Effectiveness comparison of first-line CDK4/6 inhibitors in patients with hormone-positive HER2-negative advanced breast cancer according to tumor histology: a sub-analysis of the real-world, multicenter, Italian study PALMARES-2.

1
Introduction
Invasive lobular carcinoma (ILC) is the second most common type of invasive breast cancer (BC), accounting for 15 % of diagnoses [1,2]. The majority of ILCs are hormone receptor-positive (HR+), human epidermal growth factor receptor 2-negative (HER2-), with less than 10 % being HR-negative and/or HER2-positive [3]. Compared with non-special type (NST), ILC is characterized by distinctive molecular and clinical features, including loss of E-cadherin, larger tumor size, more frequent nodal involvement, multifocal and/or bilateral disease at diagnosis, and higher incidence of pleural and/or peritoneal involvement in advanced stage [4,5]. In terms of prognostic implications, surgically resected early-stage HR+/HER2- ILC has been associated with lower risk of early recurrence counterbalanced by an increased risk of late recurrence [6,7]. However, in patients with advanced BC (aBC), the prognostic and predictive value of ILC histology remains unclear, with only retrospective studies reporting conflicting results [[8], [9], [10]].
In the past decade, cyclin dependent kinase 4/6 inhibitors (CDK4/6i), namely palbociclib, ribociclib and abemaciclib, in combination with standard endocrine therapies (ETs) revolutionized HR+/HER2-aBC treatment, leading to impressive improvements in patient progression-free survival (PFS) [[11], [12], [13], [14], [15], [16], [17]]. Beyond PFS, ribociclib also demonstrated significant overall survival (OS) benefit in both endocrine-sensitive and endocrine-resistant settings, while the combination of abemaciclib and fulvestrant improved OS in patients with endocrine-resistant disease (MONARCH 2) [[18], [19], [20]]. Despite the lack of a statistically significant OS benefit, results from PALOMA-2/3, PARSIFAL, and MONARCH 3 support palbociclib and abemaciclib as valid first-line alternatives for patients with HR+/HER2– advanced breast cancer [[21], [22], [23], [24]].
Given the lack of head-to-head randomized controlled trials (RCTs) among CDK4/6i, selecting the agent with the highest expected efficacy in specific clinical contexts remains challenging. Real-world studies therefore provide a unique opportunity to evaluate the comparative effectiveness of palbociclib, ribociclib and abemaciclib in clinical practice. However, published real-world studies comparing CDK4/6i in patients with HR+/HER2-aBC have yielded conflicting results [[25], [26], [27], [28], [29], [30], [31], [32], [33], [34]]. For instance, PALMARES-2 suggested greater effectiveness of abemaciclib and ribociclib compared with palbociclib, whereas P-VERIFY, found no significant differences among CDK4/6i25,27,28.
Although patients with ILC were included in RCTs evaluating CDK4/6i efficacy, the prognostic and predictive role of ILC was not specifically analyzed and remains poorly characterized in real-world studies. Therefore, tumor histology is not currently considered when selecting systemic therapy for HR+/HER2-aBC in international guidelines [35,36].
In this sub-analysis of the real-world PALMARES-2 study, we evaluated the impact of tumor histology on survival outcomes in patients receiving first-line ET and CDK4/6i, comparing palbociclib, ribociclib, and abemaciclib in terms of real-world PFS (rwPFS), time to chemotherapy or death (TTC-D), and OS by histologic subtype (ILC vs. NST) [25].

2
Material and methods
PALMARES-2 (NCT06805812) is a retrospective/prospective, multicenter, observational, Italian study that is collecting data on the effectiveness of first and subsequent treatment lines in HR+/HER2-aBC. Eligibility criteria, study design details and data collection are reported in the Supplementary material and in the previously published article [25].
2.1
Study objectives and endpoints
The primary objective of this pre-planned exploratory analysis was to compare the real-world effectiveness of palbociclib, ribociclib or abemaciclib in patients with ILC versus NST.
The primary endpoint was rwPFS, as defined as the time interval between ET plus CDK4/6i initiation and the detection of disease progression in the judgement of the treating physician based on radiological, clinical or biochemical criteria, or patient death, whichever occurred first.
TTC-D and OS were secondary endpoints, defined respectively as the interval between ET plus CDK4/6i initiation and the initiation of first-line chemotherapy for metastatic disease or patient death from any cause, or between treatment initiation and patient death from any cause. Patients not undergoing rwPFS, TTC-D or OS events were censored at the time of data cut-off or last medical activity, if the latter occurred before data cut-off.

