Introduction

Tyrosine kinase inhibitors (TKIs) greatly improved prognosis and clinical outcomes of a majority of chronic myeloid leukemia (CML) patients, but at the same time the use of these drugs presents cardiovascular risks.1-4 Indeed, since the introduction of imatinib, the first generation TKI (1GTKI), the annual mortality in CML has decreased from 10%–20% to 1%–2%.5 However, approximately 40%–50% of patients treated with imatinib require treatment with a second or third-generation TKI (2GTKI or 3GTKI) to achieve an optimal response.6 These newer drugs, when compared with imatinib, entail increased cardiovascular risks.1-4,7-9 In our previous study, which included patients with CML treated with imatinib, 22 patients (8.24%) experienced cardiovascular adverse events (CAEs),8 even though it seems to be generally accepted that the impact of imatinib on cardiovascular risks is minimal.7,8 Importantly, overwhelming evidence exists that documents an increase in the risk of cardiovascular events in patients with CML treated with nilotinib,1-4,9,10 with incident rates reaching 10.6% at 5 years and 24.8% at 10 years in the ENEST (Evaluating Nilotinib Efficacy and Safety in Clinical Trials–Newly Diagnosed Patients) study.11 Although bosutinib, another 2GTKI, was initially supposed to produce neutral effects on the cardiovascular system,12 Cortes et al13 concluded that vascular events occurred in 8.9% of the patients treated with this TKI. Finally, in terms of ponatinib, the PACE (Ponatinib Ph+ ALL and CML Evaluation) trial14 showed that a 5-year cumulative incidence rate of arterial occlusive events in patients with CML treated with ponatinib was as high as 31%. This study was recently reassessed by an independent adjudication committee. Adjudicated adverse outcome events (AOEs) were determined based on events that met the specified criteria for each end point. These criteria included several different factors, such as revascularization, shifts in cardiac biomarkers, and diagnostic substantiation based on imaging methods, such as computed tomography or magnetic resonance imaging. The frequency of adjudicated AOEs in patients with chronic-phase CML (CP-CML) was lower when compared with nonadjudicated AOEs (21%; 57/270), but still very high, with 95% (54/57) of them classified as serious AOEs.15 Asciminib, the first allosteric inhibitor of TKI that targets the myristol pocket of the Abelson kinase, was expected to be less toxic than ponatinib.16 The ASCEMBL study (Efficacy of CML-CP Patients Treated with ABL001 versus Bosutinib, Previously Treated with 2 or More TKIs)17 revealed that the frequency of arterial occlusive events was 5.1% (n  =  8) following treatment with asciminib and 1.3% (n = 1) after bosutinib.

The detrimental effects of TKIs on cardiovascular homeostasis could be at least partially attributed to the effects of these drugs on endothelial function. In fact, in vitro studies in human endothelial cells revealed that nilotinib upregulated proatherogenic adhesion proteins (intracellular adhesion molecule 1, E-selectin, vascular cell adhesion molecule 1), and suppressed endothelial cell proliferation, migration, and tube formation, while in in vivo studies nilotinib augmented atherosclerosis in apolipoprotein E –/– mice and blocked reperfusion and angiogenesis in a hind-limb-ischemia model of arterial occlusion.18 In the same experimental settings, imatinib produced no negative effects. In several other studies, ponatinib- and other TKI-related effects, including disturbed cell permeability, migration, wound closure, tube formation, and cell viability on the endothelium, were reported.19 Finally, although ponatinib inhibited platelet activation and aggregate formation under increased shear stress,20 this TKI induced prothrombotic angiopathy in mice21 and stroke in a zebrafish model.22 Taken together, it seems that ponatinib and other TKIs induce arterial thrombosis through endothelial dysfunction.23

Despite emerging evidence from preclinical studies concerning the negative effects of TKIs on endothelial function,1-4,9 clinically relevant data evaluating the endothelial function in CML patients treated with various TKIs in a prospective manner are lacking.

Therefore, the aim of this study was to characterize endothelial function profiles in patients diagnosed with CP-CML and treated with various TKIs in relation to cardiovascular risk factors. Our comprehensive approach included endothelial function assessment in the conduit arteries and in microcirculation based on flow-mediated dilation (FMD), reactive hyperemia peripheral arterial tonometry (RH-PAT), and flow-mediated skin fluorescence (FMSF).

