Key results
This study assessed the effectiveness of CDK 4/6i within the Danish population using retrospective data. Abemaciclib and ribociclib in first line treatment showed a prolonged PFS compared to palbociclib, remaining significant in an adjusted multivariable model. Age, sites of metastases, endocrine backbone, and endocrine sensitivity influenced first-line PFS. For OS, there were no statistically significant differences between the inhibitors, despite ribociclib having a longer median OS in both first and second-line patients.
Strengths and limitations
In terms of strengths and limitations, several aspects deserve consideration. First, Table1 reveals the increasing use of abemaciclib. However, it is noteworthy that the palbociclib group stands larger and with a longer follow-up duration attributed to its earlier approval compared to abemaciclib. Despite this, efforts ensured a follow-up of at least 12 months for all patients, maintaining a relatively substantial abemaciclib sample size. Limited follow-up for abemaciclib, and ribociclib in second line, hindered accurate PFS and OS estimates as well as confidence intervals. Due to the time-varying introductions of the CDK 4/6i to the market, there is a clear discrepancy between the three treatment groups in this study regarding follow-up. In efforts to make comparisons between the inhibitors easier we were able to calculate the survival probability at 24 months for all three inhibitors. Secondly, due to the novelty of CDK 4/6i treatment some patients missed initial first-line treatment due to their time of disease recurrence, influencing second-line ratios with seemingly fewer patients towards 2021. This suggests that the treatment has become more established as first-line treatment for metastatic BC.
Furthermore, there were 78 patients with data missing on adjuvant endocrine therapy that were deemed endocrine sensitive due to 6 or more years between primary diagnosis and recurrence. Some of these cases could be misclassified although we argue that it has negligible result impact. Finally, the choice of CDK 4/6i is influenced by the individual clinician’s considerations. It is also the individual clinician who verifies progression, thus there is no way of knowing if there exist cases where the patient has been deemed in progression due to solely clinical progression, i.e. not verified by scans. We argue that these cases are few since the clinicians follow national guidelines as a rule. In addition, regional differences in inhibitor choice introduce unaccounted selection bias, given the absence of guidelines recommending a certain inhibitor above the others during the study period.
The current study boasts several strengths in addition to acknowledging its limitations. It is founded on a population-based cohort of considerable size, minimizing the risk of type II errors, and holds real-world Danish applicability. The retrieval of up-to-date follow-up data further solidifies the integrity of this comprehensive cohort.
Interpretation
Study population and treatment characteristics
As displayed in Supplementary Fig.2, the distribution of patients switching CDK 4/6i within first-line treatment varied across the treatment groups. The introduction of palbociclib in Denmark in 2017 as the sole option might explain its lower rate of patients switching. Conversely, the abemaciclib group had a larger number of patients opting for a second choice of CDK 4/6i given the availability of both palbociclib and ribociclib. Additionally, the abemaciclib group had more than double the number of treatment discontinuations due to toxicities than both palbociclib and ribociclib (Supplementary Table3). Many patients likely transitioned to a different CDK 4/6i due to distinct toxicity profiles. Abemaciclib especially has toxicities related to the gastro-intestinal system whereas palbociclib and ribociclib have toxicities related to myelosuppression [5, 16]. Our findings support previous research which has shown that abemaciclib prompts more toxicity related treatment discontinuations [5]. Thus, it seems that palbociclib and ribociclib provoke adverse events that are primarily abnormalities shown on lab results without necessarily affecting the quality of life whereas for abemaciclib, the gastro-intestinal toxicities can be highly intolerable.
Lastly, the patients that received palbociclib treatment in first line had more visceral metastases and the multivariable analysis confirmed the significance of sites of metastases on PFS. Since palbociclib was the first CDK 4/6i type to be used in Denmark, it is likely that clinicians were stricter with the use of it and that it was chosen for patients with more serious conditions whereas the use is more loose today. Even though our results showed a significant effect of sites of metastases on PFS, we adjusted for this in our multivariable analysis and type of CDK 4/6i was still significant, which reduces the potential confounder.
