Cardiovascular event burden in women with breast cancer treated with CDK4/6 inhibitors: a retrospective matched cohort study

Abstract

Background

Cyclin-dependent kinase (CDK) 4/6 inhibitors improve outcomes in hormone receptor–positive, HER2-negative breast cancer, but real-world data on cardiovascular (CV) toxicity are limited.

Methods

In this single-centre retrospective matched cohort study, women with breast cancer treated with a CDK4/6 inhibitor at University Hospitals Leuven between 2016 and 2022 constituted the index cohort. Age- and BMI-matched women without active cancer, selected from the population-based FLEMENGHO study, constituted the reference cohort. CV events were identified from medical records, and follow-up in the reference cohort was truncated to the maximum follow-up duration of the index cohort to improve comparability.

Results

Among 536 matched pairs, the index cohort had a significantly higher observed incidence of CV events than the reference cohort (67.7 vs 17.2 per 1,000 person-years; P < 0.0001), including the cardiac composite (28.0 vs 9.6 per 1,000 person-years; P < 0.0001). Venous thromboembolism (11.6% vs 0.75%) and atrial fibrillation/flutter (4.7% vs 1.5%) were more frequent in the index cohort. Coronary events, stroke, and heart failure did not differ significantly. After adjustment for measured baseline CV risk factors, index-cohort membership remained associated with higher risks of CV events (HR, 3.63; 95% CI, 2.57–5.13) and cardiac events (HR, 3.25; 95% CI, 1.97–5.35).

Conclusions

Women with breast cancer receiving CDK4/6 inhibitors had a substantially higher observed burden of venous thromboembolism and atrial arrhythmias than the age- and BMI-matched population reference cohort without active cancer. These findings support cardiovascular awareness in this population. However, the comparison cannot isolate a drug-specific effect because active malignancy, previous cancer therapies, baseline differences, and surveillance intensity may also contribute.

Trial registration

This retrospective single-centre matched cohort study was approved by the Ethics Committee UZ/KU Leuven (approval numbers s68763 and s64406). As a retrospective observational study, prospective trial registration was not required.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s40959-026-00536-5.

Background

Breast cancer is the most diagnosed malignancy in women and a leading cause of cancer-related mortality worldwide. The hormone receptor-positive (HR +), human epidermal growth factor receptor 2-negative (HER2-) subtype constitutes the majority of cases. While endocrine therapy remains the cornerstone of treatment, acquired resistance presents a significant clinical challenge. The discovery of cyclin-dependent kinase (CDK) 4/6 inhibitors has revolutionized the treatment paradigm, improving progression-free survival in advanced HR +, HER2- breast cancer. Three CDK4/6 inhibitors—palbociclib, ribociclib, and abemaciclib—are now approved for use in combination with endocrine therapy, demonstrating substantial survival benefits [1–4].

The mechanism of action of CDK4/6 inhibitors is based on blocking the phosphorylation of the retinoblastoma tumor suppressor protein, thereby preventing cancer cell proliferation [3, 5]. While these agents are generally well tolerated, growing evidence suggests they may be associated with significant cardiovascular (CV) adverse events [2, 3, 6–9]. QTc prolongation is a well-documented risk with ribociclib, leading to recommendations for electrocardiographic monitoring [2, 3, 6, 10, 11]. Additionally, all three inhibitors have been linked to an increased incidence of venous thromboembolism (VTE), with the strongest signal generally observed for abemaciclib [1, 4, 12–17]. Emerging data also suggest potential risks for arterial thrombotic events [1, 4, 12, 15, 16, 18, 19], heart failure [3, 6, 9, 18, 20] and hypertension [2, 6, 9, 21].

As CDK4/6 inhibitors continue to be integrated into earlier-stage breast cancer treatment and explored for other malignancies, a thorough understanding of their long-term CV safety is essential [5, 22]. The aim of this retrospective matched cohort study was therefore to describe CV events in women treated with CDK4/6 inhibitors for breast cancer and to compare their incidence with that observed in a historical cohort of community-dwelling women without active cancer. We sought to characterize the clinical burden of CV events in this population rather than to establish a direct causal drug-specific effect.

Methods

Study design

This single-centre retrospective matched cohort study was approved by the Ethics Committee UZ/KU Leuven (approval numbers s68763 and s64406). Inclusion criteria for the index cohort were women with early or metastatic breast cancer treated with a CDK4/6 inhibitor for at least 3 months at the University Hospitals in Leuven, a tertiary care cancer referral centre in Belgium, between January 2016 and January 2022. Of 1,037 patients screened for eligibility, 595 met the inclusion criteria and comprised the CDK4/6 inhibitor cohort (Fig. 1). Twenty-six patients received more than one CDK4/6 inhibitor, for whom we considered only the first line of CDK4/6 inhibitor treatment. Data were collected from the electronic patient record.

Fig. 1.

