CDK 4/6 inhibitors are associated with a high incidence of thrombotic events in women with breast cancer in real-world practice

Abstract

Purpose:

Cyclin dependent kinase (CDK) 4/6 inhibitors are integral treatment for advanced hormone receptor positive breast cancer, however venous thromboembolic events (VTE) occurred in 1–5% of clinical trial patients. Thrombosis rates in the real-world setting remain unclear. We aimed to define the rate of thromboembolic events, risk factors for thrombosis on CDK 4/6 inhibitors and evaluate the Khorana VTE risk score as a predictive tool for VTE in patients on CDK 4/6 therapy.

Methods:

Multicenter retrospective analysis of adult breast cancer patients prescribed palbociclib, ribociclib or abemaciclib. The primary endpoint was thrombosis during treatment or within 30 days of CDK discontinuation. Cox regression models compared variables between those with and without thrombosis.

Results:

266 patients were included (89% palbociclib, 14% abemaciclib, 7% ribociclib). 29 (10.9%) thrombotic events occurred in 26 women. Of these, 66% were venous, 34% were arterial, and 10% experienced >1 event. The incidence of thrombosis with ribociclib was 8.3%, palbociclib 10.9%, and abemaciclib 4.8%. Hemoglobin less than 10 g/dL was a statistically significant predictor of thrombosis (HR 3.53, p 0.01). Khorana scores ranged 0–3, with the majority between 0–2 and was not predictive of VTE. Overall survival was reduced in those with thrombosis: extended Kaplan Meier median OS of 7.3 months (95% CI: 1.0–43.7) versus 35.7 months (95% CI: 29.4–45.0) if no thrombosis.

Discussion:

VTE in our analysis is higher than reported in clinical trials and arterial thrombosis comprised over one-third of events. The highest incidence was with palbociclib, followed by ribociclib. Khorana score did not predict VTE risk. Larger, real-world studies are needed. The role for prophylactic anticoagulation is yet to be defined in this patient population.

Introduction

Breast cancer is the most common cancer diagnosed in women, and the second leading cause of cancer mortality (1). Up to 10% of breast cancer is metastatic at diagnosis and up to 30% of invasive breast cancer will become metastatic. Most cases of breast cancer are hormone receptor (HR+) positive and human epidermal growth factor receptor (HER 2-) negative (1). Cyclin dependent kinase (CDK) 4/6 inhibitors in combination with endocrine therapy are key front-line therapeutics for the treatment of metastatic, HR+/HER2- breast cancer (2–4). Amplification of cyclin D1 and CDK 4 is common in HR+/HER2- breast cancer (3; 4) and inhibiting this complex allows the tumor suppressor Rb to remain hypophosphorylated, halting abnormal tumor proliferation (5; 6).

Currently there are three CDK 4/6 inhibitors approved by the U.S. Food and Drug Administration (FDA) for the treatment of metastatic HR+/HER2- setting: palbociclib, ribociclib and abemaciclib. When combined with aromatase inhibitors (AI), these agents have demonstrated a doubled progression free survival (PFS) ranging 24–28 months against the control arm of AI alone (7–11).

A unique toxicity of CDK4/6 inhibitors is an increased risk of venous thromboembolic events (VTE). In the PALOMA trials leading to palbociclib approval, VTE rates were initially 5% vs 0% in the control arm. (12). Subsequent larger trials reported a lower VTE risk: PALOMA-2 trial did not demonstrate an increased VTE risk(7), although had excluded patients with a prior history of thrombosis unlike PALOMA-1. The PALOMA-3 trial reported a 1.4% VTE incidence versus 0% in the control arm (9).

Trials evaluating ribociclib found a similarly increased risk of VTE. The final overall survival analysis of MONALESSA-7 reported a rate of pulmonary embolism (PE) in 2.7% in the ribociclib arm versus 0.9% in the placebo arm (10; 11; 13). In the MONARCH trial, VTE were reported in 16 patients (4.9%) in the abemaciclib arm, with two deaths attributed to thromboembolism while on abemaciclib or within 30 days of discontinuation. The rate of VTE in the placebo arm was 0.6% (14; 15). This led to the FDA black box warning for VTE with this medication (16).