Progression free survival compared to randomized trials
Our study revealed longer PFS results compared to randomized phase III studies. The PALOMA-2 study found a median PFS for palbociclib and letrozole of 27.6 months (95% CI: 22.4–30.3) [17] whereas our study found a median PFS for palbociclib with AI or fulvestrant of 32.0 months (95% CI: 28.9–35.3) in first-line patients. Unlike PALOMA-2, our study included both post- and premenopausal women, varied endocrine backbones (letrozole, anastrozole, exemestane, fulvestrant), and real-world patients. The real-world context was anticipated to yield a shorter PFS, especially since the current study included more palbociclib patients with visceral metastases than PALOMA-2 (71.5% vs. 48.2%). Additionally, our multivariable model noted higher progression risk with anastrozole and fulvestrant compared to letrozole. However, previous research has found fulvestrant to be equally beneficial to an AI and international guidelines for first-line treatment recommend both as options [10, 18]. Nevertheless, our longer median PFS compared with PALOMA-2 hints at the promise of real-world outcomes, although one should take note of the overlapping confidence intervals between PALOMA-2 and our study and the fall-pits of cross-study comparison.
The MONALEESA-2 study, a phase III randomized study, found a median PFS for ribociclib and letrozole of 26.4 months (95% CI: 23.0–30.3) [19] where our results showed a median PFS of 42.4 (95% CI: 35.1–52.9) which was significantly higher. The differences were similar as those described between the current study and the PALOMA-2 study. Looking at abemaciclib, the MONARCH-3 study showed a median PFS of 28.18 months [20], whereas our study found a survival probability of 65.0% at 24 months. This suggests a potential median PFS aligning with MONARCH-3, although it is necessary to extend the follow-up period for abemaciclib in the current study to make a conclusive analysis. These findings emphasize the favorable real-world effectiveness of CDK 4/6i treatment regarding PFS.
Comparing abemaciclib, palbociclib, and ribociclib in the current study, ribociclib had a significantly longer median PFS than palbociclib. However, several baseline differences existed, for instance the palbociclib group having more visceral metastases. Despite recognizing the influence of the sites of metastases on PFS, the multivariable model continued to show a significant difference in PFS between the inhibitors when adjusting for factors like metastatic sites. Furthermore, when comparing the survival probability at 24 months in first-line patients, the difference between treatment groups bordered the significance level with a lower survival probability for palbociclib. A possible explanation for the shorter PFS for palbociclib could be that evaluation scans were performed more frequently, i.e. every 3 months. Because ER-positive breast cancer evolves more slowly the clinicians have begun to evaluate every 4 or 6 months after the first year, if the patient experienced no symptoms. This means that for ribociclib and abemaciclib progression is potentially detected later, causing potential bias.
Progression free survival compared to real world studies
In recent years, a handful of studies have established the benefits of CDK 4/6i in a real-world population compared to AI monotherapy [21, 22]. A small retrospective study found 140 patients receiving CDK 4/6i treatment between 2014 and 2019 [23]. They had a median age of 65 similar to our study population and similarly a majority of recurrent breast cancer with visceral metastases. Furthermore, discontinuation of treatment was primarily due to progression which correlates to our findings. They found a median PFS of 6.3 months (95% CI: 5.1–7.4) which was significantly lower than ours. It should be noted that the majority of the patients in the study by West et al. had received treatment for metastatic breast cancer before treatment with CDK 4/6i. Thus, they did not receive CDK 4/6i treatment in first line, which makes the comparison to the current study problematic.
To the best of our knowledge, studies have so far established the benefits of CDK 4/6i, but the large part of the studies include primarily palbociclib and thus far, we have found no real-world studies that compare the three CDK 4/6i to each other.
