Background
An estimated 13% of women will be diagnosed with invasive breast cancer in their lifetime, and 3% will die from the disease [
1]. Metastatic breast cancer (MBC; i.e., breast cancer that has spread to distant sites) is an incurable disease with a 5-year survival rate of 28.1% among women [
2]. Current treatment guidelines recommend the addition of a cyclin-dependent kinase 4/6 (CDK4/6) inhibitor in combination with endocrine therapy for the treatment of patients with hormone receptor–positive (HR+)/human epidermal growth factor receptor 2–negative (HER2−) MBC based upon multiple prospective randomized phase III trials in the first- and second-line setting [
3‐
5].
Palbociclib, the first-in-class CDK4/6 inhibitor, is approved for the treatment of HR+/HER2− MBC in combination with an aromatase inhibitor or fulvestrant [
6]. The approval of palbociclib in the USA by the Food and Drug Administration was based on findings from 3 pivotal clinical trials: PALOMA-1 [
7,
8] and PALOMA-2 [
9,
10] evaluated palbociclib in combination with letrozole versus letrozole alone or placebo plus letrozole as the initial treatment for postmenopausal women in the advanced setting; PALOMA-3 [
11‐
15] evaluated combination treatment with fulvestrant versus fulvestrant alone in premenopausal or postmenopausal women who had progressed following endocrine therapy. The final overall survival (OS) results from PALOMA-1 showed that palbociclib plus letrozole had a longer median OS than letrozole alone, although the improvement was not statistically significant (37.5 vs 34.5 months; hazard ratio, 0.90 [95% CI, 0.62–1.29];
P = 0.281) [
8]. Similarly, although not statistically significant, palbociclib plus fulvestrant was associated with a longer median OS than placebo plus fulvestrant in PALOMA-3 (34.9 vs 28.0; hazard ratio, 0.81 [95% CI, 0.64–1.03];
P = 0.09), preserving the 7-month improvement in progression-free survival (PFS) [
15]. The OS data for PALOMA-2 are not yet available.
Real-world data are increasingly being used to understand the safety and effectiveness of new drug regimens in actual clinical practice [
16,
17]. Inherent limitations of real-world analyses using data collected during routine care include the lack of random treatment assignment, variations in follow-up and lack of uniform assessment of disease progression, missing data or erroneous data entry, and incomplete capture of comorbid conditions and performance status. Published real-world studies of palbociclib have shown an efficacy and safety profile consistent with that observed in clinical trials; however, these studies were limited by small sample sizes, the lack of a control group, and/or short durations of follow-up [
18‐
20]. The goal of this study was to provide real-world evidence regarding the effectiveness of palbociclib plus letrozole versus letrozole alone in a large cohort of patients from across the USA in routine clinical practice.
Methods
Study design and data source
This retrospective analysis of electronic health records (EHRs) used de-identified patient data from the Flatiron Health Analytic Database, a longitudinal database that includes structured and unstructured EHRs from > 280 cancer clinics, including approximately 800 sites of care, and represents 2.4 million patients with cancer actively being treated in the USA. A patient attrition diagram is presented in Additional file
1. Patients included in this analysis met the following inclusion criteria: women aged ≥ 18 years at MBC diagnosis with HR+/HER2− MBC before or up to 60 days after the metastatic diagnosis date; a date of first prescription (index date) for palbociclib plus letrozole or letrozole alone as first-line therapy for MBC beginning on the date of the US Food and Drug Administration approval of palbociclib, February 3, 2015, and extending 4 years to February 28, 2019; with a minimum potential follow-up for ≥ 3 months from the index date to the study cutoff date of May 31, 2019. Patients were excluded if they received prior treatment with CDK4/6 inhibitors, aromatase inhibitors, fulvestrant, tamoxifen, raloxifene, or toremifene in the metastatic setting; had a first structured activity (a recording of vital information, a medication administration, a non-canceled drug order, or a reported laboratory test/result) > 90 days after the MBC diagnostic date; or received a CDK4/6 inhibitor as part of a clinical trial.