2.2
Statistical methods and analyses
Median follow-up was estimated with inverse Kaplan–Meier method. Baseline characteristics were summarized in tables by tumor histology and CDK4/6i type. Differences between histologic or treatment groups were assessed using the Kruskal-Wallis test for continuous variables and Pearson's Chi-square or Fisher's exact test for categorical variables, as appropriate.
Survival analyses were performed using Cox proportional hazards models and illustrated as survival curves through the Kaplan-Meier method providing median survival outcomes and corresponding 95 % confidence intervals (CIs). Associations between tumor histology, type of CDK4/6i and outcomes were evaluated through multivariable Cox proportional hazards regression models adjusting for clinically relevant covariates, i.e. age, Eastern Cooperative Oncology Group performance status (ECOG PS), menopausal status, estrogen receptor (ER) expression, progesteron receptor (PgR) expression, Nottingham histologic grade, Ki67 expression, HER2 expression, de novo advanced disease at diagnosis, presence of lung, liver, bone and pleura or peritoneum metastases, and endocrine resistance (see Supplementary material) [37]. To avoid overfitting, ET backbone was not included in the model as it largely reflected endocrine resistance status. These analyses provided adjusted hazard ratios (aHR) and corresponding 95 % CIs and p-values (P) for each covariate. An interaction term in the Cox model was used to test whether the effectiveness of palbociclib, ribociclib and abemaciclib on outcomes differed by tumor histology.
Inverse probability of treatment weighting (IPTW) was used to balance clinico-pathological characteristics across CDK4/6i groups [38]. The propensity scores used in IPTW were estimated through a generalized boosted model adopting the same covariates included in Cox regression models. Estimated weights were applied in weighted Cox models to calculate median survival, hazard ratios and 95 % CIs. Covariates balance was assessed using standardized differences (d-value), with d-values <0.1 considered negligible. ER, PgR and Ki67, expression, tumor grade and HER2 status used in multivariable and IPTW models were defined based on the most recent assessment available before CDK4/6i initiation (metastatic tissue when available, otherwise primary tumor).
Covariates missing in ≤15 % of the whole study cohort were imputed using median and for numerical and categorical variables, respectively. All statistical tests were two-tailed, and P < 0.05 was considered significant. Statistical analyses were performed using R software (v.4.1.2) using the following packages: dplyr, tidyr, survival, survminer, ggplot2, WeightIt.

3
Results
3.1
Study population and patients’ characteristics
At data cutoff (January 2024), 1982 patients who initiated first-line ET plus CDK4/6i between January 2016–September 2023 were enrolled in PALMARES-2. NST tumor histology was detected in 1481 patients (74.7 %), ILC in 367 patients (18.5 %) and mixed or rare histologies in 104 (5.3 %) and 30 (1.5 %) patients (Supplementary Fig. S1).
Compared to NST, patients with ILC were older, more frequently postmenopausal, more often treated with abemaciclib and had higher prevalence of luminal A-like tumors with lower Ki67 expression in metastatic samples. ILC patients had fewer lung metastases, but more frequently presented with multiple metastatic sites and pleural/peritoneal involvement. ER, Ki67, and HER2 status was largely concordant between primary and metastatic samples, while PgR expression was lower in metastatic disease (20 % vs. 60 %). Tumor grade in metastatic samples was mostly unavailable (Table 1).

3.2
Survival outcomes according to tumor histology
Median follow-up was 31.3 months (interquartile rage [IQR]: 16.9–49.8), 29.8 months, (IQR: 15.3–47.3) in the ILC cohort and 31.2 months (IQR: 17.1–50.7) in the NST cohort, with 174 and 629 rwPFS events, respectively.
Patients with ILC had significantly shorter rwPFS than those with NST (25.6 vs. 35.5 months; HR: 1.19, 95 % CI:1.01–1.42, P = 0.042), as confirmed by multivariable analysis (aHR: 1.24, 95 %CI: 1.04–1.47, P = 0.017) (Fig. 1). Mixed and rare histologic subtypes showed no significant rwPFS differences versus NST. ILC was also associated with worse TTC-D (38.8 vs. 49.9 months, HR: 1.23, 95 %CI: 1.02–1.50, P = 0.003) and a trend towards worse OS (64.9 vs. 65.9 months, HR: 1.22; 95 %CI: 0.97–1.53, P = 0.096) compared with NST, as confirmed by multivariable analyses for both TTC-D (aHR: 1.38, 95 %CI: 1.13–1.69, P = 0.002) and OS (aHR: 1.26, 95 %CI: 0.99–1.60, P = 0.064) (Supplementary Figs. S2–S3)

3.3
Effectiveness comparison of CDK4/6i according to tumor histology
Interaction analyses between CDK4/6i type and tumor histology for rwPFS indicated that abemaciclib effectiveness was significantly affected by ILC histology (P for interaction = 0.009), whereas no significant interaction was found for ribociclib (P = 0.072) or palbociclib (P = 0.553).
