Long term outcome data from the EORTC 75111-10114 ETF/BCG randomized phase II study: Pertuzumab and trastuzumab with or without metronomic chemotherapy for older patients with HER2-positive metastatic breast cancer, followed by T-DM1 after progression.

INTRODUCTION: Older patients are at higher risk of chemotherapy-induced toxicity, raising interest in less toxic anti-HER2 regimens for older persons with HER2-positive (HER2+) metastatic breast cancer (MBC). PATIENTS AND METHODS: This phase II study randomized (1:1) patients with HER2+ MBC, aged 70+ or frail 60+, to first line chemotherapy with metronomic oral cyclophosphamide (M) + Trastuzumab (T) and Pertuzumab (P) or TP alone. T-DM1 was offered in case of progression.
RESULTS: In total, 39 and 41 patients were randomized to TP and TPM arm respectively. Median follow-up is 54.0 months. 24-month PFS was 18.7% (95% CI 8.2-32.4) and 28.7% (95% CI 15.8-43.0), respectively. A total of 49 (61.3%) patients died of whom 37 (75.5%) from disease progression; number of deaths per arm was 27 (69.2%) for TP and 22 (53.7%) for TPM. There was no significant difference in OS between the two arms (median OS TP vs TPM: 32.1 vs 37.5 months, p 0.25). Among the 40 patients who have started T-DM1 after disease progression on TP/TPM, PFS rate at 6 months after start of T-DM1 was 43.6% (95% CI: 27.7-58.5) and grade 3 or higher AE occurred in 18 pts (45%).
CONCLUSIONS: Metronomic chemotherapy-based dual blockade (TPM), followed by T-DM1 after progression, provides an active and relatively well tolerated treatment option in an older/frail HER2+ MBC population, with a median survival of over 3 years. Nevertheless, the majority of this older/frail population died from breast cancer, highlighting the need for well tolerated and efficacious treatments in these patients.

Introduction

Approximately one third of all breast cancers are diagnosed in women over 70 years of age. This number is expected to rise due to the increasing life expectancy of the general population [1,2]. More than half of deaths due to breast cancer occur in patients over 65 years of age [[3], [4], [5]].

There is a considerable heterogeneity in the general health status and life expectancy of older patients. Guidelines published by the International Society of Geriatric Oncology (SIOG) recommend screening for frailty in patients aged ≥70 years, and tailoring treatment based on whether patients are grouped as fit, susceptible (pre-frail) or frail [6].

Although the proportion of HER2 positive tumors is generally thought to be lower in older patients, still 10–15% present with breast cancers with HER2 overexpression [[7], [8], [9]]. The prognosis of patients with metastatic HER2+ breast cancer has dramatically improved over the past decade, mainly due to the introduction of more effective HER2-directed therapies [10]. This includes the addition of pertuzumab to trastuzumab and docetaxel as first line treatment and the introduction of trastuzumab emtansine (T-DM1), an antibody drug conjugate targeting HER2, in second and later lines [[11], [12], [13]]. Unfortunately, older women were underrepresented in the pivotal trials for these agents (CLEOPATRA: 16% > 65 y; EMILIA: 14% > 65 y) and those that were included were highly selected and not representative of the wider population of older patients with breast cancer [14].

In the EORTC 75111-10114 study, we investigated the addition of metronomic oral cyclophosphamide to trastuzumab-pertuzumab as first line treatment in older women with HER2+ metastatic breast cancer. The results of the primary analysis have been published previously [15]. The trial met its primary endpoint, with an estimated progression free survival at 6 months of 46·2% (95% CI 30·2–60·7) with trastuzumab and pertuzumab versus 73·4% (56·6–84·6) with trastuzumab and pertuzumab plus cyclophosphamide (hazard ratio [HR] 0·65 [95% CI 0·37–1·12], p = 0·12). Here, we present the final analysis of the long-term outcomes, including outcomes of patients who crossed over on T-DM1 as part of protocol treatment after disease progression on first line treatment.

Methods

Study design and participants

EORTC 75111-10114 was an open-label, 1:1 randomised, investigator-initiated, phase 2, selection trial which involved 30 institutions from eight countries in Europe. The study design has been previously described [15].

