (Ref)	-
	≥ 65 yr	0.23 (0.05 to 0.77)	0.017	0.34 (0.08 to 1.26)	0.109
Germline BRCA status
	No mutation	1 (Ref)	-	1 (Ref)	-
	Mutation	1.89 (0.48 to 10.41)	0.376	1.96 (0.45 to 11.67)	0.387
	Unknown	0.93 (0.40 to 2.24)	0.868	0.90 (0.34 to 2.51)	0.832
Baseline histopathology
	Ductal	1 (Ref)	-	1 (Ref)	-
	Others	0.25 (0.02 to 1.58)	0.142	0.23 (0.02 to 1.82)	0.162
Baseline histologic grade
	G1–2	1 (Ref)	-	1 (Ref)	-
	G3	2.01 (0.97 to 4.18)	0.06	1.31 (0.01 to 42.88)	0.889
Baseline nuclear grade
	G1–2	1 (Ref)	-	1 (Ref)	-
	G3	2.15 (1.03 to 4.51)	0.041	1.15 (0.03 to 215.61)	0.942
Baseline HER2 IHC
	Null	1 (Ref)	-	1 (Ref)	-
	Ultralow	1.13 (0.33 to 3.61)	0.837	0.78 (0.20 to 2.76)	0.701
	Low	1.37 (0.38 to 4.69)	0.616	1.13 (0.29 to 4.22)	0.859
Baseline Ki67
	< Median (75%)	1 (Ref)	-	1 (Ref)	-
	≥ Median (75%)	3.26 (1.74 to 6.23)	< 0.001	2.84 (1.45 to 5.66)	0.002
Baseline TIL
	< 30%	1 (Ref)	-	1 (Ref)	-
	30%–49%	1.69 (0.71 to 4.30)	0.237	1.60 (0.60 to 4.57)	0.352
	≥ 50%	2.81 (1.15 to 7.77)	0.022	2.22 (0.82 to 6.76)	0.117
Clinical stage
	Stage II	1 (Ref)	-	1 (Ref)	-
	Stage III	0.58 (0.28 to 1.23)	0.157	0.49 (0.21 to 1.12)	0.092
Immune-related adverse events
	No	1 (Ref)	-	1 (Ref)	-
	Yes	1.52 (0.63 to 4.01)	0.363	1.47 (0.56 to 4.18)	0.436

OR = odds ratio; CI = confidence interval; HER2 = human epidermal growth factor receptor 2; IHC = immunohistochemistry; TIL = tumor-infiltrating lymphocyte.

Safety

The median number of treatment cycles was eight (range, 2–8), and 92.0% (n = 160) completed eight pre-planned cycles of neoadjuvant treatment. Among the 14 patients (8.0%) who discontinued treatment, eight (4.6%) discontinued treatment owing to toxicity (including seven immune-related events), four (2.3%) because of disease progression, and one because of the patient’s refusal. Toxicities that led to treatment discontinuation included adrenal insufficiency (n = 4, 2.3%) and hypothyroidism, severe skin reactions, pneumonitis, infection, and stroke (n = 1, 0.6% each). Dose delay and dose reduction were assessed for each chemotherapy agent, with the corresponding data provided in the Supplementary Table 3.

The incidences of irAEs and hematologic adverse events during treatment are summarized in Table 4. Overall, irAEs of any grade occurred in 25 patients (14.4%). Grade 3 or higher adverse events occurred in eight patients (4.6%). Hypothyroidism was the most frequently observed adverse event (n = 8, 4.6%), followed by adrenal insufficiency (n = 7, 4.0%). Pneumonitis occurred in one patient (0.6%, grade 3). Other irAEs, including hyperthyroidism, hepatitis, proctitis, arthralgia, and rash, were reported in one patient each, and all were grades 1–2.

Table 4. Hematologic adverse events and key immune-related adverse events.

Adverse events		Any grade	Grade ≥ 3
Immune-related adverse events		25 (14.4)	8 (4.6)
	Infusion reaction	1 (0.6)	0 (0.0)
	Hypothyroidism	8 (4.6)	2 (1.1)
	Hyperthyroidism	1 (0.6)	0 (0.0)
	Severe skin reaction	2 (1.1)	2 (1.1)
	Adrenal insufficiency	8 (4.6)	3 (1.7)
	Pneumonitis	1 (0.6)	1 (0.6)
	Hepatitis	1 (0.6)	0 (0.0)
	Proctitis	1 (0.6)	0 (0.0)
	Arthralgia	1 (0.6)	0 (0.0)
	Rash	1 (0.6)	0 (0.0)
Hematologic adverse event
	Anemia	127 (73.0)	26 (14.9)
	Thrombocytopenia	78 (44.8)	10 (5.7)
	Neutropenia	118 (67.8)	72 (43.4)
	Febrile neutropenia	22 (12.6)	22 (12.6)

Values are presented as number (%).

Regarding hematologic adverse events, grade ≥ 3 neutropenia occurred in 72 patients (43.4%), and febrile neutropenia occurred in 22 (12.6%). Information on the use of granulocyte colony-stimulating factors was not collected in this study, and their administration was determined at the discretion of the treating physician. No treatment-related fatalities were observed in this study.