Outcomes
These comparative analyses were conducted to assess real-world PFS (rwPFS) as the primary outcome and OS as the secondary outcome in patients with MBC treated with palbociclib plus letrozole versus letrozole alone as first-line therapy [
21,
22]. rwPFS was defined as the time in months from the start of palbociclib plus letrozole or letrozole alone therapy to death or disease progression. Disease progression was determined by the recorded assessment of the treating clinician based on radiology, pathology, clinical assessment, or laboratory evidence. Patients who did not die or have disease progression were censored at the date of initiation of the next line of therapy for those with ≥ 2 lines of therapy or at their last visit during the study period of February 2015 to May 2019 for patients with only 1 line of therapy. The duration of follow-up was defined as the time in months from the start of palbociclib plus letrozole or letrozole alone to death or the data cutoff date of May 31, 2019, whichever came first.
Overall survival was defined as the number of months from the start of treatment with palbociclib plus letrozole or letrozole alone to death due to any cause as recorded by Flatiron in the data extract. The date of death was acquired from a recent mortality dataset generated by combining multiple data sources and benchmarked against the National Death Index [
23]. Patients who did not die were censored at the study cutoff date (May 31, 2019).
Statistical analyses
Means, standard deviations, medians, and interquartile ranges (IQRs) were calculated for continuous variables. Counts and percentages were reported for dichotomous and polychotomous variables. T-tests or chi-square tests, as appropriate, were performed to compare treatment groups for baseline demographic and clinical characteristics. The Kaplan-Meier method and 95% CIs were used to estimate medians for rwPFS and OS. Cox proportional regression analyses with a robust sandwich estimator were used to estimate hazard ratios and 95% CIs for rwPFS and OS events. In subgroup analyses, a subgroup by treatment interaction term was included.
Comparative analyses were conducted and presented by 3 methods: unadjusted (without controlling for confounders), stabilized inverse probability treatment weighting (sIPTW) method (the primary analysis that controlled for observed confounders), and finally the propensity score matching (PSM) method (a sensitivity analysis that assessed the robustness of the primary analysis results). Propensity score calculation and its application as either matching (PSM) or weights (sIPTW) were conducted to balance baseline demographic and clinical characteristics and to adjust for differences in observed potential confounders between the two cohorts. The sIPTW approach was used as the primary analysis. A multivariable binomial logistic regression model was used to generate the propensity scores [
24‐
27]. Patients in the two study cohorts were matched by the propensity score matching (PSM) method. The nearest neighbor method (without replacement and with a caliper of 0.01 [
24]) was used to select the matched samples. PSM of baseline demographic and clinical characteristics at a ratio of 1:1 then identified 464 patients in each treatment group. This approach was used as a sensitivity analysis to determine the robustness of the findings from the primary analyses [
28].
To mitigate the effect of missing survival data, a major confounding factor in real-world oncology analyses, and in the absence of reliable information on the cause of missing data, calculation of percentages always included a “missing” category presented by treatment. For some categorical variables, “missing” was included as another level, e.g., Eastern Cooperative Oncology Group (ECOG) performance status included a level of “not documented.”
Discussion
With only 3% of patients with breast cancer participating in randomized clinical trials [
29], real-world evidence is essential to evaluate the generalizability of clinical trial findings across a heterogeneous patient population and among patients who would not have met clinical trial inclusion criteria [
16]. Using electronic health records from the large multicenter Flatiron Health Analytic Database, we found that first-line palbociclib plus letrozole was associated with significantly longer rwPFS and OS than letrozole alone after sIPTW. A landmark OS analysis at 2 and 3 years of follow-up demonstrated significantly higher OS rates in the palbociclib combination group. A consistent benefit of palbociclib plus letrozole was observed across multiple subgroups. Results from the PSM sensitivity analysis also supported the results from the sIPTW primary analysis.
