Abstract
Background
The KEYNOTE-522 trial showed that neoadjuvant chemotherapy (NAC) plus adjuvant pembrolizumab improved overall survival, event-free survival (EFS), and pathological complete response (pCR) in high-risk early-stage triple-negative breast cancer. As treatments evolve, evaluating real-world (RW) effectiveness is key to understanding trial generalizability. This study benchmarked RW efficacy endpoints in early-stage triple-negative breast cancer patients treated with NAC.
Materials and methods
This retrospective study used RW data from United States community practices abstracted by oncology data specialists. Eligible patients received NAC regimens similar to the KEYNOTE-522 control arm. Control arm patients received no adjuvant therapy, while the RW cohort could receive adjuvant capecitabine. Two RW endpoints were assessed: rwpCR (ypT0/Tis ypN0 or clinical pCR with in situ disease) and rwEFS (time from NAC start to first event). Outcomes were compared with those from the KEYNOTE-522 control arm.
Results
In the RW cohort (n = 128), rwpCR was 37.5%, compared with 51.2% pCR in the KEYNOTE-522 control arm (n = 390). rwEFS over 36 months was comparable: 75.0% (95% confidence interval 67.1% to 83.8%) in the RW cohort versus 76.8% (95% confidence interval 72.2% to 80.7%) in the KEYNOTE-522 control arm (log-rank P = 0.97). The incidence rates of first events were also similar (22.0% RW versus 23.8% trial).
Conclusions
Although pCR rates were higher in the KEYNOTE-522 control arm compared with the RW cohort, rwEFS was comparable with EFS in the KEYNOTE-522 control arm. This study highlights the value of combining structured data with custom abstraction to assess RW endpoints and support future research.
Key words: early-stage triple-negative breast cancer, real-world clinical endpoints, pathologic complete response, event free survival
Highlights
-
•
RW outcomes in esTNBC patients treated with NAC were compared with KEYNOTE-522.
-
•
Endpoints included rwPCR1 and rwPCR2 and rwEFS.
-
•
RwpCR (37.5%) was lower than the KEYNOTE-522 control arm pCR (51.2%).
-
•
36-Month rwEFS was comparable with similar incidence of first events versus KEYNOTE-522.
-
•
Value of combining structured data with custom abstraction to assess RW endpoints.
Introduction
Triple-negative breast cancer (TNBC) refers to tumors lacking expression of human epidermal growth factor 2 (HER2), estrogen, and progesterone receptors.1 Approximately 15% of all breast cancers in the United States are TNBC and 40% of all breast cancer-related deaths are attributed to TNBC.1, 2, 3 This form of breast cancer is generally more aggressive than other forms. Early-stage TNBC (esTNBC) is associated with a higher risk of recurrence, while advanced stage TNBC is associated with a poorer prognosis than other subtypes.2 TNBC is attributed to lower 4-year survival rates at 77% compared with hormone receptor (HR)-positive/HER2-negative (92.5%) and HR-negative/HER2-positive (82.7%) subtypes.3,4 Patients with esTNBC are treated with curative intent, thus defining early real-world (RW) endpoints are important to understand treatment effectiveness early.
The rise in expedited Food and Drug Administration (FDA) drug approvals, often based on fewer randomized, controlled trials (RCTs), shorter submission timelines, and increasing post-approval monitoring requirements, has created an evolving opportunity to leverage RW data in regulatory decision-making.5 RW endpoints based on electronic health record (EHR) data have previously been developed in late-stage cancers. Specifically, some progression-based RW endpoints for non-small-cell lung cancer have been shown to correlate with overall survival (OS) and align with outcomes observed in clinical trials in the late-stage setting.6, 7, 8 Therefore, defining RW early clinical endpoints would provide support in advancing understanding and management of esTNBC. While RCTs provide rigorous evidence on treatment efficacy, results often reflect a highly controlled environment with selective patient populations.9 As such, the results from RCTs are not always generalizable to more diverse RW settings.9 Early clinical endpoints are critical for evaluating treatment efficacy in early-stage cancers, as reliance on OS can delay assessment and hinder timely access for patients to effective therapies.10 Early clinical endpoints for early-stage cancers are often composite endpoints, consisting of the combination of several different events into one measure. As such, it is crucial to develop clear definitions and rules for their application in the real world to ensure these endpoints are measured consistently.9 Comparison of RW endpoints with RCT data, the gold standard for data quality, provides valuable context on their reliability and relevance within RW settings.9 As the treatment landscape for esTNBC expands, there is a critical need to understand the effectiveness of therapies outside of an RCT setting.
