This study was a multicenter, non-interventional, retrospective study conducted in five cancer centers across China. The study included patients aged over 18 who were newly diagnosed with HER2-positive breast cancer that was unresectable or metastatic between January 1, 2020 and August 31, 2022, and who had received at least one round of systemic treatment. Patients with other malignancies, or participation in unblinded clinical trials were excluded. The de-identified patient data were retrospectively derived from electronic medical records (EMRs). This study has been registered on ClinicalTrials gov (NCT05769751).

The primary outcome was real-world treatment patterns, defined as the distribution and sequence of various systemic therapy regimens across different treatment lines. The secondary outcome was, real-world progression-free survival (rwPFS), assessed the time from the commencement of the current treatment line to documented disease progression or death, whichever occurred first. rwPFS data for patients without disease progression or death as of the last follow-up date were censored at the time of the last tumor imagining evaluation. A line of treatment was defined as one regimen, possibly a combination of several drugs, given from the date of initiation until failure to control the disease, treatment discontinuation, changing of anti-HER2 drugs, or patient participation in a clinical trial. Other study variables collected included demographic and clinical-pathological characteristics (age, gender, metastatic lesions, Eastern Cooperative Oncology Group Performance status, hormone receptor status, HER2 status, prior therapies), as well as post-treatment outcomes (disease progression, subsequent treatments, vital status and date of last known contact/death).

Descriptive analyses were performed on demographic and clinical characteristics, treatment patterns, and treatment outcomes in the overall population and in subgroup of patients with brain metastasis. Number and percentage of patients were calculated for dichotomous and polychotomous variables. Means, standard deviations (SDs), medians, interquartile ranges (IQRs), minimum and maximum were provided for continuous variables. The distribution of 1L treatment choices in patient groups categorized by their time-to-relapse after completing anti-HER2 neo(adjuvant) therapy were compared using Fisher’s exact test. Sankey diagrams were created to visualize the sequence of systemic therapy regimens between treatment lines. Logistic regression analysis was performed to identify the factors influencing first-line therapy choices, with the quantitative association represented by odds ratios (ORs) with 95% confidence intervals (CIs). rwPFS along was estimated using the Kaplan-Meier methods and the survival curve was plotted. Cox proportional hazard models were used to evaluate factors impacting treatment outcomes. Variables included in the final multivariate model were selected according to the statistical significance in univariate analysis (cut-off p value < 0.05). Results are presented as hazard ratios (HRs) with 95% CIs. Statistical significance was set at a two-tailed p value < 0.05. All analyses were conducted using R software (Version 4.2.1s).

A total of 865 patients with newly diagnosed unresectable or metastatic breast cancer were included in the study. The median age of the patients was 53 years (IQR, 45.67–58.54 years), and 12.02% (n = 104) were aged ≥65 years. More than half of the patients were postmenopausal (55.72%). 62.89% of the patients were diagnosed with recurrent mBC, while 37.11% were de novo mBC. Of the 861 patients with metastases, 63.65% had visceral metastases, and 54.70% had more than 1 metastatic site. The most common metastatic sites were bone (n = 366, 42.31%) and lung (n = 344, 39.77%), followed by liver (n = 304, 35.14%) and brain (n = 95, 10.98%). At baseline, the HR status was positive in 322 (54.03%) patients, with HER2 3+ accounting for 74.33% (n = 443) of them. The median follow-up time was 12.84 months (IQR: 6.08–19.96 months). A complete summary of patient characteristics was presented in Table 1 .

Regarding the proportion of each systemic therapy regimen across treatment lines, 865 (100.00%) received at least 1 line of systemic therapy for mBC. In our data set, 38.15% had 2L treatment information, and 15.03% had post-2L treatments. Considering the complexity of real-world usage, the treatment regimens were categorized into six groups based on the type of anti-HER2 therapy: dual anti-HER2 blockade monoclonal antibodies (Dual anti-HER2 mAB), single anti-HER2 blockade mAB-based therapy (Single anti-HER2 mAB), single Tyrosine Kinase Inhibitor (TKI)-based therapy (Single TKI), TKI + anti-HER2 blockade mAB-based therapy (TKI + anti-HER2 mAB), antibody-drug conjugates (ADCs), and non-anti-HER2 therapy. The prominent medications were trastuzumab (83.39% of single anti-HER2 mAB), trastuzumab plus pertuzumab (98.45% of dual anti-HER2 mAB), and pyrotinib (95.33% of single TKI). Detailed medications of each regimen can be found in Supplementary Table S1 . 94.9% of the anti-HER2 mAB or TKI regimens utilized in the metastatic setting, and 96.7% in the 1L setting, were combined with chemotherapy (data not shown). The most common 1L treatment regimens were dual anti-HER2 mAB (36.07%), followed by single anti-HER2 mAB (21.97%), and single TKI (19.19%). In the 2L setting, single TKI was most common (35.45%), followed by TKI + anti-HER2 mAB (16.36%) and single anti-HER2 mAB (15.15%) ( Figure 1A ). Treatment selection in 3L and later lines showed considerable variability, as shown in Supplementary Table S2 . Approximately 12.00% of the total patients included were enrolled in unblinded clinical trials in the real world ( Supplementary Table S3 ). The transition patterns of anti-HER2 therapy from 1L to 2L were shown in Figure 2 . The predominant treatment option following single or dual anti-HER2 mAB was TKI. Post-TKI therapy typically involved either anti-HER2 mABs (dual or single) or ADCs. The detailed regimen transition, as well as the corresponding medications, from 1L to 2L, and subsequently to 3L, were shown in Supplementary Figure S1 .