Baseline characteristics were balanced across CDK4/6i groups in the ILC cohort but not in the NST one, where palbociclib or abemaciclib were associated with more common endocrine resistant diseases, visceral involvement, bone and liver metastases, higher metastatic burden, lower PgR expression, rarer de novo aBC presentation, older age and worse ECOG PS (Table 2). The palbociclib group had longer median follow-up (45.7 months, IQR: 28.0–59.6) compared with ribociclib (25.2 months, IQR: 12.7–44.6) or abemaciclib (22.4 months, IQR: 12.2–33.8) groups.
In the ILC cohort, the three CDK4/6i showed similar rwPFS (29.0, 26.8 and 22.6 months with palbociclib, ribociclib and abemaciclib, respectively; ribociclib vs. palbociclib: HR: 0.97, 95 %CI: 0.68–1.40, P = 0.881; abemaciclib vs. palbociclib: HR: 1.34, 95 %CI: 0.93–1.93, P = 0.119, abemaciclib vs. ribociclib: HR: 1.38, 95 %CI: 0.92–2.06, P = 0.123) while, in the NST cohort, abemaciclib and ribociclib were associated with significantly longer rwPFS compared to palbociclib (46.6, 47.2 and 27.6 months; abemaciclib vs. palbociclib: HR: HR: 0.77, 95 %CI: 0.61–0.96, P = 0.018; ribociclib vs. palbociclib: HR: 0.67, 95 %CI: 0.55–0.80, P<0.001; abemaciclib vs ribociclib: HR: 1.14, 95 %CI: 0.90–1.45, P = 0.281) (Supplementary Fig. S4).
After IPTW adjustment, patient characteristics were mostly balanced across CDK4/6i in both ILC and NST cohorts, except for small differences in PgR and Ki67 expression in ILC (d = 0.15 and d = 0.11) (Supplementary Fig. S5).
In the adjusted ILC cohort, we found no rwPFS differences among CDK4/6i (ribociclib vs. palbociclib aHR: 1.01, 95 %CI: 0.67–1.45, P = 0.949; abemaciclib vs. palbociclib: aHR: 1.13, 95 %CI: 0.75–1.71, P = 0.551, abemaciclib vs. ribociclib: aHR: 1.15, 95 %CI: 0.73–1.80, P = 0.549) (Fig. 2A). Conversely, in the NST cohort, both abemaciclib and ribociclib were associated with longer rwPFS compared to palbociclib (ribociclib vs. palbociclib: aHR: 0.78, 95 %CI: 0.65–0.94, P = 0.009, abemaciclib vs. palbociclib: aHR: 0.68, 95 %CI: 0.53–0.86, P = 0.002), with no significant differences between abemaciclib and ribociclib (aHR: 0.84, 95 %CI: 0.65–1.09, P = 0.191) (Fig. 2B). These results were consistent in multivariable models (Supplementary Fig. S6). Due to different approval in Italy, yearly CDK4/6i prescriptions patterns changed over the study period, with an increasing use of ribociclib and abemaciclib, paralleled by a decreasing use of palbociclib (Supplementary Fig. S7). To mitigate follow-up imbalances, we performed a sensitivity analysis including only patients starting CDK4/6i from 2020. Although differences were smaller, palbociclib (N = 403) retained a moderately longer median follow-up than ribociclib (N = 336) and abemaciclib (N = 554) (27.5 vs. 19.5 and 20.5 months). After IPTW adjustment, small residual imbalances persisted across CDK4/6i in the ILC cohort (PgR d = 0.12, Ki67 d = 0.15, HER2 d = 0.13, de novo aBC, d = 0.10 and endocrine resistance d = 0.11), whereas covariates were balanced in the NST cohort (Supplementary Fig. S8). In patients with ILC rwPFS did not differ among CDK4/6i (ribociclib vs. palbociclib: aHR 1.17, 95 % CI 0.68–2.00, P = 0.566; abemaciclib vs. palbociclib: aHR: 1.20, 95 % CI: 0.70–2.13, P = 0.486; abemaciclib vs. ribociclib: aHR: 1.04, 95 % CI: 0.65–1.68, P = 0.864) while in the NST cohort, ribociclib and abemaciclib were associated with numerically longer rwPFS versus palbociclib (ribociclib: aHR: 0.79, 95 %CI 0.58–1.07, P = 0.133; abemaciclib: aHR: 0.71, 95 %CI: 0.51–0.98, P = 0.042; abemaciclib vs. ribociclib: aHR 0.79, 95 %CI 0.56–1.21, P = 0.468), although statistical significance was not reached for ribociclib (Supplementary Fig. S9).

