Main

Human epidermal growth factor receptor 2–positive (HER2+) breast cancer accounts for between 15% and 20% of all cases of breast cancer1,2. As many as 50% of patients with HER2+ advanced/metastatic breast cancer (mBC) develop brain metastases (BMs), which are associated with a poorer prognosis compared to patients who do not have BMs3,4,5,6.

Local therapy (including surgical resection, stereotactic radiosurgery (SRS), stereotactic radiotherapy and/or whole-brain radiation therapy (WBRT)) is recommended for BMs7; however, central nervous system (CNS) progression typically occurs within 6–12 months of treatment, and no extracranial benefit is conferred8,9,10,11. WBRT, currently recommended for treatment of multiple BMs, is associated with cognitive deterioration7,12; as some patients with HER2+ breast cancer and BMs can survive for several years, this is of particular concern13. As such, additional systemic treatment options for patients with BMs are needed. Trastuzumab-based therapy has long been the mainstay of systemic therapy for patients with HER2+ mBC, and several additional HER2-directed therapies have been investigated for the treatment of HER2+ mBC with BMs, including tucatinib14,15,16,17. Despite this, a large proportion of patients receiving treatment, including those with an initial response, eventually experience disease progression (commonly in the CNS)17.

Trastuzumab deruxtecan (T-DXd) is an antibody–drug conjugate (ADC) composed of a humanized immunoglobulin G1 monoclonal antibody specifically targeting HER2, a tetrapeptide-based cleavable linker and a potent topoisomerase I inhibitor payload18,19. On the basis of results from the randomized phase 3 DESTINY-Breast03 study, T-DXd is approved for the treatment of adult patients with unresectable or metastatic HER2+ breast cancer who have received a prior anti-HER2–based regimen in the metastatic setting or who have received a prior anti-HER2–based regimen in the neoadjuvant or adjuvant setting and developed disease recurrence during or within 6 months of completing therapy20,21.

Promising preliminary evidence of T-DXd intracranial efficacy was reported in a retrospective, exploratory pooled analysis of DESTINY-Breast01, 02 and 03. Patients with HER2+ mBC and stable (n = 104) or active (n = 44) BMs were treated with T-DXd22. The intracranial objective response rate (ORR) was 45.2% in patients with stable BMs and 45.5% in patients with active BMs. Median (95% confidence interval (CI)) CNS progression-free survival (CNS PFS) was 12.3 months (11.1–13.8) in patients with stable BMs and 18.5 months (13.6–23.3) in patients with active BMs22. Encouraging intracranial responses in patients with active (untreated or previously treated and progressing) BMs were also reported in the phase 1b/2 DESTINY-Breast07 study (n = 35); in the ongoing, five-cohort phase 2 DEBBRAH study (n = 13); in ROSET‑BM, a multicenter, retrospective, medical chart review study (n = 67); in the prospective, single-arm, single-center, phase 2 TUXEDO‑1 study (n = 15); and in a retrospective cohort analysis of heavily pretreated patients with BMs (n = 10)23,24,25,26,27.

Here we report results from the phase 3b/4 DESTINY-Breast12 study (NCT04739761), a non-comparative study that evaluated the efficacy and safety of T-DXd in patients with HER2+ mBC from two separate cohorts of patients with and without baseline BMs. DESTINY-Breast12 is, to our knowledge, the largest prospective study of T-DXd in patients with HER2+ mBC with previously treated and stable or active (untreated or previously treated and progressing) BMs.