Briefly, eligible patients had histologically proven HER2-positive metastatic breast cancer and were either 70 years or older or 60 years or older with functional restrictions according to the Instrumental Activities of Daily Living (IADL), Activities of Daily Living (ADL) or the Charlson Comorbidity Index (CCI) score. Prior chemotherapy for metastatic disease was not allowed, however patients could have received up to one line of anti-HER therapy (trastuzumab or lapatinib) as well as endocrine therapy in case of hormone sensitive metastatic breast cancer.

Patients were also required to have measurable disease as per Response Evaluation Criteria In Solid Tumors, version 1.1 (RECIST 1.1) or evaluable disease; a performance status according to world health organization scale (WHO) of 0–3; a left ventricular ejection fraction (LVEF) of 50% or greater; no history of significant cardiac disease and adequate bone marrow, liver, and renal function.

Procedures

Patients were randomised 1:1 to receive either intravenous trastuzumab (loading dose of 8 mg/kg, followed by 6 mg/kg every 3 weeks) and intravenous pertuzumab (loading dose of 840 mg, followed by 420 mg every 3 weeks) (TP) versus the same treatment in combination with metronomic oral cyclophosphamide 50 mg per day without interruption (TPM) until disease progression or unacceptable toxicity. After disease progression, patients were given the option of receiving intravenous trastuzumab emtansine (T-DM1) as part of the protocol treatment at the registered dose of 3·6 mg/kg every 3 weeks. Neither patients nor investigators were masked to treatment allocation.

The study was done in accordance with the protocol, good clinical practice guidelines, and the provisions stated in the Declaration of Helsinki. All patients provided written informed consent. Further procedure information is included in Appendix B.

Outcomes

The primary endpoint was investigator-assessed progression-free survival at 6 months from the date of randomisation. Other secondary endpoints included overall survival, breast-cancer specific survival and overall response. The purpose of this follow-up analysis is to provide long term follow-up data on the endpoints reported in the primary analysis and to report on endpoints on T-DM1 treatment, including progression free survival on T-DM1 (defined as the time from the start of T-DM1 to further disease progression or death), tumour response on T-DM1 and toxicity. Finally, we also report on prognostic factors. Data on the health related quality of life (HRQoL) are reported separately, based on the database used for the primary analysis [16].

Statistical analysis

The trial followed a Sargent and Goldberg screening design with 2 arms: TP versus TPM. For the primary analysis, both treatment arms were compared for the progression free survival rate at 6 months, with the aim of assessing whether one of the two treatments seemed superior to the other one. If the difference in estimate of 6-month PFS was 10% or more, the better arm would be selected.

Efficacy analyses were done on the intention-to-treat population (all randomised patients) and safety analyses on the safety population (all patients who received at least one dose of study treatment). Efficacy and safety analyses for T-DM1 were performed in the T-DM1 population (all patients who started T-DM1 as part of protocol treatment).

For this final report, the same statistical analysis plan as for the primary analysis was used including the definition of endpoints and statistical methods. Prognostic factor analysis related to PFS, OS and BCSS were performed on the intention to treat population and included baseline geriatric assessment (G8, IADL, ADL, social situation and SPBB frailty index), together with age, WHO performance status, ER status, PgR status, previous HER2 treatment and organ involvement. Further details on statistical analysis are available in Appendix B.

All statistical analyses were performed using SAS software version 9.4 (SAS Institute).

Results

Between July 2, 2013, and May 10, 2016, 80 patients were randomly assigned 1:1 to receive trastuzumab and pertuzumab (n = 39) or trastuzumab and pertuzumab plus metronomic oral cyclophosphamide (n = 41). All patients started their allocated treatment and could thus be analysed for efficacy and safety. After disease progression, 40 patients started T-DM1 as part of the protocol treatment (50%), 18 patients in the TP arm (46.1%) and 22 patients in the TPM arm (53.7%).

