Integration of pre‑existing cardiovascular comorbidity into CDK4/6 inhibitor selection for breast cancer.

Introduction
Cyclin-dependent kinase 4/6 (CDK4/6) inhibitors, including palbociclib, ribociclib, and abemaciclib, have significantly advanced the treatment of hormone receptor+, HER2− advanced breast cancer by consistently improving survival when combined with endocrine therapy.
1
Despite their effectiveness, CDK4/6 inhibitors have shown varying cardiovascular (CV) safety profiles that may influence their clinical optimal selection for patients with existing cardiovascular disease (CVD) or cardiometabolic risk factors.
2
Ribociclib, in particular, has been associated with a greater risk of CV adverse events compared to palbociclib and abemaciclib. Evidence from both randomized controlled trials and real-world data has linked ribociclib with QT interval prolongation and increased risk of major adverse CV events.3
–5 A recent systematic review and network meta-analysis further corroborated these safety differences, identifying ribociclib as having the highest CV risk profile among the CDK4/6 inhibitors.
6
By contrast, abemaciclib has more often been associated with venous thromboembolism, whereas palbociclib generally shows a comparatively lower CV signal.
7
Together, these contrasts frame the clinical context for agent selection.
Despite established differences in CV risk among CDK4/6 inhibitors, it remains unclear how pre-existing CVD and related risk factors (e.g., hypertension, hyperlipidemia, diabetes) influence their selection in real-world practice. In particular, whether patients with cardiometabolic comorbidities are more likely to receive inhibitors with safer CV profiles is unknown. This gap is critical, as prescribing higher-risk agents to patients with HR+/HER2– breast cancer may increase the likelihood of CV adverse events and associated healthcare costs. Focusing on initial drug selection, rather than downstream outcomes, addresses a modifiable, upstream decision point that complements prior event-rate studies by revealing whether prescribers align agent choice with baseline CV risk. Moreover, understanding real-world prescribing patterns also enables targeted interventions for multiple stakeholders: clinicians can integrate CV risk stratification into CDK4/6 selection algorithms; health systems can embed decision-support tools into electronic prescribing; and policymakers and payers can develop risk-based formulary policies that balance therapeutic access with CV safety.
In this context, we hypothesized that ribociclib, given its higher risk of CV adverse events, would be prescribed less frequently to patients with breast cancer who have pre-existing CVD or cardiometabolic risk factors. Accordingly, we examined patterns of CDK4/6 inhibitor selection among patients with breast cancer and pre-existing CVD or cardiometabolic conditions (e.g., hypertension, hyperlipidemia, diabetes).

Methods

Study design and data source
We conducted a retrospective cohort study using the 2017–2021 Merative® MarketScan™ Commercial Claims and Encounters Database. This de-identified database contains medical and pharmacy claims for approximately 30 million employer-sponsored beneficiaries each year. The Institutional Review Board at The University of Texas at Austin determined that this study does not involve human subjects and is exempt from IRB review.

Study population
The study population comprised women with breast cancer who initiated CDK4/6 inhibitors. Patients were included if they met all of the following criteria: (1) age ⩾18 years; (2) female sex; (3) documented breast-cancer diagnosis (International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) C50.xx); (4) first prescription for a CDK4/6 inhibitor after the earliest breast-cancer diagnosis regardless of line of therapy; and (5) at least 12 months of continuous enrollment prior to the CDK4/6 inhibitor initiation date. Patient selection of the final analytic cohort is provided in Supplemental Table 1.

Measures
The outcome was the specific CDK4/6 inhibitor prescribed (palbociclib, ribociclib, or abemaciclib) identified by National Drug Codes (NDCs) between January 1, 2018 and December 31, 2021.
Factors were CVD and cardiometabolic risk factors. CVD was defined as the presence of any of the following conditions: atrial fibrillation, coronary artery disease, cardiomegaly, cardiomyopathy, heart failure, peripheral artery disease, stroke, myocardial infarction, angina, or arrhythmia. Cardiometabolic risk factors included hypertension, hyperlipidemia, and diabetes. Both CVD and cardiometabolic risk factors were assessed during the 12-month look-back period preceding the initiation of CDK4/6 inhibitors. Relevant ICD-10 codes are available in Supplemental Table 2.
Additional variables were pre-specified and included age, geographic region, and insurance plan type at the time of CDK4/6 inhibitor initiation, as well as receipt of chemotherapy, endocrine therapy, targeted therapy, surgery, radiation, and metastasis status during the look-back period. Relevant ICD-10-CM, Healthcare Common Procedure Coding System (HCPCS), and NDC codes are available in Supplemental Tables 3–5.
To ensure a 12-month baseline period for all patients, we limited CDK4/6 inhibitor initiation to 2018–2021; 2017 claims were used exclusively to construct the 12-month look-back for patients initiating CDK4/6 inhibitors in 2018. Patients who initiated a CDK4/6 inhibitor in 2017 were excluded because 2016 claims were unavailable to support a complete baseline period. Breast cancer diagnosis could occur in 2017 and precede CDK4/6 initiation; such patients were included if they met all inclusion criteria.