To understand the factors influencing physicians’ treatment decisions, univariate and multivariate analyses were conducted, focusing on the relationship between baseline characteristics and 1L treatment regimens selection. Two comparisons based on predominant 1L and 2L treatment options were included in the analysis: TKI-based vs. non-TKI therapies and dual mAB vs. non-dual mAB therapies. Variables with statistically significance association on univariate analysis ( Supplementary Table S4 ) were included in the multivariate logistic regression model ( Table 2 ). Results of the multivariate analysis showed that the likelihood of opting for TKI-based regimen over non-TKI therapies in 1L was significantly higher in patients previously treated with anti-HER2 therapy in (neo)adjuvant setting [vs. HER2 treatment naïve, OR: 2.43 (1.61–3.69), p < 0.001], those with brain metastasis [vs. without brain metastasis, OR: 2.79 (1.76–4.46), p < 0.001], and patients who experienced a recurrence within 6 months [vs. ≥6 months, OR: 0.41 (0.26–0.65), p < 0.001] of post-(neo)adjuvant treatment. A higher preference for dual mAB regimen over non-dual mAB therapies in 1L was observed among patients with de novo mBC [recurrent mBC vs. de novo BC, OR: 0.32 (0.19–0.51), p < 0.001] at initial diagnosis, patients who experienced a recurrence post 6 months [vs. <6 months, OR: 2.60 (1.44–4.94), p = 0.002] of (neo)adjuvant treatment, those without brain metastasis [with vs. without brain metastasis, OR: 0.35 (0.18–0.61), p < 0.001], and patients aged less than 35 years [age ≥65 years vs. <35 years, OR: 0.48 (0.24–0.95), p = 0.04].

Previous treatment history was a key factor in the choice of 1L therapy by physicians. Figure 1B showed the distribution of 1L treatment choices. Among the 571 patients naïve to anti-HER2 treatment, dual anti-HER2 mAB was the most common choice (42.91%, n = 245). As a comparison, of the 294 patients previously treated with anti-HER2 therapy, most (n = 116, 39.46%) opted for single TKI, followed by dual anti-HER2 mAB (n = 68, 23.13%). Table 3 showed the 1L treatment choices in patients who were previously treated with anti-HER2 agents and relapsed. Of 217 patients who underwent single anti-HER2 mAB regimen in the (neo)adjuvant phase, 60.42% (n = 29 of 48) chose single TKI as 1L treatment following a relapse within 6 months. In contrast, for relapse occurring after more than 12 months, there was a higher preference for dual anti-HER2 therapy (37.88%, n = 50), followed by single TKI (23.48%, n = 31). In case of relapsing after dual anti-HER2 mAB, a higher preference for TKI was noted in 56.00% (n = 42 of 75) of the cases, regardless of prior efficacy. There was a significant difference in the distribution of 1L treatment choices in patients who received single anti-HER2 mAB (neo)adjuvant therapy (Fisher’s exact test, p < 0.001), but not in the group with dual anti-HER2 mAB (neo)adjuvant therapy (Fisher’s exact test, p = 0.09) ( Table 3 ).

Of the 865 patients who received 1L treatment, the median rwPFS was 11.04 (95% CI 10.19–12.03) months. In terms of specific therapy regimen, the median rwPFS of patients treated with dual anti-HER2 mAB was 13.57 (95% CI 12.03–17.52) months, followed by TKI+anti-HER2 mAB with a median rwPFS of 12.98 (95% CI 11.04–20.08) months ( Figure 3 ). The median rwPFS of each regimen in each treatment line after the 1L treatment were shown in Supplementary Figure S2 and Supplementary Table S5 . The longest estimated median rwPFS2 was observed among those who received a dual anti-HER2 mAB regimen in 1L and switched to a single TKI in 2L (24.32 months), albeit in a small sample ( Supplementary Table S6 ).

The association between baseline characteristics, therapy regimens and treatment outcomes of 1L was further evaluated. We first performed a univariate analysis and identified several factors significantly associated with treatment outcomes, including dual anti-HER2 mAB vs. non-dual anti-HER2 mAB, prior exposure to dual anti-HER2 therapy, and time from the end of (neo)adjuvant treatment to recurrence ( Supplementary Table S7 ). These factors were included in the multivariate cox regression model ( Table 4 ), with treatments categorized as TKI vs. non-TKI and dual mAB vs. non-dual mAB. A longer median rwPFS was associated with patients who experienced a recurrence more than 6 months post-(neo)adjuvant treatment compared to those recurrence within 6 months, in both two analysis group (HR: 0.65, p = 0.008, adjusted for TKI-based regimen; HR: 0.70, p = 0.03, adjusted for dual mAB-based regimen). Additionally, patients who did not receive a dual anti-HER2 mAB regimen had a significantly higher risk of progression or death compared to those receiving a dual anti-HER2 mAB-based regimen (HR: 1.60, 95% CI: 1.26–2.05, P < 0.001).