The real-world risk of VTE due to CDK 4/6 inhibitors, and the mechanisms leading to thrombosis, have yet to be fully determined, but it appears to be higher than reported in clinical trials. A single institution retrospective review of 424 patients with metastatic HR+/HER2- breast cancer on CDK 4/6 inhibitors found a 9% VTE rate at 18.5 month follow up (17). This study was limited in terms of providing an overview of the VTE risk with all CDK4/6 inhibitors as the population was predominantly (92%) receiving palbociclib. VTE was associated with worse PFS and OS in their study population.

Importantly, there has been little reported in the large clinical trials of CDK4/6 inhibitors or the retrospective cohort mentioned above on arterial thrombotic risk. Recent reviews focusing on cardiac toxicity related to CDK4/6 inhibitors have found higher rates of DVT, PE, atrial fibrillation, and myocardial infarction. Interestingly, the increased rates were most frequent with ribociclib (18; 19). There is also little reported on what unique factors such as smoking, or body mass index (BMI) that may further predispose breast cancer patients to VTE or arterial thrombosis when they receive CDK4/6 therapy.

The Khorana score is a VTE risk assessment tool used to stratify ambulatory solid tumor cancer patients into risk groups based on pretreatment lab values and clinical parameters. Those with high risk scores of 2 or greater are recommended for consideration of prophylactic anticoagulation (20; 21). This model has been validated in several studies however has not been studied specifically to predict thrombosis risk in CDK4/6 therapy (20–23). In this multicenter cohort study, we sought to further describe the real-world incidence of VTE and arterial thrombosis with all CDK 4/6 inhibitors. We further sought to identify risk factors associated with thrombosis, evaluate the Khorana score as a predictive tool, and describe clinical outcomes associated with thrombosis.

Methods

This study was approved by the Institutional Review Board of the Oregon Health and Science University (OHSU) prior to initiation (Study #00021050). We performed a retrospective analysis of all patients aged 18 years or older who had a diagnosis of breast cancer and were prescribed a CDK4/6 inhibitor including palbociclib, ribociclib, and abemaciclib. Utilizing review of a joint electronic medical record (EMR), data from six participating outpatient oncology practices was collected (the Knight Cancer Institute at Oregon Health & Science University and five affiliated outpatient community oncology practices in Beaverton, Oregon; East Portland, Oregon; Gresham, Oregon; Northwest Portland, Oregon; and Tualatin, Oregon), between February 2015 and March 2020. All data was obtained from the EMR.

We identified 331 total patients prescribed CDK4/6 inhibitor during the above time frame. Included patients received a CDK4/6 inhibitor for a minimum of 1 month. Thrombosis was considered secondary to CDK4/6 inhibitors if it occurred during treatment or within 30 days of discontinuation. We included symptomatic and asymptomatic deep venous thrombosis, central line associated thrombosis, pulmonary embolism, mesenteric thrombosis, stroke, transient ischemic attack, myocardial infarction, and peripheral arterial ischemia. We did not include superficial venous thrombosis. We excluded patients without a breast cancer diagnosis, those on CDK4/6 inhibitor for less than one month, and those who were lost to follow up. Our final included patient number after screening out patients who did not meet inclusion criteria or who met and exclusion criteria was 266 patients.

We collected patient data on age, ethnicity, types of thromboembolism, anticoagulant prescribed, whether the thromboembolism was symptomatic and required hospitalization, prior history of thromboembolism, BMI and Eastern Cooperative Oncology Group (ECOG) score. We collected data on patient platelet count, white blood cell count, and hemoglobin prior to initiation of CDK4/6 inhibitor. For patients with venous thrombosis only, we calculated a Khorana Risk Score for Venous Thromboembolism. This was based on pre-treatment values as follows: 0 point was given for breast cancer malignancy, 1 point for platelet count ≥ 350×10⁹/L, 1 point for hemoglobin level < 10g/dL or using red blood cell growth factors, 1 point for pre-chemotherapy leukocyte count ≥11×10⁹/L, and 1 point for BMI≥ 35kg/m² (20).

