Adverse outcomes in chronic myeloid leukemia patients treated with tyrosine kinase inhibitors: Follow‐up of patients diagnosed 2002–2017 in a complete coverage and nationwide agnostic register study

Abstract Tyrosine kinase inhibitors (TKIs) have profoundly improved the clinical outcome for patients with chronic myeloid leukemia (CML), but their overall survival is still subnormal and the treatment is associated with adverse events. In a large cohort‐study, we assessed the morbidity in 1328 Swedish CML chronic phase patients diagnosed 2002–2017 and treated with TKIs, as compared to that in carefully matched control individuals. Several Swedish patient registers with near‐complete nationwide coverage were utilized for data acquisition. Median follow‐up was 6 (IQR, 3–10) years with a total follow‐up of 8510 person‐years for the full cohort. Among 670 analyzed disease categories, the patient cohort showed a significantly increased risk in 142 while, strikingly, no category was more common in controls. Increased incidence rate ratios/IRR (95% CI) for more severe events among patients included acute myocardial infarction (AMI) 2.0 (1.5–2.6), heart failure 2.6 (2.2–3.2), pneumonia 2.8 (2.3–3.5), and unspecified sepsis 3.5 (2.6–4.7). When comparing patients on 2nd generation TKIs vs. imatinib in a within‐cohort analysis, nilotinib generated elevated IRRs for AMI (2.9; 1.5–5.6) and chronic ischemic heart disease (2.2; 1.2–3.9), dasatinib for pleural effusion (11.6; 7.6–17.7) and infectious complications, for example, acute upper respiratory infections (3.0; 1.4–6.0). Our extensive real‐world data reveal significant risk increases of severe morbidity in TKI‐treated CML patients, as compared to matched controls, particularly for 2nd generation TKIs. Whether this increased morbidity may also translate into increased mortality, thus preventing CML patients to achieve a normalized overall survival, needs to be further explored.

Whether this increased morbidity may also translate into increased mortality, thus preventing CML patients to achieve a normalized overall survival, needs to be further explored.

| INTRODUCTION
Chronic phase (CP) chronic myeloid leukemia (CML) is characterized by the fusion gene BCR-ABL1, coding for an oncoprotein that acts as a tyrosine kinase. 1 With the clinical introduction of imatinib, and subsequently other targeted tyrosine kinase inhibitors (TKIs), during the recent 20 years, the prognosis for CML patients has improved dramatically. The 5-year relative survival rates for CML patients younger than 70 years of age are now approaching those of the general population. [2][3][4] Compared to imatinib, recommended doses of the newer 2nd-and 3rd-generation TKIs (dasatinib, nilotinib, bosutinib, and ponatinib) generally induce more rapid and deeper molecular responses, as assessed by quantitative RT-PCR of the fusion-transcript. [5][6][7][8][9] However, this has not translated into any clear differences between the various TKIs regarding overall survival. 10 Most TKIs are considered safe and well-tolerated, based mainly on short-term tolerability data from randomized trials. 11 Dose-limiting toxicities typically reach only grade 2 on the Common Terminology Criteria for Adverse Events (CTC-AE) scale. Since some of the TKIs have been available for less than 10 years and the number of patients diagnosed with CML is comparatively small, systematic studies of long term-toxicity have been scarce. We have previously used data from the Swedish CML Register to study TKI-related cardiovascular toxicity. 12 The issue was highlighted by reports on an increased incidence of mainly arterial events. 13,14 It has been suggested that this type of toxicity may help explain why the new generation of TKIs have not demonstrated superiority over imatinib in terms of overall survival. In our previous publication, we were limited to patients diagnosed between 2002 and 2012 with relatively limited duration of follow-up in the patients treated with the newer TKIs nilotinib and dasatinib, and a very scarce use of bosutinib and ponatinib. There is a persistent paucity of data on the full range of morbidity in CML patients, which limits the possibility of tailoring TKI therapy to individual patients both with regards to short-term response and to toxicity. With this background, we set out to explore the range of possible adverse events in a large group of CML patients treated with TKIs in a real-world setting during extended time periods and to elaborate upon relations of these events to specific TKIs.  16 All diagnoses and causes of death were coded using the 10th revision of the international classification of disease (ICD-10) throughout the study period.
Data on current TKI treatment was ascertained using a twopronged approach. Since 2005, all dispensed prescriptions were available through linkage with the nationwide Prescribed Drug Register. 18 Earlier data, from 2002 to 2005, were instead extracted from the CML Register, where treatment is recorded both at diagnosis and during follow-up. Of note, treatment data from the CML Register were also available for patients enrolled in clinical trials where drugs were supplied by the sponsor and thus not recorded in the Prescribed Drug Register. In subjects receiving multiple TKIs, follow-up for each drug started at the date of the first dispensation of each treatment and continued until the day before the end-date of the next treatment or at the end of follow-up, whichever occurs first.

