Pathological Complete Response and Safety of Neoadjuvant Pembrolizumab-Based Chemoimmunotherapy in Triple-Negative and Hormone Receptor-Low Breast Cancer: A Real-World Cohort Study

Abstract

Background

Neoadjuvant pembrolizumab-based chemoimmunotherapy is standard of care for early-stage triple-negative breast cancer (TNBC). Real-world data remain limited outside trial populations, particularly in the Middle East, and hormone receptor (HR)-low tumors (ER/PR 1–10%) are undercharacterised in this setting.

Methods

We conducted a retrospective cohort study of 149 patients with non-metastatic TNBC or HR-low breast cancer at King Hussein Cancer Center, Jordan (April 2021–September 2024). Primary endpoint was pCR (ypT0/Tis ypN0), analysed across all 149 treatment-initiated patients with the 7 who did not undergo surgery classified as non-pCR; Firth penalised logistic regression identified factors independently associated with pCR. Safety was assessed in all 149.

Results

The overall pCR rate was 59.1% (88/149). On Firth penalised multivariable logistic regression, three variables were independently associated with pCR: grade 3 histology (aOR 4.44, 95% CI 1.61–13.66, p = 0.004), pathogenic germline mutation (aOR 3.19, 95% CI 1.26–9.11, p = 0.014), and age younger than 40 years (aOR 2.82, 95% CI 1.20–7.10, p = 0.017). HR-low tumors (n = 15) achieved 66.7% pCR, comparable to HR-negative disease (58.2%). Immune-related adverse events occurred in 44 patients (29.5% of 149); grade ≥3 events in 9 (20.5% of irAE patients; 6.0% of the cohort); permanent immunotherapy discontinuation for treatment-limiting toxicity in 6 (13.6% of irAE patients; 4.0% of the cohort). Hospitalization occurred in 33.6%, predominantly due to febrile neutropenia. At a median follow-up of 27.3 months, 137 of 149 patients (91.9%) were alive and 136 (91.3%) were recurrence-free; all 13 recurrences were distant. Among surgical patients, disease-free survival was 96.6% in the pCR group versus 85.2% with residual disease (log-rank p = 0.011).

Conclusion

In this real-world Middle Eastern cohort, pembrolizumab-based chemoimmunotherapy achieved pCR rates consistent with pivotal trial data. Grade 3 histology, pathogenic germline mutation, and age younger than 40 years were independently associated with pCR; the HR-low findings are exploratory. These results support prospective biomarker-driven trials in early TNBC across underrepresented populations.

Introduction

Triple-negative breast cancer (TNBC), defined by the absence of estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) amplification, accounts for approximately 15–20% of all breast cancers worldwide.¹ It is characterised by aggressive clinical behaviour, higher rates of recurrence, and poorer overall survival compared with other breast cancer subtypes.² TNBC disproportionately affects younger women, those of African ancestry, and carriers of germline BRCA1/2 mutations,³ underscoring its biological heterogeneity and global health burden.⁴ Historically, treatment was limited to cytotoxic chemotherapy, which, while effective in a subset of patients, offered modest long-term survival benefit and substantial toxicity; unlike hormone receptor (HR)-positive or HER2-positive disease, early-stage TNBC lacked approved targeted therapies applicable across the broad population.⁵

The advent of immune checkpoint inhibitors (ICIs) targeting the programmed death-1 (PD-1)/programmed death-ligand 1 (PD-L1) axis has fundamentally reshaped this landscape.⁶ The landmark KEYNOTE-522 trial established chemoimmunotherapy as a standard of care in early-stage TNBC, demonstrating that adding pembrolizumab to standard neoadjuvant chemotherapy significantly improves pathological complete response (pCR), event-free survival (EFS), and overall survival (OS), irrespective of PD-L1 status.^7–9 Achieving pCR has long been considered a validated surrogate for favourable long-term outcomes, with patients who attain it carrying substantially lower risks of distant recurrence and death.^10,11 Yet immunotherapy has simultaneously exposed the limitations of pCR as a sole endpoint: in the GeparNuevo trial,¹² neoadjuvant durvalumab improved invasive disease-free survival and OS on long-term follow-up despite no significant increase in pCR.¹³ This dissociation implies that immune-mediated suppression of micrometastatic disease and durable immune surveillance may confer long-term benefit independently of primary tumour eradication, and that pCR alone may underestimate the clinical value of immunotherapy in biologically responsive subgroups.

The broader landscape of neoadjuvant immunotherapy trials has further refined understanding of the determinants of benefit in early breast cancer. In HR-positive disease, the Phase III CheckMate 7FL and KEYNOTE-756 studies extended checkpoint blockade beyond TNBC, demonstrating increases in pCR that were most pronounced among tumours with low HR expression and other high-risk features.^14,15 In TNBC, the phase III IMpassion031 trial showed that adding atezolizumab to standard neoadjuvant chemotherapy significantly improved pCR, although mature survival data remain less conclusive than those observed with pembrolizumab-based regimens.¹⁶ By contrast, the NeoTRIPaPDL1 trial did not demonstrate a significant improvement in pCR; a key design difference was the omission of anthracyclines from the neoadjuvant phase, with these agents administered only postoperatively, potentially delaying the immunogenic cell death and antigen release required to prime antitumour immunity effectively.¹⁷

Taken together, these divergent outcomes suggest two recurring principles relevant to immunotherapy efficacy in early breast cancer: early activation of the immune response during neoadjuvant treatment, and the use of an immunogenic chemotherapy backbone incorporating anthracycline and platinum agents.^16,17 Trials that introduced ICIs concurrently with taxane-based chemotherapy, most notably KEYNOTE-522 and GeparNuevo, reported larger pCR gains than studies in which immunotherapy exposure or key chemotherapy components were deferred.^7,12 Taxanes may enhance antitumour immunity through immunogenic cell death, improved antigen presentation, and increased tumour-infiltrating lymphocyte (TIL) density, creating a window of maximal immune priming while tumour burden remains high.^18–20 Administering immunotherapy during this window may exploit that immunogenic vulnerability, whereas deferring it may forfeit the opportunity.²¹ These mechanistic interpretations remain hypotheses generated by cross-trial comparison rather than conclusions from randomised head-to-head evaluation.