At the time of final data cut-off (March 26, 2021), all patients in the TPM group had discontinued their first line treatment, while 1 patient in the TP group remained on treatment. The main reason for treatment discontinuation was progressive disease (65.8% in the TP group and 60.0% in the TPM group). In the T-DM1 population, 1 patient was still on treatment with T-DM1. Most patients in the T-DM1 population discontinued treatment because of progressive disease (71.8%). The median follow-up from randomisation in the intention to treat population has increased from 20·7 months (IQR: 12·5–30·4) at the time of the primary analysis to 54.0 months (IQR: 39.6–58.2) at final data cut-off. Median follow-up in the T-DM1 population was 33.7 months (IQR: 26.4–38.0) from the start of T-DM1.

Baseline characteristics were generally well balanced between the treatment groups (Table 1). The median age was 76.7 years (range 61.4–91.4), 19 patients had a WHO performance status of 2–3 (23.8%). Geriatric assessment showed a G8-score≤14 in 55 patients (68.8%) and an SBBP score≤9 in 57 patients (71.3%). In patients who discontinued treatment, the median number of cycles of trastuzumab and pertuzumab was 6 in the TP group (range 1–81) and 13.5 in the TPM group (range 1–74). The median number of cycles of cyclophosphamide was 13 in the TPM group (range 1–70) and the median number of cycles of T-DM1 in the T-DM1 population was 7 (range 1–50).

Table 1

Baseline characteristics.

Baseline characteristics	TP (n = 39)	TPM (n = 41)
Median age, years (range)	76.2 (61.4–91.4)	77.3 (67.7–89.6)
Hormone receptor positivity
ER and PgR negative	12 (30.8)	13 (31.7)
ER and/or PgR positive	27 (69.2)	28 (68.3)
Prior (neo)adjuvant chemo (±antiHER2) therapy
No	29 (74.4)	36 (87.8)
Yes	10 (25.6)	5 (12.2)
Prior anti-HER2 therapy for metastatic disease
No	36 (92.3)	37 (90.2)
Yes	3 (7.7)	4 (9.8)
Visceral involvement
No	1 (2.6)	4 (9.8)
Yes	38 (97.4)	36 (87.8)
WHO performance status
0	10 (25.6)	17 (41.5)
1	17 (43.6)	17 (41.5)
2	8 (20.5)	7 (17.1)
3	4 (10.3)	0 (0.0)
G8 score at baseline
≤14	27 (69.2)	28 (68.3)
>14 (normal)	12 (30.8)	12 (29.3)
Missing	0 (0.0)	1 (2.4)
SPPB score at baseline
Frail (SPPB≤7)	20 (52.6)	17 (41.5)
Pre-frail (7<SPPB≤9)	9 (23.1)	11 (26.8)
Normal (9<SPPB ≤ 12)	5 (12.8)	8 (19.5)
Unknown	5 (12.8)	5 (12.2)

Data are n (%) unless otherwise stated. TP = trastuzumab, pertuzumab. TPM = trastuzumab, pertuzumab, metronomic cyclophosphamide. ER = oestrogen receptor. PgR = progesterone receptor. HER2 = human Epidermal growth factor receptor 2. G8 = G8 geriatric assessment screening tool. SPPB=Short Physical Performance Battery.

In the updated primary analysis, the estimated progression free survival at 6 months was 43.1% (95% CI 27.1–58.1) in the TP group versus 73.0% (55.8–84.3) in the TPM group, corresponding to a 29.9% absolute difference with the addition of metronomic oral cyclophosphamide and thus still reaching the 10% threshold defined in the protocol. Progression free survival at 12 months was 33.7% (19.3–48.8) in the TP group and 51.9% (34.7–66.5) in the TPM group, while at 24 months this was 18.7% (8.2–32.4) and 28.7% (15.8–43.0) respectively (Fig. 1A). Overall response rate was 44% (16/36) in the TP group and 53% (19/36) in the TPM group, in 72 patients with measurable disease at baseline (Appendix C).

Fig. 1

A. Progression free survival and B. Overall survival after randomisation to trastuzumab/pertuzumab or trastuzumab/pertuzumab with metronomic cyclophosphamide.