Statistical Analysis

Patient, disease, and treatment characteristics are summarized in Table 1, stratified by whether or not a patient had a venous or arterial thrombotic event while taking a CDK4/6 inhibitor or up to 30 days after discontinuation. The primary outcome measure of time-to-thrombosis (starting from CDK4/6 inhibitor initiation) was modeled using Cox proportional hazards regression, with outcome censoring applied 30 days after the CDK4/6 inhibitor end date or at the data cutoff date (3/30/2020), whichever occurred first. Only those characteristics that were documented at the time of CDK4/6 inhibitor initiation were considered as Cox model predictors when assessing thrombosis risk. The p-values from univariable Cox models with these baseline characteristics entered as the lone regressor are provided in Table 1. When modeling time-to-thrombosis, categorical variables were constructed from continuous measures using the thresholds from Khorana score components (21). As there were no observed deaths while a patient was still taking a CDK4/6 inhibitor, there was no need to consider death as a competing risk; consequently, cumulative incidence rates of thrombosis were estimated as 1 minus the Kaplan-Meier estimate.

Table 1.

Patient characteristics and Khorana Risk Score for Venous Thromboembolism

Characteristic	No Thrombosis (n=240)	Thrombosis (n=26)	Total (n=266)	P-Value*
Gender, female	238 (99%)	26 (100%)	264 (99%)
Age, years
<46	32 (13%)	2 (8%)	34 (13%)	<ref.>
46–55	43 (18%)	3 (12%)	46 (17%)	0.938
56–65	73 (30%)	9 (35%)	82 (31%)	0.414
>65	92 (38%)	12 (46%)	104 (39%)	0.352
Ethnicity
White	220 (92%)	21 (81%)	246 (91%)
Asian	7 (3%)	0 (0%)	7 (3%)
Black	4 (2%)	2 (8%)	6 (2%)
Hispanic	3 (1%)	0 (0%)	3 (1%)
Other	5 (2%)	3 (12%)	8 (3%)
BMI, kg/m2	28 (23 – 33)	30 (25 – 35)	28 (24 – 33)	0.301
BMI ≥35 kg/m2	36 (15%)	6 (23%)	42 (16%)	0.186
History of thrombotic event	38 (16%)	7 (27%)	45 (17%)	0.060
CBC measures
WBC, K/μL	5.8 (4.7 – 7.4)	5.9 (4.6 – 7.4)	5.8 (4.7 – 7.4)	0.904
WBC >11 K/μL	10 (4%)	2 (8%)	12 (5%)	0.306
Platelets, K/μL	235 (188 – 284)	251 (196 – 284)	236 (188 – 284)	0.811
Platelets ≥350 K/μL	26 (11%)	2 (8%)	28 (11%)	0.422
Hemoglobin	12.6 (11.6 – 13.8)	12.5 (11.0 – 13.5)	12.6 (11.5 – 13.8)	0.008
Hgb <10 g/dL	13 (5%)	4 (15%)	17 (6%)	0.014
Tumor features
HER2 positive	6 (3%)	1 (4%)	7 (3%)
Patient exposure during CDK4/6 therapy
Aspirin	62 (26%)	7 (27%)	69 (26%)
Smoker	22 (9%)	4 (15%)	26 (10%)
Medication-related
CDK4/6 type†‡
Abemaciclib	34 (14%)	3 (12%)	37 (14%)
Palbociclib	213 (89%)	23 (88%)	236 (89%)
Ribociclib	15 (6%)	3 (12%)	18 (7%)
Duration of CDK4/6 therapy, months	9.4 (3.9 – 19.5)	12.8 (7.1 – 32.5)	9.9 (4.0 – 19.5)
Endocrine therapy†
Anastrozole	28 (12%)	5 (19%)	33 (13%)
Letrozole	123 (51%)	17 (65%)	140 (53%)
Exemestane	12 (5%)	1 (4%)	13 (5%)
Fulvestrant	95 (40%)	7 (27%)	102 (38%)
Tamoxifen	5 (2%)	0 (0%)	5 (2%)
Khorana Risk Score	No VTE (n=247)	VTE (n=19)	Total (n=266)	P-Value
0	169 (68%)	12 (63%)	181 (68%)
1	66 (27%)	6 (32%)	72 (27%)
2	7 (3%)	1 (5%)	8 (3%)
3	5 (2%)	0 (0%)	5 (2%)
4	0	0	0
5	NA	NA	NA
6	NA	NA	NA

Frequency count (percentage) is displayed for categorical variables whereas median (interquartile range) is displayed for continuous variables.