3.4
Impact of individual CDK4/6i on other clinical outcomes according to tumor histology
As for TTC-D, we found significant interactions between ILC histology and both abemaciclib and ribociclib (P = 0.001 and 0.003), but not palbociclib use (P = 0.212); conversely, we found no significant interactions affecting OS between tumor histology and abemaciclib (P = 0.132) ribociclib (P = 0.326) or palbociclib (P=0.907).
In the IPTW-adjusted ILC cohort, we found no significant difference across CDK4/6i for TTC-D (ribociclib vs. palbociclib, aHR: 1.37, 95 %CI: 0.88–2.14, P=0.166; abemaciclib vs palbociclib, aHR: 1.41, 95 %CI: 0.87–2.27, P=0.161; abemaciclib vs. ribociclib aHR: 1.03, 95 %CI: 0.63–1.68, P=0.916) or OS (ribociclib vs. palbociclib: aHR: 0.78, 95 % CI: 0.46–1.34, P = 0.380; abemaciclib vs. palbociclib: aHR: 1.12, 95 % CI: 0.63–1.97, P = 0.714; abemaciclib vs. ribociclib: aHR: 1.43, 95 % CI: 0.76–2.67, P=0.267) (Supplementary Fig. S10A–S11A).
In the IPTW-adjusted NST cohort, ribociclib and abemaciclib were associated with longer TTC-D versus palbociclib (ribociclib vs. palbociclib aHR: 0.75, 95 %CI: 0.61–0.94, P=0.011; abemaciclib vs palbociclib aHR: 0.67, 95 %CI: 0.50–0.89, P=0.006; abemaciclib vs. ribociclib: aHR: 0.89, 95 %CI: 0.65–1.21, P=0.455, while only ribociclib was associated with better OS (ribociclib vs palbociclib: aHR: 0.72, 95 %CI 0.60–0.99, P = 0.049; abemaciclib vs palbociclib: aHR: 0.72, 95 %CI: 0.51–1.02, P=0.065; abemaciclib vs ribociclib: aHR: 0.93, 95 %CI 0.64–1.35, P = 0.698) (Supplementary Fig. S10B–S11B). Similar trends were observed in univariable and multivariable analyses (Supplementary Fig. S12–S15).

4
Discussion
CDK4/6i in combination with ET are the cornerstone of first-line treatment for patients with HR+/HER2-aBC [35,36]. However, selecting the most appropriate CDK4/6i for individual patients to balance expected efficacy and potential toxicities remains challenging in clinical practice. Although ILC represents the second most frequent histological subtype among HR+/HER2-aBC, the only RCT to specifically report the CDK4/6i plus ET efficacy in this group was PALOMA-2 [11]. Pooled analyses of PALOMA-2/3, MONALEESA-2/3/7, and MONARCH 2/3 trials confirmed that the addition of CDK4/6i to ET significantly improves PFS in patients with ILC [39,40]. Given the lack of head-to-head comparisons in RCTs, real-world studies provide a unique opportunity to assess the comparative effectiveness of palbociclib, ribociclib, and abemaciclib in patients with HR+/HER2-aBC across histological subtypes.
In this sub-analysis of the PALMARES-2 study, ILC was independently associated with shorter rwPFS and TTC-D, with also a trend towards worse OS, when compared with NST tumors in a cohort of 1982 patients with HR+/HER2-aBC treated with fist-line CDK4/6i plus ET, as revealed by both univariate and multivariable analysis.
Palbociclib showed similar real-world effectiveness in ILC and NST patients, abemaciclib was less effective in ILC, while ribociclib showed a non-significant trend toward reduced (P = 0.072). Overall, the lower effectiveness of CDK4/6i in ILC patients appears to be driven by ribociclib and abemaciclib.
These findings are clinically relevant given the conflicting evidence on the prognostic and predictive impact of tumor histology in patients with HR+/HER2-aBC [[8], [9], [10]]. A previous retrospective analysis reported no rwPFS or OS difference in patients with HR+/HER2- ILC vs NST; however, 93 % of patients in that cohort received palbociclib, whose effectiveness was not affected by tumor histology in our study [10].