In recent years, therapeutic options for esTNBC have evolved and now include chemotherapy, immunotherapy, and targeted agents such as poly(ADP-ribose) polymerase (PARP) inhibitors.11 In 2021, pembrolizumab was approved for use by the FDA in patients with esTNBC in combination with chemotherapy as neoadjuvant chemotherapy (NAC) before surgical resection and then continued as a single agent as adjuvant therapy after surgery based on the results from KEYNOTE-522. Pembrolizumab is the only immune checkpoint inhibitor approved for use in esTNBC.12,13 The KEYNOTE-522 trial compared NAC in combination with pembrolizumab with NAC in combination with placebo.12 Key endpoints of interest were pathological complete response (pCR), event-free survival (EFS), and OS. Patients who received NAC in combination with pembrolizumab demonstrated improved pCR compared with NAC alone. Additionally, improvements in EFS and OS were observed among esTNBC patients who received pembrolizumab.12
The current study is a retrospective cohort study using a United States secondary database to compare patients with esTNBC on NAC regimens similar to those in the KEYNOTE-522 control arm [neoadjuvant carboplatin + paclitaxel followed by an anthracycline (doxorubicin or epirubicin) + cyclophosphamide].12 In the study, RW endpoints including RW pCR (rwpCR) and RW EFS (rwEFS) were characterized and benchmarked against KEYNOTE-522 trial outcomes to understand how RW effectiveness compares with that observed in a trial. pCR has been assessed in similar, recent studies including a retrospective database review in China,14 and another RW comparison to the treated population of the KEYNOTE-522 trial.15 pCR was defined in these studies using similar parameters to that used in the current study (detailed in Materials and Methods) and statistical analyses were used to determine prognostic factors relating to achievement of pCR. There is currently a paucity of RW data on EFS in esTNBC patients, suggesting further research is required in this area.
Objectives
The objective of this study was to characterize RW clinical endpoints for patients with esTNBC receiving care in the United States, using longitudinal EHR data from the N-Power Medicine’s (NPM) Breast Analytical Dataset (ADS). The study benchmarked RW endpoints (rwpCR and rwEFS) against outcomes observed in the KEYNOTE-522 control arm.
Materials and methods
Study design
This retrospective database study, supplemented with a manual chart review, utilized data from the breast cohort from the NPM ADS, a longitudinal database integrating multiple sources of patient care information from United States community practices. Specifically, cancer registries captured diagnostic workup and staging; EHRs and laboratory reports provided longitudinal clinical outcomes and treatment data; and EHR physician notes documented clinical context for treatment decisions. These data were supplemented by custom abstraction in which a targeted chart review was conducted by the oncology data specialists (ODS). The study aimed to compare the outcomes from the KEYNOTE-522 control arm with RW data.
Study population
The RW patients were selected from the NPM ADS, a proprietary database including data from patients diagnosed with cancer from large community health systems based in the United States, including managed care delivery networks across 25 states, 457 hospitals, and >1300 oncologists. Patients with esTNBC included in the ADS were diagnosed with breast cancer (ICD-O-3 codes C50.0-C50.6, C50.8-C50.9; behavior code /2 or /3), confirmed by pathology and with two or more distinct clinical encounters after diagnosis. The criteria for patient selection were applied to identify RW patients whose characteristics resembled the KEYNOTE-522 control arm. Patients included in the study were adults (≥18 years of age) who had an esTNBC diagnosis between 1 January 2016 and 31 December 2021. The study end date was 31 December 2022, allowing for a minimum potential follow-up of 12 months. Eligible patients had documentation of at least two medical encounters, with a minimum of one after initial diagnosis.
To confirm TNBC status at diagnosis, patients were required to have HER2-negative disease defined by one of the following: immunohistochemistry (IHC) result of 0, 1+ or 2+, with negative in situ hybridization (ISH) results or nonamplified ISH results alone or annotated as negative with estrogen and progesterone receptor negative status (≤1% positivity). Early-stage disease was defined per the American Joint Committee on Cancer (AJCC) TNM (tumor–node–metastasis) clinical staging as having a combined primary tumor (T) and regional lymph node (N) classification of T1c N1-N2 or T2-T4 N0-N2, and no evidence of distant metastases (M0).16 Included patients were required to have an Eastern Cooperative Oncology Group performance status (ECOG PS) of 0, 1 or unknown.