Despite these advances, critical questions persist regarding real-world implementation: the generalisability of trial outcomes to broader patient populations; which tumour or patient characteristics best identify those most likely to achieve pCR; whether HR-low tumours (ER and/or PR 1–10%) respond comparably to HR-negative TNBC under immunotherapy-based regimens; whether the total number of immunotherapy cycles delivered influences response; and what the real-world safety profile of this regimen looks like outside controlled trial populations, including in under-represented geographic regions.^22,23

Consequently, this study evaluates the real-world safety and efficacy of neoadjuvant pembrolizumab-based chemoimmunotherapy in a large, consecutive Middle Eastern cohort (n = 149), with specific exploratory attention to the distinct clinical behaviour of the HR-low subgroup (ER and/or PR 1–10%) and to the association between germline mutation status and pathological complete response.

Patients and Methods

Study Design and Population

This retrospective cohort study included 149 female patients with histologically confirmed non-metastatic TNBC or HR-low breast cancer who received at least one cycle of neoadjuvant chemoimmunotherapy at King Hussein Cancer Center (KHCC), Amman, Jordan, between April 2021 and September 2024. This study included patients with early-stage HER2-negative breast cancer displaying low or absent hormone receptor expression (ER ≤10% and PR ≤10%). Within this study population, absolute TNBC was defined as ER <1% and PR <1% per ASCO/CAP guidelines, while tumors with ER and/or PR expression of 1–10% were classified as HR-low and managed according to the institutional TNBC pathway.

Exclusion criteria included metastatic disease, bilateral breast cancer, prior breast cancer diagnosis, and insufficient data on immunotherapy regimen, dosage, or duration. All 149 treatment-initiated patients constituted both the primary pCR analysis population and the safety population. Of these, 142 (95.3%) underwent surgery; the 7 who did not (died during neoadjuvant treatment, n = 2; progressive or metastatic disease, n = 3; patient refusal, n = 2) were classified as not achieving pCR. Median follow-up was 27.3 months (IQR 20.7–33.0; range 5.2–54.7 months) across all 149 patients and 27.4 months (IQR 21.3–32.9; range 11.3–54.7 months) among the 142 surgical patients; follow-up data were complete for all 149 patients.

Ethics Approval and Informed Consent

This study was approved by the Institutional Review Board of King Hussein Cancer Center (approval number: 24 KHCC 86) and was conducted in accordance with the Declaration of Helsinki. Individual informed consent was waived by the IRB given the retrospective design and use of de-identified data. Reporting follows the STROBE guidelines.

Treatment Regimen

The neoadjuvant regimen consisted of pembrolizumab (200 mg IV), carboplatin (AUC 5 IV), and paclitaxel (80 mg/m² IV weekly) for 12 weeks, followed by pembrolizumab (200 mg IV), doxorubicin (60 mg/m² IV), and cyclophosphamide (600 mg/m² IV) every 3 weeks for 4 cycles, in accordance with the KEYNOTE-522 protocol.⁷ One patient (0.7%) received an alternative pembrolizumab-containing chemotherapy backbone at the discretion of the treating team. Dose adjustments and discontinuations were implemented according to toxicity profile and clinical judgment. Adjuvant treatment was administered according to institutional protocols and physician discretion.

KHCC Institutional Treatment Protocol

At KHCC, the institutional approach to early-stage TNBC and HR-low breast cancer follows a clinically pragmatic framework that reflects both the available evidence and the resource constraints of routine practice in a regional cancer centre. For the purposes of this analysis, early-stage breast cancer was defined as T1–T3 N0–N1 M0 disease (up to and including T3N1), whereas locally advanced disease encompassed T3N2, T3N3, T4 (any N), or any T with N2–N3 involvement. Tumours with low hormone receptor expression (ER and/or PR 1–10%) are managed according to the TNBC pathway rather than the luminal protocol, consistent with the growing recognition of their shared biology and comparable chemosensitivity.^24,25 Proliferative activity, assessed by Ki-67 immunohistochemistry, is classified at KHCC into three tiers: low (Ki-67 < 20%), intermediate (Ki-67 20–30%), and high (Ki-67 > 30%).

Adjuvant treatment is stratified by pathological response at surgery. Under the current institutional protocol, patients who achieve pCR (ypT0/Tis ypN0) are observed without further systemic therapy, reflecting their favourable prognosis and the absence of evidence that adjuvant intensification improves outcomes in this group.^10,11 Patients with residual disease continue adjuvant pembrolizumab for an additional nine cycles. In this population, the use of additional cytotoxic or targeted agents is discussed individually with each patient, as neither capecitabine nor olaparib is currently reimbursed through public funding at KHCC and must be self-funded. For patients able to meet this cost, capecitabine is offered for eight cycles to address chemotherapy-resistant residual disease, based on the CREATE-X trial.²⁶ Olaparib for one year is discussed as an alternative in patients with confirmed pathogenic germline BRCA1/2 mutations, given the demonstrated survival benefit in this molecularly defined subgroup.²⁷ All treatment decisions in the residual-disease setting are made through shared decision-making that explicitly accounts for financial toxicity, treatment burden, and individual patient preference, reflecting real-world constraints that frequently shape care delivery in resource-limited settings outside the clinical trial environment.

Importantly, this protocol changed during the study period. In the earlier part of the cohort, adjuvant pembrolizumab was administered to all patients who had received neoadjuvant pembrolizumab, irrespective of pathological response, in keeping with the KEYNOTE-522 regimen as originally specified. The institutional policy was subsequently revised to restrict adjuvant pembrolizumab to patients with residual disease. This evolution in practice explains why the number of patients receiving adjuvant pembrolizumab (n = 90) exceeds the number with residual disease (n = 54), and it should be taken into account when interpreting post-neoadjuvant treatment exposure in this cohort.