Study flow diagram. Selection of the index and reference cohorts and the 1:1 matching procedure.Abbreviations: BMI, body mass index; CDK4/6 inhibitor, cyclin-dependent kinase 4/6 inhibitor. FLEMENGHO, Flemish Study on Environment, Genes and Health Outcomes

Next, we derived a reference cohort from the Flemish Study on Environment, Genes and Health Outcomes (FLEMENGHO) [23], a prospective population study with random family-based recruitment. From the FLEMENGHO database of 2345 technical examinations conducted between 2005 and 2017, we excluded 1157 examinations on men, and 14 examinations on women with a cancer diagnosis within one year after the examination. Of the remaining 1174 examinations, none of the participants were treated with a CDK4/6 inhibitor in the past or at the time of the technical examination. This is consistent with the 2005–2017 timeframe of the selected FLEMENGHO examinations, during which CDK4/6 inhibitors were not yet, or were only rarely, used in routine breast cancer care.

We applied iterative nearest-neighbour matching to pair each participant in the index cohort with one participant in the reference cohort, allowing a maximum age difference of 4 years and requiring the same BMI category. Starting with the eldest participant in the index cohort (because hardest to match), an eligible reference participant was selected without replacement. This procedure yielded 536 matched pairs, while 59 participants from the index cohort remained unmatched and were included in sensitivity analyses. Matching was deliberately limited to age and BMI because of their clinical relevance to cardiovascular and thromboembolic risk and to preserve an adequate number of matched pairs. Remaining measured baseline differences were addressed through multivariable adjustment.

Baseline clinical information

For the index cohort, excessive alcohol intake was defined as a documented history of alcohol misuse or consumption of more than one alcoholic drink per day before treatment initiation. Hypertension was defined as a documented diagnosis or use of antihypertensive medication before treatment. Hypercholesterolaemia was defined as a documented diagnosis, use of lipid-lowering medication, or a low-density lipoprotein cholesterol level > 115 mg/dL in the last sample obtained before treatment.

The same definitions were applied to the reference cohort using medical and biochemical data collected during an in-person FLEMENGHO study visit [23].

Collection of cardiovascular events

For the index cohort, the index date was initiation of the first CDK4/6 inhibitor. Events were collected from the electronic health record and were classified as occurring during active treatment or after discontinuation. For the reference cohort, the index date was the FLEMENGHO study examination; diseases and procedures were recorded through standardised questionnaires during in-person visits and telephone contacts and were confirmed or supplemented by general-practitioner and hospital records [23]. Only events after the index date were counted as incident outcomes; pre-index events contributed to medical history. To improve comparability, reference-cohort follow-up was truncated at 107 months, the maximum follow-up in the index cohort.

CV and non-CV deaths were coded and time-stamped. The cardiac event composite included coronary events (defined as acute coronary syndrome or chronic coronary syndrome requiring new coronary revascularisation), heart failure or cardiomyopathy (defined as heart failure symptoms with systolic and/or diastolic dysfunction on echocardiography), cardiac tamponade, valvular heart disease (defined as at least moderate valvular dysfunction), pacemaker or implantable cardioverter-defibrillator implantation, and new-onset atrial fibrillation (AF) or flutter (documented by ECG when available; no protocolised rhythm-screening schedule was used). AF/flutter was therefore included in the cardiac composite and was also reported separately. QTc prolongation, defined as a QTc interval > 470 ms, was assessed only in the index cohort and was therefore not included in either the cardiac or the overall CV event composite.

The overall CV composite comprised cardiac and other CV events. Other CV events included stroke, transient ischaemic attack, peripheral artery disease, and VTE. VTE was defined as deep-vein thrombosis, including catheter-related thrombosis, and/or pulmonary embolism, including asymptomatic pulmonary embolism. Events were extracted from source records without blinded independent adjudication.

Statistical methods

Python version 3.12 (Python Software Foundation, Wilmington, DE, USA) was used for data management and statistical analysis. All tests were two-sided, and P < 0.05 was considered statistically significant. Baseline characteristics were summarised for the matched reference cohort and the matched and unmatched index cohorts. Means and proportions were compared using two-sample t tests and chi-squared tests, respectively.

Survival analyses. We listed all fatal and nonfatal events observed during follow-up across the matched cohort groups and made between-group comparisons using a Fisher’s Exact test with correction for multiple testing using the Benjamini–Hochberg FDR method [24]. We reported CV and cardiac event rates per group, considering only the first event occurring during follow-up. Kaplan–Meier plots were constructed for the incidence of CV events per group, applying the log-rank test for statistical comparison. Using Cox proportional hazards modelling, we calculated multivariable-adjusted hazard ratios (HR) reflecting the risk for a CV or cardiac event associated with membership in the matched index cohort. HRs were adjusted for age, BMI class (as ordinal variable), excessive alcohol intake, smoking, diabetes mellitus, hypertension, statin use, history of CV disease, and eGFR class (as ordinal variable). In sensitivity analyses, we accounted for the matched-pair structure by applying Cox proportional hazards models with robust variance estimation clustered by matched pair. We repeated the Cox modelling in the full index cohort (with unmatched and matched patients) to identify predictors of adverse events in patients undergoing CDK4/6 inhibitor therapy. In order to account for competing risks, we estimated cumulative incidence functions for CV events and non-CV death using the Aalen-Johansen estimator. Differences in the cumulative incidence functions between matched index and reference cohort were evaluated using the Gray’s test, which statistically compares the subdistribution hazards while accounting for the competing event of death not related to a CV disease [25].