^*Since the thromboembolic event time was considered in our modeling, p-values are not from 2-group Fisher’s exact or Wilcoxon tests. Rather, p-values are from univariable Cox regression models of time-to-thrombosis for those patient characteristics that (i) were determined before a patient initiated CDK4/6 inhibitor therapy and (ii) have enough patients in each category for stable regression estimates.

<ref.> denotes the reference level category in the Cox regression model. The reference level was compared to each of the other categories in a pairwise fashion.

^†Variable category percentages can add up >100% because a patient can belong to >1 category.

^‡Since 21 patients received more than 1 CDK4/6 inhibitor during the study period, patients can be included in >1 group. The frequency counts and relative percentages of our population in terms of each patient’s first CDK4/6 inhibitor are: Palbociclib = 231 (87%); Ribociclib = 14 (5%), Abemaciclib = 21 (8%).

NA = not applicable as breast malignancy carries a Khorana score of 0, making the maximum score possible 4.

Time-to-VTE was modeled with Cox regression to evaluate the correlation between computed Khorana scores (using variables whose values were known at the time of CDK4/6 inhibitor start) and the risk of venous thromboembolism during CDK4/6 inhibitor, since arterial events were not included in the development of this score. Per the original Khorana thrombosis risk score paper (20), low-risk, intermediate-risk, and high-risk groups were defined by patient scores of 0, 1–2, and ≥3, respectively. To increase group size, a binary version of Khorana thrombosis risk was also considered, with low- and high-risk categories represented by scores of 0 and ≥1, respectively.

Each patient’s survival time was measured from the CDK4/6 inhibitor start date until the death, last known alive, or data cutoff date, whichever occurred first. Study follow-up time was estimated using these raw values as well with the “reverse” Kaplan-Meier approach. Since thrombosis was considered a time-varying predictor, the effect of a venous or arterial clot on survival was displayed using the extended Kaplan-Meier estimator and quantified with a hazard ratio from a Cox model that allows a patient’s thrombosis status to change at each observed death time in the population (24). All other patient characteristics considered as predictors of overall survival were known and fixed at the CDK4/6 inhibitor start date. A multivariable model of overall survival was built by first considering all patient features with univariable Cox regression p-values < 0.20 and then applying an automated backward elimination procedure.

Statistical analyses were performed, and the extended Kaplan-Meier curve plot was created with R version 4.0.2. Univariable or multivariable model p-values < 0.05 were considered statistically significant.

Results

Two hundred and sixty-six patients were treated with CDK4/6 inhibitors for breast cancer during the study period. Patient demographics of the 26 women who developed a thrombosis are described in Table 1. The majority of patients were postmenopausal, aged over 56 years (70%), white (91%) with an Eastern Cooperative Oncology Group (ECOG) performance status of 0–1 (87%). The majority received palbociclib (89%) in combination with an aromatase inhibitor (71%) or fulvestrant (38%). Several patients received multiple CDK 4/6 inhibitor therapy and several received multiple endocrine therapies while on CDK 4/6 inhibitor. Only 2% total received combination tamoxifen therapy. Prior history of thrombosis was present in 17%. One quarter of patients were on concurrent aspirin (26%). Median time on CDK 4/6 inhibitor therapy was 9.9 months.

Twenty-nine thrombotic events were identified in 26 total patients, comprising both venous and arterial thromboembolism while on CDK4/6 inhibitor therapy or within the 30 days after discontinuation. Three of the 29 events were second thrombotic event during the study period. Of the thrombotic events, 66% were venous, 34% were arterial, and 10% experienced >1 event. Palbociclib had the highest incidence of thrombosis (10.9%), followed by ribociclib (8.3%) then abemaciclib (4.8%). Most thromboses were symptomatic (76%) and resulted in disruption of cancer treatment in 34% of patients. The specific clinical features of the 29 thrombotic events are listed in Table 2.

Table 2.