EFS

At a median follow-up duration of 18.4 months (95% CI, 17.1 to 19.0), EFS events occurred in seven patients (4.0%): six distant recurrences and one local recurrence. The 12-month EFS rate in the overall population was 97.4% (95% CI, 94.9 to 99.9). The 24-month EFS rate was 93.9% (95% CI, 88.5 to 99.7) (Figure 1A).

Figure 1. Event-free survival according to pathologic response and residual cancer burden.

EFS in (A) total population, (B) by pCR, and (C) by RCB.

EFS = event-free survival; pCR = pathologic complete response; RCB = residual cancer burden.

A significant difference in EFS was observed according to pCR status; the 12-month EFS rates were 100% (95% CI, 100 to 100) among those with pCR, and 93.1% (95% CI, 86.9 to 99.9) among those with non-pCR (p = 0.007). The corresponding 24-month EFS rates were 96.3% (95% CI, 89.4 to 100) for pCR and 89.9% (95% CI, 81.5 to 99.2) for non-pCR (Figure 1B). Among those with residual disease, EFS differed by RCB: the 12-month EFS rates were 100% (95% CI, 100 to 100) for RCB-I, 93.3% (95% CI, 84.6 to 100) for RCB-II, and 71.4% (95% CI, 44.7 to 100) for RCB-III. The 24-month EFS rates were 92.9% (95% CI, 80.3 to 100) for RCB-I, 93.2% (95% CI, 84.6 to 100) for RCB-II, and 71.4% (95% CI, 44.7 to 100) for RCB-III (Figure 1C).

Patterns of adjuvant systemic therapy

Adjuvant systemic therapy was administered to 130 (74.7%) patients. In the overall population, pembrolizumab was the most commonly used treatment regimen (n = 93, 53.4%). Among the 108 patients who achieved pCR, 67 (62.0%) received pembrolizumab, and 41 (38.0%) did not receive any adjuvant therapy. Among the 66 patients who did not achieve pCR, 26 (39.4%) received pembrolizumab, 20 (30.3%) received capecitabine, and eight (12.1%) received both (Supplementary Table 4).

DISCUSSION

In our study of 174 Korean patients with early-stage TNBC treated with the neoadjuvant KEYNOTE-522 regimen, pembrolizumab plus chemotherapy achieved a pCR rate of 62.1%, which is consistent with the original trial results. Treatment was generally well tolerated, with 92.0% of the patients completing eight pre-planned cycles of treatment. To our knowledge, this is one of the largest real-world Asian cohort studies to date.

In the KEYNOTE-522 trial, Asian patients comprised 18% of the study population (n = 216). Among them, 136 received pembrolizumab plus chemotherapy and 80 received placebo plus chemotherapy across four Asian countries; approximately 40% of these patients were enrolled in Korea [12]. Our study provides real-world data from a large cohort of 174 Korean patients who received the same regimen in routine clinical practice, with baseline characteristics largely comparable to those reported in the original KEYNOTE-522 trial.

In addition, the pCR rate of 62.1% in our cohort was highly consistent with the 64.8% rate reported in the original trial; however, the distribution of RCB classes among patients with residual disease was comparable. In our cohort, RCB classes I, II, and III comprised 13.2%, 19.0%, and 5.7% of the cohort, respectively, consistent with the KEYNOTE-522 trial findings of 8.8%, 18.5%, and 5.1%, respectively [21]. Although the follow-up duration was limited, EFS differed significantly based on RCB status, which is consistent with previous studies [21,22,23]. Of the seven EFS events, six (85.7%) were distant metastases, underscoring the unmet need for effective therapies in patients with residual disease after neoadjuvant immunochemotherapy [24,25].

An additional distinction between our real-world cohort and the KEYNOTE-522 trial population is the inclusion of patients with clinical N3 disease who were ineligible for enrollment in the KEYNOTE-522 trial. In our study, 19 patients presented with cN3 disease, of whom 47.4% achieved pCR following neoadjuvant pembrolizumab-based therapy, which was notably lower than that of patients who met the KEYNOTE-522 inclusion criteria (63.9%). Furthermore, a higher proportion of patients in this subgroup had a substantial residual tumor burden, classified as RCB-II (26.3% vs. 18.1%) and RCB-III (15.8% vs. 4.5%).

The lower rates of pCR among patients with N3 disease are consistent with findings reported before the pembrolizumab era [26]. However, although limited by the small sample size, the pCR rate observed in our cohort was numerically higher than that in the historical data [26], and no patient with clinical N3 disease who achieved pCR experienced an EFS event. These findings support the need for further investigations into the potential benefits of pembrolizumab in this high-risk population.