Results

Patients

A total of 504 patients were treated across 78 sites between June 2021 and February 2024; 263 patients had baseline BMs, and 241 patients had no baseline BMs (Fig. 1). Of patients with baseline BMs, 157 had previously treated and stable BMs, and 106 had active BMs (39 had untreated BMs; 67 had previously treated BMs that were progressive at study entry (hereafter termed previously treated/progressing BMs), with no clinical indication for immediate retreatment with local therapy). Demographics and baseline disease characteristics for both cohorts are summarized in Table 1. Patients with baseline BMs received a median of 1.0 regimen (range, 0–4) of previous anti-cancer therapy in the metastatic setting, and 158 patients (60.1%) received prior intracranial radiotherapy (including 40 patients (15.2%) who had WBRT and 15 (5.7%) who had SRS; the type of intracranial radiotherapy was not always recorded by investigators, and only WBRT and SRS intracranial radiotherapy were reported). Median time from last intracranial radiotherapy to treatment initiation in patients with prior intracranial radiotherapy was 164 d (range, 9–2,115) overall and 116.5 d (range, 9–1,798) and 214.5 d (range, 15–2,115) in the stable (n = 90) and active (n = 68) BMs subgroups, respectively. The median follow-up duration in this cohort was 15.4 months (range, 0.1–30.0), and 118 patients (44.9%) were continuing to receive treatment at final data cutoff (8 February 2024). The most common reasons for discontinuation of study treatment included progressive disease (PD; 30.8%) and adverse events (AEs; 11.8%) (Fig. 1). Patients with no baseline BMs received a median of 1.0 regimen (range, 0–4) of previous anti-cancer therapy in the metastatic setting and had a median follow-up duration of 16.1 months (range, 0.8–28.4); 95 patients (39.4%) were continuing to receive treatment at final data cutoff. Primary reasons for discontinuation of study treatment included PD (35.7%) and AEs (7.5%) (Fig. 1).

Fig. 1: Patient disposition.
figure 1

COVID-19, coronavirus disease 2019; DCO, data cutoff.

Full size image
Table 1 Demographics and baseline clinical characteristics for patients with and without baseline BMs
Full size table

Response and progression in both cohorts were assessed by independent central review (ICR) per Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST 1.1).

Overall efficacy in the baseline BMs cohort

Overall PFS at 12 months was 61.6% (95% CI: 54.9–67.6) in all patients with BMs (Fig. 2a and Table 2) and 62.9% (95% CI: 54.0–70.5) and 59.6% (95% CI: 49.0–68.7) in patients with stable and active BMs, respectively (Table 2). Within the active BMs subgroup, PFS at 12 months was 47.0% (95% CI: 29.6–62.7) and 66.7% (95% CI: 53.4–76.9) in patients with untreated and previously treated/progressing BMs, respectively (post hoc analysis). Overall, 89 patients (33.8%) were free of progression at the time of the analysis, and median PFS (post hoc analysis) was 17.3 months (95% CI: 13.7–22.1). Time to progression (time from first dose until documented disease progression) data were immature, and the median was not calculated. Time to second progression (PFS2; time from first dose to second progression or death) data were immature, and the median was not reached. PFS2 at 12 months was 83.1% (95% CI: 77.5–87.4). Overall survival (OS) data were immature at the time of analysis (16.3% maturity); 12-month OS was 90.3% (95% CI: 85.9–93.4) (Fig. 2b and Table 2). In the baseline BMs full analysis set, confirmed ORR was 51.7% (95% CI: 45.7–57.8) (Table 2). A total of 11 patients (4.2%) had a complete response, and 125 (47.5%) patients had a partial response. Most responses (121/136) were reported by 6 months; at the time of analysis, response was ongoing in more than 50% of patients (n = 134 (51.0%)), and, therefore, median duration of response (DOR) was not calculated (Extended Data Fig. 1a). The ORR in patients with stable and active BMs was 49.7% (95% CI: 41.9–57.5) and 54.7% (95% CI: 45.2–64.2), respectively (Table 2). In a post hoc analysis restricted to patients with measurable disease at baseline (n = 198), confirmed ORR was 64.1% (95% CI: 57.5–70.8) overall and 67.0% (95% CI: 58.1–75.8) and 60.7% (95% CI: 50.5–70.8) in patients with stable (n = 109) and active (n = 89) BMs, respectively (Table 2). The best percentage change in target lesion size is shown in Fig. 3a.