Patients were required to receive NAC regimens similar to those used in the KEYNOTE-522 control arm (neoadjuvant carboplatin + paclitaxel followed by an anthracycline + cyclophosphamide). Eligible regimens for the RW cohort included:
• Carboplatin + taxane, followed by an anthracycline (either doxorubicin or epirubicin) + cyclophosphamide
• Anthracycline + cyclophosphamide, followed by carboplatin + taxane
• Carboplatin + taxane followed by another regimen
Exclusion criteria were aligned with the KEYNOTE-522 control arm. Patients were excluded if they had received immunotherapy, had an unknown surgical history, or were participating in a clinical trial (for full exclusion criteria see Figure 1).
Figure 1.
Selection of eligible patients in the database with esTNBC. AJCC, American Joint Committee on Cancer; ECOG PS, Eastern Cooperative Oncology Group performance status; esTNBC, early-stage triple-negative breast cancer; hx, history; ICI, immune checkpoint inhibitor; NAC, neoadjuvant chemotherapy; ODS, oncology data specialist; RT, radiotherapy.
aEligibility criteria were confirmed during manual data curation.
Endpoints and analyses
Descriptive summaries of patient characteristics and NAC were contrasted with those in the KEYNOTE-522 control arm.
rwpCR was derived from clinician evaluation of resected breast specimen and sampled regional lymph nodes from first surgery following receipt of NAC. rwpCR was primarily determined using pathology reports evaluating residual disease in the breast and regional lymph nodes, with clinician documentation used if pathology reports were unavailable. Two definitions of rwpCR were used: rwpCR definition 1 (rwpCR1) covered post-NAC stage ypT0/Tis ypN0 (no tumor detected/carcinoma in situ, lymph node-negative) or clinical assessment of pCR with in situ disease explicitly noted. rwpCR definition 2 (rwpCR2) included post-NAC pathologic stage ypT0 ypN0 (no tumor detected, lymph node-negative) or clinical assessment of pCR. rwpCR1 was the key outcome compared with pCR outcomes within the KEYNOTE-522 control arm using Fisher’s exact test.
rwEFS was defined as the length of time from NAC initiation to the first event to occur. Events for the rwEFS analysis included: disease progression during the NAC period that precluded surgery and/or resulted in surgical margins with residual disease; local or distant recurrence at any site; diagnosis of a second primary cancer; death from any cause. rwEFS was estimated from the initiation of NAC using the Kaplan–Meier product-limit method. The Kaplan–Meier estimator was used to plot and compare the KEYNOTE-522 control arm with the distributions of rwEFS events for the first 36 months. No formal statistical analyses were conducted for comparison of rwEFS across subgroups and analyses represent numerical comparisons only.
Ethics
This study involved only retrospective analysis of de-identified patient data from an existing United States database, with no direct interaction with human participants. As such, institutional review board approval and informed consent were not required, in accordance with applicable regulations and ethical guidelines.
Results
Of the 27 875 patients in the Enriched Breast Cancer Cohort, 128 patients were eligible for inclusion in this study as the RW cohort. These patients had ODS-confirmed esTNBC and were confirmed to have received an NAC regimen similar to the control group of the KEYNOTE-522 study (Figure 1).
The baseline characteristics are presented in Supplementary Table S1, available at https://doi.org/10.1016/j.esmorw.2026.100698. A key difference between the RW cohort and the KEYNOTE-522 control arm was staging at diagnosis, where a higher proportion of patients in the KEYNOTE-522 control arm had stage II disease compared with the RW cohort (74.6% versus 43.8%, respectively), while a greater percentage of patients in the RW cohort had stage III disease (54.7% versus 25.1%). The RW cohort predominantly had grade 3 and ductal carcinoma (78.1% and 92.2%, respectively), whereas these data were unavailable for KEYNOTE-522.
Neoadjuvant chemotherapies
In the RW cohort, the most commonly used NAC regimen was doxorubicin + cyclophosphamide followed by carboplatin + taxane (63%). The second most common regimen was carboplatin + taxane followed by doxorubicin + cyclophosphamide (26%). Additionally, 10% of the RW cohort received carboplatin + taxane followed by another regimen (for all RW treatments see Table 1). In the RW cohort, most patients received mastectomy (66%) as their first primary site surgery followed by breast-conserving surgery (32%). Only two patients in the RW cohort refused primary surgery.
Table 1.
NAC regimens and first primary surgery site
| NAC regimen | RW cohort, n (%) n = 128 | KEYNOTE-522 control arm n (%) n = 390 |
|---|---|---|
| Doxorubicin + cyclophosphamide → carboplatin + taxane | 81 (63) | |
| Carboplatin + taxane → doxorubicin + cyclophosphamide | 33 (26) | 247 (63)a |
| Carboplatin + taxane → regimen other than doxorubicin + cyclophosphamide | 11 (9) | — |
| Carboplatin + taxane → epirubicin + cyclophosphamide | 1 (1) | 122 (31)a |
| Docetaxel + carboplatin → regimen other than doxorubicin + cyclophosphamide | 2 (2) | — |
| First primary site surgery | ||
| Mastectomy | 85 (66) | — |
| Breast-conserving surgery | 41 (32) | — |
| Surgery refused | 2 (2) | — |
Percentages represent the percentages with data and may not add up to 100% because of rounding.