Results

Baseline characteristics

Table 1 shows the baseline clinical characteristics. Of all 595 index patients, 528 (88.7%) received a CDK4/6 inhibitor for a metastatic setting, 33 (5.5%) as neo-adjuvant and 34 (5.7%) as adjuvant therapy. Treatment was predominantly palbociclib (469/595 [78.8%]), with smaller proportions receiving abemaciclib (72/595 [12.1%]) and ribociclib (54/595 [9.1%]); 326 participants (54.8%) had received anthracyclines. Median treatment duration was 18 months (10th–90th percentile, 5–53) in the matched index cohort and 16 months (10th–90th percentile, 5–39.4) in the unmatched index cohort. Median follow-up for the cardiac-event analysis was 107 months in the matched reference cohort, 40 months in the matched index cohort, and 30 months in the unmatched index cohort. For the overall CV-event analysis, the corresponding median follow-up durations were 107, 37, and 28 months, respectively. The matched index and reference cohorts were similar in age, BMI class, hypertension, beta-blocker use, and statin use, but important imbalances remained: the matched index cohort had lower prevalences of current or former smoking and hypercholesterolaemia, but higher prevalences of excessive alcohol intake, diabetes, prior CV disease, and renal impairment (all P < 0.05).

Table 1.

Baseline characteristics of the index and reference cohorts

Variable	Matched reference cohort (n = 536)	Matched index cohort (n = 536)	Unmatched index cohort (n = 59)
Age, years	59.4 ± 12.3	60.7 ± 12.6	73.4 ± 8.9*†
BMI class			*†
Normal, n (%)	265 (49.4%)	265 (49.4%)	58 (98.3%)
Overweight, n (%)	168 (31.3%)	168 (31.3%)	1 (1.7%)
Obesity, n (%)	103 (19.2%)	103 (19.2%)	0 (0.0%)
Smoking		*	*
Never, n (%)	248 (46.3%)	401 (74.8%)	48 (81.4%)
Current, n (%)	91 (17.0%)	50 (9.3%)	6 (10.2%)
Past, n (%)	197 (36.8%)	85 (15.9%)	5 (8.5%)
Excessive alcohol intake, n (%)	5 (0.9%)	73 (13.6%)*	8 (13.6%)*
Medical history and therapy
Hypertension, n (%)	186 (34.7%)	167 (31.2%)	28 (47.5%)†
Beta-blocker use, n (%)	99 (18.5%)	118 (22.0%)	17 (28.8%)
Hypercholesterolaemia, n (%)‡	352 (65.7%)	205 (38.2%)*	23 (39.0%)*
Statin use, n (%)	114 (21.3%)	102 (19.0%)	10 (16.9%)
Diabetes mellitus, n (%)	35 (6.5%)	75 (14.0%)*	5 (8.5%)
History of cardiac disease, n (%)	33 (6.2%)	63 (11.8%)*	15 (25.4%)*†
History of CVD, n (%)	59 (11.0%)	100 (18.7%)*	19 (32.2%)*†
History of cancer, n (%)	43 (8.0%)	536 (100.0%)*	59 (100.0%)*
eGFR class, n (%)		*	*
≥ 90 mL/min/1.73 m2, n (%)	300 (56.0%)	241 (45.0%)	15 (25.4%)
60–90 mL/min/1.73 m2, n (%)	204 (38.1%)	212 (39.6%)	34 (57.6%)
45–60 mL/min/1.73 m2, n (%)	21 (3.9%)	51 (9.5%)	7 (11.9%)
30–45 mL/min/1.73 m2, n (%)	9 (1.7%)	31 (5.8%)	3 (5.1%)
< 30 mL/min/1.73 m2, n (%)	2 (0.4%)	1 (0.2%)	0 (0.0%)
Breast cancer details
Left-sided breast cancer, n (%)	-	283 (52.8%)	30 (50.8%)
CDK4/6 inhibitor, n (%)
Abemaciclib	-	70 (13.1%)	2 (3.4%)
Palbociclib	-	417 (77.8%)	52 (88.1%)
Ribociclib	-	49 (9.1%)	5 (8.5%)
Time treated with CDK4/6 inhibitor, months	-	18.0 (5.0–53.0)	16.0 (5.0–39.4)
Antihormonal therapy, n (%)
Aromatase inhibitor	-	322 (60.1%)	37 (62.7%)
Gonadotropin-releasing hormone agonist + aromatase inhibitor	-	20 (3.7%)	0 (0.0%)
None, other or unknown	-	6 (1.1%)	2 (3.4%)
Selective oestrogen receptor degraders	-	188 (35.1%)	20 (33.9%)
Anthracyclines, n (%)	-	306 (57.1%)	20 (33.9%)
Trastuzumab or pertuzumab, n (%)	-	11 (2.1%)	1 (1.7%)
Radiotherapy, n (%)	-	378 (70.5%)	41 (69.5%)
Left-sided radiotherapy, n (%)	-	198 (36.9%)	21 (35.6%)