Statistical analyses
Descriptive statistics summarized demographic and clinical characteristics across CDK4/6 inhibitor groups. Chi-square tests were applied for categorical factors and variables, and analysis of variance was used for continuous variables.
Multinomial logistic regression models were used to estimate odds ratios (OR) for initiating palbociclib or abemaciclib (vs ribociclib) among patients with breast cancer, based on their history of CVD and cardiometabolic risk factors. We prespecified ribociclib as the reference category to align with our hypothesis regarding its higher CV risk profile and to directly quantify the odds of initiating alternative agents relative to ribociclib. As a sensitivity analysis, we re-estimated models using palbociclib as the reference.
Unadjusted models were first estimated without additional variables, followed by adjusted models, controlling for demographic and clinical variables. All analyses were conducted using SAS version 9.4, with two-tailed tests and a significance level set at p < 0.05. The reporting of this observational cohort study conforms to the STROBE statement.
8

Results
Table 1 summarizes baseline characteristics for 5002 women who initiated a CDK4/6 inhibitor (palbociclib n = 3734; ribociclib n = 328; abemaciclib n = 940). Palbociclib initiators were significantly older than ribociclib and abemaciclib users (mean ± SD = 60.1 ± 12.8 vs 56.3 ± 12.0 and 56.9 ± 12.1 years, respectively; p < 0.001). Comorbidity burden did not differ among groups (Charlson Comorbidity Index: 8.8 ± 2.1, 8.7 ± 1.8, and 8.7 ± 2.2 for palbociclib, ribociclib, and abemaciclib, respectively; p = 0.486). Overall CVD prevalence was similar, except for atrial fibrillation (p = 0.028), heart failure (p = 0.011), and arrhythmia, which was most common with abemaciclib (37.9%) and least common with palbociclib (27.4%; p < 0.001). Among cardiometabolic risk factors, hypertension was highest with ribociclib (46.7%) and lowest with abemaciclib (40.4%; p = 0.024), whereas hyperlipidemia and diabetes frequencies were comparable across groups (p = 0.190 and p = 0.235, respectively).
Figure 1 shows the odds of initiating palbociclib (a) or abemaciclib (b) versus ribociclib among women with breast cancer, stratified by pre‑existing CVD and cardiometabolic risk factors. In the unadjusted models (top rows), no risk factor reached statistical significance; however, hypertension displayed a non‑significant trend toward lower initiation of palbociclib (OR 0.94; 95% confidence interval (CI) 0.75–1.18) and abemaciclib (OR 0.78; 95% CI 0.60–1.00). After adjustment for other demographic and clinical variables (bottom rows), hypertension emerged as the only significant factor: patients with hypertension had 33% lower odds of receiving palbociclib (adjusted odds ratio (AOR) 0.67; 95% CI 0.51–0.87) and 40% lower odds of receiving abemaciclib (AOR 0.60; 95% CI 0.45–0.81) compared to ribociclib. Neither a composite history of CVD nor individual conditions such as hyperlipidemia or diabetes significantly influenced CDK4/6 inhibitor selection once additional variables were accounted for. Supplemental Tables 6 and 7 show the estimated models without and with adjustment for additional variables, respectively. As a sensitivity analysis, Supplemental Tables 8 and 9 show the estimated models using palbociclib as the reference, without and with adjustment for additional variables, respectively.

Discussion
In this retrospective cohort study of 5002 women with breast cancer who initiated a CDK4/6 inhibitor, we assessed whether pre-existing CVD and CV risk factors influenced the choice of agent. Unexpectedly, ribociclib, the CDK4/6 inhibitor associated with a higher risk of CV adverse events, was more frequently prescribed to patients with hypertension, even after adjusting for demographic and clinical characteristics. Specifically, patients with hypertension had a significantly lower likelihood of initiating palbociclib (AOR 0.67; 95% CI 0.51–0.87) or abemaciclib (AOR 0.60; 95% CI 0.45–0.81) relative to ribociclib. No statistically significant associations were observed between inhibitor selection and other conditions (documented CVD, hyperlipidemia, or diabetes).
These findings suggest a potential disconnect between clinical evidence and real-world prescribing practices. Recent studies have shown that ribociclib is associated with QT-interval prolongation and higher CV risk compared to palbociclib and abemaciclib.
9
Our results indicate that these CV-related safety differences may not be fully considered when choosing therapy for patients with CV risk factors, including those with a history of CVD. This is particularly concerning, as patients with hypertension are more likely to experience treatment-related CV complications.
A recent real-world effectiveness data suggesting improved survival outcome with ribociclib compared to palbociclib may influence prescribing behavior, potentially overriding CV safety considerations.
10
In addition, provider familiarity with monitoring protocols for ribociclib’s cardiac effects may create confidence in managing these risks.
11
While other considerations, such as patient preferences, institutional protocols, or formulary differences, may also play a role, the finding is particularly concerning given that palbociclib actually has broader insurance coverage and established patient assistance programs. The consistent inverse association between hypertension and the use of palbociclib or abemaciclib suggests a need for better integration of CV risk assessment into prescribing decisions.
Lower initiation of CDK4/6 inhibitors with safer CV profiles in patients with hypertension may lead to higher healthcare utilization and adverse outcomes. CV events occur in 24% of patients receiving CDK4/6 inhibitors, with a median onset of 2.3 months.
4
They also elevate mortality risk, with hazard ratios of 4.89 for cardiomyopathy/heart failure and 5.88 for atrial fibrillation.
5
These complications, such as arrhythmias and treatment interruptions, can significantly increase hospitalizations and monthly healthcare costs for hypertension and for major adverse events.5,12