NAC, neoadjuvant chemotherapy; RW, real-world.
Source: Schmid et al., 2020 (Supplement).12
Clinical outcomes
pCR
Across the two patient cohorts, a significantly higher proportion of patients in the KEYNOTE-522 control arm achieved pCR compared with those receiving a similar regimen in the RW cohort (Fisher’s exact test, P < 0.05). This was observed for both rwpCR1 (ypT0/Tis N0: 51.2% versus 37.5%) and rwpCR2 (ypT0 N0: 45.3% versus 32.0%) (Figure 2).
Figure 2.
Comparison of pCR achieved in the RW cohort versus the control arm of KEYNOTE-522. pCR1, pathologic complete response definition 1 (ypT0/Tis N0); pCR2, pathologic complete response definition 2 (ypT0 N0); RW, real-world; yp, post neoadjuvant therapy pathologic (T/N) value.
A subgroup analysis compared rwpCR1 rates across treatment, tumor and patient characteristics with those in the KEYNOTE-522 control arm, highlighting groups with the greatest alignment between cohorts. These subgroups included patients who completed NAC regimens (43.8% versus 51.2%), those with an ECOG performance score of 0 (49.1% versus 44.0%), and patients diagnosed at stage I or II (51.2% versus 43.1%) (Table 2).
Table 2.
pCR among RW cohort and control arm of KEYNOTE-522 in subgroups of interest
| Subgroup | RW cohort pCR1 n (%) | KEYNOTE-522 control arm pCR1 n (%) |
|---|---|---|
| Treatment characteristics | ||
| Recipients of CT + AC or AC + CT | ||
| Yes | 44 (42.3) | 103/201 (51.2) |
| No | 4 (16.7) | — |
| NAC completed | ||
| Yes, per treatment plan | 39 (43.8) | — |
| No, or unknown if completed | 9 (23.1) | — |
| Tumor characteristics | ||
| Nodal status | ||
| Positive | 22 (31.9) | 45/102 (44.1) |
| Negative | 26 (44.1) | 58/99 (58.6) |
| Primary tumor classification | ||
| T1 to T2 | 38 (44.7) | 84/149 (56.4) |
| T3 to T4 | 10 (23.3) | 19/52 (36.5) |
| Stage at diagnosis | ||
| I or II | 25 (43.1) | — |
| III | 23 (32.9) | — |
| Patient characteristics | ||
| Age, years | ||
| <65 | 42 (39.3) | 95/176 (54.0) |
| ≥65 | 6 (28.6) | 8/25 (32.0) |
| ECOG PS | ||
| 0 | 33 (44.0) | 85/173 (49.1) |
| 1 | 5 (21.7) | 18/28 (64.3) |
| Unknown | 10 (33.3) | — |
AC, doxorubicin + cyclophosphamide; CT, carboplatin + taxane; ECOG PS, Eastern Cooperative Oncology Group performance status; NAC, neoadjuvant chemotherapy; pCR, pathologic complete response; pCR1, abstracted pCR1 defined as ypT0N0 or ypTisN0 or clinician statement the patient achieved pCR with in situ disease explicitly noted; RW, real-world; T1 to T2, tumor stages 1 to 2; T3 to T4, tumor stages 3 to 4; yp, post-neoadjuvant therapy pathological (T/N) value.
rwEFS
The distribution of rwEFS during the first 36 months of follow-up in the RW cohort and KEYNOTE-522 control arm was similar at 75.0% (95% confidence interval 67.1% to 83.8%) and 76.8% (95% confidence interval 72.2% to 80.7%), respectively (long-rank P = 0.97) (Supplementary Figure S1, available at https://doi.org/10.1016/j.esmorw.2026.100698).
Additional estimates of 36-month rwEFS were conducted through subgroup analysis. These comparisons are descriptive without formal statistical testing. The 36-month rwEFS was numerically higher among patients who did not receive carboplatin + taxane followed by doxorubicin + cyclophosphamide OR doxorubicin + cyclophosphamide followed by carboplatin + taxane regimens. It was also numerically higher in those who did not receive adjuvant capecitabine and in patients who completed adjuvant treatment. Additionally, numerically higher rwEFS was observed among patients with negative nodal status, had T1/T2 tumors, or had unknown ECOG PS. No differences in 36-month rwEFS were observed across age groups and only minor differences were identified in patients with an ECOG PS of 0 versus 1 (Table 3).