Values are mean ± standard deviation, median (10th-90th percentile) or count (percentage)

^*P < 0.05 vs. matched reference cohort

^†P < 0.05 vs. matched index cohort

^‡LDL measurements were unavailable for 190 index cohort participants (31.9%)

Incidence of CV events

Table 2 presents cardiac and CV event rates across the matched cohort groups. The CV event rate was 67.7 per 1,000 person-years in matched index cohort and 17.2 per 1,000 person-years in the reference cohort. The cardiac-event rate was 28.0 versus 9.6 per 1,000 person-years, respectively. Cumulative incidence of both outcomes was higher in the index cohort (log-rank P < 0.0001; Fig. 2).

Table 2.

Cardiovascular event rates by cohort

	Matched reference cohort (n = 536)	Matched index cohort (n = 536)	Unmatched index cohort (n = 59)
Cardiovascular events
Number of events	73	117	14
Person-years of follow-up	4251.1	1729.3	166.2
Event rate, per 1000 person-years	17.2	67.7	84.3
Cardiac events
Number of events	42	52	12
Person-years of follow-up	4384.6	1856.8	169.2
Event rate, per 1000 person-years	9.6	28.0	70.9

Fig. 2.

Cumulative incidence of cardiovascular and cardiac events. Kaplan–Meier curves depicting the cumulative incidence of (A) overall CV events and (B) cardiac events in matched index cohort (red line) versus matched reference cohort (blue line). Log-rank P < 0.0001 for both outcomes

Cardiac events were defined as coronary events, heart failure or cardiomyopathy, cardiac tamponade, valvular heart disease, pacemaker or ICD implantation and new-onset AF/flutter. Cardiovascular events included cardiac events, stroke, transient ischaemic attack, peripheral artery disease and venous thromboembolism.

Supplemental Table 1 summarises fatal and nonfatal events. VTE was the most common CV event in the matched analysis, occurring in 62 participants (11.6%) in the matched index cohort and four (0.75%) in the matched reference cohort (Supplemental Fig. 1). In the full index cohort, 67 patients (11.3%) experienced a VTE, comprising 23 non-catheter-associated DVTs, seven catheter-associated DVTs, 32 isolated pulmonary embolisms, and five combined DVT and pulmonary embolism events.

In the phase-specific first-event analysis, no patients were excluded. During treatment, 24 first VTEs occurred over 1,147.4 person-years, corresponding to an incidence rate of 20.9 per 1,000 person-years (95% CI, 13.4–31.1). Among patients who remained event-free during treatment, 43 first VTEs occurred over 844.2 person-years of follow-up after treatment discontinuation, corresponding to an incidence rate of 50.9 per 1,000 person-years (95% CI, 36.9–68.6). The post-treatment incidence rate was significantly higher than the during-treatment rate (incidence rate ratio, 2.44; 95% CI, 1.45–4.20; P < 0.001; Supplemental Table 2A). Two patients experienced VTEs in both periods; only their first, during-treatment event was included.

During-treatment VTE incidence rates differed among CDK4/6 inhibitors (global P = 0.023; Supplemental Table 2B). The rates were 19.9 per 1,000 person-years for palbociclib (17 first events over 856.0 person-years), 0 per 1,000 person-years for ribociclib (no events over 136.4 person-years), and 45.2 per 1,000 person-years for abemaciclib (seven first events over 155.0 person-years). These findings should be interpreted cautiously because no pairwise comparisons or adjustments for confounding were performed, and the treatment groups differed considerably in size and treatment duration.

New-onset AF/flutter was also more frequent in the matched index cohort than in the matched reference cohort (25 [4.7%] vs eight [1.5%]; Supplemental Table 1 and Supplemental Fig. 1). In the full index cohort, 27 first events occurred among 557 patients without pre-existing AF/flutter. No events were excluded from the phase-specific analysis. Eleven events occurred during treatment, yielding an incidence rate of 9.9 per 1,000 person-years (95% CI, 5.0–17.8), whereas 16 occurred after treatment discontinuation, yielding an incidence rate of 18.8 per 1,000 person-years (95% CI, 10.8–30.6). This difference was not statistically significant (incidence rate ratio, 1.90; 95% CI, 0.83–4.52; P = 0.120; Supplemental Table 2A).