At the time of the final analysis, 27 (69.2%) patients in the TP group had died, in comparison to 22 (53.7%) in the TPM group. Median overall survival was 32.1 months in the TP group and 37.5 months in the TPM group, with no significant difference between the two treatment groups (HR 0.72, 95% CI 0.41–1.26, p = 0.25) (Fig. 1B). Breast cancer specific survival was also similar (HR, 0.92, 95% CI 0.49–1.71) (Appendix D). Amongst the patients who had died, 37 patients (75.5%) died due to disease progression, while the other deaths were attributed to toxicity in one patient (2%), unrelated cardiovascular or other chronic disease in 2 patients (4.1%) and other (non breast cancer related) causes in 7 patients (14.3%).

In the patients who started T-DM1, progression free survival was 43.6% (95% CI 27.7–58.5) at 6 months and 34.5% (20.2–49.3) at 12 months (Fig. 2). An objective response was observed in 25% of patients with measurable disease (9/36), all of which were partial responses.

Fig. 2

Progression free survival after starting T-DM1.

Analysis population restricted to patients who started T-DM1 after disease progression on trastuzumab and pertuzumab with or without metronomic cyclophosphamide (n = 40). Progression-free survival on T-DM1 was estimated by the interval-censored method from the start of T-DM1.

Grade 3-5 adverse events (AE) during first line treatment occurred in 21 (53.8%) patients in the TP group and 24 (58.5%) patients in the TPM group (Table 2). Cardiac toxicity was observed in one (2.6%) patient in the TP group and 4 (9.8%) patients in the TPM group. No new safety signal was detected in comparison to the primary analysis. Three (7.9%) patients in the TP group and 7 (17.5%) patients in the TPM group discontinued treatment because of (non-hematological) toxicity (Appendix E). Additionally, the majority of patients in the TPM group needed a treatment interruption of cyclophosphamide (n = 23, 56.1%), mainly due to non-hematological AE (n = 13, 31.7%) and in 12 patients (29%) cyclophosphamide was discontinued before trastuzumab and pertuzumab.

Table 2

Toxicity of TPM, TP, and T-DM1.

Toxicity	TP (n = 39)		TPM (n = 41)		T-DM1 (n = 40)
	All grade	Grade 3-5	All grade	Grade 3-5	All grade	Grade 3-5
All adverse events	39 (97.4)	21 (53.8)	41 (100)	24 (58.5)	35 (87.5)	18 (45)
Gastro-intestinal
Diarrhea	23 (59)	4 (10.3)	29 (70.7)	5 (12.2)	9 (22.5)	1 (2.5)
Nausea	10 (25.6)	0	20 (48.8)	1 (2.4)	10 (25)	0
Mucositis oral	8 (20.5)	0	10 (24.4)	0	5 (12.5)	0
Constipation	6 (15.4)	1 (2.6)	13 (31.7)	0	10 (25)	0
General disorders
Fatigue	25 (64.1)	3 (7.7)	33 (80.5)	2 (4.9)	20 (50)	3 (7.5)
Anorexia	14 (35.9)	0	17 (41.5)	2 (4.9)	14 (35)	2 (5)
Pain	10 (25.6)	2 (5.1)	14 (34.1)	2 (4.9)	6 (15)	0
Hypertension	9 (23.1)	6 (15.4)	9 (22)	5 (12.2)	5 (12.5)	2 (5)
Respiratory
Dyspnea	9 (23.1)	2 (5.1)	12 (29.3)	5 (12.2)	5 (12.5)	1 (2.5)
Epistaxis	8 (20.5)	0	5 (12.2)	0	10 (25)	1 (2.5)
Cough	6 (15.4)	0	13 (31.7)	0	3 (7.5)	0
Liver and kidney function
AST increase	15 (38.4)	0	18 (43.9)	0	31 (77.5)	1 (2.5)
Serum creatinine increase	13 (33.3)	1 (2.6)	17 (41.5)	0	14 (35)	0
ALT increase	11 (28.2)	1 (2.6)	10 (24.4)	1 (2.4)	21 (52.5)	1 (2.5)
Hematological
Neutropenia	6 (15.4)	0	9 (21.9)	0	10 (25)	1 (2.5)
Lymphopenia	19 (48.7)	2 (5.1)	36 (87.8)	17 (41.5)	25 (62.5)	6 (15)
Anaemia	22 (56.4)	0	33 (80.5)	1 (2.4)	24 (60)	0
Thrombocytopenia	4 (10.3)	0	8 (19.5)	0	19 (47.5)	1 (2.5)
Special interest
Heart failure	1 (2.6)		3 (7.3)		0
Decrease of LVEF (10% and to below 50%)	1 (2.6)		1 (2.4)		0
Falls	0	0	0	1 (2.4)	4 (10)	1 (2.5)
Peripheral sensory neuropathy	1 (2.6)	0	5 (12.2)	1 (2.4)	2 (5)	0