Fig. 2: Kaplan–Meier analysis of key efficacy endpoints in patients with baseline BMs.
figure 2

a, Overall PFS. b, OS. c, CNS PFS per RECIST 1.1 as assessed by ICR. Tick marks indicate censored data. Analysis was based on the full analysis set.

Full size image
Table 2 Overall anti-tumor activity in patients with and without baseline BMs
Full size table
Fig. 3: Best percentage change in target lesions.
figure 3

a, Best percentage change from baseline in target lesion size in patients with baseline BMs and measurable disease at baseline (full analysis set). b, Best percentage change from baseline in CNS target lesion size in patients with baseline BMs and measurable CNS disease at baseline. c, Best percentage change from baseline in target lesion size, in patients with no baseline BMs and measurable disease at baseline (full analysis set). All patients had at least one post-baseline scan. Responses were assessed per RECIST 1.1 by ICR. A value of +20% was imputed as best percentage change from baseline if best percentage change could not be calculated because of missing data in the following situations: a patient had a new lesion or progression of non-target lesions or target lesions or a patient had withdrawn because of PD and had no evaluable target lesion data before or at PD. The dashed line indicates a 30% decrease in tumor size (partial response). Asterisks indicate imputed values.

Full size image

CNS efficacy in the baseline BMs cohort

Overall, CNS (including intracranial) progression was reported in 101 (38.4%) patients treated with T-DXd; CNS PFS at 12 months was 58.9% (95% CI: 51.9–65.3) (Fig. 2c and Table 2) and 57.8% (95% CI: 48.2–66.1) and 60.1% (95% CI: 49.2–69.4) in patients with stable and active BMs, respectively (Table 2). In total, 138 (52.5%) patients had measurable CNS disease at baseline (stable BMs: n = 77; active BMs: n = 61). The proportion of these patients with confirmed CNS ORR overall was 71.7% (95% CI: 64.2–79.3) and 79.2% (95% CI: 70.2–88.3) and 62.3% (95% CI: 50.1–74.5) in patients with stable and active BMs, respectively (Table 2). Within the active BMs subgroup, CNS ORR was reported in 19 out of 23 patients (82.6% (95% CI: 67.1–98.1)) and in 19 out of 38 patients (50.0% (95% CI, 34.1–65.9)) with untreated and previously treated/progressing BMs, respectively (post hoc analysis). The best percentage change in CNS target lesion size is shown in Fig. 3b.

Overall efficacy in the no baseline BMs cohort

The proportion of patients in the no baseline BMs full analysis set with confirmed ORR was 62.7% (95% CI: 56.5–68.8) (Table 2). A total of 23 patients (9.5%) had a complete response, and 128 (53.1%) patients had a partial response (Table 2). Most responses (140/151) were reported by 6 months; at the time of analysis, response was ongoing in more than 50% of patients (n = 151 (62.7%)), and, therefore, median DOR was not calculated (Extended Data Fig. 1b). In a post hoc analysis of patients with measurable disease at baseline (n = 215), confirmed ORR was 68.4% (95% CI: 62.2–74.6) (Table 2). The best percentage change in target lesion size is shown in Fig. 3c. OS data were immature at the time of analysis (17.0% maturity); 12-month OS was 90.6% (95% CI: 86.0–93.8) (Table 2). At 12 months, 72.1% (95% CI: 65.4–77.8) of patients had not experienced progression; time to progression data were immature, and the median was not calculated (Extended Data Fig. 2). Only four patients developed new symptomatic CNS metastases (incidence rate 0.017%; 95% CI: 0.00452–0.04250) (Table 2).