During-treatment incidence rates of AF/flutter were 8.7 per 1,000 person-years for palbociclib, 7.6 for ribociclib, and 17.7 for abemaciclib, with no significant overall difference among agents (global P = 0.475; Supplemental Table 2B).

QTc data were available for 310 of 595 patients (52.1%), of whom 19 (6.1%; 3.2% of the full cohort) developed QTc prolongation. Sixteen of these patients had another potential contributing cause (other QTc prolonging drug or electrolyte disorder), whereas three had no other identified cause. QTc prolongation was observed in 12 patients treated with palbociclib, five treated with abemaciclib, and two treated with ribociclib. Among patients with available QTc data, the corresponding proportions were 5.3%, 12.8%, and 4.3%, respectively. Given the substantial and differential proportion of missing QTc data, these findings are descriptive and no formal between-agent comparison was performed. QTc prolongation was not included in the cardiac or overall CV event composites because it was not assessed in the reference cohort.

We did not observe significant differences between matched index and reference cohort for coronary events, heart failure/cardiomyopathy, pacemaker or defibrillator implantation, valvular disease, stroke, or peripheral artery disease.

Association between CV events and index-cohort membership

After adjustment for measured baseline CV risk factors, membership in the matched index cohort was associated with CV events (adjusted HR, 3.63; 95% CI, 2.57–5.13) and cardiac events (adjusted HR, 3.25; 95% CI, 1.97–5.35) (Table 3). The higher risk in the matched index cohort was confirmed in sensitivity analyses accounting for matched pairs using clustered robust variance estimation (HR for CV events, 3.88; 95% CI, 2.76–5.46; HR for cardiac events, 3.62; 95% CI, 2.19–5.98). Older age, hypertension, absence of statin therapy, and a history of CV disease were associated with adverse events. Non-CV mortality was substantially higher in the matched index cohort than reference cohort (270/536 [50.4%] vs 36/536 [6.7%]). Nevertheless, cumulative incidence of CV events remained higher in the treated cohort after accounting for non-CV death (Gray's test P < 0.0001; Supplemental Fig. 2).

Table 3.

Multivariable associations with cardiovascular and cardiac events

	CV events		Cardiac events
Covariable	HR (95% CI)	P value	HR (95% CI)	P value
Cohort status (matched index versus reference)	3.63 (2.57–5.13)	< 0.001	3.25 (1.97–5.35)	< 0.001
Age, per 10 years increase	1.48 (1.22–1.63)	< 0.001	1.63 (1.34–2.16)	< 0.001
BMI class (ordinal)	1.12 (0.93–1.35)	0.25	1.26 (0.97–1.62)	0.08
Excessive alcohol intake, yes vs no	1.18 (0.73–1.92)	0.50	1.28 (0.64–2.55)	0.49
Current smoking, yes vs no	1.37 (0.90–2.08)	0.14	1.52 (0.90–2.59)	0.12
Diabetes mellitus, yes vs no	0.99 (0.64–1.53)	0.97	0.85 (0.47–1.51)	0.58
Hypertension, yes vs no	1.41 (1.00–1.97)	0.05	1.69 (1.08–2.66)	0.02
Statins, yes vs no	0.61 (0.42–0.89)	0.01	0.46 (0.28–0.75)	< 0.001
History of CVD, yes vs no	1.59 (1.12–2.26)	0.01	2.01 (1.29–3.13)	< 0.001
eGFR class (ordinal)	1.05 (0.87–1.27)	0.60	1.20 (0.95–1.51)	0.12

Hazard ratios (HRs) and 95% confidence intervals express the multivariable-adjusted risk of a CV or cardiac event. Participants in the unmatched index cohort (n = 59) were excluded. Cardiac events comprised coronary events, heart failure or cardiomyopathy, cardiac tamponade, valvular heart disease, pacemaker or ICD implantation, and new-onset AF/flutter. Cardiovascular events comprised cardiac events, stroke, transient ischaemic attack, peripheral artery disease, and venous thromboembolism

In sensitivity analyses in the index cohort, only age independently predicted CV events, but a trend was observed between higher risk for CV events and the use of selective oestrogen receptor degraders or previous use of anthracyclines (P = 0.06 for both) (Supplemental Table 3).

Discussion

In this retrospective matched cohort study, women with breast cancer treated with a CDK4/6 inhibitor had higher observed rates of overall CV and cardiac events than age- and BMI-matched women from a population-based reference cohort without active cancer. The excess was driven mainly by VTE and new-onset AF/flutter; coronary events, heart failure or cardiomyopathy, stroke, valvular disease, and device implantation did not differ significantly between cohorts. Although non-CV mortality was substantially higher in the index cohort, the cumulative incidence of CV events remained higher after competing-risk analysis (Gray's test P < 0.0001). These findings demonstrate a clinically relevant CV burden in the treated population, but they do not establish a causal effect of CDK4/6 inhibition because the comparison also captures differences related to active cancer, metastatic burden, previous and subsequent treatments, and intensity of medical surveillance.