Data are given as n (%). LVEF = left ventricular ejection fraction. ALT = alanine aminotransferase. AST = aspartate aminotransferase. Adverse events occurring in ≥20% of patients in one treatment group regardless of treatment attribution, as well as adverse events of special interest are described.

In the T-DM1 population, grade 3–5 AE were seen in 18 (45%) patients during treatment with T-DM1 (Table 2), with the only toxicities occurring in more than one patient being lymphopenia (n = 6), fatigue (n = 3), anorexia (n = 2), hypertension (n = 2) and AST/ALT increase (n = 2). Two patients (5.1%) experienced a grade 5 AE: one death was considered as related to cachexia and tumour progression; the other to acute pneumonia and renal failure. Treatment interruption occurred in 17 (42.5%) patients, while 12 patients needed at least one dose reduction of T-DM1 (30%). Two patients (5.1%) discontinued T-DM1 due to toxicity.

On multivariate analysis, only the WHO performance status was prognostic for progression free survival (p = 0.037), while the G8 score (p = 0.029) and SPPB score (p = 0.032) were prognostic for overall survival and the G8 score (p = 0.005) and social situation (p = 0.019) for breast cancer specific survival. Details can be found in Appendix F.

Discussion

The combination of trastuzumab, pertuzumab and a taxane has been established as the preferred first line treatment for metastatic HER2+ breast cancer since the publication of the CLEOPATRA trial [11]. Although treatment with paclitaxel and docetaxel is generally feasible in fit older patients, it can still be associated with severe acute and long term toxicities, limiting its applicability in those patients that are more frail [14,17]. Alternative treatment options for these patients are needed.

The extended follow-up of the EORTC 75111-10114 trial confirms the activity and safety of a taxane free regimen in this setting, with a greater efficacy when adding metronomic chemotherapy to trastuzumab and pertuzumab in comparison to dual anti-HER2 therapy alone. PFS at 6 months is improved by the addition of metronomic cyclophosphamide, with an absolute difference of almost 30%, and this benefit persists both at 12 and at 24 months. OS and BCSS were similar between both regimens, however it should be noted that this trial was not formally powered for these analyses. Strikingly, even in this generally frail population with multiple comorbidities (41% severe according to the CCI), still the majority of deaths (75.5%) were related to breast cancer, indicating the unmet need for well tolerated and efficacious therapies in this population. Notably, both PFS and OS are lower in the TPM arm than was seen with docetaxel in CLEOPATRA, although cross trial comparisons are difficult since we included a frailer population.

Both regimens were relatively well tolerated, with no apparent increase in high grade toxicity by the addition of metronomic cyclophosphamide. However, premature treatment discontinuation due to toxicity was more frequent in the TPM arm, and a substantial proportion of patients needed at least a temporary interruption of cyclophosphamide. As described in a separate manuscript, HRQoL was similar between the two treatment arms, and so did not appear to be impacted by these interruptions and/or discontinuations [16]. Cardiac toxicity was low, although more frequent in the TPM arm. Cyclophosphamide can cause reversible direct myocardial toxicity and exacerbate underlying myocardial dysfunction, and this risk is greater in older patients [18]. However, this was mainly demonstrated for higher doses of cyclophosphamide (>120–170 mg/kg or 1.55 mg/m2 per day) and has not yet been observed with the metronomic approach.