Overall survival compared to randomized trials
We found a median OS for patients that received any CDK 4/6i in first line of 50.8 months (95% CI: 46.9–54.4). Deluche et al. investigated real-world outcomes for ER-positive/HER2-negative metastatic BC, reporting a median OS of 43.3 months (95% CI: 42.5–44.5) for 22000 patients treated between 2008 to 2016, predating CDK 4/6i use [3]. The current study’s real-world data, encompassing 1554 receiving a CDK 4/6i as first-line treatment for metastatic BC, suggests significant treatment benefits.
The current study’s OS outcomes stratified by CDK 4/6i type differed from randomized trials. In the MONARCH-3 interim OS analysis, Goetz et al. reported a median OS of 67.1 months (confidence intervals not published yet) with 70.2 months median follow-up for abemaciclib and a non-steroidal AI [7]. The current study found 37.8 months median OS (32.5–NA) with 24.8 months median follow-up, a high censoring rate preventing upper CI estimation for abemaciclib. For palbociclib, with 55.1 months median follow-up, our median OS was 48.7 months (95% CI: 44.7–54.1), similar to PALOMA-2’s 53.9 months (95% CI: 49.8–60.8) for palbociclib and letrozole [9]. For ribociclib we were able to obtain a median follow-up of 47.9 months, which gave a median OS of 54.4 months (47.9-NA). MONALEESA-2 reported a median OS of 63.9 months (95% CI: 52.4–71.0) with a longer follow-up of 6.6 years [8]. The limited CI in the current study hinders precise cross-study comparison.
No significant difference was observed in OS between the treatment groups. Extending follow-up, especially for abemaciclib, is essential for more accurate OS estimates. PALOMA-2 reported no significant difference in OS between the palbociclib group and the placebo group while MONARCH-3 and MONALEESA-2 reported benefit with, respectively, abemaciclib and ribociclib compared to placebo groups [7,8,9]. Further investigations are needed to validate if palbociclib’s OS benefit is in fact inferior to abemaciclib and ribociclib. Although the OS for ribociclib was notably prolonged compared to palbociclib in the current study, the difference is not statistically significant, likely due to the duration of follow-up. Extended follow-up would clarify the significance more accurately. Additionally, information on subsequent treatment after progression on CDK 4/6i treatment could provide valuable knowledge in the context of overall survival.
Endocrine sensitivity
The current study revealed statistically significant PFS improvement in first-line endocrine sensitive patients vs. first-line endocrine resistant patients (Supplementary Fig.1a). The influence of endocrine sensitivity on PFS was significant in both the univariate end the multivariate models (Supplementary Table2). This aligns with established research indicating that long term ET induces acquired endocrine resistance [24].
PALOMA-3 investigated the efficacy of palbociclib and fulvestrant in patients who had relapsed or progressed during prior ET [25]. Per ESMO guidelines, these patients would be categorized as endocrine resistant [12]. Their PFS of 9.2 months (95% CI: 7.5 to not estimable) was shorter than results from PALOMA-2 based on mainly endocrine sensitive patients. The current study is in accordance with those findings, showing shorter PFS in endocrine resistant patients as well as in second-line patients compared with first-line patients, indicating the need to address endocrine resistance in metastatic BC. Similarly, MONARCH-2 found a median OS of 16.4 months in endocrine resistant patients – significantly shorter than MONARCH-3 results in endocrine sensitive patients [26].
In general, fulvestrant was used more commonly as endocrine backbone in the palbociclib groups in first and second line compared to ribociclib and abemaciclib (Table1 and Supplementary Table1). Palbociclib was the initial CDK 4/6i in Denmark and patients were likely treated with endocrine monotherapy before it, making fulvestrant the natural endocrine backbone choice for second-line patients. Alternatively, physician preference might drive backbone selection considering AI and fulvestrant are considered equally beneficial, as previously discussed [10, 18].
Furthermore, patients treated with first line CDK 4/6i differed significantly in endocrine sensitivity (Supplementary Table4a). The palbociclib group had more endocrine sensitive patients who received fulvestrant. It is important to recognize that a significant difference in endocrine backbone between the treatment groups makes it difficult to correctly compare the subgroups of endocrine backbone regarding endocrine sensitivity.