In our ILC cohort, all CDK4/6i were associated with similar clinical outcomes. By contrast, in patients with NST, abemaciclib and ribociclib were associated with longer rwPFS than palbociclib, in line with the overall PALMARES-2 cohort results, as well as a similar Danish study [25,26]. Differences in CDK4/6i effectiveness across tumor histologies remained consistent after adjustment for acknowledged prognostic covariates, suggesting that these findings may reflect ILC intrinsic biological features, or other clinical factors not captured in the real-world setting. In this context, ILC exhibits distinct mechanisms of endocrine resistance and different genomic landscape compared to NST, including more frequent ERBB2 and NF1 mutations and PI3K-pathway alterations, all characteristics associated with poorer outcomes in patients treated with ET with or without CDK4/6i [[41], [42], [43], [44], [45], [46]]. Furthermore, CDK4/6i differ in pharmacokinetic and pharmacodynamic properties, with ribociclib and abemaciclib showing stronger CDK4 inhibition, and abemaciclib exhibiting additional off-target activity against other CDKs [47]. Last, palbociclib is the CDK4/6i with best manageability and safety profile [48,49]. Identifying patients in whom palbociclib provides comparable benefit as ribociclib or abemaciclib is therefore highly relevant in clinical practice.
To the best of our knowledge, this is the first large real-world study specifically assessing the prognostic and predictive role of tumor histology in patients with HR+/HER2-aBC treated with first-line ET plus CDK4/6i. While our findings regarding ILC are novel, our conclusions on real-world effectiveness in NST, which represented the 74.1 % of our population, partially contrast with those of the P-VERIFY trial (N = 9146) and other studies, which revealed no differences in OS and rwPFS among CDK4/6i27-34. Although PALMARES-2 and P-VERIFY applied relatively similar methodologies, they included distinct patient populations, as P-VERIFY included only US patients receiving aromatase inhibitors, and it did not account for several variables considered in PALMARES-2, including tumor histology [25,27,28]. Differences in baseline features and covariate selection for IPTW may also explain the longer rwPFS observed in our NST cohort compared with other real-world studies and RCTs [[11], [12], [13], [14], [15], [16], [17]].
Since palbociclib, ribociclib and abemaciclib were similarly effective the ILC cohort, these results may guide clinicians to personalize CDK4/6i selection based on tolerability, comorbidities, manageability, potential drug-drug interactions and patient preferences in this specific context.
Our study has limitations. Firstly, due to its retrospective nature, our analysis is unavoidably affected by confounding factors, including ones related to CDK4/6i selection. Despite Cox-regression and IPTW adjustment for clinically relevant patient- and tumor-related features, the lack of random treatment assignment remains an inherent limitation of any retrospective analysis. Secondly, although these analyses were pre-specified in the study protocol, no specific statistical plan was previously elaborated, therefore results should be interpreted as exploratory. Thirdly, centralized histopathological review was not performed; however, histological concordance between pathologists has been reported to be high, especially after the widespread introduction of E-cadherin expression by immunohistochemistry [50]. Fourthly, tumor progression was investigator-assessed based on imaging (i.e., CT/PET scan), clinical (i.e, ECOG PS; clinical tumor measurements) and biochemical (i.e., CA15.3 blood concentration) evaluations rather than RECIST1.1 criteria, introducing potential heterogeneity in rwPFS estimates [51]. Additionally, higher toxicity and discontinuation rates with abemaciclib and ribociclib may influence assessment patterns and follow-up intensity, potentially affecting rwPFS. Nevertheless, although this inherent limitation of real-world studies must be highlighted, rwPFS is a widely accepted endpoint reflecting real-world effectiveness of systemic treatments and it has been validated as a reliable surrogate for survival outcomes in RCTs [52,53]. Furthermore, the consistency between rwPFS and TTC-D, an endpoint independent from investigator assessment, supports the robustness of our results. Fifthly, the shorter follow-up for abemaciclib and ribociclib compared with palbociclib was only partially addressed by the conduction of a sensitivity analysis in patients initiating CDK4/4i treatment from 2020. Finally, due relatively short follow-up, survival outcomes, especially OS, remain immature, limiting the strength of our conclusions. Longer follow-up and ongoing patient accrual will provide a more definitive comparison of CDK4/6i effectiveness in the overall population and ILC cohort.
In conclusion lobular histology was associated with worse clinical outcomes among patients with HR+/HER2-aBC treated with first-line ET plus CDK4/6i. The three CDK4/6i demonstrated similar real-world effectiveness in patients with ILC, while abemaciclib and ribociclib were associated with longer rwPFS in the NST cohort. Our findings highlight the importance of evaluating patient- and tumor-related features, including histology, to personalize first-line treatment for HR+/HER2-aBC, balancing real-world effectiveness, safety, and patient preferences. In this perspective, palbociclib, the safest and most manageable CDK4/6i, can be considered a highly effective treatment option for patients with HR+/HER2-aBC and lobular tumor histology.