For patients with ER + HER2+ breast cancer (+- 70% of our population), the addition of an endocrine treatment to dual anti-HER2 blockade could be another alternative to taxane based chemotherapy. This approach has been studied in the PERTAIN trial in an age-unselected population (33% > 65 y), and has shown promising survival outcomes along with manageable toxicity (≥ grade 3 AE in 50.4% of patients for the combination of trastuzumab, pertuzumab and an aromatase inhibitor) [19]. A direct comparison with the results from our trial is again difficult due to differences in patient populations, and as the results of PERTAIN were not known at the time we designed our study, endocrine treatment was not included as a treatment option.

Although the efficacy (in terms of PFS) of T-DM1 after disease progression on first line treatment in our population appears slightly lower to what was observed in the pivotal EMILIA trial [12], it is generally in line with the efficacy of T-DM1 in the control arm of more recent trials [20,21]. The rate of grade 3–5 AE we observed was comparable to EMILIA (≥ grade 3 45 vs. 40.8%), with a wide range of AE seen in our trial, also outside those typically associated with T-DM1.

The relatively small sample size is a limitation of this study and highlights the difficulty to set up large randomized phase III trials in an older population. Obstacles are patient and physician worries about trial participation, but certainly also limited financial support by the pharmaceutical industry for trials targeting this specific population. On the other hand, a major strength of our study was that the majority had clear signs of frailty (representing a very poorly studied population), and the long-term follow-up providing insight in death rate and cause of death.

Anticancer drugs are not well tolerated in all older patients with cancer, and extrapolating results from either younger or highly selected fit older patients included in general oncology trials can lead to erroneous conclusions about the safety and efficacy of these treatments in the general older population. Several methods to improve the evidence base for older patients have been suggested. These include study-design elements that promote participation of older adults, such as more inclusive eligibility criteria, and new and composite endpoints relevant to them, for example overall treatment utility [22,23]. Additionally, our study as well as others prove the feasibility of conducting clinical trials specifically in an older frail patient population [[24], [25], [26]]. Moreover, we confirm the prognostic value of geriatric screening instruments in this setting, so that these may potentially be used as stratification factors in future trials.

In conclusion, the combination of trastuzumab-pertuzumab with metronomic cyclophosphamide followed by T-DM1 on progression provides a relatively safe and effective treatment option for HER2+ metastatic breast cancer, especially in frail older patients who might not tolerate a taxane based regimen.

Role of the funding source

F Hoffmann-La Roche provided the study drugs and provided financial support, but had no other role in the study. The as the sponsor of the study was involved in protocol development, data collection, and statistical analysis.

Declaration of interest

Hans Wildiers: Hans Wildiers's institution received financial compensation on his behalf for advisory boards, lecture fees and/or consultancy fees from Immutep Pty, MSD, Astrazenca Ireland, Daiichi, AbbVie, Lilly, PSI CRO AG, KCE, EISAI, Astrazeneca, Roche. Hans Wildiers's institution received an unrestricted research grant on his behalf from Roche and he received travel support from Pfizer and Roche.

Thomas Meyskens: none.

Sandrine Marreaud: none.

Lissandra Dal Lago: Consulting fees from Merck. Educational grants from Novartis, Roche.,

Peter Vuylsteke: Consulting fees from MSD, Roche, Pfizer, Novartis, Lilly, Astra Zeneca. Travel grants from Pfizer, Mundipharma, Roche.

Giuseppe Curigliano: Consulting fees from BMS, Roche, Pfizer, Novartis, Lilly, Astra Zeneca, Daichii Sankyo, Merck, Seagen, Ellipsis.

Simon Waters: none.

Barbara Brouwers: none.

Bart Meulemans: none.

Berta Sousa: none.

Coralie Poncet: none.

Etienne Brain: Receipt of travel supports from Pfizer, Sandoz. Receipt of consultation fees from Daiichi, Eli Lilly, Pfizer, Sandoz. Receipt of honoraria from Eli Lilly, Pfizer, Sandoz, Seagen.