Patients and methods

Patients

This was a prospective cross-sectional study of 48 adult patients (25 women, 23 men; median [interquartile range, IQR] age, 52 [44.75–61.75] years) diagnosed with CP-CML and treated with a 1GTKI (imatinib), 2GTKIs (bosutinib, nilotinib), 3GTKI (ponatinib), and asciminib. Patients treated with dasatinib were excluded from the study as a nonrepresentative group. Medical data were extracted from medical records. Basic laboratory blood tests were performed and Systematic Coronary Risk Estimation 2 (SCORE2) and SCORE2-Older Persons (SCORE2-OP) algorithms were calculated for each patient. Endothelial function was assessed between January and June 2022. The study was approved by the Bioethical Committee of the Jagiellonian University, Kraków, Poland (1072.6120.122.2021), and conducted in accordance with the Declaration of Helsinki. The patients provided their written informed consent to participate in the study.

Noninvasive assessment of endothelial function

On the vascular testing day, the patients were examined between 7:30 AM and 10:30 AM while they were fasting and had abstained from caffeine, vitamins, supplements, and exercise. Tests were performed to obtain a hemodynamic steady state while the patient was in a supine position in a quiet, semidarkened, and temperature-controlled (22–25 °C) room after a 20-minute rest.

Flow-mediated dilation

FMD was examined in the brachial artery (diameter, 3–5 mm) of the right arm, according to a standard protocol,24 using an ultrasound probe of 14 MHz (Siemens Acuson S2000, Warszawa, Poland). To ensure stability and fixed position of the ultrasound probe during recording that started 1 minute before the artery occlusion and ended 4 minutes after the occlusion release, the ultrasound probe was fixed with a stereotactic probe holding device (Quipu Studio, Pisa, Italy). Offline analysis of FMD changes was performed with commercially available software (Cardiovascular Suite, Quipu, Pisa, Italy), and a number of parameters describing FMD responses were used as shown in Supplementary material, Figure S1.

Reactive hyperemia peripheral arterial tonometry

Reactive hyperemia was evaluated using the noninvasive RH-PAT on the index finger according to a standard protocol (EndoPAT 2000; Itamar Medical, Caesarea, Israel). RH-PAT was measured simultaneously with FMD upon release of a blood pressure cuff during the reactive hyperemia phase. The following parameters were measured: heart rate (HR, bpm), reactive hyperemia index (RHI), natural logarithm RHI (LnRHI), augmentation index (AI, %), and AI normalized to HR of 75 bpm (AI75, %).25 The parameters were calculated automatically using Endo-PAT 2000 software (version 3.2.4) provided by the manufacturer (EndoPAT2000).

Flow-mediated skin fluorescence

FMSF was quantified using the AngioExpert device (Angionica, Łódź, Poland)26-28 after concurrent FMD and RH-PAT measurements and subsequent 15-minute acclimatization period according to a previously described protocol.26-28 Offline analysis of FMSF response using Angionica software included a number of parameters as exemplified in Supplementary material, Figure S2.

Systematic Coronary Risk Estimation 2 and Systematic Coronary Risk Estimation 2-Older Persons

The SCORE2 and SCORE2-OP (jointly termed SCORE2 in this study) were assessed according to the published guidelines.29,30 Eight patients with cardiovascular comorbidities, including 4 with type 2 diabetes, were not assessed according to the SCORE2 algorithm, in accordance with the 2021 European Society of Cardiology guidelines.29 Five patients younger than 40 years old were also not assessed with this algorithm. Of 4 European regions stratified on the standardized cardiovascular disease (CVD) mortality risk (low, moderate, high, and very high), Poland belongs to the high-risk region, and all patients included in the study came from Poland. In the analysis of the relationship between SCORE2 and endothelial parameters, we used a subgroup of patients eligible for SCORE2 stratification. We divided this subpopulation into 2 groups of high to very high risk and of moderate to low risk.