Incidence of thrombosis

Thrombotic Event information	Number (%) or Percentage (95% CI)
Number of patients analyzed	266
Total number of thrombotic events during CDK4/6 inhibitor therapy	29
Total arterial events	10
Total venous events	19
Total arterial/venous events	2
Number of patients with at least 1 thrombotic event during CDK4/6i	26 (9.8%)
Number of patients with 2 events	3 (1.1%)
12-month cumulative incidence rate of thrombosis, by CDK4/6 inhibitor
Abemaciclib (n = 21)	4.8% (0.0% – 14.1%)
Palbociclib (n = 232)	10.9% (5.8% – 16.0%)
Ribociclib (n = 13)	8.3% (0.0% – 24.7%)
Site of event
First thrombotic event	n = 26
CVA	2 (6.9%)
CVA + DVT	1 (3.5%)
CVA + NSTEMI	1 (3.5%)
DVT	10 (34.5%)
NSTEMI	1 (3.5%)
PE	4 (13.8%)
PE + DVT	1 (3.5%)
PE + PVT	1 (3.5%)
Port associated DVT	1 (3.5%)
Retinal vein thrombosis	1 (3.5%)
TIA	3 (10.3%)
Second thrombotic event	n = 3
DVT	1 (3.5%)
PE + NSTEMI	1 (3.5%)
TIA	1 (3.5%)
Thrombotic event characteristic or result
Symptomatic clot	22 (75.9%)
Patient was hospitalized	16 (55.2%)
Interrupted cancer treatment	10 (34.5%)

Abbreviations: CVA, cerebrovascular accident; DVT, deep vein thrombosis; NSTEMI, non-ST-elevation myocardial infarction; PE, pulmonary embolism; PVT, portal vein thrombosis; TIA, transient ischemic attack.

Multivariable regression evaluating risk factors for thrombosis is described in Table 3. Elevated hazard ratios for development of thrombosis while on CDK 4/6 inhibitor included a prior history of thrombosis (HR 2.26, p 0.06), elevated BMI greater or equal to 35 kg/m2 (HR 1.84, p 0.186), WBC count less than 11 K/μL (HR2.10, p 0.306), and hemoglobin less than 10 g/dL (HR 3.53, p 0.014). Of the above variables, only hemoglobin <10 g/dL was a statistically significant predictor of thrombosis with p = 0.014. Nearly all patients had a low risk Khorana risk score of 0–1 (95%), 8 patients had a score of 2 (3%), and 5 patients (2%) had a Khorana score of 3. None had scores of 4 or greater. Khorana scores did not predict which patients experienced thrombosis (Table 1).

Table 3.

Predictors of thrombosis events

Variable / model predictor	Variable type	Categories or summary stats	Sample size (%)	Hazard Ratio (95% CI)†	p-value†
Thrombosis prior to CDK4/6	binary (2 missing)	No^	219 (82%)
		Yes	45 (17%)	2.26 (0.94 –5.42)	0.060
Age, years	4 groups	< 46 ^	34 (13%)
		46 – 55	46 (17%)	0.93 (0.15 –5.62)	0.938
		56 – 65	82 (31%)	1.90 (0.41 –8.79)	0.414
		> 65	104 (39%)	2.04 (0.46 –9.11)	0.352
BMI, kg/m2	binary (5 missing)	< 35^	219 (82%)
		≥ 35	42 (16%)	1.84 (0.73 –4.62)	0.186
WBC, K/μL	continuous (5 missing)	median (range): 5.8 (1.1 – 41.0)	N/A	1.02 (0.99 –1.06)	0.904
WBC, K/μL	2 groups (5 missing)	≤ 11^	249 (94%)
		> 11	12 (5%)	2.10 (0.49 –8.95)	0.306
Platelets, K/μL	continuous (5 missing)	Median (range): 236 (52 – 501)	N/A	0.99 (0.95 – 1.04) increments of 10 K/μL	0.811
Platelets, K/μL	2 groups (5 missing)	< 350^	233 (88%)
		≥ 350	28 (11%)	0.56 (0.13 –2.37)	0.422
Hgb, g/dL	continuous (5 missing)	median (range): 12.6 (2.5 – 16.1)	N/A	0.78 (0.66 –0.93)	0.008
Hgb, g/dL	2 groups (5 missing)	< 10	17 (6%)	3.53 (1.21 −10.31)	0.014
		≥ 10^	244 (92%)

^†Hazard Ratio (HR) and p-value from a univariable Cox regression model.

^{^}= reference group for HR.

The continuous version of BMI was not included in the above table because it violated the proportional hazards assumption. Categorical versions of continuous variables were based on the Khorana scoring system (BMI, WBC, platelets, hemoglobin) for predicting venous thromboembolism; note that arterial thromboses were also counted as clots in the models whose output is displayed in the above table).