Author contributions

Hans Wildiers: Conceptualization, Methodology, Investigation, Writing – original draft, Supervision, Funding acquisition, Thomas Meyskens: Writing – original draft, Visualization, Sandrine Marréaud: Methodology, Investigation, Writing – review & editing, Supervision, Lissandra Dal Lago: Investigation, Writing – review & editing, Peter Vuylsteke: Investigation, Writing – review & editing, Giuseppe Curigliano: Investigation, Writing – review & editing, Simon Waters: Investigation, Writing – review & editing, Barbara Brouwers: Conceptualization, Methodology, Investigation, Writing – review & editing, Bart Meulemans: Data curation, Berta Sousa: Investigation, Coralie Poncet: Formal analysis, Writing – original draft, Visualization, Etienne Brain: Conceptualization, Methodology, Investigation, Writing – review & editing

Best overall response	TP (n = 36)	TPM (n = 36)	T-DM1 (n = 36)
Complete response	1 (2.8)	1 (2.8)	0
Partial response	15 (41.7)	18 (50.0)	9 (25.0)
Stable disease	12 (33.3)	12 (33.3)	17 (47.2)
Progressive disease	4 (11.1)	4 (11.1)	4 (11.1)
Early death	2 (5.6)	0	3 (8.3)
Not evaluable	2 (5.6)	1 (2.8)	3 (8.3)

Main reason for treatment discontinuation	TP (n = 38)	TPM (n = 40)	T-DM1 (n = 39)
Progressive disease	25 (65.8)	24 (60.0)	28 (71.8)
Toxicity	3 (7.9)	7 (17.5)	2 (5.1)
Patient's decision (unrelated to toxicity)	5 (13.2)	4 (10.0)	3 (7.7)
Other malignancy	3 (7.9)	0	0
Death unrelated to malignancy/toxicity	1 (2.6)	2 (5.0)	1 (2.6)
Other	1 (2.6)	2 (5.0)	4 (10.3)
Lost to follow-up	0	1 (2.5)	1 (2.6)

Potential prognostic factors	P-value (Wald-test) Adjusted for treatment	Selection for the full multivariate model: 10% threshold (Yes/No)
Age	0.055	Yes
WHO performance status	0.037	Yes
Estrogen receptor status	0.225	No
Progesteron receptor status	0.503	No
Prior anti-HER2 treatment	0.132	No
Social Situation	0.149	No
GDS-4 score	0.510	No
G8 score	0.231	No
CCI score	0.580	No
ADL score	0.104	No
IADL score	0.055	Yes
SPPB score	0.276	No
Lymph node involvement	0.162	No
Soft tissue involvement	0.965	No
Visceral involvement	0.596	No
Skeletal involvement	0.164	No

	Full multivariate interval censored regression model	Backward procedure selection (Yes/No)	Final multivariate interval censored regression model
	HR (95% CI)		HR (95% CI)	P-value (Wald-test)
Treatment
TP	1.0	Yes	1.0	0.085
TPM	0.58 (0.35; 0.96)		0.65 (0.40; 1.06)
Age category		No
≤ 69	1.0
69–75	0.43 (0.15; 1.19)
≥ 75	0.40 (0.15; 1.03)
WHO performance status
0	1.0	Yes	1.0	0.037
1	1.73 (0.97; 3.08)		1.74 (0.99; 3.08)
2–3	1.24 (0.47; 3.25)		2.28 (1.18; 4.38)
IADL score		No
≤ 3	1.0
3–5	0.82 (0.31; 2.20)
≥ 6	0.48 (0.18; 1.33)

Only the performance score has a significant effect on progression-free survival (p = 0.037).

Potential prognostic factors	P-value (Wald-test)Adjusted for treatment	Selection for the full multivariate model: 10% threshold (Yes/No)
Age	0.705	No
WHO performance status	0.007	Yes
Estrogen receptor status	0.188	No
Progesteron receptor status	0.912	No
Prior anti-HER2 treatment	0.018	Yes
Social Situation	0.003	Yes
GDS-4 score	0.502	No
G8 score	0.002	Yes
CCI score	0.019	Yes
ADL score	0.135	No
IADL score	0.020	Yes
SPPB score	0.001	Yes
Lymph node involvement	0.783	No
Soft tissue involvement	0.271	No
Visceral involvement	0.081	Yes
Skeletal involvement	0.590	No

Among the 80 randomized patients, two patients had missing data on some covariates and were therefore not included in the multivariate models.