Statistical analysis

Quantitative variables were presented as median with IQR. The Wilcoxon rank-sum test was used for comparisons. When comparing more than 2 categories, the Kruskal–Wallis test was used with the pairwise Wilcoxon rank-sum test comparisons, and P value was adjusted using the Holm method. Qualitative variables were expressed as numbers (percentage), and the Fisher test or the χ2 test was used for comparisons, as appropriate. The Spearman correlation coefficient was calculated to assess the relationship between quantitative variables. Two-sided P values below 0.05 were considered significant. Statistical analysis was performed with the R Project for Statistical Computing software version 4.2.1 (R Foundation for Statistical Computing, Free Software Foundation Inc., Vienna, Austria).

Results

Clinical characteristics of the patients

The clinical characteristics of 48 patients included in the study are presented in Table 1. The median (IQR) age of all patients at the study entry was 52 (44.75–61.75) years, and 52.1% of them were women. As many as 17 patients were treated with imatinib, 6 with bosutinib, 12 with nilotinib, 6 with ponatinib, and 7 with asciminib. The participants treated with bosutinib were older (median, 68.5 [66.25–70] years) than those treated with imatinib (median, 52 [42–57] years) (P = 0.04). Seventeen patients (35%) had cardiovascular comorbidities at the study entry. Eight of them had CVD with high cardiovascular risk (4 patients had type 2 diabetes, 3 coronary artery disease, and 1 chronic kidney disease). Three individuals (6.3%) suffered from CAEs during TKI therapy, 2 of them developed hypertension (1 was treated with asciminib and 1 with ponatinib), and 1 person treated with imatinib suffered from deep vein thrombosis. Blood test results in the CML patients treated with TKIs are presented in Supplementary material, Table S1.

Table 1. Clinical characteristics of patients diagnosed with chronic myeloid leukemia and treated with tyrosine kinase inhibitors

Parameter

All TKIs (n = 48)

Asciminib (n = 7)

Ponatinib (n = 6)

Nilotinib (n = 12)

Bosutinib (n = 6)

Imatinib (n = 17)

P value

Women

25 (52.1)

4 (57.1)

3 (50)

6 (50)

4 (66.7)

8 (47.1)

0.97

Age at study entry, y

52 (44.75–61.75)

52 (48.5–65)

54.5 (48.5–59)

48.5 (39.75–55)

68.5 (66.25–70)

52 (42–57)

0.04a

Any comorbidities

34 (70.8)

4 (57.1)

6 (100)

8 (66.7)

4 (66.7)

12 (70.6)

0.53

Cardiovascular comorbidities

17 (35.4)

2 (28.6)

3 (50)

2 (16.7)

4 (66.7)

6 (35.3)

0.1

Dose of TKI at study entry, mg

NA

80 (80–80)

18.75 (15–39.38)

500 (400–800)

500 (350–500)

400 (400–400)

NA

Overall time of TKI treatment, mo

98.52 (43.06–169.3)

47.6 (43.11–66.65)

94.22 (45.1–115.18)

100.57 (61–178.82)

102.62 (47.93–147.85)

110 (53–171)

0.68

SCORE2 / SCORE2-OP

High / very high

22 (45.8)

1 (14.3)

3 (50)

5 (41.7)

4 (66.67)

9 (52.9)

0.45

Low / intermediate

13 (27.1)

3 (42.9)

1 (16.7)

4 (33.3)

0

5 (29.4)

Not assessed with SCORE2 / SCORE2-OP

<⁠40 years oldb

5 (10.4)

1 (14.3)

0

2 (16.7)

0

2 (11.8)

0.45

Disease associated with high riskc

8 (16.7)

2 (28.6)

2 (33.3)

1 (8.3)

2 (33.3)

1 (5.9)

Data are presented as number (percentage) of patients or median (interquartile range).

a Significant difference between the bosutinib and imatinib groups (P = 0.04)

b Patients below 40 years old were not assessed with the SCORE2/SCORE2-OP.

c Disease disqualifying from SCORE2/SCORE2-OP evaluation based on the ESC guidelines.29

Abbreviations: NA, not applicable; SCORE2, Systematic Coronary Risk Estimation 2; SCORE2-OP, Systematic Coronary Risk Estimation 2-Older Persons; TKI, tyrosine kinase inhibitor

Endothelial function in high / very high vs low / intermediate cardiovascular risk patients according to the Systematic Coronary Risk Estimation 2 algorithm

In our study group, 35 patients were eligible for SCORE2 calculation. Five of them had very high cardiovascular risk, 17 had high cardiovascular risk, and in 13 remaining patients this risk was low to moderate.