Table 3.
36-Month EFS in the RW cohort in subgroups of interest
| Subgroup | rwEFS % (95% CI) |
|---|---|
| Treatment characteristics | |
| Recipients of CT + AC or AC + CT | |
| Yes (n = 104) | 74.0 (65.0-84.3) |
| No (n = 24) | 78.9 (64.1-97.2) |
| Adjuvant capecitabine | |
| Yes (n = 42) | 59.5 (45.1-78.5) |
| No (n = 84) | 84.2 (76.1-93.1) |
| NAC completed | |
| Yes, per treatment plan (n = 89) | 75.8 (66.2-86.7) |
| No, or unknown if completed (n = 39) | 72.8 (59.4-89.1) |
| Tumor characteristics | |
| Nodal status | |
| Positive (n = 69) | 73.1 (62.7-85.3) |
| Negative (n = 59) | 77.1 (65.5-90.6) |
| Primary tumor classification | |
| T1 to T2 (n = 85) | 85.0 (76.8-94) |
| T3 to T4 (n = 43) | 54.7 (40.1-74.5) |
| Patient characteristics | |
| Age, years | |
| <65 (n = 107) | 75.4 (66.9-85) |
| ≥65 (n = 21) | 74.7 (57.6-96.7) |
| ECOG PS | |
| 0 (n = 75) | 74.7 (64.1-86.9) |
| 1 (n = 23) | 67.7 (50.4-91) |
| Unknown (n = 30) | 80.8 (66.6-98) |
AC, doxorubicin + cyclophosphamide; CI, confidence interval; CT, carboplatin + taxane; ECOG PS, Eastern Cooperative Oncology Group performance status; EFS, event-free survival; NAC, neoadjuvant chemotherapy; pCR, pathologic complete response; pCR1, abstracted pCR1 defined as ypT0N0 or ypTisN0 or clinician statement the patient achieved pCR with in situ disease explicitly noted; RW, real-world; T1 to T2, tumor stages 1 to 2; T3 to T4, tumor stages 3 to 4; yp, post-neoadjuvant therapy pathological (T/N) value.
Adjuvant capecitabine subgroup
Adjuvant capecitabine was administered to 42 (32.8%) patients in the RW cohort but was not permitted in KEYNOTE-522. Of patients in the RW cohort who did not achieve rwpCR1 (n = 77), 53.2% received capecitabine and 45.5% did not. Although these results are comparable, a detailed comparison shows that 97.6% of patients who received adjuvant capecitabine did not achieve rwpCR1, while only 41.7% of patients who did not receive capecitabine failed to achieve rwpCR1. As such, only 1 in 42 patients (2.4%) who received capecitabine achieved rwpCR1 and 47 in 84 (56.0%) patients who did not receive capecitabine achieved rwpCR1. This may explain the 36-month rwEFS of 59.5% for patients treated with capecitabine (versus 84.2% among those who did not receive adjuvant capecitabine) (Supplementary Figure S2, available at https://doi.org/10.1016/j.esmorw.2026.100698).
Any first event
The incidence of any first event was similar between the RW cohort and the KEYNOTE-522 control arm, at 22.0% and 23.8% percent, respectively. Events including recurrence (local and distant), progression, secondary primary cancer, positive surgical margins, and death were generally comparable across both groups, with only minor variations (Table 4).
Table 4.
Summary of first events at 36 months in the RW cohort and KEYNOTE-522 control arm
| Variable | RW cohort, n (%) n = 128 | KEYNOTE-522 control arm, n (%) n = 390 |
|---|---|---|
| Any first event, n (%) | 28 (22.0) | 93 (23.8) |
| Distant recurrence | 15 (12.0) | 51 (13.1) |
| Local recurrence | 4 (3.1) | 17 (4.4) |
| Distant progression in neoadjuvant period | 4 (3.1) | 0 |
| Disease progression precluding surgery | 1 (0.8) | 15 (3.8) |
| Secondary primary cancer | 3 (2.3) | 4 (1.0) |
| Positive surgical margin (R1) | 1 (0.8) | 0 |
| Death | 0 | 6 (1.5) |
RW, real-world.