Endpoints

The primary endpoint was pCR, defined as the absence of invasive residual tumor in the breast and axillary lymph nodes (ypT0/Tis ypN0) at surgical resection. Secondary endpoints included safety (hospitalization rate and irAEs), surgical outcomes, and rate of breast-conserving surgery (BCS). Disease-free survival (DFS) was defined as the time from date of cancer diagnosis to first documented recurrence or death from any cause; patients without an event were censored at the date of last follow-up.

Data Collection

Clinical, pathological, treatment, and outcome data were extracted from electronic medical records. Baseline variables included demographics, ECOG performance status, clinical stage, nodal status, histological subtype and grade, HR and HER2 status, Ki-67 proliferation index, and germline genetic testing results. Germline testing was performed using next-generation sequencing (NGS) multigene panels at a CLIA-certified, CAP-accredited laboratory. Treatment variables included neoadjuvant regimen details, number of completed cycles, hospitalizations, irAEs, surgical management, and adjuvant therapies.

Immune-Related Adverse Events

Immune-related adverse events were identified by retrospective medical record review and categorized by organ system. Severity was graded according to CTCAE version 5.0. Immunotherapy discontinuation due to toxicity was documented.

Statistical Analysis

Descriptive statistics were used to summarize baseline and outcome data. Continuous variables are reported as medians with interquartile ranges (IQRs); categorical variables as frequencies and percentages. Associations between clinicopathological variables and pCR were evaluated by univariable logistic regression and Firth penalised multivariable logistic regression across all 149 treatment-initiated patients, with the 7 patients who did not undergo surgery classified as non-pCR. Firth penalisation reduces first-order bias in maximum likelihood estimation and is recommended when reference categories are small or sparse data are present.^28,29 Odds ratios are reported with 95% profile penalised-likelihood confidence intervals and penalised likelihood-ratio p-values; the same inferential approach is used for all univariable and multivariable estimates.

Variables were selected for multivariable inclusion based on established clinical significance and univariable trends (p < 0.10). Events per variable (EPV) was calculated using the limiting outcome class (non-pCR, n = 61 in the full cohort), yielding EPV = 12.2 for the final five-parameter model, exceeding the recommended minimum of 10. Model discrimination was assessed by the optimism-corrected area under the receiver operating characteristic curve (AUC) using a 1000-iteration bootstrap (Harrell method). Multicollinearity was examined using variance inflation factors, all of which were below 1.2. The primary pCR analysis followed an intention-to-treat-style approach including all 149 treatment-initiated patients, with the 7 patients who did not undergo surgery (due to death, disease progression, or patient refusal) classified as non-pCR, since these represent treatment failures rather than random missing data. The corresponding estimate restricted to the 142 surgically evaluable patients is reported alongside for comparability with neoadjuvant trial literature, and pre-specified sensitivity analyses confirmed that model estimates were stable across both denominators. Receptor status classification followed ASCO/CAP guidelines. Follow-up time was calculated from date of cancer diagnosis to date of last contact. Survival outcomes were illustrated by Kaplan–Meier curves and compared using the Log rank test. All tests were two-sided; p < 0.05 was considered significant. Missing Ki-67 data (33.6%) precluded its inclusion as a covariate. One surgical patient with an unavailable histological grade (0.7% of the cohort) was omitted from the multivariable model run to ensure complete-case analysis for the final covariates. Statistical analyses were performed using R (v4.3; logistf package for Firth penalised regression; rms package for bootstrap optimism correction; survival package for Kaplan–Meier analyses) and jamovi (Version 2.7.28) for descriptive statistics and univariable logistic regression.

Results

Patient Characteristics

A total of 149 female patients with non-metastatic TNBC or HR-low breast cancer were included. The median age was 45 years (IQR 38–52); 28% were younger than 40 years and 62% were aged 40–60 years. Most patients were non-smokers (76%) with ECOG performance status 0 (78%). T2 tumors predominated (54%), clinically node-positive disease was present in 62%, and 72% presented with locally advanced disease. Grade 3 tumors accounted for 84.6% of cases. HR-low expression (ER and/or PR 1–10%) was identified in 15 patients (10.1%). HER2-low status was present in 24.8%. Ki-67 was elevated (>20%) in 87 of 99 assessable patients (87.9%); data were unavailable in 50 patients (33.6%).

Germline genetic testing was performed in 143 patients (96.0%); as proportions of the full cohort (n = 149), pathogenic mutations were identified in 35 (23.5%), variants of uncertain significance in 26 (17.4%), and results were negative in 82 (55.0%). Among mutation carriers, BRCA1 (48.6%) and BRCA2 (37.1%) were most frequent. Baseline characteristics are summarized in Table 1. The patient flow is illustrated in Figure 1.

Table 1.

Baseline Patient Characteristics (N = 149)

Characteristic	N (%) or Median (IQR)
Demographics
Age at diagnosis (years)	45 (38–52)
Age group	<40: 42 (28%) │ 40–60: 92 (62%) │ >60: 15 (10%)
Smoking status	Non-smoker: 113 (76%) │ Smoker: 36 (24%)
ECOG performance status	0: 116 (78%) │ 1: 30 (20%) │ 2: 3 (2%)
Clinical Staging
Clinical T stage	T1: 6 (4%) │ T2: 80 (54%) │ T3: 48 (32%) │ T4: 15 (10%)
Clinical nodal status	Negative: 57 (38%) │ Positive: 92 (62%)
Stage group	Early stage: 41 (28%) │ Locally advanced: 108 (72%)
Histopathology
Histological grade	Grade 2: 22 (14.8%) │ Grade 3: 126 (84.6%) │ Not available: 1 (0.7%)
Receptor Status & Tumor Biology
ER status	<1%: 139 (93.3%) │ Low (1–10%): 10 (6.7%)
PR status	<1%: 143 (96.0%) │ Low (1–10%): 6 (4.0%)
HR-low expression (ER and/or PR 1–10%)a	15 (10.1%)
HER2 status	HER2 0: 112 (75.2%) │ HER2 1+: 26 (17.4%) │ HER2 2+/FISH−: 11 (7.4%)
Ki-67 proliferation index	High (>20%): 87/99 assessable (87.9%) │ Not assessed: 50 (33.6%)
Genetic Testing
Germline result	Negative: 82 (55.0%) │ Pathogenic mutation: 35 (23.5%) │ VUS: 26 (17.4%) │ Not performed: 6 (4.0%)
Pathogenic mutations (n = 35)	BRCA1: 17 (48.6%) │ BRCA2: 13 (37.1%) │ CHEK2/PALB2/RAD51D (multi-gene): 2 (5.7%) │ RAD51D, RAD51C or BARD1 (single gene): 3 (8.6%)

Notes: ^aOne patient had both ER low (1–10%) and PR low (1–10%), accounting for a combined HR-low group of 15 patients.