CRediT authorship contribution statement
Giacomo Mazzoli: Writing – review & editing, Writing – original draft, Visualization, Software, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Leonardo Provenzano: Writing – review & editing, Writing – original draft, Supervision, Project administration, Methodology, Investigation, Data curation, Conceptualization. Maria Vittoria Dieci: Writing – review & editing, Supervision, Investigation. Giuseppe Curigliano: Writing – review & editing, Supervision, Investigation. Mario Giuliano: Writing – review & editing, Supervision, Investigation. Andrea Botticelli: Writing – review & editing, Supervision, Investigation. Matteo Lambertini: Writing – review & editing, Supervision, Investigation. Gianpiero Rizzo: Writing – review & editing, Investigation. Rebecca Pedersini: Writing – review & editing, Investigation. Marianna Sirico: Writing – review & editing, Investigation. Nicla La Verde: Writing – review & editing, Investigation. Alessandra Gennari: Writing – review & editing, Investigation. Alberto Zambelli: Writing – review & editing, Investigation. Angela Toss: Writing – review & editing, Investigation. Marta Piras: Writing – review & editing, Investigation. Monica Giordano: Writing – review & editing, Investigation. Barbara Tagliaferri: Writing – review & editing, Investigation. Daniele Generali: Writing – review & editing, Investigation. Donata Sartori: Writing – review & editing, Investigation. Giuseppe Fotia: Writing – review & editing, Investigation, Data curation. Matteo De Monte: Writing – review & editing, Writing – original draft, Investigation, Data curation. Francesca Ligorio: Writing – review & editing, Supervision, Investigation, Data curation. Flavia Jacobs: Writing – review & editing, Investigation, Data curation. Giovanna Armani: Writing – review & editing, Investigation, Data curation. Christian Zurlo: Writing – review & editing, Investigation. Alice Menichetti: Writing – review & editing, Investigation. Gaia Griguolo: Writing – review & editing, Writing – original draft, Investigation. Valeria Faso: Writing – review & editing, Investigation. Ambra Carnevale Schianca: Writing – review & editing, Investigation. Elisabetta Munzone: Writing – review & editing, Investigation, Data curation. Antonio Marra: Writing – review & editing, Investigation. Edoardo Chiappe: Writing – review & editing, Investigation. Simone Scagnoli: Writing – original draft, Investigation, Data curation. Simona Pisegna: Writing – review & editing, Investigation, Data curation. Camilla Capasso: Writing – review & editing, Investigation, Data curation. Carmine De Angelis: Writing – review & editing, Investigation, Data curation. Grazia Arpino: Writing – review & editing, Investigation, Data curation. Carmen Criscitiello: Writing – review & editing, Investigation, Data curation. Valentina Guarneri: Writing – review & editing, Supervision, Investigation, Data curation. Giancarlo Pruneri: Writing – review & editing, Supervision, Data curation. Luigi Mariani: Supervision, Formal analysis, Data curation. Claudio Vernieri: Writing – review & editing, Writing – original draft, Resources, Project administration, Methodology, Investigation, Funding acquisition, Conceptualization.

Availability of data and materials
The datasets generated and/or analyzed during the current study are available from the corresponding author upon a reasonable request.

Ethics approval and consent to participate
The study was conducted in accordance with the Declaration of Helsinki, and it was approved by the Institutional Review Board and Ethics Committee of the “Fondazione IRCCS Istituto Nazionale dei Tumori” of Milan (INT 101/23). Patients alive at the time of data collection and/or analysis signed a specific informed consent form for the use of their personal data for research purposes. The Ethics Committee authorized the collection and analysis of data from patients who were not alive when the study was conducted and who already authorized the use of data collected as per clinical practice for research purposes.

Consent for publication
Not applicable.

