Several retrospective studies have described the clinical manifestation of peripheral artery occlusive disease (PAOD) in patients receiving nilotinib. We thus prospectively screened for PAOD in patients with chronic phase chronic myeloid leukemia (CP CML) being treated with tyrosine kinase inhibitors (TKI), including imatinib and nilotinib. One hundred and fifty-nine consecutive patients were evaluated for clinical and biochemical risk factors for cardiovascular disease. Non-invasive assessment for PAOD included determination of the ankle-brachial index (ABI) and duplex ultrasonography. A second cohort consisted of patients with clinically manifest PAOD recruited from additional collaborating centers. Pathological ABI were significantly more frequent in patients on first-line nilotinib (7 of 27; 26%) and in patients on second-line nilotinib (10 of 28; 35.7%) as compared with patients on first-line imatinib (3 of 48; 6.3%). Clinically manifest PAOD was identified in five patients, all with current or previous nilotinib exposure only. Relative risk for PAOD determined by a pathological ABI in first-line nilotinib-treated patients as compared with first-line imatinib-treated patients was 10.3. PAOD is more frequently observed in patients receiving nilotinib as compared with imatinib. Owing to the severe nature of clinically manifest PAOD, longitudinal non-invasive monitoring and careful assessment of risk factors is warranted.
In Bcr-abl-positive chronic phase chronic myeloid leukemia (CP CML), the introduction of the second-generation tyrosine kinase inhibitors (TKI) nilotinib and dasatinib as first-line treatment options has resulted in superior response rates and a significant reduction of the risk to develop blastic phase disease as compared with therapy with imatinib mesylate, as has been demonstrated with nilotinib.1, 2 For nilotinib, recent studies have shown rapid molecular and cytogenetic responses both in imatinib-resistant and newly diagnosed CML patients.3 The phase 3 ENESTnd trial (Evaluating Nilotinib Efficacy and Safety in Clinical Trials—newly diagnosed patients) proved the superiority of nilotinib (300 or 400 mg BID) over standard therapy with imatinib (400 mg) in CML, with major molecular response rates of 44% (300 mg BID) and 43% (400 BID) versus 22% (imatinib) at 1 year.2
The approved TKI for CML have distinct toxicities that allow for the selection between these compounds based on individual patient tolerability and comorbidites. Nilotinib was shown to be well tolerated with only mild toxicities including transient skin rash (31–36%), headache (14–21%), nausea (11–19%), pruritus (15–13%) and myalgia (10%).2 Biochemical abnormalities included grade 3/4 elevations in lipase (7.2–7.6%), bilirubin (3.6–7.9%) and glucose (4.7–6.1%).4 In almost all cases, these toxicities are reversible upon dose interruptions or reductions.
Peripheral artery occlusive disease (PAOD) refers to the obstruction of large arteries excluding coronary, aortic arch or the central nervous system supplying arteries. Generally, the underlying pathological mechanism is atherosclerosis leading to stenosis or thrombus formation. Risk factors for PAOD are similar to those for atherosclerosis and include smoking, diabetes mellitus, dyslipidemia, hypertension, and also male gender, age >50 years and obesity. As per both the American College of Cardiology Foundation (ACCF) and the American Heart Association (AHA), the diagnosis of PAOD is established by a pathologic ankle-brachial index (ABI) defined as a ratio of systolic blood pressure between the upper and lower extremities below 0.9.5, 6 Additional methods such as duplex ultrasonography or angiography are used to identify the involved arteries and localize atherosclerotic lesions.