Safety: baseline BMs cohort

Median total treatment duration was 11.5 months (range, 0.1–26.9; Extended Data Table 1). The most common AEs included nausea, fatigue and constipation (Table 3). Grade 3 or higher AEs occurred in 134 (51.0%) patients, and the most common grade 3 or higher AEs were neutropenia (n = 43 (16.3%)), fatigue (n = 23 (8.7%)) and anemia (n = 19 (7.2%)). AEs led to treatment discontinuation in 40 (15.2%) patients (Extended Data Table 1); the most common AE leading to discontinuation was interstitial lung disease (ILD)/pneumonitis (n = 27 (10.3%)). Investigator-reported ILD/pneumonitis occurred in 42 patients (16.0%) with baseline BMs; most events were grade 1 (n = 26 (9.9%)), and there were six (2.3%) grade 5 events (Extended Data Tables 2 and 3). The median time to first onset of ILD/pneumonitis was 168.0 d (range, 35–646). Seven patients (2.7%) had a reported opportunistic infection (no systematic testing for infection was done); five patients had opportunistic infection reported as co-occurring with ILD/pneumonitis (ILD/pneumonitis events were grade 4 (n = 1) or grade 5 (n = 4); opportunistic infections were aspergillus (n = 1) and Pneumocystis jirovecii pneumonia (PJP)/infection (n = 4)). Two patients had opportunistic infections that were not reported to co-occur with ILD/pneumonitis (cytomegalovirus infection (n = 1) and PJP (n = 1)). Left ventricular ejection fraction decrease from baseline occurred in 31 patients (11.8%); no grade 4 or higher events were reported (Extended Data Table 2).

Table 3 Most common AEs by grouped and preferred term (≥20% of patients in either cohort)
Full size table

Safety: no baseline BMs cohort

Median total treatment duration was 12.0 months (range, 0.7–28.4; Extended Data Table 1). The most common AEs included fatigue, nausea and constipation (Table 3). Grade 3 or higher AEs occurred in 118 (49.0%) patients, and the most common grade 3 or higher AEs were neutropenia (n = 44 (18.3%)), fatigue (n = 24 (10.0%)) and anemia (n = 12 (5.0%)). AEs led to treatment discontinuation in 23 (9.5%) patients (Extended Data Table 1); the most common AE leading to discontinuation was ILD/pneumonitis (n = 13 (5.4%)). Investigator-reported ILD/pneumonitis occurred in 31 (12.9%) patients with no baseline BMs; most events were grade 1 (n = 22 (9.1%)), and there were three (1.2%) grade 5 events (Extended Data Tables 2 and 3). The median time to first onset of ILD/pneumonitis was 169.0 d (range, 24–484). Left ventricular ejection fraction decrease occurred in 26 (10.8%) patients; there were no grade 3 or higher events (Extended Data Table 2).

Discussion

DESTINY-Breast12 is, to our knowledge, the largest prospective study reporting intracranial activity of T-DXd in patients with HER2+ mBC and baseline BMs. This phase 3b/4 study was designed to collect data from settings that resemble real-world clinical practice, to provide a detailed understanding of T-DXd outcomes in patients previously treated with HER2-targeted agents.

PFS was selected as the primary endpoint in the BMs cohort because it was anticipated that a large proportion of this patient population may have no measurable disease at baseline and to minimize any potential confounding effect from prior locally directed therapy28. The 12-month overall PFS rate was 61.6% (95% CI: 54.9–67.6) in patients with baseline BMs. Overall ORR (including patients with no measurable disease at baseline) was lower in patients with stable BMs (49.7%) compared to patients with active BMs (54.7%); however, in line with clinical expectations, a post hoc analysis of ORR in patients with measurable disease at baseline revealed a higher ORR in the stable BMs subgroup (67.0%) versus the active BMs subgroup (60.7%). The different trends observed are likely explained by the imbalance of patients with measurable disease between the two subgroups.

Currently, tucatinib in combination with trastuzumab and capecitabine is the preferred systemic therapy for previously treated patients with HER2+ mBC and active BMs7,29,30. The phase 2 HER2CLIMB study investigated trastuzumab and capecitabine with either placebo or tucatinib in patients with previously treated HER2+ mBC. Patients in the tucatinib arm of the total population (N = 612) were heavily pretreated at baseline (median of three (range, 1–14) previous therapy regimens in the metastatic setting31,32). In an updated exploratory subanalysis of patients with measurable baseline BMs (n = 198), confirmed intracranial ORR was 47.3% (95% CI: 33.7–61.2) in patients with active BMs receiving tucatinib, capecitabine and trastuzumab. Median CNS PFS by investigator per RECIST 1.1 was 9.9 months (95% CI: 8.4–11.7) overall and 9.6 months (95% CI: 7.6–11.1) and 13.9 months (95% CI: 9.7–24.9) in patients with active and stable BMs, respectively32. In DESTINY-Breast12, tucatinib as a previous regimen was exclusionary, to avoid any confounding effect from a drug known to be active on CNS lesions.