Discussion
This study leveraged a large RW database of patients with esTNBC, supplemented with a custom abstraction to characterize two RW endpoints (rwpCR and rwEFS) and benchmark them against outcomes within the KEYNOTE-522 control arm. While curating an RW endpoint from a secondary source is challenging due to the lack of standardized data capture and variability in clinical documentation, this retrospective database study demonstrated that a rigorous methodology can produce RW endpoints comparable to trial-defined endpoints in routine clinical practice.6
The rwpCR1 rate observed in the RW cohort was substantially lower than the pCR1 rate reported in the KEYNOTE-522 control arm, with improved alignment observed in the RW subgroups that received the same regimen as the trial control arm (carboplatin + taxane and anthracycline + cyclophosphamide regimens) or adhered to and completed their NAC regimen. As such, differences in pCR rates between the RW and trial cohorts primarily reflected broader treatment variability in the RW setting, while the subgroup findings highlight the role of treatment standardization and adherence in achieving more comparable outcomes between the settings.
Our comprehensive rwpCR definition incorporated dual criteria, reflecting a rigorous approach to endpoint capture. By systematically capturing data using standardized definitions aligned with KEYNOTE-522 trial criteria, we applied trial emulation principles to create a trial-like cohort structure within routine clinical practice, enabling meaningful benchmarking of RW effectiveness. Our findings align with recent RW comparative analyses, including a similar esTNBC study that achieved rwpCR rates consistent with KEYNOTE-522 results.17 This consistency across independent RW studies strengthens confidence that trial emulation principles combined with rigorous data curation enable comparability between RW and trial-based endpoints.
Adjuvant capecitabine is commonly used to treat patients with esTNBC with residual invasive disease following NAC treatment.18 Accordingly, receipt of adjuvant capecitabine was associated with non-achievement rwpCR1, likely reflecting its use in the setting of residual disease.18 While the use of adjuvant capecitabine in the RW cohort more accurately reflects RW treatment patterns, this approach was not standard at the time of KEYNOTE-522, likely contributing to the heterogeneity in the results observed across the RW cohort and the KEYNOTE-522 control arm. The definition of the rwpCR1 endpoint in the RW cohort was similar to that of the trial (defined as ypT0/Tis ypN0), however, the ascertainment methods differed. For example, there may have been heterogeneity in pathology assessments in community practices versus academic centers, which may have contributed to the variation in results. For the RW cohort, rwpCR1 was abstracted from reports, differing from the trial’s methodology. The abstraction, however, was quality controlled, followed strict procedures, and based on pathology reports, ensuring robustness.
Additionally, differences in baseline characteristics between the RW cohort and the KEYNOTE-522 control arm may have influenced outcomes. The RW cohort had a greater proportion of patients with more advanced tumors (T3-T4) than the KEYNOTE-522 control arm. As demonstrated by the subgroup analysis, a lower proportion of patients with larger tumors achieved rwpCR1 in the RW cohort and in the KEYNOTE-522 control arm. This is consistent with established literature demonstrating that advanced tumors are associated with poorer clinical outcomes for patients with TNBC.19 As such, the greater prevalence of larger tumors in the RW cohort compared with the KEYNOTE-522 control arm may have impacted the rates of rwpCR1 achieved. Furthermore, the RW cohort included a greater proportion of patients with an unknown ECOG PS, reflecting limited capture of ECOG PS in clinical practice in early-stage cancers, potentially underestimating the true prevalence of poor performance at diagnosis compared with the KEYNOTE-522 control arm. Among the patients with a recorded ECOG PS score, the RW cohort had a larger proportion of patients with poorer scores than the KEYNOTE-522 control arm, which may have contributed to lower treatment response rates given the predictive value of ECOG PS.20 Therefore, observed differences in rwpCR1 and pCR1 outcomes may be more reflective of variation in RW clinical practices such as treatment duration or local assessment protocols, rather than the data curation itself. Additionally, grade 3 ductal carcinoma predominated in the RW cohort, whereas equivalent pathological data were unavailable for KEYNOTE-522, likely contributing to outcome variations.
In contrast to the variation in rwpCR1 and pCR1 observed between the studies, rwEFS closely resembled the EFS results from the KEYNOTE-522 control arm. This may have been due to variation in adjuvant treatment assigned dependent on rwpCR1 results as patients in the RW cohort not treated with adjuvant capecitabine had a lower 36-month rwEFS rate than those who received adjuvant capecitabine. It should also be noted, however, that these results closely resemble the rwEFS observed in patients based on rwpCR1 status, rather than being solely attributable to capecitabine use. Therefore, the association between capecitabine use and rwEFS may more accurately reflect capecitabine as a surrogate marker for poorer rwEFS outcomes. As such, rwEFS was significantly higher for patients who achieved rwpCR1 than those who did not achieve rwpCR1. This highlights the significance of rwpCR1 as a prognostic indicator for rwEFS. Although the use of pCR as a potential surrogate endpoint for OS and disease-free survival has been debated, rwpCR1 demonstrates prognostic value for EFS in this RW cohort.21 Similarities in EFS in the RW cohort and KEYNOTE-522 control group may also relate to the relatively low risk of an EFS event in this patient population, coupled with the moderate reduction in risk of recurrence with NAC. Overall, these findings suggest that data from large databases supplemented through custom abstraction can be used to characterize rwEFS as a feasible and clinically relevant RW endpoint.