Following an initial report by Aichberger et al,7 who noted 3 out of 24 cases (12.5%) of severe PAOD in nilotinib-treated CP CML patients, three additional retrospective analyses were published.8, 9, 10 Tefferi et al.10 reported two patients with late CP CML, of whom one developed severe PAOD requiring percutaneous transluminal angioplasty. Intrigued by these observations, the cohort of the initial report by Aichberger et al.7 was extended to three additional centers, and it identified 11 (6.1%) among 179 patients receiving nilotinib treatment with clinically manifest severe PAOD.8 Of these 11 patients, 2 received nilotinib as frontline therapy. One of these two patients had hypereosinophilic syndrome and no CML. More recently, Quintas-Cardama et al.9 reported a retrospective analysis of 233 CML patients who received nilotinib at the MD Anderson Cancer Center between 2004 and 2011. In this cohort, five cases (2.1%) of nilotinib-associated vascular events were observed. However, different from the previous reports, two of these cardiovascular events were of vasospastic nature, lacking any evidence of an atherosclerotic lesion. A phase 3 setting as in the ENESTnd trial may be the best design to capture toxicities that occur at low frequencies. The observation of seven cases of PAOD in the most recent article on the ENESTnd trial is of considerable interest.4 All of these seven cases were exclusively found in the two nilotinib study arms (four receiving 300 mg BID, three 400 mg BID), equating to a frequency of 1.2% after a median follow-up of 3 years. One of these cases was already published in our previous report.8 After a follow-up of 4 years, two additional PAOD cases have occurred in the nilotinib 400 mg BID arm.
Taken together, the frequency of nilotinib-associated PAOD in this series of reports ranges between 1.2 and 12.5%,4, 7 and depends on size and type of study cohort, definition and grade of PAOD as well as duration of nilotinib exposure. However, all previously published articles are limited by the description of only clinically manifest PAOD without screening of asymptomatic patients and without prospective monitoring. Likewise, there is no sufficient control data on CP CML patients who have not received nilotinib treatment.
We therefore screened all CP CML patients that were treated at our institution between August 2011 and November 2012 by assessment of the ABI followed by duplex ultrasonography when indicated and obtained clinical risk factors and relevant biochemical parameters. In addition, characteristics of patients from collaborating centers with clinical manifest PAOD were determined. We here present the first analysis of this prospective investigation.
Patients and methods
All CP CML patients that were treated between August 2011 and November 2012 at our institution were included in this study and constitute the screening group for the purpose of this study. About half of the patients were either previously or currently treated within multicenter clinical trials including four with nilotinib: ENESTnd (NCT00471497, n=18), ENEST1st (NCT00519090, n=15), CAMN107A2101 (NCT00109707, n=17), ICORG-08-02 (NCT00809211, n=6); one with dasatinib: DASISION (NCT00481247, n=2), and one with ponatinib: PACE (NCT01207440, n=14) (Table 1). All trials were conducted in accordance with the applicable regulatory requirements. Patients gave their written informed consent to participate in a clinical trial or to have their data analyzed retrospectively. All procedures were followed in accordance with the Helsinki Declaration and were approved by the local ethics committee. Alternative TKI included imatinib, dasatinib, bosutinib, ponatinib or bafetinib (INNO-406). Patients with clinically manifest PAOD were included in a separate analysis and contained patients of the screening group, including one patient from Berlin with hypereosinophilic syndrome being treated within the CAMN107A2101 study, as well as patients reported from collaborating centers.
Clinical evaluation of the screening group
All screening group patients were regularly seen and treated at our center. Normally, CP CML patients are seen at 3–6 months intervals, if not requested otherwise by a study protocol. Few patients were seen more frequently due to toxicities. During their visits, patients once filled in a questionnaire on cardiovascular risk factors including history of cardiovascular diseases, diabetes mellitus, nicotine abuse, age, size, weight and walking distance.
In all screening group patients, a blood sample was taken after at least 4 weeks of the most recent therapy, once in fasting state, to screen for diabetes mellitus (blood glucose levels and HbA1c) as well as for dyslipidemia (cholesterol, low-density lipoprotein (LDL), high-density lipoprotein and triglycerides).
Non-invasive assessment of PAOD
To screen for PAOD, ABI was taken at the cardiovascular department of our institution and was typically carried out through systolic blood pressure determination on the upper limb with the clinical RR method and on the lower limb through a duplex-assisted technology (Elcat, handydop, www.elcat.de). According to a recently updated guideline of the American College of Cardiology Foundation and the American Heart Association Task Force on Practice Guidelines, an ABI of 1.00–1.40 is considered normal, whereas values <0.90 are abnormal and values of 0.91–0.99 are considered ‘borderline’. Outside a clinical trial, ABI testing should be considered in individuals aged ⩾65.11
In patients with pathological ABI values, duplex ultrasonography of the lower limb was performed to identify the involved arteries as well as atherosclerotic plaques. All vascular assessments were carried out under the supervision of board-certified cardiovascular specialists. In patients with pathological findings by duplex ultrasonography, an angiographic investigation was considered as clinically indicated.