Because of the decreased quality of life and poor prognosis observed in patients with BMs3,4,5,6, additional treatment options for this patient population are needed, particularly for later lines of therapy. CNS activity with tucatinib (a small-molecule tyrosine kinase inhibitor) is well established; however, questions remain regarding the intracranial efficacy of ADCs, including T-DXd, in patients with active BMs. Despite its large molecular size, promising CNS activity for T-DXd was previously reported. The 12-month PFS for the overall BMs population in DESTINY-Breast12 was similar to that observed in an exploratory analysis of patients with stable and active BMs enrolled in DESTINY-Breast03 (72.0% (95% CI: 55.0–83.5); n = 43), a phase 3, randomized, open-label study that investigated T-DXd versus trastuzumab emtansine (T-DM1) in patients with HER2+ mBC previously treated with trastuzumab and a taxane33.

Promising CNS activity in patients with active BMs treated with T-DXd was observed in small prospective studies24,26. In the phase 2 DEBBRAH (n = 13) and TUXEDO-1 (n = 15) studies, intracranial ORR per Response Assessment in Neuro-Oncology (RANO)-BM criteria was 46.2% (95% CI: 19.2–74.9) and 73.3% (95% CI: 48.1–89.1), respectively24,26. In an interim analysis of the dose-expansion phase of the ongoing DESTINY-Breast07 study (n = 35), which assessed T-DXd monotherapy in patients with HER2+ mBC and active BMs in the first-line or second-line setting, PFS at 12 months was 75.0% (80% CI: 63.5–83.4), and median PFS was 19.5 months (80% CI: 19.4–24.3) (Anders et al.23). Results from DESTINY-Breast12 extend these observations to a larger group of patients with active BMs (12-month PFS: 59.6% (95% CI: 49.0–68.7)), including those with untreated BMs (47.0% (95% CI, 29.6–62.7)) and previously treated/progressing BMs (66.7% (95% CI, 53.4–76.9)).

CNS ORR in our study was 71.7% overall and 79.2% and 62.3% in patients with stable and active BMs, respectively. Within the active BMs subgroup, CNS ORR was 82.6% in patients with untreated BMs and 50.0% in patients with previously treated/progressing BMs. These results are numerically higher than those observed in the pooled DESTINY-Breast01, 02 and 03 analysis of patients with treated/stable (45.2%; n = 104) or active (45.5%; n = 44) BMs22. This may be reflective of the heavily pretreated population included in the pooled analysis (median 3.0 prior treatment regimens in the metastatic setting versus 1.0 for DESTINY-Breast12). In DESTINY-Breast12, responses in patients with baseline BMs were durable, despite a relatively short follow-up duration (15.4 months).

Results from clinical studies have also been corroborated by real-world evidence. In the retrospective ROSET-BM study, 12-month PFS was 62.0% (95% CI: 47.8–73.4) in patients with active BMs (n = 67) and 71.4% (95% CI: 33.7–90.1) in patients with stable BMs (n = 12), in line with the results of the current study25. Patients with leptomeningeal metastases (LM) were excluded from DESTINY-Breast12. However, T-DXd showed sustained activity in two small retrospective studies in patients with mBC and LM: ROSET-BM (n = 19) and a small case series (n = 8)25,34. Further investigation is needed to confirm the efficacy of T-DXd in this patient population.