Values of the main endothelial parameters, that is, FMD, RHI, hyperemic response index (HR index), reactive hyperemia response (RHR), and normoxia oscillatory index (NOI) were not significantly different in the low / moderate and high / very high cardiovascular risk groups as assessed with the SCORE2 algorithm (Table 2; Supplementary material, Figure S3). Only a few exceptions differentiating these 2 groups were found among ancillary parameters in the FMD and FMSF assays. The patients with high / very high cardiovascular risk according to the SCORE2 had lower values of shear rate maximum (SR Max), SR at 30 and 60 seconds of the hyperemic response, neurogenic component at rest, hypoxia sensitivity (HS), logarithmic HS (logHS), myogenic component at reperfusion, and a parameter representing flow motion during the reperfusion phase (Fmindex[R]), in comparison with the low / moderate risk patients (Table 2).

Table 2. Endothelial function parameters in relation to Systematic Coronary Risk Estimation 2 algorithm

Risk category according to SCORE2

Low / intermediate risk

High / very high risk

P value

FMD

Evaluable patients, n

12

18

–

FMD, %

5.7 (3.93–12.44)

5.53 (2.18–8.93)

0.6

FMDr, %

4.54 (4.02–7.27)

3.32 (1.08–5.85)

0.31

Brachial artery diameter, mm

Basic diameter

3.64 (3.34–4.1)

4.04 (3.57–4.86)

0.17

Maximum diameter

4.06 (3.69–4.35)

4.33 (3.78–4.95)

0.29

Recovery diametera

3.63 (3.35–4.12)

4.5 (3.66–4.81)

0.05

At 30 s of HR

3.91 (3.74–4.17)

3.92 (3.74–4.79)

0.64

At 60 s of HR

4.05 (3.94–4.32)

4.26 (3.74–4.97)

0.68

Shear rate

Basic, s-1

232.65 (184.48–335.7)

210.2 (159–248.33)

0.2

Maximum, s-1

1189.6 (1082.53–1540.05)

865.9 (751.42–1127)

0.02

SR area, a.u.

2 328 724 (309 471–4 764 725)

1 254 437 (345 258–3 663 692)

0.79

SR area to Max, a.u.

523 552 (54 086–1 871 737)

624 815 (153 360–1 417 380)

0.88

SR at 30 s of HR, s-1

586.7 (554.2–624.9)

445.8 (338.75–537.12)

0.01

SR at 60 s of HR, s-1

417.1 (359.72–446.75)

284.55 (232.45–378.17)

0.01

SR area to 30 s of HR, a.u.

412 273 (36 601–834 367)

356 049 (43 194–788 535)

0.82

SR area to 60 s of HR, a.u.

1 112 072 (54 072–1 871 737)

628 532 (183 978–1 473 313)

0.95

RH-PAT

Evaluable patients, n

13

21

–

RHI

2.02 (1.93–2.24)

2.29 (1.65–2.45)

0.64

Heart rate, bpm

70 (66–81)

68 (65–75)

0.49

AI, %

4 (1–12)

12 (3–17)

0.32

AI 75, %

4 (0–7)

6 (–2 to 15)

0.38

LnRHI

0.77 (0.66–0.89)

0.83 (0.5–0.9)

0.84

FMSF

Evaluable patients, n

13

21

–

IR index, %

10.1 (5.59–11.49)

9.83 (6–12.98)

0.97

HR index, %

10.53 (8.41–11.8)

7.28 (6.74–10.3)

0.05

IR max, %

12.89 (8.46–14.23)

12.97 (9.02–16.87)

0.85

HR max, %

18.23 (17.62–24.1)

17.75 (15.88–20.95)

0.22

RHR, %

30.04 (24.85–32.41)

25.07 (16.62–33.51)

0.15

NOI, %

72.24 (54.21–86.15)

77.55 (64.67–86.58)

0.55

MR, %

79.02 (73.34–86.17)

72.86 (61.87–80.1)

0.17

HS

38.07 (27.3–57.48)

14.64 (4.19–34)

0.02

LogHS

1.58 (1.44–1.76)

1.17 (0.62–1.53)

0.02

PSD × 106, MSA

93.49 (22.06–115.46)

23.39 (13.77–58.16)

0.12

PSD(R) × 106, MSA

92.89 (71.79–174.54)

54.63 (15.74–75.65)

0.06

Fm index, a.u.