Clinical definition of PAOD
Typical peripheral ulcerations as well as lesions identified by duplex ultrasonography or alternative imaging techniques were defined as clinically manifest PAOD. In addition, patients presenting after treatment of an acute event of PAOD in an external medical center were also considered as clinical manifestation of PAOD.
Descriptive statistical analysis was performed using Fisher’s exact test and the one-way ANOVA test with Bonferroni post test where appropriate. A P-value <0.05 was considered statistically significant. All tests were two-sided.
Overall, 159 patients of the screening group were evaluable and categorized into five groups: (I) patients on first-line imatinib (n=54), for most comparisons considered as reference, (II) patients on first-line nilotinib (n=33), (III) patients with previous imatinib exposure and on second-line nilotinib (n=33), (IV) patients with previous nilotinib exposure and currently receiving an alternative TKI (post-nilotinib, n=25) and (V) nilotinib-naive patients not receiving imatinib (n=14).
About half of the patients were or had been treated within clinical trials at the time of data collection. Best responses were complete hematologic remission in 10 (6.2%), major cytogenetic remission in 20 (12.4%), complete cytogenetic remission in 17 (10.7%), and major or complete molecular remission in 114 (71.7%, data not shown).
CML-associated baseline parameters including prior treatment and disease duration until assessment of ABI are shown in Table 2. Of note is a statistically significant different median CML duration between patients on first-line imatinib (102 months, range 16–250) as compared with patients on first-line nilotinib (30 months, range 7–62, P<0.0001), which results in statistically significant different median treatment durations between patients on first-line imatinib (102 months, range 16–250) and patients on first-line nilotinib (30 months, range 7–62, P<0.0001).
Clinical risk factors for PAOD
Clinical risk factors such as age, gender, smoking, arterial hypertension, a history of coronary heart disease, diabetes mellitus and body mass index are shown in Table 3. There is a difference in the distribution of male versus female patients in some of the groups. When all other risk factors were compared with patients on first-line imatinib, none of these factors, with the exception of smoking in nilotinib-naive patients (P=0.0149), showed a statistically significant difference.
Biochemical risk factors for PAOD
Across the five treatment groups of the screening cohort, there were no statistically significant differences with regard to glucose levels, HbA1c, triglycerides or high-density lipoprotein (Table 4). However, statistically significant elevations of both cholesterol and LDL were observed in nilotinib-treated patients. In detail, median cholesterol levels were statistically significant higher in the first-line nilotinib (209 mg/dl, P<0.0001), second-line nilotinib (222 mg/dl, P<0.0001) and post-nilotinib (193 mg/dl, P<0.034) cohorts as compared with that in the first-line imatinib cohort (164 mg/dl), but not the nilotinib-naive cohort (194 mg/dl, P<0.7). Similarly, median LDL levels were statistically significant higher in the first-line nilotinib (135 mg/dl, P<0.0003), 2nd-line nilotinib (139 mg/dl, P<0.0001) and post-nilotinib (117 mg/dl, P<0.03) groups as compared with that in the first-line imatinib (95 mg/dl) group (Table 4).
Screening for PAOD
Ankle-brachial indices (ABI) were obtained in 129 of 159 patients (81%). A pathological ABI, defined as values <0.9, was observed in 24 of 129 (18.6%) examined patients (Table 5). Fourteen (58.3%) of these were unilateral and 10 (41.7%) were bilateral. The distribution of pathological ABIs across the different treatment groups was 3 of 48 (6.3%) in the first-line imatinib group, 7 of 27 (26%) in the first-line nilotinib group, 10 of 28 (35.7%) in the second-line nilotinib group, 3 of 18 (16.6%) in the post-nilotinib group and 1 of 8 (12.5%) in the nilotinib-naive group (Table 5). Both first- and second-line nilotinib groups had statistically significant higher frequencies as compared with the first-line imatinib group (P=0.0297 and P=0.0029, respectively). Pathological ABIs in the first-line nilotinib group were observed after 21, 27, 37, 40, 45, 51 and 56 months, respectively. Six of these patients were in complete molecular remission, while one patient was in complete cytogenetic remission (data not shown).