ORR was chosen as the primary endpoint for the cohort of patients without baseline BMs as it is an early indicator of treatment effect and allowed for early assessment of T-DXd benefit. In this cohort, overall efficacy was in line with prior reports; however, the proportion of patients with complete responses (9.5%) was lower than that reported in DESTINY-Breast03 (21%)35. CNS as a site of symptomatic progression was very uncommon in the non-BMs cohort of DESTINY-Breast12.

Overall, the safety profile of T-DXd was consistent with previous reports22,23, with no new safety signals identified. Discontinuation rates due to AEs were low (15.2% and 9.5% for patients with and without BMs, respectively). Regarding the rates of decreased left ventricular ejection fraction (11.8% and 10.8% for patients with and without BMs, respectively), most cases were grade 1 or grade 2, with only two grade 3 or higher events reported in the BMs cohort.

Lack of an adjudication committee in DESTINY-Breast12 limits direct comparison of ILD/pneumonitis rates with those from previous clinical studies that included an adjudication committee within the protocol. However, rates of ILD/pneumonitis events observed in both cohorts (16% (grade 5: 2%) and 13% (grade 5: 1%) of patients with and without baseline BMs, respectively) were consistent with T-DXd data for patients with HER2+ mBC in the DESTINY-Breast01 (16%; grade 5: 3%), DESTINY-Breast02 (10%; grade 5: <1%) and DESTINY-Breast03 (15%; grade 5: 0) studies35,36,37. Most cases of ILD/pneumonitis were mild or moderate; however, six deaths in the baseline BMs cohort and three deaths in the no baseline BMs cohort were judged by investigators to be caused by ILD/pneumonitis. Although opportunistic infections were not systematically tested, five cases of opportunistic infection were reported as co-occurring with ILD/pneumonitis (one grade 4 event and four grade 5 events) in the baseline BMs cohort. Clinical and radiologic features of drug-induced ILD/pneumonitis can resemble infectious etiology38; in patients with co-occurring opportunistic infection and ILD/pneumonitis, differentiating the underlying cause of pulmonary toxicity can be challenging, and drug-induced ILD is a diagnosis of exclusion39. These results highlight the need to consider PJP prophylaxis in patients taking chronic corticosteroids. Prompt initiation of steroidal treatment in patients with suspected ILD/pneumonitis is required in accordance with current guidelines, and T-DXd should be interrupted as a precaution until the etiology is confirmed39,40. Delays in providing this treatment (for example, waiting for results of blood culture tests) should be avoided where possible to minimize worsening of ILD/pneumonitis and associated fatalities in this patient population. Where ILD/pneumonitis is suspected, the possibility of infectious etiology should be explored subsequent to immediate treatment to inform future treatment decisions.

Ongoing studies are further defining the potential CNS efficacy of T-DXd in settings beyond HER2+ mBC, including HER2-low breast cancer in the phase 2 TUXEDO-4 study41. After the pan-tumor approval of T-DXd in HER2+ solid tumors42, exploring CNS efficacy of T-DXd outside of breast cancer may be informative. Other ADCs are being tested in prospective clinical studies, including datopotamab deruxtecan in patients with breast cancer and BMs or LM (TUXEDO-2 and DATO-BASE) and patritumab deruxtecan in patients with breast cancer and BMs, non-small cell lung cancer and BMs or solid tumors and LM (TUXEDO-3)43,44,45.

Limitations of DESTINY-Breast12 include the open-label, single-arm study design and exclusion of patients with LM. Few relevant historical cohorts for comparison were available at the time of study protocol development in 2019. Efficacy conclusions relied on single-arm time-to-event efficacy analyses. The immaturity of the final dataset makes cross-trial comparisons challenging, and no long-term follow-up is planned. For the non-BMs cohort, ORR was the primary endpoint despite including patients with no measurable disease at baseline, and PFS was not investigated. A proportion of patients with stable BMs and those with active BMs who were previously treated and progressing had prior intracranial radiotherapy, which may have impaired assessment of target lesions. Patients without baseline BMs did not undergo regular brain imaging; therefore, only incidence of symptomatic CNS metastases could be investigated in that cohort. Patients with Black and Asian ethnicities were underrepresented in the treated population. Patient-reported and neurocognitive outcomes were recorded as part of the study, and these analyses will be reported in future reports.