94.66 (25.19–118.16)

26.49 (16.67–59.53)

0.12

Fm index(R), a.u.

107.47 (82.23–176.35)

70.27 (20.97–86.16)

0.03

Contribution of relative components of microcirculation oscillationsb, %

Endoc

34.88 (19.23–40.24)

48.86 (23.68–59.26)

0.36

Neuroc

27.38 (20.3–41.22)

27.7 (19.43–37.95)

0.71

Myoc

31.44 (15.99–46.34)

22.45 (13.21–36.67)

0.23

Endo(R)d

20.42 (6.68–34.73)

36.81 (32.18–45.7)

0.05

Neuro(R)d

27.67 (11.35–38.03)

26.84 (15.36–34.68)

0.93

Myo(R)d

40.82 (35.27–56.77)

31.65 (24.19–41.15)

0.05

Fraction of flowmotion at rest (Fm index) and during the reperfusion phase (Fm index [R])b

Endoe

18.27 (5.56–50.85)

9.47 (4.12–18.21)

0.24

Neuroe

21.94 (11.31–28.44)

7.98 (3.41–11.4)

0.03

Myoe

13.69 (6.43–21.52)

6.28 (2.09–17.32)

0.11

Endo(R)f

16.38 (12.49–26.71)

22.25 (5.64–35.04)

0.99

Neuro(R)f

19.81 (16.45–32)

10.46 (6.7–18.96)

0.13

Myo(R)f

38.82 (27.3–57.48)

14.64 (4.19–34)

0.02

Data are presented as median (interquartile range) with the Wilcoxon rank-sum test P value.

P values <⁠0.05 were considered significant.

a Recovery diameter of the brachial artery in the last 30 s of FMD examination (usually 4 min after the end of occlusion)

b Oscillations at baseline and at reperfusion were grouped into 3 different frequency intervals: ≤ 0.021 Hz, 0.021–0.052 Hz, and 0.052–0.15 Hz, corresponding to endothelial, neurogenic, and myogenic activity, respectively.

c Endo, neuro, and myo denote contribution of the endothelial, neurogenic, and myogenic component at rest.

d Endo(R), neuro(R), and myo(R) denote contribution of the endothelial, neurogenic, and myogenic component at reperfusion stage.

e Endo, neuro, myo denote the fraction of endothelial, neurogenic, and myogenic activity at rest.

f Endo(R), neuro(R), myo(R) denote the fraction of endothelial, neurogenic, and myogenic activity at reperfusion stage.

Abbreviations: AI, augmentation index; AI 75, augmentation index normalized to heart rate of 75 bpm; a.u., arbitrary unit; FMD, flow mediated dilation; FMDr, FMD for recovery diameter; Fm index, basal flowmotion at rest; Fm index(R), flowmotion during reperfusion; FMSF, flow-mediated skin fluorescence; HR index, hyperemic response index; HR max, maximum hyperemic response; HS, hypoxia sensitivity; IR index, ischemic response index; IR max, maximum IR; log HS, logarithm of HS, LnRHI, natural logarithm of RHI; RHI, reactive hyperemia index; MR, metabolic recovery; MSA, mean squared amplitude; NOI, normoxia oscillatory index; PSD, power spectra density at rest; PSD(R), PSD at reperfusion; RHI, reactive hyperemia index; RH-PAT, reactive hyperemia peripheral arterial tonometry; RHR, reactive hyperemia response; SR, shear rate; SR area to Max, SR area to maximum diameter; SR Max, maximum shear rate; others, see Table 1

Due to a lack of established cutoffs for the FMD, RH-PAT, and FMSF parameters, we used the first quartile (Q1) as a cutoff to define the population of the studied patients displaying impaired endothelial function. Medians and IQRs of those parameters were as follows: 5.92% (4.36%–9.02%) for FMD, 2.01 (1.64–2.42) for RHI, 10.06% (7.16%–11.18%) for HR index, 25.2% (18.02%–32.96%) for RHR, and 74.65% (62.2%–86.37%) for NOI. Higher SCORE2 risk was not significantly associated with worse (<⁠Q1) values of the endothelial parameters. Rather unexpectedly, the percentage of patients with worse endothelial function described by NOI was higher in the low-to-moderate risk group (Supplementary material, Table S2).