Duplex ultrasonography and frequencies of PAOD
Overall, 30 duplex ultrasound investigations were performed. In eight of these patients (26.7%), atherosclerotic lesions typical of PAOD were identified (Table 5). Across the five treatment groups, pathological duplex ultrasonographies were found in one patient in the first-line imatinib group, two patients in the first-line nilotinib group, two patients in the second-line nilotinib group and three patients in the post-nilotinib group (Table 5).
Clinically manifest PAOD, defined as typical peripheral ulcerations or an acute event of PAOD, was identified in five patients. Of these, one was in the in the first-line nilotinib group, three in the second-line nilotinib group and one in the post-nilotinib group (Table 5).
Comparing patients on first-line imatinib and first-line nilotinib only, a pathological ABI (3 versus 7 events), a pathological duplex ultrasonography (1 versus 2 events) or clinically overt PAOD (0 versus 1 event) was more frequent under nilotinib (P<0.0001, P=0.0319 and P=0.0360, respectively, data not shown). With a total drug exposure of 409.8 patient-years in the first-line imatinib group and 93.2 patient-years in the first-line nilotinib group, the relative risk for a pathological ABI and the occurrence of a pathological duplex sonography with or without a PAOD event in nilotinib-treated patients as compared with imatinib-treated patients was 10.3 (95% CI, 2.3–61.5) and 13.2, respectively (95% CI, 1.1–692.5).
Characteristics of patients with clinically manifest PAOD
Altogether, 27 patients with clinically manifest PAOD were identified. All but one patient with hypereosinophilic syndrome had CP CML (Table 6). These patients differed from the screening group as they developed clinically manifest PAOD. None of these patients had a history of preexisting PAOD, but all had cardiovascular risk factors. Data on six of these patients have previously been published,8 and three of them are being described in parallel publications.12, 13 Of these 27 patients, only 1 was nilotinib-naive. This patient was a 77-year-old male with a long history of CML (13.3 years), with a long treatment duration of first-line imatinib (10 years) and with numerous cardiovascular risk factors including smoking, arterial hypertension, obesity and dyslipidemia.
CML-related clinical parameters and cardiovascular characteristics are summarized in Table 6. Twenty (74%) patients were either on first- or second-line nilotinib when PAOD was diagnosed. In five patients, nilotinib had already been discontinued or had been switched to another TKI (ponatinib) for different reasons when PAOD was noted. Atherosclerotic lesions were primarily located in the lower limbs. Most frequently, the femoral (44.4%) and tibial (37%) arteries were affected. In one patient, acute loss of vision occurred and upon intensive ophthalmological work-up, the diagnosis of an ischemic neuropathy was established.
Clinical interventions that were performed included percutaneous transluminal angioplasty (33.3%), stent implantation (22.2%), amputation of single digits or lower limbs (22.2%) and surgery (18.5%). In 11 (40.7%) patients, the diagnosis of PAOD was established by radiographic imaging techniques, but without the clinical indication for an invasive procedure.
Management of patients in whom PAOD was diagnosed during nilotinib therapy included switch to another TKI, discontinuation and dose reduction of nilotinib.
Cardiovascular toxicities caused by TKI used in the treatment of malignant diseases are a rare but emerging phenomenon.14 Prior to discussing specific cardiovascular toxicities of these drugs, the limitations of even well-designed and monitored prospective trials in recognizing these cardiovascular toxicities and in assessing the safety profile of these compounds must be addressed. First, many of these studies have follow-up intervals of <5 years, and thus, will rarely capture late toxicities such as the development of atherosclerotic lesions. In addition, exclusion of significant cardiovascular morbidity will result in skewed study populations in many of these trials as is also indicated by the significantly younger age of these patients. Second, as many of the cardiac events constitute diseases by themselves, physicians may misinterpret them as independent events occurring in their patients independently of their malignancy. Thus, underreporting may be another mechanism leading to underestimation of cardiovascular side effects. Third, in the case of symptomatic PAOD, claudication may falsely be diagnosed as unspecific musculoskeletal pain without any further diagnostic work-up. Both, the difficulties to adequately diagnose PAOD in a given patient and to connect this event to nilotinib treatment may therefore explain the relatively recent description of these toxicities.