Baseline CV vulnerability remained important. Older age, hypertension, prior CV disease, and statin non-use were independently associated with events in the matched analysis, whereas only age remained an independent predictor within the index cohort. The association with statin non-use should not be interpreted as evidence of a protective treatment effect, because indication, adherence, and residual confounding could not be assessed. Overall, the results support a multifactorial model in which conventional risk factors interact with cancer burden and anticancer treatment.

Overall cardiovascular burden in context

An increased CV burden was already recognised in breast cancer populations before CDK4/6 inhibitors entered routine practice. In a matched US cohort of 91,473 older breast cancer survivors, Matthews et al. reported a substantially higher risk of deep-vein thrombosis and pericarditis and smaller increases in arrhythmia, heart failure, valvular disease, and sudden cardiac arrest compared with cancer-free women [26]. Hamood et al. observed 466 CV events among 2,165 women with non-metastatic breast cancer, with a 13-year cumulative incidence of 33.6% [27]. A systematic review and meta-analysis similarly found increased risks of CV death, heart failure, and atrial fibrillation, but no consistent excess of coronary artery disease, myocardial infarction, or stroke [28]. This pattern is broadly consistent with our higher global event rate despite the absence of significant differences in several arterial and myocardial outcomes.

Direct comparisons remain limited by differences in disease stage, endpoint definitions, and follow-up. Most women in our cohort had metastatic disease, and VTE was included in the overall CV composite. Moreover, 57.1% of the matched index cohort had previously received anthracyclines and 70.5% radiotherapy. In a population-based cohort of 70,230 breast cancer survivors, left-sided radiotherapy after mastectomy was associated with increased CV risk, while anthracycline-based chemotherapy increased heart-failure risk [29]. The observed burden is therefore biologically and epidemiologically plausible even without a direct drug-specific effect.

Venous thromboembolism

VTE was the main contributor to the difference between cohorts, occurring in 11.6% of the matched index cohort and 0.75% of the matched reference cohort. In the full index cohort, 11.3% experienced a first incident VTE. The incidence rate was higher after treatment discontinuation than during treatment (50.9 vs 20.9 per 1,000 person-years; incidence rate ratio, 2.44). This pattern argues against a purely on-treatment pharmacological effect and suggests that the increased VTE burden may be driven predominantly by the underlying malignancy and its progression rather than by CDK4/6 inhibitor exposure itself. However, it does not exclude a treatment-related contribution. Treatment discontinuation commonly coincides with disease progression, subsequent therapy, hospitalisation, declining performance status, and continued use of venous access devices, all of which may further increase thrombotic risk.

Real-world studies have generally reported higher VTE rates during CDK4/6 inhibitor treatment than registration trials. In a cohort of 424 patients with metastatic breast cancer, predominantly treated with palbociclib, Gervaso et al. reported 38 VTEs and a 1-year cumulative incidence of 6.3% [17]. West et al. observed VTE in approximately 7% of 266 treated patients and a 1-year incidence of any venous or arterial thrombosis of 10.4%; notably, the Khorana score did not predict thrombotic events [4]. In an abemaciclib-specific cohort of 364 patients, Watson et al. reported VTE in 4.7%, corresponding to 91 events per 1000 person-years, with a median time to VTE of 2.8 months [30]. Pharmacovigilance analyses have likewise identified increased VTE reporting, particularly a deep-vein thrombosis signal for abemaciclib, although such databases cannot provide incidence estimates or establish causality [1].

Our overall VTE proportion of 11.3% lies at the upper end of these real-world estimates. However, our observation period also included events after treatment discontinuation, whereas previous studies generally restricted events to active treatment or a short period thereafter. During-treatment incidence rates were 45.2 per 1,000 person-years with abemaciclib, 19.9 with palbociclib, and 0 with ribociclib (global P = 0.023). Although compatible with the recognised VTE signal for abemaciclib, these comparisons remain exploratory because treatment groups differed considerably in size and exposure duration, no pairwise or confounder-adjusted comparisons were performed, and only 54 patients received ribociclib. Consequently, these findings cannot establish comparative thrombotic safety or support routine thromboprophylaxis based solely on CDK4/6 inhibitor exposure.

Atrial fibrillation and flutter

Evidence regarding atrial arrhythmias during CDK4/6 inhibitor treatment remains limited. In our study, new-onset AF/flutter occurred more frequently in the matched index cohort than in the matched reference cohort (4.7% vs 1.5%). Incidence rates were numerically higher after treatment discontinuation than during treatment (18.8 vs 9.9 per 1,000 person-years), although this difference was not statistically significant. Similar to VTE, this pattern suggests that the increased arrhythmic burden may be more closely related to the underlying malignancy and its clinical course than to a direct pharmacological effect of CDK4/6 inhibition, although a treatment-related contribution cannot be excluded. Malignancy-related inflammation, autonomic imbalance, older age, pre-existing cardiac substrate, previous cardiotoxic therapies, and clinical deterioration may all contribute.