Abbreviations: ECOG, Eastern Cooperative Oncology Group; FISH, fluorescence in situ hybridization; VUS, variant of uncertain significance.

Figure 1.

STROBE-compliant patient flow diagram. Of 149 patients enrolled with non-metastatic TNBC or HR-low breast cancer, all received at least one cycle of neoadjuvant chemoimmunotherapy; 129 (86.6%) completed all eight planned chemotherapy cycles. A total of 142 patients (95.3%) underwent surgery, of whom 88 achieved pCR and 54 had residual disease. Seven patients did not undergo surgery (died during neoadjuvant treatment, n = 2; progressive or metastatic disease, n = 3; patient refusal, n = 2) and were classified as failures to achieve pCR in the primary analysis, giving an overall pCR rate of 59.1% (88/149); the corresponding rate among surgically evaluable patients was 62.0% (88/142). Adjuvant treatment exposure and survival outcomes are reported separately in the Results (Surgical Management and Adjuvant Therapy) rather than in this diagram, because those treatments are not mutually exclusive and do not map onto a single branch of the flow.

Abbreviations: HR-low, hormone receptor expression 1–10%; pCR, pathological complete response (ypT0/Tis ypN0); TNBC, triple-negative breast cancer.

Treatment Delivery

All 149 patients were planned to receive eight cycles of neoadjuvant chemoimmunotherapy. The majority completed all eight planned cycles of chemotherapy (86.6%) and immunotherapy (81.9%). Sixteen patients (10.7%) received ≤6 cycles of immunotherapy, of whom 14 proceeded to surgery. Within this subgroup, immunotherapy was permanently discontinued for treatment-limiting toxicity in 3 patients and 2 patients did not reach surgery because of disease progression; the reasons for early cessation in the remaining patients were not consistently coded in the retrospective record and could not be reconstructed reliably. This heterogeneity of reasons is clinically relevant, because patients stopping for toxicity and patients stopping for progression represent different biological trajectories, and it is a further reason the cycle-count analysis is treated as exploratory.

Primary Endpoint: Pathological Complete Response

The overall pCR rate (ypT0/Tis ypN0) was 59.1% (88 of 149 patients); the corresponding rate restricted to the 142 surgically evaluable patients was 62.0% (88/142). Table 2 presents univariable screening results for all nine candidate variables (age group, smoking status, germline result, disease stage, histological grade, HR-low status, HER2-low status, irAE occurrence, and immunotherapy cycle count) together with the final multivariable model. Variables meeting the pre-specified univariable threshold (p < 0.10) together with clinical relevance criteria were carried forward; the final Firth penalised model retained age group, germline result, and histological grade (5 estimated parameters excluding the intercept; n = 148, since one patient had unavailable histological grade; 61 events in the limiting outcome class, EPV = 12.2). Three variables were independently associated with pCR. Grade 3 histology conferred the strongest association (pCR 63.5% vs 31.8% for grade 2; aOR 4.44, 95% CI 1.61–13.66, p = 0.004). Pathogenic germline mutation was associated with a pCR rate of 80.0% versus 53.7% among patients with negative or VUS results (aOR 3.19, 95% CI 1.26–9.11, p = 0.014).

Table 2.

Univariable and Multivariable Firth Penalised Logistic Regression for pCR (Full Cohort, N = 149)

Variable	Category	No pCR n(%)	pCR n(%)	Univariable OR (95% CI, p)	Multivariable OR (95% CI, p)
Age group	40–60 (Ref)	44 (47.8)	48 (52.2)	Reference	Reference
	<40 years	9 (21.4)	33 (78.6)	3.24 (1.46–7.73), p = 0.003	2.82 (1.20–7.10), p = 0.017
	>60 years	8 (53.3)	7 (46.7)	0.81 (0.27–2.36), p = 0.697	0.89 (0.28–2.90), p = 0.851
Smoking	Smoker (Ref)	14 (38.9)	22 (61.1)	Reference	Not retained
	Non-smoker	47 (41.6)	66 (58.4)	0.90 (0.42–1.91), p = 0.790	Not retained
Germline result	Negative/VUS (Ref)	50 (46.3)	58 (53.7)	Reference	Reference
	Not tested	4 (66.7)	2 (33.3)	0.48 (0.08–2.26), p = 0.355	0.47 (0.07–2.79), p = 0.408
	Pathogenic mut.	7 (20.0)	28 (80.0)	3.28 (1.41–8.49), p = 0.005	3.19 (1.26–9.11), p = 0.014	3.19 (1.26–9.11), p = 0.014
Stage	Early stage (Ref)	14 (34.1)	27 (65.9)	Reference	Not retained
	Locally advanced	47 (43.5)	61 (56.5)	0.68 (0.32–1.42), p = 0.308	Not retained
Histological grade	Grade 2 (Ref)	15 (68.2)	7 (31.8)	Reference	Reference
	Grade 3	46 (36.5)	80 (63.5)	3.58 (1.43–9.68), p = 0.006	4.44 (1.61–13.66), p = 0.004	4.44 (1.61–13.66), p = 0.004
HR-low (1–10%)	No (Ref)	56 (41.8)	78 (58.2)	Reference	Not retained
	Yes	5 (33.3)	10 (66.7)	1.37 (0.48–4.38), p = 0.562	Not retained
HER2-low status	No (Ref)	43 (38.4)	69 (61.6)	Reference	Not retained
	Yes	18 (48.6)	19 (51.4)	0.66 (0.31–1.39), p = 0.272	Not retained
irAE occurrence	No (Ref)	43 (41.0)	62 (59.0)	Reference	Not retained
	Yes	18 (40.9)	26 (59.1)	1.00 (0.49–2.04), p = 0.993	Not retained
Immuno. cycles	≤6	8 (50.0)	8 (50.0)	Reference	Not retained
	>6	53 (39.8)	80 (60.2)	1.50 (0.54–4.20), p = 0.430	Not retained