| Outcomes
Using data from the Patient and Cause of Death registers, we extracted information for all patients on events requiring care at a hospital or events registered as the cause of death, coded as International Classification of Disease (ICD) revision 10, and categorized these into 670 disease categories, excluding malignancies, hematological diseases, external causes of morbidity and mortality, as well as symptom-based ICD-codes. Details of how the disease categories were categorized are available in Table S1.

| Study design and statistical analyses
In the first analysis, we compared the incidence of all disease categories to the matched control population. Here, patients were followed from the date of CML diagnosis until the date of first incident event in each disease category investigated, emigration, death, progression to accelerated phase or blast crisis, or until the last day of follow-up (December 31st, 2018), whichever occurred first. To remove effects of disease progression not captured by specified progression date, we also terminated follow-up 5 months before allogeneic hematopoietic stem cell transplantation. Similarly, the control population was followed from date of diagnosis of matched CML patient, until the first incident event in each disease category, emigration, death, or December 2018, whichever occurred first. treatment to change in the model as the patient received different treatments, using the publicly available Stratify macro. 21 In recognition of it being a strictly explorative and due to likely insufficient power, the main comparisons were based on raw p values are presented. However, results are also presented using Bonferroni adjustment.
To limit effects of health outcomes seen primarily in the initial untreated and initiation-phase of TKI therapy in CML, and effects of the likely increased surveillance of patients at the time of their diagnosis, where they typically undergo frequent blood testing and diagnostic workups, we also performed a sensitivity analysis where start of follow-up was delayed until 6 months after CML diagnosis. As this analysis included the same hypothesis test and covariates from the main analysis, but only investigating the sub-population of FDR significant disease categories, Bonferroni-adjustment was conducted to handle sub-population adjustment for multiple testing not accounted for using FDR.
All statistical computations were conducted using SAS (SAS Institute v 9.4, California, USA).

| Ethical considerations
This study has been approved by the Swedish Ethical Review Authority (ref. nr: 2020-05425).
As is shown in  Figure 1 and Table S2 and Figure S1). Disease categories with increased risks were observed particularly among diseases of the cir-  (Table 3 and Figure S1). For nilotinib, the risk was statistically significantly elevated for several cardiovascular outcomes, most notably acute myocardial infarction (IRR, 2.9; 95% CI, 1. In conclusion, data from this large "real-world" cohort study with close to full nationwide coverage extensively and specifically identify and pinpoint a number of severe adverse events linked to CML patients subjected to prolonged treatment with TKI (in particular with 2nd generation drugs), as compared to outcomes for carefully matched controls. The full clinical implications of these findings are still somewhat unclear, but they may well provide important novel insights into the long-term morbidity and mortality of CML patients. In general, we believe that explorative studies using this or similar methodologies, capturing a large set of event types in large patient cohorts, constitute a fruitful approach to identify and quantify clinical risks associated with novel therapeutics, perhaps in particular offtarget effects by newly introduced precision-based treatment regimens.

ACKNOWLEDGMENTS
We are grateful to everyone involved with the maintenance of the Swedish CML Register as well as in the care of patients diagnosed with CML.

CONFLICT OF INTEREST
Dr Dahlén reports funding outside the scope of this work from Novartis. Dr Edgren reports funding from outside this work from Celgene.

AUTHOR CONTRIBUTIONS
Torsten Dahlén had full access to all the data in the study and takes

PATIENT CONSENT
The current study is supported by the general consent given for national quality-of-care registers in Sweden, where patients are informed of participation in quality-of-care registers and the possibility to actively withdraw from such register.

DATA AVAILABILITY STATEMENT
Due to laws guarding the integrity of the Swedish citizens, no patientlevel data can be made available. Upon request, aggregate data with different levels of resolution may be possible to provide.