The distribution of first events, including distant recurrence, local recurrence, and secondary primary cancer, were also comparable between the RW cohort and the KEYNOTE-522 control arm. This similarity suggests that the methodology used of combining structured RW data from a large database with custom abstraction enabled the reliable capture of key clinical outcomes. As such, this supports the feasibility and clinical relevance of rwEFS as an RW endpoint and reinforces the utility of the same data curation approach in future studies that aim to report rwEFS as an outcome.
Limitations
While the current study provides valuable insights, there are limitations to consider. The RW cohort was not subject to randomization like the KEYNOTE-522 trial cohort, which may have introduced biases in estimating treatment effects. Additionally, the sample size of the RW cohort was relatively small (n = 128) compared with the KEYNOTE-522 control arm (n = 390), primarily due to the strict exclusion criteria applied (see Figure 1). This may have influenced the robustness of comparisons between the two groups as the smaller sample size can increase potential for variability or bias. The broader inclusion criteria for the RW cohort meant there were differences in tumor and patient characteristics in the RW cohort compared with the KEYNOTYE-522 control arm. While broader criteria were used in the RW cohort to better reflect true RW treatment practices for patients with esTNBC in the United States community practice setting, this may hinder the comparability of results. Additionally, missing clinical data in the RW cohort, such as ECOG PS and HER2 status, may further impact comparability. The unavailability of tumor grade and histology data for KEYNOTE-522 limits cohort comparability and may contribute to observed outcome variations. Finally, there may be unmeasured differences between the cohorts, for example, patients’ distance to care facility and treatment choices. As a result, the comparability of results between the RW cohort and the KEYNOTE-522 control arm should be interpreted with caution.
Conclusions
A greater proportion of patients in the KEYNOTE-522 control arm achieved pCR1 compared with the RW cohort. This was despite aligned definitions of pCR1 or rwpCR1 used across both the trial and the RW study. The reduced rwpCR1 rates in the RW cohort observed were likely due to variation in treatment practices rather than the use of data from a secondary source. rwEFS and the first events from the RW study closely approximated the results from KEYNOTE-522, likely due to relatively low absolute risk of an EFS event in this population. This study demonstrates that combining database-derived structured data with systematic abstraction using trial emulation principles provides a robust framework for benchmarking RW effectiveness against trial results, with potential applications to future comparative effectiveness research.
Acknowledgments
Funding
This work was supported by Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., 126 East Lincoln Ave., P.O. Box 2000, Rahway, NJ 0706, USA (no grant number).
Disclosure
XH, AH, JRE, RK, and WP are employees of Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA, who may own stock and/or hold stock options in Merck & Co., Inc., Rahway, NJ, USA. GIC, HM, TP, and AH contributed to the study and received support from Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA for their contributions to the research and manuscript preparation.
Data sharing
The full datasets generated and analyzed to inform the conclusions drawn within this manuscript during the current study are available from the corresponding author upon reasonable request.