∗PALMARES-2 study group
Giampaolo Bianchini21,32, Chiara Corti33, Laura Boldrini5,6, Pier Paolo Berton Giachetti5,6, Jalissa Katrini5,6, Maria Silvia Cona14, Vincenza Cantile7, Angela Grieco7, Miriam Pirolo7, Maria Zappulo7, Maria Anna Rachele De Giglio7, Marta Laganà12, Deborah Cosentini12, Ugo De Giorgi34,35, Andrea Vingiani6,30, Antonino Belfiore30, Caterina Sposetti33, Valentina Bianchessi1, Alice Abate1, Daniela Miliziano1, Anthea Iacobucci1,6, Giulia Coppola1,6, Beatrice Fratini2, Cristina Ferraris38, Gabriele Martelli38, Secondo Folli38, Paolo Baili36, Ilaria Cavallo36, Gianfranco Scaperrotta37, Catherine Depretto37, Angioletta Lasagna11, Ornella Ponzoni19,20, Claudia Leli23, Carla Strina25, Silvia Coccato26, Rosalinda Coviello22. 21 Department of Medical Oncology, IRCCS San Raffaele Hospital, Milan, Italy; 32 Faculty of Medicine and Surgery, Università Vita-Salute San Raffaele, Milan, Italy; 5 Division of Early Drug Development for Innovative Therapy, IRCCS European Institute of Oncology, Milan, Italy; 6 Department of Oncology and Hematology-Oncology, University of Milan, Milan, Italy; 14 Department of Oncology, Polo Ospedaliero Luigi Sacco, ASST Fatebenefratelli Sacco, Milan, Italy; 7 Department of Clinical Medicine and Surgery, Università degli Studi di Napoli Federico II, Naples, Italy; 12 Medical Oncology Department, ASST Spedali Civili of Brescia, Brescia, Italy; 34 Department of Experimental Medicine, University of Salento, Lecce, Italy; 35 Clinical and Experimental Oncology Unit, “Vito Fazzi” Hospital, Lecce, Italy; 30 Department of Diagnostic Innovation, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy; 33 Breast Oncology Center, Dana-Farber Cancer Institute, Boston (MA), United States; 1 Department of Medical Oncology, Fondazione IRCCS Istituto Nazionale Tumori, Milan, Italy; 2 IFOM ETS, the AIRC Institute of Molecular Oncology, Milan, Italy; 38 Surgical Oncology Department, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy; 36 Analytical Epidemiology and Health Impact Unit, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy; 37 Breast Radiology Unit, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy; 11 Medical Oncology Unit, Fondazione IRCCS Policlinico San Matteo, Pavia, Italy; 19 Department of Oncology and Hematology, Azienda Ospedaliero-Universitaria di Modena, Modena, Italy; 20 Department of Medical and Surgical Sciences, University of Modena and Reggio Emilia, Modena, Italy; 23 Maugeri Clinic and Scientific Institute IRCCS, Pavia, Italy; 25 Breast Cancer Unit, Azienda Socio Sanitaria Territoriale di Cremona, Cremona, Italy; 26 Oncology Unit Dolo-Mirano, AULSS3 Veneziana, Mirano, Italy; 22 SC Oncologia, Asst-Lariana, Como, Italy.

Declaration of competing interest
The authors declare the following conflicts of interest related to this article:
L.P.: personal fee as invited speaker MSD, Pfizer and Novartis.
M.V.D.: personal fees for consultancy/advisory role from: Eli Lilly, Pfizer, Novartis, Seagen, Gilead, MSD, Exact Sciences, AstraZeneca, Daiichi Sankyo, Roche.
G.C.: consulting fees from Roche, Novartis, Lilly, Pfizer, Astra Zeneca, Daichii Sankyo, Ellipsis, Veracyte, Exact Science, Celcuity, Merck, BMS, Gilead, Sanofi, Menarini; Payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing or educational events by Lilly, Pfizer, Relay, Gilead, Novartis and support for attending meetings and/or travel from Daichii Sankyo.
M.G.: consulting/advisory role for Roche, AstraZeneca, Lilly, Daichii Sankyo, Novartis, Pfizer, Seagen, MSD, Eisai; honoraria from Novartis, Pfizer, Lilly, AstraZeneca, Daichii Sankyo; travel, accommodation, expenses from Lilly, Pfizer, AstraZeneca.
M.L.: reported having an advisory role for Roche, Lilly, Novartis, AstraZeneca, Pfizer, Seagen, Gilead, MSD, Pierre Fabre, Menarini and Exact Sciences; receiving speaker honoraria from Roche, Lilly, Novartis, Pfizer, Sandoz, Libbs, Daiichi Sankyo, Takeda, Knight, Ipsen, Menarini and AstraZeneca; receiving travel grants from Gilead, Roche, and Daiichi Sankyo; receiving research funding (to his institution) from Gilead; and having nonfinancial interests as the chair of the European Society for Medical Oncology (ESMO) Young Oncologists Committee (YOC) and as a member of the national council of the Italian Association of Medical Oncology.
N.L.V.: grant from Eisai; speaker bureau from GSK; travel expenses for conference from Gentili, Celgene, and Pfizer; advisory role from Novartis and Celgene; advisory role, travel expenses for conference from Pfizer; advisory board from MSD, Roche, Novartis, Astrazeneca, and Daiichi Sanyo.