Abbreviations: BMI, body mass index; WBC, white blood cell; Hgb, hemoglobin; history of thrombosis, platelet count, smoking status.

Median overall survival (OS) using the extended Kaplan-Meier approach was worse in those who developed thrombotic events: median OS 35.7 months (95% CI: 29.4 – 45.0) in those without thrombosis compared with 7.3 months (95% CI: 1.0 – 43.7) in those with thrombotic events (24). (Figure 1).

Figure 1.

Overall Survival extended Kaplan-Meier curves

Discussion

In this cohort study of 266 women with breast cancer receiving CDK 4/6 inhibitors, we found a high incidence of VTE compared to published clinical trial data (7%), consistent with other published real-world cohorts (17). Unique to this study, we also noted a significant incidence of arterial events, bringing the total thrombosis rate to 10.9%. This occurred despite a 25% aspirin utilization rate in the cohort, which may have mitigated some of the thrombotic risk. Thrombotic risk was highest with palbociclib in our study, followed by ribociclib. The elevated incidence of thrombosis with ribociclib is discordant with published CDK 4/6 inhibitor data which reported the highest thrombotic risks with palbociclib and abemaciclib. We note however, that ribociclib and abemaciclib comprised a minority of our total study population, often given as a subsequent CDK4/6 inhibitor therapy (21%). Similar to a prior cohort study, our analysis suggests that the development of a thrombotic event impacted overall survival and further, disrupted cancer directed therapy in 34% of patients who experienced a thrombotic event (17).

The thrombotic risk associated with CDK 4/6 inhibitor therapy is becoming increasingly documented as post-trial data continues to accumulate. Outcomes in real-world practice can often uncover adverse events not observed in controlled trials where patient selection, limited follow up time, and other confounders can skew the translatability of the analysis. In an evaluation of targeted cancer therapies, 39% of serious adverse events described in updated drug labels were not reported in the initial phase 3 trials leading to the drug approval (25). The real-world incidence of adverse events is likely even higher as these events are often underreported (26).

Recent meta-analyses evaluating CDK 4/6 inhibitors have described increased rates of venous and arterial events, as well as other cardiac toxicities, noting increased risk ratios for PE (3.5), deep vein thrombosis (DVT) (2.5), atrial fibrillation (2.5) and QT prolongation (2.3) (19). Another analysis utilizing the FDA adverse events reporting system found that cardiovascular events encompassed 2.9% of all reported CDK 4/6 adverse events, of which myocardial infarction and atrial fibrillation were the most common. The highest rates of cardiac events in both studies were found with ribociclib, followed by the other CDK 4/6 inhibitors (18). Most arterial thromboses in our population were comprised of TIA and CVA events. It is unknown if arrhythmia contributed to any of the events observed in our study population. Although ribociclib comprised only 7% of our total study population, the increased rates of thrombosis we observed may reflect an emerging trend.

The Khorana score was not predictive of thrombosis in our population. Current clinical practice guidelines recommend ambulatory solid tumor cancer patients with Khorana scores of 2 or greater be considered for thromboprophylaxis prior to initiation of systemic chemotherapy (27). A meta-analysis encompassing over 34,000 ambulatory solid tumor patients found the Khorana score predicted VTE risk at 6 months at the following rates: a score of 0 carried 5% risk, scores of 1–2 had 6.6% risk, and scores of 3 or greater had 11.0% VTE risk (23). Our study cohort had higher VTE thrombosis rates for similar Khorana scores: Khorana scores of 0 with a 6.6% VTE risk (12/181), score of 1–2 with 8.8% risk (7/80), scores of 3 had no thrombosis (0/5), and none with scores of 4. Scores of 5 and 6 were not possible given breast malignancy is assigned 0 points in the Khorana risk score. Interestingly, our rates of VTE were similar to the 6.6% at which prophylactic anticoagulation is currently recommended, however was observed largely in our Khorana score 0 population. High rates of thrombotic events in our cohort despite low risk Khorana scores suggest a prothrombotic class effect of the CDK 4/6 inhibitors. In patients with Khorana scores of 2 or higher, thromboprophylaxis with apixaban 2.5mg twice daily has been shown protective against VTE, with VTE rates of 4.2% compared to the placebo control group VTE rate of 10.2% (28). Whether thromboprophylaxis would be beneficial in women on CDK 4/6 inhibitors has yet to be established.