An increased atrial arrhythmia burden was already observed in breast cancer populations studied largely before routine CDK4/6 inhibitor use. Abdel-Qadir et al. reported a slightly higher 10-year cumulative incidence of atrial fibrillation among 68,113 women with early-stage breast cancer than among the matched cancer-free reference cohort (7.4% vs 6.8%), with the greatest relative risk during the first year after diagnosis [31]. Among patients specifically treated with CDK4/6 inhibitors, Fradley et al. reported AF/flutter in 68 of 1,376 patients (4.9%), corresponding to 65 events per 1,000 person-years [6]. This rate was substantially higher than in our cohort, which may reflect differences in patient characteristics, comorbidity and cancer burden, surveillance intensity, and event identification using diagnosis codes rather than detailed medical-record review. A more recent real-world cohort of 2,782 CDK4/6 inhibitor-treated patients reported cumulative incidences of new-onset atrial arrhythmia of 1.1% at 1 year and 2.8% at 5 years, without significant differences among the individual agents [32]. Together, these findings suggest that atrial arrhythmias may represent an under-recognised complication in patients with breast cancer receiving contemporary systemic therapy. As episodes may be paroxysmal or asymptomatic, greater awareness of palpitations, dyspnoea, dizziness, an irregular pulse, or otherwise unexplained functional decline may facilitate earlier recognition in routine care.

Other cardiovascular events

Unlike VTE and AF/flutter, coronary events, heart failure or cardiomyopathy, stroke, and peripheral artery disease did not differ significantly between the matched cohorts. However, the small number of individual events limits statistical power and does not exclude clinically relevant associations.

Real-world evidence concerning heart failure during CDK4/6 inhibitor therapy remains heterogeneous. Fradley et al. reported heart failure or cardiomyopathy in 80 of 1,376 CDK4/6 inhibitor-treated patients (5.8%), corresponding to 78 events per 1,000 person-years [6]. This rate was substantially higher than observed in our cohort, possibly because Fradley et al. used diagnosis codes from a large healthcare database and applied broader outcome definitions. Pharmacovigilance studies have also frequently reported cardiac failure, although these databases cannot estimate incidence or establish causality [9]. Heart failure is additionally a recognised complication of breast cancer treatment independent of CDK4/6 inhibition. A meta-analysis reported a pooled incidence of 4.44 per 1,000 person-years and a 21% higher risk among breast cancer survivors than among the matched cancer-free reference cohort [28]. Previous anthracycline exposure and thoracic radiotherapy may have contributed to this background risk in our cohort [29].

Although we found no significant differences in coronary events, stroke, or peripheral artery disease, arterial thrombosis has been described in real-world CDK4/6 inhibitor cohorts. In the study by West et al., arterial events accounted for approximately one-third of 29 thrombotic events among 266 treated patients [4]. Among 364 patients receiving abemaciclib, Watson et al. documented arterial thrombosis in nine patients (2.5%), including four ischaemic strokes [30]. Pharmacovigilance analyses have also identified arterial thrombotic signals, particularly with ribociclib, but are susceptible to reporting bias and cannot determine absolute risk [1]. Conversely, studies of breast cancer populations largely predating routine CDK4/6 inhibitor use found no consistent increase in myocardial infarction, coronary artery disease, or stroke compared with cancer-free populations [26, 28]. The available evidence therefore remains inconclusive, and our findings cannot exclude uncommon agent-specific arterial events.

QTc prolongation was documented in 19 patients (3.2%), but QTc data were missing in 47.9%, and 16 of the 19 affected patients had another potential contributor. A dedicated PALOMA-2 substudy found no clinically relevant QTc effect of palbociclib [33], whereas a meta-analysis of randomised trials found an increased risk with palbociclib and ribociclib, with a more pronounced effect and grade 3 events confined to ribociclib [34]. Given the predominance of palbociclib and limited ribociclib exposure in our cohort, reliable agent-specific estimates could not be obtained.

Despite the higher overall CV event burden, CV mortality was low: only one of 271 deaths in the matched index cohort was cardiovascular, whereas 256 were attributed to cancer. This contrasts with older, long-term breast cancer survivor populations, in which CV disease accounted for a substantially greater proportion of deaths [35]. The difference likely reflects the predominantly metastatic disease in our cohort, with cancer progression remaining the principal competing cause of mortality.

Strengths and limitations

This study has several strengths. It included a large, unselected real-world cohort of patients receiving CDK4/6 inhibitors, with detailed retrospective assessment of clinically documented fatal and nonfatal cardiovascular events. The population-based reference cohort enabled comparison with women of similar age and BMI, while the phase-specific analyses provided insight into event rates during treatment and after discontinuation. First-event analyses, agent-specific estimates, and competing-risk methods further strengthened the assessment, particularly given the high cancer-related mortality in the index cohort.