The results of the DESTINY-Breast12 study indicate the CNS efficacy of T-DXd in a large, prospective patient cohort. Without a direct comparison between T-DXd and the tucatinib, trastuzumab and capecitabine regimen, treatment selection for previously treated patients with HER2+ mBC and BMs should be balanced between efficacy and toxicity considerations on an individual basis.

In conclusion, T-DXd showed substantial and durable overall and intracranial clinical activity in patients with HER2+ mBC, including a large cohort with stable and active BMs. No new safety signals were identified. ILD/pneumonitis remains an important identified safety risk of T-DXd. These results support the use of T-DXd for previously treated patients with HER2+ mBC, including those with stable and active BMs.

Methods

Inclusion and ethics

This study was approved by the institutional review board or ethics committee at each investigational site before initiation (Supplementary Information). This study was performed in accordance with International Council for Harmonisation Good Clinical Practice guidelines, the Declaration of Helsinki and local regulations on the conduct of clinical research. An independent data monitoring committee was responsible for monitoring patient safety during the study. Patients provided written informed consent before participating in the study. Patients were eligible for inclusion regardless of sex or gender.

Study design and treatment

We conducted a prospective, open-label, single-arm, multicenter, international phase 3b/4 study involving patients with pathologically documented HER2+ advanced or metastatic breast cancer with or without baseline BMs. HER2+ expression was locally confirmed as determined by American Society of Clinical Oncology–College of American Pathologists guidelines46.

Patients were eligible if they were aged 18 years or older, had disease progression on one or more prior anti-HER2–based regimens, received no more than two prior therapy regimens in the metastatic setting (had to be tucatinib naive) and had an Eastern Cooperative Oncology Group performance status (ECOG PS) score of 0 or 1. Patients with known or suspected LM were excluded.

Patients were allocated to one of two cohorts: those with baseline BMs (previously treated stable BMs and active (untreated or previously treated and progressing) BMs) and those with no evidence of BMs at baseline. Patients with no measurable disease at baseline were permitted to enroll. Stable BMs were defined as BMs radiographically stable for ≥4 weeks since completion of treatment; active BMs were defined as untreated BMs with lesions ≤2 cm or BMs that had progressed since local CNS therapy, with no clinical indication for immediate retreatment with local therapy. Washout periods before the first day of dosing were ≥7 d and ≥21 d for SRS or gamma knife and WBRT, respectively. Patients who received local therapy for isolated CNS progression in either cohort could continue study treatment until a second progression (brain or body) was observed, upon which study treatment was discontinued. Concomitant use of ≤3 mg of dexamethasone daily or equivalent was permitted in patients with baseline BMs. T-DXd was administered intravenously every 3 weeks (21-d cycle) at a dose of 5.4 mg per kg of body weight until RECIST 1.1-defined47 disease progression outside the CNS, unless there was unacceptable toxicity or withdrawal of consent or another criterion for discontinuation was met.

Tumor assessments of the chest, abdomen (including the entire liver and both adrenal glands) and pelvis used images from computed tomography (CT) or magnetic resonance imaging (MRI; with intravenous (IV) contrast) collected at screening/baseline and every 6 weeks for the first 48 weeks and 9 weeks thereafter during study intervention. For patients with baseline BMs, MRI (with and without IV contrast) or contrast-enhanced CT images of the brain were collected for all patients at baseline and every 6 weeks for the first 48 weeks and 9 weeks thereafter during study intervention. Patients with active and measurable BMs had intracranial lesions included as target lesion(s) for RECIST 1.1 CNS assessments.