Limitations
Several limitations should be considered when interpreting these findings. First, administrative claims data do not include clinical details such as blood pressure values, disease severity, or laboratory results. As a result, we could not evaluate the degree of hypertension control or CVD severity, both of which may influence treatment decisions. Second, the dataset lacks information on drug tolerability, patient adherence, and prescribing rationale, including potential influences such as patient preferences or insurance coverage. Third, although we adjusted for multiple measured variables, unmeasured and residual confounding may remain. Potential unmeasured confounders include provider specialty, clinical experience, prescribing preferences, and local drug availability. In addition, several measured variables present in claims data (e.g., copay, coinsurance, deductible) were not incorporated into our adjusted models. These unmeasured and residual factors may substantially influence CDK4/6 inhibitor selection independently of CV safety profiles. Fourth, we analyzed CVD as a composite measure rather than examining individual CVD subtypes separately because several categories had small sample sizes that would have yielded unstable estimates. This composite approach may mask heterogeneous associations. Future studies with larger cohorts should examine whether specific CVD subtypes differentially influence CDK4/6 inhibitor selection. Fifth, generalizability is limited because the MarketScan database primarily captures commercially insured populations. Therefore, our findings may not generalize to Medicaid, uninsured patients, or to non-U.S. settings. Lastly, we could not distinguish first-line from later-line CDK4/6 inhibitor therapy due to inherent limitations in claims data. Accordingly, analyses should be interpreted as associations with the initial agent selected, irrespective of treatment line. However, CV safety considerations apply across treatment lines, and this inclusive approach provides a comprehensive assessment of how baseline CV comorbidity influences real-world CDK4/6 inhibitor prescribing. Taken together, these limitations indicate that this analysis is exploratory and associational rather than predictive or causal. Our findings should be viewed as hypothesis-generating observations that describe prescribing patterns and identify potential gaps between documented safety evidence and clinical practice.
Despite these limitations, the study provides meaningful insights. The findings support the development of clinical guidelines and decision-making tools that incorporate CV risk into CDK4/6 inhibitor selection. Aligning treatment choices with individual comorbidity profiles, particularly in patients with hypertension, may help prevent avoidable CV events and improve overall treatment outcomes.

Conclusion
In conclusion, CDK4/6 inhibitor prescribing did not consistently reflect CV risk profiles, especially among patients with hypertension. Despite its higher CV risk, ribociclib was more often selected for these patients. Incorporating CV risk prediction and comorbidities into CDK4/6 inhibitor selection could reduce preventable CV complications and associated costs, improving the alignment of treatment with patient risk.

Supplemental Material

sj-docx-1-taw-10.1177_20420986251414588 – Supplemental material for Integration of pre-existing cardiovascular comorbidity into CDK4/6 inhibitor selection for breast cancer

Supplemental material, sj-docx-1-taw-10.1177_20420986251414588 for Integration of pre-existing cardiovascular comorbidity into CDK4/6 inhibitor selection for breast cancer by Chanhyun Park, Kiyoung Kim, Nora B. Abifaraj, Sydney D. Bryant and Ji Haeng Heo in Therapeutic Advances in Drug Safety

sj-docx-2-taw-10.1177_20420986251414588 – Supplemental material for Integration of pre-existing cardiovascular comorbidity into CDK4/6 inhibitor selection for breast cancer

Supplemental material, sj-docx-2-taw-10.1177_20420986251414588 for Integration of pre-existing cardiovascular comorbidity into CDK4/6 inhibitor selection for breast cancer by Chanhyun Park, Kiyoung Kim, Nora B. Abifaraj, Sydney D. Bryant and Ji Haeng Heo in Therapeutic Advances in Drug Safety