Supplementary data
References
- 1.Howard F.M., Olopade O.I. Epidemiology of triple-negative breast cancer: a review. Cancer J. 2021;27(1):8–16. doi: 10.1097/PPO.0000000000000500. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Howard F.M., Pearson A.T., Nanda R. Clinical trials of immunotherapy in triple-negative breast cancer. Breast Cancer Res Treat. 2022;195(1):1–15. doi: 10.1007/s10549-022-06665-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Wu Q., Siddharth S., Sharma D. Triple negative breast cancer: a mountain yet to be scaled despite the triumphs. Cancers (Basel) 2021;13(15):3697. doi: 10.3390/cancers13153697. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Howlader N., Cronin K.A., Kurian A.W., Andridge R. Differences in breast cancer survival by molecular subtypes in the United States. Cancer Epidemiol Biomarkers Prev. 2018;27(6):619–626. doi: 10.1158/1055-9965.EPI-17-0627. [DOI] [PubMed] [Google Scholar]
- 5.Darrow J.J., Avorn J., Kesselheim A.S. FDA approval and regulation of pharmaceuticals, 1983-2018. JAMA. 2020;323(2):164–176. doi: 10.1001/jama.2019.20288. [DOI] [PubMed] [Google Scholar]
- 6.Griffith S.D., Miksad R.A., Calkins G., et al. Characterizing the feasibility and performance of real-world tumor progression end points and their association with overall survival in a large advanced non-small-cell lung cancer data set. JCO Clin Cancer Inform. 2019;3:1–13. doi: 10.1200/CCI.19.00013. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Ma X., Bellomo L., Magee K., et al. Characterization of a real-world response variable and comparison with RECIST-based response rates from clinical trials in advanced NSCLC. Adv Ther. 2021;38(4):1843–1859. doi: 10.1007/s12325-021-01659-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Stewart M., Norden A.D., Dreyer N., et al. An exploratory analysis of real-world end points for assessing outcomes among immunotherapy-treated patients with advanced non-small-cell lung cancer. JCO Clin Cancer Inform. 2019;3:1–15. doi: 10.1200/CCI.18.00155. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Blonde L., Khunti K., Harris S.B., Meizinger C., Skolnik N.S. Interpretation and impact of real-world clinical data for the practicing clinician. Adv Ther. 2018;35(11):1763–1774. doi: 10.1007/s12325-018-0805-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Lux M.P., Ciani O., Dunlop W.C.N., Ferris A., Friedlander M. The impasse on overall survival in oncology reimbursement decision-making: how can we resolve this? Cancer Manag Res. 2021;13:8457–8471. doi: 10.2147/CMAR.S328058. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Lee J. Current treatment landscape for early triple-negative breast cancer (TNBC) J Clin Med. 2023;12(4):1524. doi: 10.3390/jcm12041524. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Schmid P., Cortes J., Pusztai L., et al. Pembrolizumab for early triple-negative breast cancer. N Engl J Med. 2020;382(9):810–821. doi: 10.1056/NEJMoa1910549. [DOI] [PubMed] [Google Scholar]
- 13.Cortes J., Rugo H.S., Cescon D.W., et al. Pembrolizumab plus chemotherapy in advanced triple-negative breast cancer. N Engl J Med. 2022;387(3):217–226. doi: 10.1056/NEJMoa2202809. [DOI] [PubMed] [Google Scholar]
- 14.Guan D., Jie Q., Wu Y., Xu Y., Hong W., Meng X. Real-world data on breast pathologic complete response and disease-free survival after neoadjuvant chemotherapy for hormone receptor-positive, human epidermal growth factor receptor-2-negative breast cancer: a multicenter, retrospective study in China. World J Surg Oncol. 2022;20(1):326. doi: 10.1186/s12957-022-02787-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Krishnan J., Patel A., Roy A.M., et al. Real-world efficacy and adverse events of neoadjuvant immunotherapy in patients with early-stage triple-negative breast cancer: a multi-center experience. J Clin Oncol. 2024;42(suppl 16) [Google Scholar]
- 16.Amin M.B., Greene F.L., Edge S.B., et al. The eighth edition AJCC cancer staging manual: continuing to build a bridge from a population-based to a more “personalized” approach to cancer staging. CA Cancer J Clin. 2017;67(2):93–99. doi: 10.3322/caac.21388. [DOI] [PubMed] [Google Scholar]
- 17.Arnaud E., Vaflard P., Escalup L., et al. Treatment toxicities and pathological response through the evolution of neoadjuvant regimens in early triple-negative breast cancer. ESMO Real World Data Digit Oncol. 2025;9 doi: 10.1016/j.esmorw.2025.100157. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Ma Y., Zhu M., Zhang J., et al. Efficacy of adjuvant capecitabine in triple-negative breast cancer with residual disease after neoadjuvant therapy: a real-world study. Breast. 2025;81 doi: 10.1016/j.breast.2025.104477. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Liu Y., He M., Zuo W.J., Hao S., Wang Z.H., Shao Z.M. Tumor size still impacts prognosis in breast cancer with extensive nodal involvement. Front Oncol. 2021;11 doi: 10.3389/fonc.2021.585613. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Chalker C., Voutsinas J.M., Wu Q.V., et al. Performance status (PS) as a predictor of poor response to immune checkpoint inhibitors (ICI) in recurrent/metastatic head and neck cancer (RMHNSCC) patients. Cancer Med. 2022;11(22):4104–4111. doi: 10.1002/cam4.4722. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Conforti F., Pala L., Sala I., et al. Evaluation of pathological complete response as surrogate endpoint in neoadjuvant randomised clinical trials of early stage breast cancer: systematic review and meta-analysis. BMJ. 2021;375 doi: 10.1136/bmj-2021-066381. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.