Notes: Analyses include all 149 treatment-initiated patients, with the 7 who did not undergo surgery classified as non-pCR. Grade totals sum to 148 because one patient had unavailable histological grade; that patient is excluded from grade-containing models. Odds ratios are Firth penalised estimates with profile penalised-likelihood 95% confidence intervals and likelihood-ratio p-values. The final multivariable model retained age group, germline result, and histological grade (5 parameters excluding the intercept; 61 events in the limiting outcome class; EPV = 12.2). Variables marked “Not retained” did not meet the pre-specified univariable screening threshold (p < 0.10) together with clinical relevance criteria and were not entered into the final model.

Abbreviations: OR, odds ratio; CI, confidence interval; VUS, variant of uncertain significance; HR, hormone receptor; irAE, immune-related adverse event.

Age younger than 40 years was also independently associated with pCR (pCR 78.6% vs 52.2% in the 40–60 year reference group; aOR 2.82, 95% CI 1.20–7.10, p = 0.017), whereas age above 60 years was not (aOR 0.89, 95% CI 0.28–2.90, p = 0.851). The model demonstrated acceptable discrimination (apparent AUC 0.728; bootstrap optimism 0.027; optimism-corrected AUC 0.701), indicating minimal residual overfitting. The three associations were directionally and quantitatively stable across sensitivity analyses, including restriction to the 142 surgically evaluable patients (141 of whom had grade recorded; age < 40 aOR 2.73, p = 0.025), retention of all 149 patients using a grade-unknown indicator (aOR 2.82, p = 0.017), and unpenalised maximum-likelihood estimation (aOR 3.03, p = 0.018). Further pre-specified sensitivity analyses — exclusion of HR-low tumors, exclusion of genetically untested patients, restriction to patients completing more than six immunotherapy cycles, and alternative coding of age as a continuous variable — left the direction and approximate magnitude of all three associations unchanged. Smoking status, disease stage, HR-low expression, HER2-low status, irAE occurrence, and immunotherapy cycles were not significantly associated with pCR. Subgroup pCR rates are illustrated in Figure 2.

Figure 2.

Pathological complete response (pCR) rates across clinically relevant subgroups (full cohort, n = 149). (A) pCR by age group. (B) pCR by germline mutation status. (C) pCR by histological grade. (D) pCR by receptor status. All rates are observed proportions across all 149 treatment-initiated patients, with the 7 who did not undergo surgery classified as non-pCR. The subgroup analysis by histological grade includes 148 patients, excluding one patient with unavailable grade data. Adjusted odds ratios (aOR) shown beneath (A–C) are from the Firth penalised multivariable logistic regression model, with profile penalised-likelihood 95% confidence intervals and likelihood-ratio p-values. In (D) neither HR-low status nor HER2-low status was retained in the final model; the bars are shown as descriptive subgroup rates only. Overall cohort pCR was 59.1% (88/149). Subgroup estimates for HR-low tumours (n = 15), grade 2 histology (n = 22), and patients without germline testing (n = 6) are based on small numbers and are exploratory; the corresponding confidence intervals are wide, and these comparisons should not be interpreted as establishing equivalence or difference. Bar colours denote variables independently associated with pCR in the final multivariable model, reference groups, the exploratory HR-low subgroup, and categories not reaching statistical significance, as indicated in the key.

Abbreviations: aOR, adjusted odds ratio; HR-low, hormone receptor expression 1–10%; pCR, pathological complete response; VUS, variant of uncertain significance.

With a median follow-up of 27.3 months (IQR 20.7–33.0; range 5.2–54.7 months), 137 of 149 patients (91.9%) were alive and 136 of 149 (91.3%) were recurrence-free at last contact; all 13 recurrences (8.7%) were distant, with no locoregional-only events. Counting recurrence or death from any cause as the disease-free survival (DFS) event, 132 of 149 patients (88.6%) were event-free. These are point-prevalence proportions, not Kaplan–Meier probabilities; time-to-event analyses were restricted to the 142 surgical patients. Among surgical patients, 85 of 88 in the pCR group (96.6%) and 50 of 54 with residual disease (92.6%) were alive (log-rank p = 0.25). DFS events occurred in 3 of 88 pCR patients (event-free 96.6%) and 8 of 54 with residual disease (event-free 85.2%; log-rank p = 0.011). Given only 17 DFS events in total, this difference should be regarded as preliminary.

Safety

Hospitalizations

During neoadjuvant treatment, 50 patients (33.6%) required at least one hospital admission (median 1, range 1–5); 74% had a single admission. Infection was the predominant cause (80.0%), with febrile neutropenia accounting for the majority (38 episodes). irAE-related admissions accounted for 22.0% (11 patients). Remaining admissions (6.0%) were attributable to non-immune, non-infectious causes. Admission category percentages sum to more than 100% because individual patients could have more than one concurrent cause. Hospitalization data are presented in Table 3.

Table 3.

Hospital Admissions During Neoadjuvant Treatment (N = 50 Patients)

Cause of Admission	N (%)a	Specific Cause	N
Infection	40 (80.0%)	Febrile neutropenia Thigh abscess Wound infection	38 1 1
irAEsb	11 (22.0%)	Colitis Rash Elevated liver enzymes Diabetic ketoacidosis Central adrenal insufficiency	9 1 1 1 1
Other (non-immune)	3 (6.0%)	Atrial fibrillation Allergic rash (non-immune)	1 2

Notes: ^aPercentages are of 50 admitted patients and may exceed 100% as individual patients could have more than one concurrent cause of admission. ^b11 patients with irAE-related admissions; 13 specific irAE episodes recorded, as 2 patients had more than one concurrent contributing irAE.