Influence of tyrosine kinase inhibitor–based treatment on endothelial function

More than a half of the patients (n = 31; 64.5%) were treated with multiple TKIs. Total median (IQR) treatment duration was 98.52 (43.06–169.3) months, and it ranged from 8 to 244.83 months (Table 1). Most patients (n = 47; 97.9%) had a history of treatment with imatinib and for almost all of them (n = 46; 95.8%), it was the first-line treatment. Nilotinib was the most commonly used TKI in the second-line treatment (n = 20; 41.7%). The most frequently used third-line TKI was bosutinib (n = 7, 14.6%). Specific data on the duration and sequence of TKI treatment are shown in Supplementary material, Table S3.

Several endothelial parameters correlated with the duration of treatment with TKIs: SR Max, RHR, LogHS, and myogenic activity at reperfusion stage (Myo[R]). The correlation was the strongest for RHR (Spearman ρ = –0.46; P = 0.001) (Table 3; Supplementary material, Figure S4).

Table 3. Correlations of endothelial function parameters with overall length of treatment with tyrosine kinase inhibitors

Parameter

Spearman ρ

P value

FMD (n = 41)

FMD

–0.07

0.65

FMDr

–0.34

0.05

Basic diameter

0.11

0.49

Maximum diameter

0.11

0.5

Recovery diameter

0.2

0.24

SR basic

–0.14

0.37

SR Max

–0.43

0.005

SR area

0.01

0.96

SR area to Max

0

1

Diameter at 30 s

0.09

0.62

Diameter at 60 s

0.2

0.27

SR at 30 s

–0.24

0.14

SR at 60 s

–0.12

0.46

SR area to 30 s

–0.12

0.47

SR area to 60 s

–0.08

0.65

RH-PAT (n = 43)

RHI

–0.03

0.87

Heart rate

–0.23

0.14

AI

0.16

0.32

AI 75

0.06

0.68

LnRHI

–0.03

0.87

FMSF (n = 47)

IR index

–0.35

0.06

HR index

–0.2

0.18

IR max

–0.33

0.08

HR max

–0.15

0.31

RHR

–0.46

0.001

NOI

0.2

0.17

MR

–0.06

0.71

HS

–0.35

0.02

LogHS

–0.35

0.02

PSD × 106, MSA

–0.21

0.16

PSD(R) × 106, MSA

–0.27

0.06

FMindex

–0.2

0.19

FMindex(R)

–0.28

0.06

Contribution of relative components of microcirculation oscillationsa

Endob

0

0.98

Neurob

0.2

0.17

Myob

–0.2

0.17

Endo(R)c

0.15

0.32

Neuro(R)c

0.16

0.29

Myo(R)c

–0.21

0.16

Fraction of flowmotion at rest (Fm index) and during the reperfusion phase (Fm index [R])a

Endod

–0.2

0.17

Neurod

–0.12

0.44

Myod

–0.25

0.09

Endo(R)e

–0.13

0.39

Neuro(R)e

–0.15

0.31

Myo(R)e

–0.35

0.02

P values <⁠0.05 were considered significant.

a Oscillations at baseline and at reperfusion were grouped into 3 different frequency intervals: ≤ 0.021 Hz, 0.021–0.052 Hz, and 0.052–0.15 Hz, corresponding to endothelial, neurogenic, and myogenic activity, respectively.

b Endo, neuro, and myo denote contribution of the endothelial, neurogenic, and myogenic component at rest.

c Endo(R), neuro(R), and myo(R) denote contribution of the endothelial, neurogenic, and myogenic component at reperfusion stage.

d Endo, neuro, myo denote the fraction of endothelial, neurogenic, and myogenic activity at rest.

e Endo(R), neuro(R), myo(R) denote the fraction of endothelial, neurogenic, and myogenic activity at reperfusion stage.