Endpoints

The primary endpoint for patients with baseline BMs was PFS (time from first dose to disease progression or death (by any cause in absence of progression)). Secondary endpoints included CNS PFS (time from first dose to CNS progression or death); OS (time from first dose to death by any cause); ORR (proportion of patients with confirmed complete or partial response); PFS2 (time from first dose to second progression or death); time to progression (time from first dose until documented disease progression); CNS ORR (proportion of patients with measurable BMs at baseline with confirmed complete or partial response of brain lesions); DOR (time from first documented confirmed response until documented progression or death by any cause); and safety. The primary endpoint for patients with no baseline BMs was ORR. Secondary endpoints included DOR, OS, time to progression, incidence of new symptomatic CNS metastases (number of new symptomatic CNS metastasis during study intervention period / total number of patients without symptomatic CNS metastasis at baseline) and safety. Response and progression in both cohorts were assessed by ICR per RECIST 1.1. Additional prespecified secondary endpoints not reported in this analysis are site of next progression (CNS versus extracranial versus both), duration of treatment on subsequent lines of therapy, patient-reported outcomes (European Organization for the Research and Treatment of Cancer 30-item core quality of life questionnaire (EORTC QLQ-C30), Neurologic Assessment in Neuro-Oncology (NANO) scale, cognitive tests and St. George’s Respiratory Questionnaire–idiopathic pulmonary fibrosis version (SGRQ-I; patients with ILD/pneumonitis only)) in both cohorts and time to new CNS lesions, CNS DOR and MD Anderson Symptom Inventory (MDASI) Symptom Diary (brain tumor-specific outcomes) in the BMs cohort only.

Safety

AEs were coded using Medical Dictionary for Regulatory Activities version 26.1 preferred terms and graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE) version 5.0. For potential cases of ILD or pneumonitis, study intervention was interrupted and a full investigation was carried out, based on the investigator’s judgment and sponsor review by medical monitor and study safety physician. Adjudication of reported ILD/pneumonitis cases by a separate committee was not conducted in this study; an ILD advisory committee reviewed the diagnosis and management of ILD/pneumonitis cases (outside the parameters of the study). Pulmonary toxicity management guidelines were described previously48.

Statistical analysis

Data analyses were completed using SAS software version 9.4. This single-arm study was not designed to test any prespecified hypothesis; therefore, no formal sample size calculation was performed. The sample size was chosen based on precision estimates for the primary endpoint in each cohort. Assuming an underlying PFS in the BMs cohort and an underlying ORR in the non-BMs cohort in line with available data at the time of study design, a sample of 250 participants in each cohort ensured that the one-sided width of a two-sided 95% CI for each endpoint would not exceed 6.3%. Efficacy analyses were conducted in the full analysis set (defined as all patients who were enrolled in the study and received at least one treatment dose), and no data were excluded. Safety data are reported for the safety analysis set (identical to the full analysis set). Analyses were performed separately by cohort, and no comparison of results between the two cohorts was planned. Safety analyses were descriptive only. PFS, OS, CNS PFS (all 12-month rates), PFS2 and DOR were analyzed by the Kaplan–Meier method. CIs for median PFS were derived based on the Brookmeyer–Crowley method. ORR was assessed using data obtained from first dose until progression, or the last evaluable assessment in the absence of progression, regardless of whether patients withdrew from therapy. CNS ORR was assessed using data obtained from first dose until brain progression, or the last evaluable assessment in the absence of brain progression, regardless of whether patients withdrew from therapy. For PFS, patients who had not progressed or had died by the time of analysis were censored at the time of the latest date of assessment from their last evaluable RECIST 1.1 assessment. Patients who progressed or died immediately after two or more consecutive missed visits were censored at the time of the latest evaluable RECIST 1.1 assessment before the two missed visits. For CNS PFS, patients who had systemic progression, but no CNS progression, were censored at the time of the progression assessment; the analysis did not account for systemic progression as a competing event. For OS, patients not known to have died at the time of analysis were censored on the last recorded date on which the patient was known to be alive. For ORR and CNS ORR, patients who stopped treatment without a response or progression, received a subsequent therapy and then responded were not included as responders. Prespecified subgroup analyses of the full analysis set were conducted for patients with active and stable BMs in the baseline BMs cohort, and descriptive statistics are provided.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.