Immune-Related Adverse Events

Immune-related adverse events were observed in 44 patients (29.5%). Endocrine toxicities were most frequent (29 patients; 65.9% of irAE patients), predominantly hypothyroidism (n = 24), with adrenal insufficiency (n = 4) and hyperthyroidism (n = 2) also recorded. Gastrointestinal toxicities occurred in 15 patients (34.1%), predominantly colitis (n = 13). Dermatological irAEs occurred in 5 patients (11.4%); cardiological (myocarditis), rheumatological (undifferentiated arthritis), and neurological irAEs were each recorded in 1 patient (2.3%). Grade ≥3 irAEs occurred in 9 patients (20.5% of those with irAEs). Across the entire neoadjuvant phase, immunotherapy was permanently discontinued because of treatment-limiting toxicity in 6 patients (13.6% of those experiencing irAEs; 4.0% of the cohort). irAE data are summarized in Table 4 and illustrated in Figure 3.

Table 4.

Immune-Related Adverse Events (irAEs; Total Affected Patients = 44 [29.5% of 149])

Category	N (%)a	Specific Toxicity	n
Any irAE	44 (29.5%)	—	—
System involved
Endocrineb	29 (65.9%)	Hypothyroidism Adrenal insufficiency Hyperthyroidism	24 4 2
Gastrointestinalc	15 (34.1%)	Colitis Esophagitis Other GI event	13 1 1
Dermatological	5 (11.4%)	Rash	5
Cardiological	1 (2.3%)	Myocarditis	1
Rheumatological	1 (2.3%)	Undifferentiated arthritis	1
Neurological	1 (2.3%)	Neurological toxicity	1
Severity (CTCAE v5.0)
Grade 1 │ Grade 2	10 (22.7%) │ 25 (56.8%)	—	—
Grade 3 │ Grade 4	8 (18.2%) │ 1 (2.3%)	—	—
Grade ≥3 (combined)	9 (20.5%)	—	—
Treatment impact
Immunotherapy discontinued	6 (13.6%)	Due to treatment-limiting toxicity	6
Continued despite irAE	38 (86.4%)	—	—

Notes: ^aPercentages in the System column are proportions of the 44 patients with any irAE. ^bOne patient had both hypothyroidism and adrenal insufficiency (30 endocrine events across 29 patients). ^cGastrointestinal events are distributed across 15 unique patients.

Abbreviations: CTCAE, Common Terminology Criteria for Adverse Events; GI, gastrointestinal.

Figure 3.

Immune-related adverse event (irAE) profile (N = 149). (A) overall irAE incidence across the full cohort; 44 of 149 patients (29.5%) experienced at least one irAE and 105 (70.5%) did not. (B) distribution of irAEs by organ system among the 44 affected patients; percentages are proportions of these 44 patients and sum to more than 100% because some patients had events in more than one organ system. (C) irAE severity by CTCAE v5.0 grade; Grade 1 (mild) 10 patients (22.7%), Grade 2 (moderate) 25 (56.8%), Grade 3 (severe) 8 (18.2%), Grade 4 (life-threatening) 1 (2.3%); combined Grade ≥3 rate 20.5% (9/44 irAE patients). Immunotherapy was permanently discontinued because of treatment-limiting toxicity in 6 of the 44 irAE patients (13.6%). Hospital admissions during neoadjuvant treatment, which occurred in 50 of 149 patients (33.6%), are presented separately in Table 3.

Abbreviations: CTCAE, Common Terminology Criteria for Adverse Events; irAE, immune-related adverse event.

Surgical Management and Adjuvant Therapy

Surgery was performed in 142 patients (95.3%). Among surgical patients, mastectomy was performed in 68.3% and breast-conserving surgery in 31.7%. Sentinel lymph node biopsy was performed in 59.2% and axillary lymph node dissection in 40.8%. Adjuvant pembrolizumab was administered to 90 of 142 surgical patients (63.4%), with the majority completing nine cycles (69/90; 76.7%). Uptake differed by pathological response: 48 of 54 patients with residual disease (88.9%) versus 42 of 88 patients achieving pCR (47.7%). This number exceeds the 54 patients with residual disease because, in the earlier part of the study period, adjuvant pembrolizumab was given to all patients who had received neoadjuvant pembrolizumab regardless of pathological response, before institutional policy was revised to restrict it to patients with residual disease (KHCC Institutional Treatment Protocol). Adjuvant capecitabine was given to 7 patients (4.9%), all of whom had residual invasive disease at surgery (non-pCR), consistent with the CREATE-X paradigm; this subset represented patients with significant residual cancer burden at the discretion of the treating physician. Radiotherapy was administered to 111 surgical patients (78.2%).

Discussion

Principal Findings

In this real-world cohort of 149 patients treated at a regional cancer center in Jordan, neoadjuvant pembrolizumab-based chemoimmunotherapy achieved a pCR rate of 59.1% across all treatment-initiated patients (62.0% among the 142 who underwent surgery), consistent with the 58–65% range reported in pivotal trials of neoadjuvant chemoimmunotherapy.^7–9,16,30 On Firth penalised multivariable logistic regression, three variables were independently associated with pCR: grade 3 histology (aOR 4.44, p = 0.004), pathogenic germline mutation (aOR 3.19, p = 0.014), and age younger than 40 years (aOR 2.82, p = 0.017). Because the analysis was conducted in the full treatment-initiated cohort rather than only in surgically evaluable patients, these estimates are not inflated by attrition of patients with progression or death before surgery.

Firth penalisation was applied because the small grade 2 reference group inflated the unpenalised maximum-likelihood estimate for grade 3 relative to the penalised estimate (OR 5.02 vs 4.44; 11.7% shrinkage). The model demonstrated fair discrimination (apparent AUC 0.728; bootstrap optimism 0.027; optimism-corrected AUC 0.701). HR-low tumors (n = 15) demonstrated pCR rates similar to HR-negative disease (66.7% vs 58.2%), though this finding is exploratory. irAEs occurred in 29.5% of patients, consistent with published trial data, and 33.6% required hospitalization. With a median follow-up of 27.3 months, 91.9% of patients were alive and 91.3% were recurrence-free.