Another important difference between our real-world cohort and the KEYNOTE-522 population was the choice of adjuvant therapy, which should be considered when interpreting survival outcomes. In the KEYNOTE-522 trial, all patients received nine cycles of adjuvant pembrolizumab. However, in clinical practice, continued pembrolizumab therapy after pCR may be limited by reimbursement constraints, and different adjuvant therapies may be selected based on the tumor response and clinical factors. Although these findings should be interpreted with caution, given the short follow-up duration, 38% of the patients who achieved pCR in our cohort did not receive any adjuvant systemic therapy; however, no EFS events were observed in this group. The results of the ongoing OptimICE-pCR trial are expected to further clarify the role of adjuvant therapy in patients who achieve pCR [27]. For patients with residual disease, alternative adjuvant strategies were frequently used, given the poorer prognosis associated with a higher residual tumor burden. In our cohort, only 39% of patients received adjuvant pembrolizumab, whereas the others were treated with alternative regimens, most commonly capecitabine (30.3%), based on the CREATE-X trial results, which were conducted before the pembrolizumab era. The 24-month EFS rates for RCB-II and RCB-III in our cohort (93.2% and 71.4%, respectively) were numerically higher than those reported in the KEYNOTE-522 RCB subgroup, which may reflect differences in adjuvant treatment patterns. Ongoing clinical trials are expected to inform further optimal adjuvant strategies for this setting [24,28,29].

In terms of safety, the overall incidence of irAEs in our study was notably lower than that reported in the KEYNOTE-522 trial. Specifically, irAEs of any grade occurred in 14.4% of patients, with grade ≥ 3 events in 4.6%, compared with 35.0% and 13.0%, respectively, in the pembrolizumab-chemotherapy arm of KEYNOTE-522 [2]. This discrepancy may partly reflect the less rigorous documentation of adverse events in routine practice than in clinical trials. In real-world clinical practice, irAEs are routinely monitored by treating physicians; however, surveillance is generally less stringent than in clinical trials. In our study population, thyroid function testing was performed in many patients, but not universally, and adrenal function testing was typically ordered only when symptoms developed. This may lead to the underdetection of subclinical diseases and result in a lower reported incidence of irAEs compared with clinical trials. Therefore, our findings are likely to reflect irAEs that were clinically significant enough to be identified and documented by treating physicians, rather than capturing all events.

However, even when considering these limitations, the treatment discontinuation rate was notably lower in our cohort, as 4.6% of patients discontinued treatment because of toxicity, compared with 23.3% in the KEYNOTE-522 trial [11]. Adverse events leading to treatment discontinuation were largely irAEs, most commonly adrenal insufficiency, highlighting the importance of early recognition and monitoring. Despite the considerable rates of grade ≥ 3 neutropenia (43.4%) and febrile neutropenia (12.6%), hematologic toxicities were generally manageable and did not lead to treatment discontinuation.

In the exploratory analyses for clinical predictors of pCR, high Ki-67 (≥ median, 75%) emerged as an independent predictor of pCR in multivariable analysis (2.84-fold increased odds). Although prior studies have demonstrated that Ki-67 thresholds ranging from > 20% to > 35% are predictive of pCR in cohorts treated with neoadjuvant chemotherapy [30,31], no previous study has specifically assessed the predictive role of Ki-67 in the context of neoadjuvant immunochemotherapy. The higher cutoff applied in our study (≥ 75%) likely reflects the need to identify highly proliferative tumors sensitive to intensive immunochemotherapy. However, the Ki-67 cutoff used in this study was arbitrary, based on the cohort median, and requires validation in larger cohorts. Additionally, the prognostic implications of high baseline Ki-67 levels warrant further investigation because high Ki-67 levels have been associated with poor survival in previous studies [32]. Additionally, high TIL levels, a known favorable prognostic factor in TNBC associated with improved survival and a higher likelihood of pCR [33], also demonstrated a favorable trend for pCR in our cohort. However, this association was not significant after adjusting for clinical variables, possibly because of the small number of patients with higher TIL levels. Although our study could not directly assess this, further studies are needed to clarify the predictive value of TILs in guiding the benefits of immunotherapy and optimizing the chemotherapy backbone in TNBC [34].

This study has several limitations beyond those inherent to its single-center retrospective design. The relatively short follow-up duration precludes a sufficient assessment of survival outcomes. Nevertheless, our results derived from one of the largest real-world Asian cohorts demonstrate the efficacy and feasibility of the KEYNOTE-522 regimen in routine clinical practice.

In conclusion, our real-world findings provide additional support for the routine implementation of neoadjuvant pembrolizumab plus chemotherapy in patients with early TNBC and valuable evidence to inform clinical practice.

SUPPLEMENTARY MATERIALS

Supplementary Table 1

Pathologic and clinical outcomes in patients with residual disease after neoadjuvant therapy

Supplementary Table 2

Baseline characteristics according to pathologic complete response status

Supplementary Table 3

Dose delay and dose reduction for each chemotherapy agent

Supplementary Table 4

Adjuvant systemic therapy

Supplementary Figure 1

Changes in Ki-67, tumor-infiltrating lymphocytes, estrogen receptor, progesterone receptor, and human epidermal growth factor receptor 2 detected by immunohistochemistry after neoadjuvant chemotherapy.