	Full multivariate Cox regression model	Backward procedure selection (Yes/No)	Final multivariate Cox regression model
	HR (95% CI)		HR (95% CI)	P-value (Wald-test)
Treatment
TP	1.0	Yes	1.0	0.214
TPM	0.75 (0.36; 1.53)		0.69 (0.38; 1.24)
WHO performance status
0	1.0	No
1	0.85 (0.34; 2.16)
2–3	1.07 (0.25; 4.60)
Social Situation
At home by myself	1.0	No
At home with someone	0.78 (0.37; 1.63)
Institutional care	0.82 (0.17; 3.99)
Unknown	7.41 (1.16; 47.94)
G8 score
≤ 14	1.0	Yes	1.0	0.029
> 14	0.50 (0.18; 1.41)		0.38 (0.16; 0.90)
Prior anti-HER2 treatment
None	1.0	No
Adjuvant/metastatic	1.75 (0.69; 4.45)
CCI score
0	1.0	No
1–2	1.56 (0.77; 3.15)
> 2	0.59 (0.15; 2.32)
IADL score
≤ 3	1.0	No
3–5	0.62 (0.17; 2.33)
≥ 6	0.45 (0.12; 1.74)
SPPB score
≤ 7	1.0	Yes	1.0	0.032
7–9	0.41 (0.15; 1.14)		0.37 (0.15; 0.90)
9–12	0.82 (0.30; 2.20)		0.67 (0.28; 1.61)
Unknown	0.61 (0.12; 3.10)		1.92 (0.84; 4.39)
Visceral involvement
No	1.0	No
Yes	1.97 (0.80; 4.82)

G8 score (p = 0.029) and SPPB score (p = 0.032) have a significant effect on overall survival.

Potential prognostic factors	P-value (Fine and Gray)Adjusted for treatment	Selection for the full multivariate model: 10% threshold (Yes/No)
Age	0.449	No
WHO performance status	0.091	Yes
Estrogen receptor status	0.123	No
Progesteron receptor status	0.769	No
Prior anti-HER2 treatment	0.106	No
Social Situation	0.0008	Yes
GDS-4 score	0.416	No
G8 score	0.002	Yes
CCI score	0.752	No
ADL score	0.143	No
IADL score	0.573	No
SPPB score	0.036	Yes
Lymph node involvement	0.453	No
Soft tissue involvement	0.108	No
Visceral involvement	0.341	No
Skeletal involvement	0.336	No

Among the 80 randomized patients, two patients had missing data on some covariates and were therefore not included in the multivariate models.

	Full multivariate Fine and Gray regression model	Backward procedure selection (Yes/No)	Final Fine and Gray model
	HR (95% CI)		HR (95% CI)	P-value
Treatment
TP	1.0		1.0	0.899
TPM	0.97 (0.46, 2.04)		0.96 (0.48; 1.89)
WHO performance status
0	1.0	No
1	0.82 (0.26; 2.60)
2–3	0.73 (0.16; 3.26)
Social Situation
At home by myself	1.0	Yes	1.0	0.019
At home with someone	1.04 (0.47; 2.28)		0.99 (0.48; 2.06)
Institutional care	2.33 (0.82; 6.69)		2.68 (1.31; 5.50)
Unknown	3.63 (0.52; 25.15)		1.97 (0.24; 16.33)
G8 score
≤ 14	1.0	Yes	1.0	0.005
> 14	0.27 (0.08; 0.85)		0.24 (0.09; 0.65)
SPPB score
≤ 7	1.0	No
7–9	0.36 (0.09; 1.42)
9–12	0.72 (0.26; 2.00)
Unknown	1.10 (0.27; 4.54)

have a significant effect on breast cancer specific survival.