Reducing the risk of thrombosis while on CDK 4/6 inhibitors may require careful identification and mitigation of preexisting risk factors prior to initiation. Our study population was comprised of 17% who had a prior history of thrombosis, and of these, 16% (7/45) developed thrombosis while on CDK 4/6 inhibitor. On Cox regression analysis, we found a statistically significant increased hazard ratio risk for thrombosis development in those with a hemoglobin value less than 10g/dL. This is consistent with previous literature, noting worse outcomes in those with anemia, and is incorporated as a risk point in the Khorana model as well. A prior history of thrombosis neared statistical significance as a predictor of thrombosis and remains an important consideration as it may convey an increased thrombosis risk.

The Khorana score was not a useful tool to predict thrombosis in our HR+ HER2 negative breast cancer population while on CDK4/6 inhibitors. Increasingly, the role of the tumor mutational profile as an independent predictor of cancer associated thrombosis is being described (29–32). For example, non-small cell lung cancers (NSCLS) driven by aberrant anaplastic lymphoma kinase (ALK) rearrangements have been shown to be at increased risk for arterial and venous thrombosis compared to non ALK rearranged NSCLC ((29; 31). Another recent publication evaluated over 14,000 pooled solid tumor samples, including 14% from breast tumors, identified 7 somatic tumor mutations independently associated with VTE risk. Mutations including CDKN2B, CDK4 and CDKN2A were associated with thrombosis (30). This is of interest, as the HR+ HER2- breast cancers often have overexpression of cyclin D1 and CDK 4, in addition to mutations within the CDK 4/6 pathway including CDKN2A loss (3; 4; 6). Mechanisms of how somatic mutations may alter thrombotic risk are just beginning to be understood and may relate to over-expression of pro-coagulant factors by the tumor, activation of platelets or leukocytes or pro-inflammatory cytokines, and/or secondary effects of tumor cells on the surrounding microenvironment. Dedicated evaluation of thrombosis risk related to mutational profiles in metastatic HR+ HER2 negative tumors may help to offer more personalized thrombotic risk prediction in this breast cancer patient population.

A future mitigation strategy may involve creating a risk stratification model specific to women taking CDK 4/6 inhibitors. Development of guidelines similar to the risk stratification guidelines used with multiple myeloma regimens involving immunomodulatory drugs (thalidomide, lenalidomide, pomalidomide) could potentially be beneficial (27; 33–35). This risk model could incorporate patient specific risk factors, treatment specific risk factors as well as incorporation of the tumor mutational risk. Further analysis is needed to define the best strategy.

Several limitations of our study are worthy of discussion. First, our study population was comprised primarily of post-menopausal women of white ethnicity, prescribed mostly palbociclib, thus limiting the generalizability of the findings. Second, no comparable control population was available for comparison. Endocrine therapy combinations, including tamoxifen in the premenopausal setting, may confer additional thrombotic risk that we were unable to evaluate fully in this study. Palbociclib comprised the majority of CDK 4/6 inhibitor in our population, making comparisons between the agents challenging. A predominantly white patient population also limits insight into any racial differences that may be present.

In summary, this study confirms the increased rate of VTE associated with CDK 4/6 inhibitors and is the first to our knowledge to describe an elevated risk for arterial events. This study further contributes to the post-approval literature demonstrating thrombotic risk is higher in real-world patient populations than reported in clinical trials and suggests that traditional risk assessment models may not be predictive in women utilizing CDK 4/6 inhibitors. Thrombosis was seen with all CDK 4/6 inhibitors in our study and was most frequent with palbociclib and ribociclib. Larger, more diverse post-approval studies including higher numbers of patients taking ribociclib and abemaciclib are needed to evaluate thrombotic risk and outcomes further. Future research evaluating the tumor mutational profile in HR+ HER2 negative tumors and correlation with thrombosis may provide additional insight into overall thrombosis risk. The role of prophylactic anticoagulation remains unknown in women treated with CDK 4/6 inhibitors.

Manuscript Significance Statement:

CDK4/6 inhibitors used in breast cancer treatment have high rates of venous and arterial thrombosis, at an incidence higher than traditional ambulatory thrombosis risk models recommend prophylaxis, and when applied, were unable to predict thrombotic events in this patient population.