Survival analyses should be interpreted cautiously because of the limited median follow-up (27.4 months) and the small number of observed events (12 deaths and 13 recurrences across all 149 enrolled patients, giving 17 composite DFS events; 7 deaths and 11 composite events among the 142 surgical patients in the time-to-event analysis). The disease-free survival advantage observed for patients achieving pCR is therefore hypothesis-generating, and longer follow-up will be required to determine whether it translates into an overall survival difference within this cohort.

Comparison with KEYNOTE-522 and Real-World Cohorts

Our pCR rate (59.1% across all treatment-initiated patients; 62.0% among surgical patients) is closely comparable to that of the KEYNOTE-522 trial (64.8%), confirming that trial-level efficacy is reproducible in a resource-limited, single-center, Middle Eastern setting.^7–9 This finding is consistent with multiple published real-world cohorts spanning Europe, Latin America, and Türkiye, which have collectively reported pCR rates between approximately 56% and 63%, supporting the reproducibility of pembrolizumab-based chemoimmunotherapy across diverse healthcare systems. A detailed cross-cohort comparison is provided in Table 5.^{7–9,31–35} Safety outcomes in our cohort were broadly aligned with these studies, with irAE rates and toxicity profiles consistent with the published literature, though direct comparisons are limited by differences in grading practices, follow-up duration, and patient selection across cohorts.^31,33,34

Table 5.

Comparison of the Present Cohort with KEYNOTE-522 and Selected Real-World Neoadjuvant Immunotherapy Studies in Early Triple-Negative Breast Cancer

Study	Setting	N	pCR (%)	Key Safety/Delivery Findings	Survival/Follow-Up	Ref.
► Current study (KHCC, Jordan)	Single center	149	59.1 (62.0 surgical)	irAEs 29.5%; grade ≥3 irAEs 20.5% of irAE patients; hospitalization 33.6%; discontinuation for toxicity 13.6% of irAE patients	Median FU 27.3 months (IQR 20.7–33.0); 91.9% alive, 91.3% recurrence-free; all 13 recurrences distant; pCR vs residual disease OS log-rank p = 0.25, DFS log-rank p = 0.011	—
KEYNOTE-522	RCT	784	64.8	irAEs: any grade ~30%; endocrine most common	5-yr OS 86.6% vs 81.7% (HR 0.66); 5-yr EFS 81.2% vs 72.2%	[7–9]
UK (ESIT, 34 centres)	Multicenter	545	56.0	irAEs 65% (20% grade ≥3); delays 59%, discontinuation 47%	Survival data not reported	[31]
Montpellier Cancer Institute, France	Single center	92	61.1	Grade 3–4 irAEs 20%; immunotherapy discontinuation 29.3%	No mature survival data	[32]
Neo-Real/GBECAM-0123 (LatAm)	Multicenter	726	62.9	Grade ≥3 AEs lower with pembrolizumab alone	Survival data maturing	[33]
Helsinki Univ. Hospital, Finland	Population-based	75	56.0	Pembrolizumab stopped preoperatively in 29%; myocarditis 14.7%	Survival maturing	[34]
Turkish Oncology Group	Multicenter	164	60.4	Weekly carboplatin associated with higher pCR	Median FU 49.2 months; pCR independently associated with EFS; OS analyses underpowered (5 deaths) and descriptive	[35]

Biological Interpretation

The association between pathogenic germline mutations and superior pCR (80.0% vs 53.7%) may partially explain the high response rates observed in genomically unstable tumors: homologous recombination deficiency is associated with increased neoantigen burden and enhanced immunogenicity, and platinum-based chemotherapy exploits defective DNA repair, thereby providing a biological framework compatible with the observed mutation-associated benefit.^27,30 Because every patient in this cohort received platinum-containing chemotherapy together with pembrolizumab, the analysis cannot distinguish a germline-mutation-associated benefit specific to immune checkpoint inhibition from enhanced platinum sensitivity. This mutation-associated benefit was not uniform across genes: BRCA1/2 carriers (n = 30) achieved a higher pCR rate (83.3%, 25/30) than carriers of other pathogenic variants (PALB2, CHEK2, RAD51D/C, BARD1; n = 5; 60.0%, 3/5), consistent with reported differences in homologous recombination deficiency and platinum sensitivity across these genes.³⁶ However, given the relatively small number of carriers (n = 35), and the especially small non-BRCA1/2 subgroup (n = 5), this gene-level comparison should be interpreted as descriptive and hypothesis-generating rather than confirmatory.

The higher pCR rates observed in patients younger than 40 years (78.6% vs 52.2%) and in grade 3 tumors (63.5% vs 31.8%) are biologically compatible with the hypothesis that TNBC arising in younger women may harbor a more immunologically active tumor microenvironment. However, antitumor immune activity cannot be inferred solely from age or histological grade, because the abundance and functional state of infiltrating CD8+ T cells, including exhausted T-cell phenotypes, may substantially influence response to immune checkpoint blockade;³⁷ the biological basis for the age- and grade-related associations observed here therefore remains uncertain from observational data alone.^18,19,30,38

Crucially, the absence of a statistically significant association between the number of immunotherapy cycles (≤6 vs >6) and pCR should be interpreted cautiously. Because the ≤6 cycle reference subgroup was small (n = 16), this sub-analysis was severely underpowered to detect small or moderate differences in response. While hypothesis-generating regarding early immune activation, these findings are strictly exploratory and should not justify deliberate clinical de-escalation of cycle count outside clinical trials.^{13,20,38–40}

In the HR-low subgroup, the observed pCR rate (66.7%) was numerically comparable to HR-negative disease (58.2%), consistent with the biological rationale that HR-low tumors may share immunological and transcriptomic features with TNBC.^21,24,25 Data from KEYNOTE-756 and CheckMate 7FL similarly showed higher pCR rates in ER-low subgroups than in unselected HR-positive populations.^14,15 The immunological basis for responses to checkpoint blockade across these tumour subtypes has been extensively reviewed elsewhere.^6,21 Given the limited number of HR-low tumors in our cohort, however, these findings should be considered exploratory and require confirmation in prospective studies with adequate statistical power.