Abbreviations: see, Table 2

As the duration of treatment with TKIs might have been linked to the patient age, we characterized the endothelial parameters of all 3 methods that could be age-dependent. All parameters that correlated with the length of treatment also correlated with age (SR Max, RHR, LogHS, and Myo[R]), but some parameters correlated with age rather than with the duration of TKI treatment (Supplementary material, Table S4 and Figure S5).

To further determine if the correlation between several parameters of endothelial function (SR Max, RHR, LogHS, and Myo[R]) with the duration of TKI treatment was independent of age, linear regression models were constructed, which revealed that the length of treatment was a predictor of RHR when controlled with age (P = 0.004). After correction for age, SR Max was weakly associated with the duration of treatment (P = 0.04). When adjusted for age, no correlation between LogHS and Myo(R) was found (P = 0.05). Thus, out of the endothelial parameters, RHR showed the strongest correlation with the duration of TKI treatment with effects that were independent of age (Supplementary material, Tables S5 and S6).

The patients treated with ponatinib displayed lower values of the FMSF-based parameter (Myo[R]) (Supplementary material, Table S7). There was no difference in RH-PAT parameters between TKIs. The ponatinib-treated patients displayed lower values of SR at 60 seconds, as compared with the nilotinib-treated individuals; however, the differences did not reach statistical significance in comparison with other TKIs (Supplementary material, Table S7).

Discussion

The patients suffering from CML treated with some of the TKIs have an increased risk of cardiovascular events.1-4 CML itself has a detrimental impact on endothelial function, and treatment with TKIs can lead to further deterioration of vascular status and serious, life-threatening cardiovascular complications, such as intravascular thrombosis.23,31 Despite the evidence that CAEs during TKI therapy could be promoted by endothelial dysfunction, only a few studies analyzing the endothelial function in CML are available, and they are based either on biomarkers collected from the peripheral blood of CML patients32,33 or analysis of endothelial function performed with a single method (RH-PAT),34 which recently has shown rather negative results. In fact, Kaneko et al34 reported that CML patients did not display differences in the baseline characteristics between the low RHI (<⁠1.67; n = 10), borderline RHI (≥1.67 and <⁠2.10; n = 14), and normal RHI (≥2.10; n = 6), and treatment responses among the 3 groups did not show important differences.

In the present work, we comprehensively evaluated the endothelial function in CML patients treated with TKIs, and assessed it in the conduit arteries and microcirculation with the FMD, RH-PAT, and FMSF methods. Based on this approach, we demonstrated, to the best of our knowledge for the first time, that endothelial dysfunction in patients with CML treated with TKIs was not related to cardiovascular risk based on SCORE2 algorithm but depended on CML-specific factors, including total duration of TKI treatment and ponatinib treatment alone. In our study, we observed distinct changes in the parameters measured with FSFM in the skin microcirculation of patients with CML. These alterations are not linked to age-dependent changes. Instead, they are associated with TKI treatment. Our findings suggest that these changes in skin microcirculation could serve as a valuable indicator. Therefore, monitoring skin microcirculation could potentially provide insights into the impact of TKI treatment on vascular health. This approach appears promising and warrants further investigation. In contrast, alterations in endothelial function in the conduit artery measured by FMD or microcirculation of the finger measured by RH-PAT may not provide as specific insights into the effects of TKIs on blood vessels. These results are in line with the notion that alterations in the skin microcirculation may precede changes in endothelial function in the conduit vessels in people with CVDs,35 and that skin microcirculation is affected by myeloproliferative diseases.36 Accordingly, skin microcirculation tests can be regarded as representative and sensitive enough to detect generalized microvascular dysfunction.37

To assess whether endothelial dysfunction in patients with CML treated with TKIs was related to cardiovascular risk, we characterized the patients using 2 appropriate systematic coronary risk estimation scales, namely SCORE2 and SCORE2-OP.29,30 The recently introduced SCORE2 estimates the individual 10-year risk of fatal and nonfatal CVD (myocardial infarction, stroke) in apparently healthy people with risk factors that are untreated or have been stable for several years. This test was intended for people between 40 and 69 years old, while the SCORE2-OP was intended for risk assessment in people between 70 and 89 years old.29,30 Obviously, for the SCORE2 and SCORE2-OP analysis, we have selected eligible patients from our study group, that is, those who did not have severe cardiovascular comorbidities and were older than 40 years, resulting in a subgroup of 35 CML patients (73% of the initial study group).