These findings from a large cohort of real-world patients are complementary to the clinical efficacy findings of palbociclib plus endocrine therapy observed in randomized clinical trials [
7,
10,
12]. The rwPFS benefit of palbociclib plus letrozole versus letrozole alone (hazard ratio, 0.58 [95% CI, 0.49–0.69];
P < 0.0001) observed in the current study is consistent with PFS results from the PALOMA-2 trial (hazard ratio, 0.56 [95% CI, 0.46–0.69];
P < 0.0001) [
9,
10]. Patients in the present (unadjusted) analysis were older than those enrolled in PALOMA-2, and a higher percentage of patients across both cohorts had bone-only disease at baseline compared with PALOMA-2. The current real-world data analysis also demonstrated a significant OS benefit with palbociclib plus letrozole versus letrozole alone (hazard ratio, 0.66 [95% CI, 0.53–0.82]). The OS benefit was generally consistent across most subgroups. However, it is unknown if the OS benefit observed in this real-world analysis would be consistent with PALOMA-2 results as OS data are not yet available.
Although there are intrinsic limitations to retrospective analyses of real-world data, the large size and geographic distribution of the Flatiron Database is a strength of this study. Real-world effectiveness endpoints for this analysis include rwPFS and OS; a recent publication by Bartlett et al. validated the real-world effectiveness endpoints as defined by Flatiron in comparison with randomized clinical trial data [
30], and data from Flatiron have been used to support the label expansion for palbociclib to include men [
31]. In addition, the OS endpoint from the Flatiron database includes external data sources (i.e., the National Death Index, US Social Security Death Index, obituaries, and commercial death data) in addition to health records and has been validated [
23]. As described previously, the absence of randomization in treatment assignment common to observational data is an important factor that must be accounted for through appropriate statistical tools to adjust for differences in baseline and clinical characteristics. The statistical approach used in this analysis to address observed confounders included the use of PSM to balance patient demographic and clinical characteristics; PSM results were consistent with sIPTW analyses.
Unobserved variables cannot be fully addressed through sIPTW or PSM and thus may confound these findings; however, this analysis adjusted for the known clinical confounders that were most likely to have a major impact on the outcomes of the study. The 5-year survival for female patients with MBC remains < 30% [
2]; thus, the impact of those potential variables on overall survival as an outcome would have to be profound to affect the findings of the current study. Some subgroups, such as younger patients (aged ≤ 50 years), may have insufficient cohort size to detect significant differences in outcomes. At the end of the study period (data cutoff), over a third of patients remained on treatment, and both cohorts had a high frequency of a CDK4/6 inhibitor-containing regimen as a subsequent second-line therapy. The use of CDK4/6 inhibitors beyond the first-line setting may have affected the overall survival results by overestimating the overall survival in the letrozole alone cohort, therefore biasing the results towards the null. However, definitive analysis of the impact of subsequent treatments on overall survival was limited by the small percentage of patients with available data. The efficacy of subsequent CDK4/6 inhibitor treatments beyond initial disease progression is an active area of investigation, and additional studies, including analysis of rwPFS after subsequent therapy in real-world clinical practice, are needed to understand the impact of treatment sequencing. Disease progression was based on the individual treating physician’s clinical assessment or interpretation of radiographic or pathologic results and not on standard criteria, such as Response Evaluation Criteria in Solid Tumors. Additionally, findings from the Flatiron Database may not be generalized to other patient populations. Finally, although median OS was reached in the letrozole alone group, significant censoring in the OS analysis highlights the need for subsequent evaluation with longer follow-up.
Conclusions
Although findings from the PALOMA trials demonstrate the efficacy of palbociclib among patients who met study inclusion and exclusion criteria, real-world evidence of effectiveness provides further support for clinical decision-making in patients encountered in routine clinical practice [
16,
32]. In this comparative effectiveness analysis, first-line palbociclib plus letrozole was associated with longer rwPFS and OS than letrozole alone in a heterogeneous population and among various patient subgroups. While overall survival data are not yet available for the PALOMA-2 randomized controlled trial, these results expand our understanding of clinical outcomes in the real-world setting and provide additional evidence regarding the use of palbociclib in combination with an aromatase inhibitor as a first-line treatment of patients with HR+/HER2− MBC.
Acknowledgements
Editorial support for the development of this manuscript was provided by Jill Shults, PhD, of ICON plc (North Wales, PA), and funded by Pfizer Inc.
Open AccessThis article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit
http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (
http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.