Clinical Implications

These findings generate hypotheses for future studies evaluating de-escalation of adjuvant chemotherapy in patients achieving pCR and intensification strategies in high-risk biologically defined TNBC subgroups. The high pCR rates observed in younger patients and germline mutation carriers warrant prospective investigation, although treatment modifications should not be inferred from these observational findings. The comprehensive germline testing rate in our cohort (96.0%) underscores the feasibility and value of routine genetic testing in TNBC; germline profiling at baseline is consistent with current guidelines and is supported by these observational data.

The mastectomy rate (68.3%) likely reflects the high prevalence of locally advanced disease (72%) at presentation and institutional surgical decision-making established before neoadjuvant therapy. Additional factors such as tumor-to-breast volume ratio, multifocal disease, and patient preference may also have contributed. Accordingly, the observed mastectomy rate should not be interpreted as evidence of inadequate treatment response. Regarding HR-low disease, management according to the TNBC pathway appears biologically reasonable on the basis of the current data, but prospective evidence from adequately powered studies is required before definitive guidance can be offered.

Strengths and Limitations

Strengths: This study offers several notable features: consecutive patient enrollment minimizes selection bias; germline testing was performed in 96.0% of patients, enabling robust analysis of mutation-associated outcomes; irAEs were characterized in detail according to organ system and CTCAE grade; the surgical completion rate was high (95.3%); and the inclusion of HR-low disease addresses an understudied and clinically relevant population. The real-world, resource-limited institutional context adds external validity relevant to oncology practice in the Middle East and LMIC settings more broadly.

Limitations: The retrospective, single-center design limits causal inference and generalizability. Subgroup analyses were constrained by small patient numbers, particularly for HR-low disease (n = 15) and the ≤6 immunotherapy cycle group (n = 16), resulting in imprecise estimates. Although Firth penalisation reduces estimation bias, the grade 3 aOR carries wide confidence intervals reflecting the small grade 2 reference group (n = 22); external validation in larger cohorts is required. One surgical patient with missing histological grade data was excluded from the multivariable model run to preserve complete-case regression integrity. The optimism-corrected AUC (0.701) indicates only modest overfitting, but external discrimination should be assessed prospectively. Because the study includes no chemotherapy-only comparator arm, none of the three associations can be interpreted as a pembrolizumab-specific predictive biomarker.

With a median follow-up of 27.4 months, a true long-term pCR-associated survival benefit has not yet been demonstrated due to highly immature survival tracking and negligible event numbers; mature 5-year survival data will be required. pCR remains an imperfect surrogate endpoint in the immunotherapy era. Ki-67 data were unavailable in 33.6% of patients, precluding inclusion in regression models. Toxicity data derived from electronic records may under-capture low-grade or subclinical events. Granular treatment-delivery metrics such as chemotherapy dose reductions, treatment delays, and relative dose intensity were not systematically captured in this retrospective dataset; such metrics have been shown to influence pCR in other real-world pembrolizumab-treated TNBC cohorts and represent an important target for prospective data collection.⁴¹ For the same reason, granulocyte colony-stimulating factor prophylaxis practices and the specific chemotherapy component implicated in each febrile-neutropenia episode could not be reported, and the reasons for omission of adjuvant pembrolizumab in surgical patients were not consistently coded and could not be reconstructed retrospectively. Tumor-infiltrating lymphocyte quantification and centralized pathology review were not performed, limiting biological characterization. Residual confounding cannot be excluded because potentially relevant biomarkers, including PD-L1 expression, tumor-infiltrating lymphocytes, and Ki-67, were unavailable or incomplete for a substantial proportion of patients. Disease-free survival was measured from date of cancer diagnosis rather than date of randomisation as used in KEYNOTE-522; this approach was adopted because a pre-treatment start date is systematically available in retrospective electronic records, though it introduces a modest lead-time component relative to trial methodology and limits direct survival comparisons with KEYNOTE-522.

Conclusions

In this real-world Middle Eastern cohort, neoadjuvant pembrolizumab-based chemoimmunotherapy achieved pCR rates consistent with pivotal trial data. Grade 3 histology, pathogenic germline mutation, and age younger than 40 years were each independently associated with pCR in the full treatment-initiated cohort. These associations should be interpreted as identifying tumours with greater sensitivity to the chemoimmunotherapy regimen as a whole rather than as validated biomarkers of pembrolizumab-specific benefit, since the study includes no non-immunotherapy comparator. HR-low tumor findings and the cycle-count analysis are hypothesis-generating and require prospective validation. These findings provide real-world evidence from an underrepresented Middle Eastern population and support prospective validation of biomarker-informed treatment strategies in early TNBC.

Funding Statement

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. Article processing charges will be supported by King Hussein Cancer Center, Amman, Jordan.

Disclaimer

The views and opinions expressed in this article are those of the authors and do not necessarily reflect the official policy or position of King Hussein Cancer Center or any affiliated institution. This study is retrospective and observational; results are hypothesis-generating and should not be used to alter clinical practice outside the context of a controlled trial.

Use of Artificial Intelligence

AI-assisted tools were used solely for language refinement and editing of author-written text. The tool used was Anthropic Claude (Claude Opus 4.1). The authors confirm that the originality and accuracy of all content has been checked and verified by the authors, and that no scientific content, data, analysis, interpretation or conclusion was generated by an AI tool. The authors have reviewed the terms of use of the specific AI tool employed and confirm that its use is suitable for and compatible with publication of this work. The authors take full responsibility for the integrity of the whole content of this article, including the accuracy of all references.

Data Sharing Statement

The datasets generated and/or analysed during the current study are available from the corresponding author, Dr Baha Sharaf (BS.13628@KHCC.JO), upon reasonable academic request, subject to institutional review board requirements, patient confidentiality protections, and applicable data protection regulations.

Ethics Statement

This study was approved by the Institutional Review Board of King Hussein Cancer Center (approval number: 24 KHCC 86) and was conducted in accordance with the Declaration of Helsinki. Individual informed consent was waived by the IRB given the retrospective design and use of de-identified data.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising, or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

The authors declare no conflicts of interest.