In our study cohort of 159 patients, we found an association of nilotinib treatment with severe PAOD, as previously described. But in addition to earlier reports, our data suggest that early PAOD can be detected more frequently in nilotinib- than in imatinib-treated patients, when cardiovascular techniques such as assessment of ABI and duplex ultrasonography are used. In this regard, the relative risk determined by ABI of 10.3 for patients on first-line nilotinib as compared with patients on first-line imatinib is remarkable.
In view of several reports showing attenuation of glucose metabolism in imatinib-treated patients, a hypothetical vasoprotective effect of imatinib may be postulated.15, 16, 17, 18 The differences between the two first-line groups would thus be explained by a lowered frequency of PAOD in imatinib-treated patients as compared with healthy individuals, while no such effect would be present in nilotinib-treated patients. However, numerous cardiovascular trials studying PAOD in the normal population demonstrated frequencies of PAOD in age-matched cohorts ranging between 1.3 and 6.7% comparable to the 1st-line imatinib group which clearly suggests an elevated cardiovascular risk by nilotinib.19, 20, 21
One of the limitations of this study is the absence of baseline cardiovascular parameters (ABI, duplex ultrasonography, biochemistry risk factors) in our patients. However, this applies to all cohorts and does not sufficiently explain the differences between imatinib- and nilotinib-treated patients which needs urgent confirmation in other CML cohorts.
Glucose elevation in nilotinib-treated patients has been observed in previous trials and has been widely accepted as one of the major biochemical alterations associated with this drug. However, no differences in glucose and HbA1c levels across our cohorts were observed (Table 4). Careful monitoring, adequate dose adjustments and medical optimization of glucose metabolism may have helped to manage this toxicity in our cohort. However, we observed significant elevation of both cholesterol and LDL in all three nilotinib-treated patient cohorts. Although hypercholesterolemia under nilotinib has been noted previously, the elevation of LDL has hitherto been undescribed.22 Although biochemistry risk factor assessment in our screening cohort was analyzed in the fasting state at least 4 weeks after the initiation of TKI therapy, longitudinal data as well as baseline biochemistry samples will be needed to fully understand the impact of hyperglycemia, hypercholesterolemia and LDL elevation in the development of PAOD under nilotinib therapy.
Taken together, the underlying mechanisms leading to the formation of atherosclerotic plaques in some of the patients receiving nilotinib are not yet known. In view of the observed biochemical alterations, an accelerated form of the metabolic syndrome may be involved, but other factors may also be contributing, given the relatively short latency in some of the affected patients. Some of these other factors may include inhibition of kinases such as DDR1, KIT or PDGFR that have been implicated in vascular cell homeostasis or alternative mechanisms leading to vascular inflammation.9, 23, 24
In our cohort, the management of patients with clinically manifest PAOD was conducted by cardiovascular experts of our institutions in accordance with standardized guidelines of the participating centers and countries. As in some of the patients, dose reductions were performed, while others were fully discontinued from nilotinib or switched to an alternative TKI, our cohort is too small to define optimal management regarding TKI therapy against a background of subclinical or clinical PAOD. In addition, there is currently no marker to distinguish between cardiovascular events related to a TKI and cardiovascular events that may have had occurred without TKI exposure, as many patients had cardiovascular risk factors at baseline. Management of cardiovascular risk factors and treatment of cardiovascular complication according to international guidelines is mandatory. Our data is consistent with that reported by Giles et al,12 in which, a retrospective cohort analysis on reported PAOD rates in 2390 patients with CP CML treated on three randomized phase 3 studies showed that nilotinib was associated with higher rates of PAOD versus imatinib.
In conclusion, with several effective TKI at hand, our data strongly indicate that cardiovascular morbidity and the risk for the development of PAOD should be considered in patients beginning treatment of CP CML. Other potential manifestations of atherosclerosis, including fatal myocardial infarction, have been attributed to imatinib, nilotinib and dasatinib. We strongly suggest to capture baseline ABI, biochemical risk factors and to monitor these parameters regularly throughout TKI therapy of CML.
FEN thanks Madeleine Etienne CRA for data collection.