Our study demonstrated that the CML patients with high / very high risk in the SCORE2 did not have statistically worse endothelial function based on the main assessed parameters than the patients at low / moderate risk. The worse (<⁠Q1) values of the endothelial parameters were not significantly associated with higher cardiovascular risk based on the SCORE2 scale. The SCORE2/SCORE2-OP algorithm has been recently used in real-life settings to assess the risk of arterial occlusive events in patients treated with ponatinib and nilotinib, and patients with a high to very high SCORE2/SCORE2-OP risk showed a higher incidence rate of these complications (69.2% vs 46.5%; P <⁠0.001).38 Our results suggest, however, that the SCORE2 calculation is not optimally suited for CML patients in terms of assessing CVD risk that is known to be linked to endothelial dysfunction. Our findings seem to indicate that in CML patients eligible for SCORE2 and SCORE2-OP analysis, endothelial dysfunction was determined mainly by CML-specific factors, and not by classic risk factors for CVDs. Given the emerging evidence from preclinical studies on the detrimental effects of TKIs on endothelial function,18-23 we tested whether endothelial function in CML patients might have been influenced by TKI treatment. The number of parameters of endothelial function measured with the FMD, RH-PAT, and FMSF methods correlated negatively with patient age. It was an intriguing finding that the skin microcirculation response (RHR) assessed with the FMSF method, rather than the FMD or RH-PAT responses, negatively correlated with the duration of TKI treatment even after adjustment for age. These results are in line with the findings from previous studies that assessed age-dependent deterioration of endothelial function using FMD, RH-PAT, or FMSF.39-41 However, when our results were adjusted for age, the correlations of treatment duration with FMSF were found independent of this variable, whereas the correlations with FMD were weaker.

Accordingly, we confirmed that many parameters of endothelial function correlated with patient age, but to the best of our knowledge, we showed for the first time that FMSF-based assessment of skin microcirculation is feasible to detect age-independent vascular effects of TKIs. Of note, FMD and RH-PAT are both well-established techniques with documented prognostic value in CVDs.42-44 The FMSF is a relatively novel and innovative technique used to detect changes in microcirculation in patients with diabetes, heart failure, and other conditions in which the RHR readout based on the FSFM method was substantially affected.45 Furthermore, several reports using other methods to assess skin microcirculation showed that it may reflect early changes in the cardiovascular system and systemic microcirculation alterations.46 Our results extend this knowledge by showing that FMSF-based assessment of skin microcirculation could detect vascular effects of TKI-based treatment, and thus this method, or perhaps some other ways of monitoring the skin microcirculation,46 may prove useful in monitoring the vascular effects of TKIs in CML patients.

Our results seem to be in line with the notion that endothelial function in CML patients might be more profoundly affected in the microcirculation than macrocirculation. The architecture of bone marrow microvessels is altered in myeloproliferative neoplasms, including CML,47 and this disease may also have a detrimental impact on microcirculation in other organs, including the retina or lungs.48 Our results based on the FMSF suggest that CML may also affect the skin microcirculation, reflecting the systemic nature of the microvascular dysfunctional state in CML patients, which could be further modulated by TKI treatment.49

This study has several limitations. We were able to include a relatively low number of patients, and our patient population was heterogeneous, as we included individuals treated with multiple TKIs. However, this approach matched the clinical reality in which patients are treated with numerous TKIs. There are no well-established cutoffs for endothelial parameters in FMSF, FMD, and RH-PAT tests we used in our study; therefore, we defined the worst endothelial function based on the lower first quartile of the parameters measured in CML patients, and we did not include a control group. In our CML cohort, a majority of the patients (73%) were eligible for SCORE2/SCORE2-OP assessment, but it remains to be established whether in a larger study group the number of CML patients and cardiovascular comorbidities would be similar.

In summary, despite these limitations, our study provides novel insights into endothelial status of CML patients treated with TKIs. Its findings indicate that FMSF-based assessment of skin microcirculation may prove useful in detecting vascular effects of TKIs and guiding the vascular safety of TKI therapy. Further studies are needed to validate our conclusions in a bigger cohort of patients before the assessment of skin microcirculation can be used as a novel diagnostic tool for patients with CML.