Several limitations should nevertheless be considered. Most importantly, the reference cohort consisted of women without active cancer rather than patients with breast cancer of comparable stage and treatment history who did not receive a CDK4/6 inhibitor. Consequently, the observed differences cannot distinguish the potential effects of CDK4/6 inhibition from those of the underlying malignancy, metastatic disease, previous and subsequent anticancer therapies, hospitalisation, immobility, or venous access devices. Matching was restricted to age and BMI because these variables were available in both cohorts, were clinically relevant, and allowed an adequate number of matched pairs to be retained. Nevertheless, important baseline differences remained, including differences in diabetes, previous cardiovascular disease, kidney function, smoking, and hypercholesterolaemia. Although these factors were included in multivariable models, residual and unmeasured confounding remain likely. Because patients treated for less than 3 months were excluded, the index cohort may also be subject to selection bias favouring patients who survived and tolerated CDK4/6 inhibitor therapy beyond the earliest treatment period, limiting inference regarding very early CV events.

Differential surveillance may also have influenced the results. Patients in the index cohort had more frequent clinical assessments, ECGs, imaging examinations, and hospital contacts than the population-based reference cohort, potentially increasing the detection of incidental VTE and asymptomatic or paroxysmal atrial arrhythmias. Conversely, the absence of protocolised rhythm monitoring may have resulted in under-recognition of AF/flutter. Events were identified retrospectively from routine medical records and were not independently adjudicated.

Finally, treatment discontinuation was not random and frequently coincided with disease progression, subsequent therapy, or declining clinical status. Comparisons between the during- and post-treatment periods are therefore susceptible to time-varying confounding and should not be interpreted causally. Agent-specific analyses were limited by unequal group sizes and small numbers of events, particularly among patients receiving ribociclib and abemaciclib. Moreover, QTc data were missing for almost half of the index cohort, precluding a systematic assessment of QTc prolongation or reliable comparisons between individual agents.

Clinical implications

These findings support risk-adapted cardio-oncology care rather than automatic attribution of events to CDK4/6 inhibition. Baseline assessment should integrate age, established CV disease and risk factors, previous anthracycline and thoracic radiotherapy exposure, concomitant treatments, and thrombotic risk. Modifiable risk factors should be managed according to contemporary cardio-oncology guidance [36]. Clinicians should remain alert to symptoms of VTE and atrial arrhythmia during treatment and after discontinuation, while ECG monitoring should reflect the agent-specific risk profile, particularly for ribociclib.

Prospective studies should use comparable control groups of patients with breast cancer at a similar disease stage and treatment line who are not receiving CDK4/6 inhibitors. Future studies with standardised outcome assessment and detailed information on disease progression and subsequent treatments are needed to distinguish cancer-related cardiovascular risk from potential treatment-related toxicity.

Conclusions

In this retrospective matched cohort study, women with breast cancer treated with CDK4/6 inhibitors experienced a higher burden of cardiovascular events than in a matched population cohort without active cancer, driven primarily by venous thromboembolism and atrial fibrillation/flutter. These findings are clinically important because they identify a patient population in whom cardiovascular surveillance may be warranted. However, the observed excess risk should be interpreted cautiously, as this study design cannot distinguish the relative contributions of CDK4/6 inhibitor therapy, underlying malignancy, prior cardiotoxic treatments, and baseline cardiovascular risk. Prospective studies with a breast cancer comparator cohort not exposed to CDK4/6 inhibitors and standardised event assessment are needed.

Abbreviations

AF/flutter

Atrial fibrillation or atrial flutter

BMI

Body mass index

CDK4/6 inhibitor

Cyclin-dependent kinase 4/6 inhibitor

CI

Confidence interval

CV

Cardiovascular

eGFR

Estimated glomerular filtration rate

FLEMENGHO

The Flemish Study on Environment, Genes and Health Outcomes

HER2-

Human epidermal growth factor receptor 2-negative

HF

Heart failure

HR

Hazard ratio

HR + 

Hormone receptor-positive

ICD

Implantable cardioverter-defibrillator

QTc

Corrected QT interval

TIA

Transient ischaemic attack

VTE

Venous thromboembolism

Authors’ contributions

LVA and VL conceived and designed the study. VL collected the data. NC performed the statistical analyses. VL and NC drafted the manuscript.VL prepared the graphical abstract and Fig. 1. NC prepared Fig. 2. LVA, HW, PN, TK critically revised the manuscript for important intellectual content. LVA supervised the study. All authors read and approved the final manuscript.

Funding

The authors received no funding for this work.

Data availability

The datasets generated and analysed during the current study are not publicly available due to patient privacy and institutional restrictions but are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This single-centre retrospective matched cohort study was approved by the Ethics Committee UZ/KU Leuven (approval numbers s68763 and s64406). The need for individual patient consent was waived by the ethics committee due to the retrospective nature of the study and use of anonymised data.

Consent for publication

Not applicable. This manuscript does not contain identifiable individual participant data.

Competing interests

The authors declare no competing interests.