A.G.: Research grants from Pharmanutra, AAA; advisory boards, activities as a speaker, travel grants, consultancy from Roche, Novartis, Pfizer, Eli Lilly, Daiichi Sankyo, AstraZeneca, MSD, Seagen, Gilead, Pierre Fabre, Eisai, Exact Sciences, Stemline.
A.Z.: honoraria for advisory board and consultancy for Roche, Novartis, Lilly, Pfizer, Seagen, Daiichi Sankyo, Astra Zeneca, Gilead, Merck, Exact Sciences, Gentili, Menarini Stemline.
A.T.: advisory role for Lilly, Novartis, AstraZeneca, Pfizer, Seagen, Gilead, and MSD; speaker honoraria from Lilly, Novartis, Pfizer; Travel Grants from Gilead, Daiichi Sankyo and Novartis.
M.P.: travel support from Pfizer and Novartis.
B.T.: speaker honoraria from Novartis, Daiichi Sankyo, Eli Lilly, Pfizer, Seagen. Travel grants from Novartis, Roche, Pfizer, Daiichi Sankyo.
D.G.: consulting/Advisor for Roche, Lilly, Novartis, Pfizer, Menarini, Stellantis; honoraria from Novartis, Pfizer, Lilly, AstraZeneca, Roche, Istituto Gentili; travel/accommodation from Novartis, Lilly, Pfizer, Roche.
F.L.: Honoraria as a speaker: Novartis, Pfizer, Eli-Lilly, Accademia di Medicina.
G.G.: invited speaker for Novartis, Eli Lilly, MSD; advisory role for Gilead, Seagen, Menarini.
E.M: received honoraria for Consulting or Advisory Role:Exact Sciences, MSD Oncology, Daiichi Sankyo, AstraZeneca, Pfizer, Seagen, Ipsen. Travel, Accommodations, Expenses: Roche, Pfizer,Lilly, Novartis, Gilead Sciences, AstraZeneca, Pierre Fabre.
A.M.: has received honoraria as a consultant, advisor, or speaker from Roche and Menarini/Stemline and has received travel and accommodation support from AstraZeneca.
S.S.: speaker fees from Novartis, Pfizer, Roche, Lilly, BMS, and MSD.
S.P.: Speaker for. Novartis, Eli Lilly, Pfizer, Roche, Gilead, Gentili, Sophos. Advisory Board for: Seagen, Daichii-Astra Zeneca.
C.C.: personal fees for consulting, advisory role, and speakers’ bureau from Lilly, Roche, Novartis, MSD, Seagen, Daiichi Sankyo, AstraZeneca, Gilead and Pfizer.
C.D.A.: advisory role for Roche, Lilly, Novartis, Astrazeneca, Pfizer, Seagen, Daicii-Sankyo, Gilead, and GSK and speaker honoraria from Roche, Lilly, Novartis, Pfizer, Seagen, GSK, GILEAD, and Daiichi-Sankyo. Travel Grants from Gilead and research support (to the Institution) from Novartis, Gilead, and Daiichi-Sankyo outside the submitted work.
G.Arp.: honoraria from Roche, Pfizer, AstraZeneca, Novartis, Celgene, Eli Lilly, Amgen, and Eisai.
V.G.: honoraria from Eli Lilly, Daiichi Sankyo, GSK, Gilead, Novartis, Exact sciences, Roche, Astra Zeneca, Menarini StemLine, Zentiva; advisory role for Eli Lilly, Daiichi Sankyo, Menarini Stemline, Exact Sciences, Gilead, Astra Zeneca, MSD, Novartis, Pierre Fabre, Pfizer-Seagen, Roche; payment for expert testimony from Eli Lilly; patents for HER2DX (Institution);
G.P.: personal fees from Foundation One, Illumina and Lilly. S.S.: speaker fees from Novartis, Pfizer, Roche, Lilly, BMS, and MSD. S.P.: Speaker for. Novartis, Eli Lilly, Pfizer, Roche, Gilead, Gentili, Sophos. Advisory Board for: Seagen, Daichii-Astra Zeneca.
C.V.: role in advisory boards for Pfizer, Novartis, Eli Lilly, Daiichi Sankyo, Astra Zeneca, Menarini Stemline, Roche; consultancy activity for Eli Lilly, Novartis, Pfizer, Gilead; honoraria as a speaker from Novartis, Eli Lilly, Pfizer, Astra Zeneca, Daiichi Sankyo, Menarini Stemline, MSD, Istituto Gentili, Accademia Nazionale di Medicina; research grants from Roche and Daiichi Sankyo/Astra Zeneca (to the Institution)