Real‐world efficacy and safety outcomes of imatinib treatment in patients with chronic myeloid leukemia: An Australian experience

Abstract Tyrosine kinase inhibitors (TKI) have revolutionized the treatment of chronic myeloid leukemia (CML), but patients still experience treatment‐limiting toxicities or therapeutic failure. To investigate the real‐world use and outcomes of imatinib in patients with CML in Australia, a retrospective cohort study of patients with CML commencing imatinib (2001–2018) was conducted across two sites. Prescribing patterns, tolerability outcomes, and survival and molecular response were evaluated. 86 patients received 89 imatinib treatments. Dose modifications were frequently observed (12‐month rate of 58%). At last follow‐up, 62 patients (5‐year rate of 55%) had permanently discontinued imatinib treatment, of which 44 switched to another TKI (5‐year rate of 46%). Within 3 months of starting imatinib, 43% (95% CI, 32%–53%) of patients experienced imatinib‐related grade ≥3 adverse drug reactions (ADRs). Higher comorbidity score, lower body weight, higher imatinib starting dose, and Middle Eastern or North African ancestry were associated with a higher risk of grade ≥3 ADR occurrence on multivariable analysis (MVA). Estimated overall survival and event‐free survival rates at 3 years were 97% (95% CI, 92%–100%) and 81% (95% CI, 72%–92%), respectively. Cumulative incidence of major molecular response (MMR) at 3 years was 63% (95% CI, 50%–73%). On MVA, imatinib starting dose, ELTS score, BCR‐ABL1 transcript type, pre‐existing pulmonary disease, and potential drug–drug interactions were predictive of MMR. In conclusion, imatinib induced deep molecular responses that translated to good survival outcomes in a real‐world setting, but was associated with a higher incidence of ADRs, dose modifications and treatment discontinuations than in clinical trials.


| INTRODUC TI ON
Imatinib, a BCR-ABL1 tyrosine kinase inhibitor (TKI), has significantly changed the treatment landscape of chronic myeloid leukemia (CML).
In the landmark IRIS trial, imatinib induced complete cytogenetic response (CCyR) in 85.2% of patients by 18 months, compared to only 22.1% with interferonα plus low-dose cytarabine, and was also better tolerated with significantly superior survival outcomes. 1 Despite widespread introduction of second and third-generation TKIs, imatinib used in first-line is associated with a lower incidence of adverse drug reactions (ADRs) and similar long-term survival outcomes. 2 A significant proportion of patients receiving imatinib for CML management do not achieve major molecular response (MMR) or deep molecular response (DMR) on long-term treatment (5-year cumulative incidence of 60% and 42%, respectively 3 ), whilst others develop resistance to treatment or intolerable ADRs necessitating treatment discontinuation. 3,4 About one-third of patients achieve sustained deep molecular response (sDMR, 8-year cumulative incidence of 37% 5 ); considered the gateway to obtaining treatment-free remission (TFR).
A precision medicine approach is required to improve the utilization of this lifesaving drug, to reduce the risk of ADRs and treatment failure, to restore and maintain good health-related quality of life, and potentially to achieve a cure at an affordable cost. 6 It is important to understand the gap between patients enrolled in clinical trials and the real-world setting, with only 6% of patients diagnosed with a new cancer between 2018 and 2019 in Australia participating in a clinical trial. 7 The FDA recently published a framework highlighting the importance of using real-world observational data as a supplement to clinical trial data, to provide a more complete picture of tolerability and effectiveness of a drug. 8 Clinical trials employ strict patient inclusion and exclusion criteria, which can lead to populations in clinical trials that differ significantly from patients found in real-world clinical practice. 9,10 A review of eligibility criteria for cancer clinical trials submitted as investigational new drug applications to the FDA in 2015 found that 74% of trials excluded patients with a history of cardiovascular disease, 70% excluded patients with known hepatitis, 32% excluded patients with autoimmune diseases, and 29% excluded patients with gastrointestinal disorders. 11 Clinical trials also exclude patients on certain concomitant medicines for chronic health conditions, which have the potential for pharmacokinetic (PK) or pharmacodynamic (PD) drug-drug interactions with the investigational drug. Furthermore, approximately 60% of oncology clinical trials require an Eastern Cooperative Oncology Group (ECOG) performance status (PS) of 0 or 1 (or an equivalent Karnofsky PS of ≥70%), resulting in the exclusion of patients with poorer prognosis. 11 These exclusion criteria indirectly result in significant age disparities and differences in concomitant medication burden compared to patients who will ultimately receive the drug in practice. 12 Therefore, there is uncertainty as to the extent which findings from oncology clinical trials can be extrapolated (or generalized) to the heterogenous population of patients with comorbid conditions that are treated in real-world routine clinical practice. 10 The aim of this study was to investigate the real-world use, tolerability, and efficacy outcomes of imatinib in patients with CML treated in Australian hematology clinical practice.

| Data collection
Individual medical records were retrospectively reviewed, and demographic characteristics, disease characteristics, treatment details, prescribing patterns, tolerability outcomes, and efficacy outcomes were collected. Data were re-abstracted and verified by a second investigator in 30% of randomly selected patients. Variables and endpoints are defined in Appendix S1, Supplementary Methods. Demographic characteristics included geographic ancestry, comorbidities, kidney and hepatic function, age, sex, and total body weight at the time of CML diagnosis and at imatinib commencement.
Geographic ancestry was assigned based on information contained on patient registration forms and in the medical record, taking into consideration documented self-reported geographic ancestry or inference of geographic ancestry (using birthplace, family name or maiden name, language[s] spoken, and religion). [13][14][15] The Charlson Comorbidity Index (CCI) score was derived from comorbidities noted in patients' medical records and included their diagnosis of CML. 16 Any concomitant medicines that the patient used regularly during their TKI treatment (for ≥2 weeks) were documented, including All documented adverse events during imatinib treatment were evaluated for causality to imatinib using the Naranjo algorithm. 22 Adverse events classified as possible, probable, and definite were termed as imatinib-related ADRs. 23 The type and severity grade of adverse events were defined using the Common Terminology Criteria for Adverse Events version 5.0 (CTCAE v5). 24 The time to occurrence and management of adverse events (i.e. dose modification, hospitalization) were also recorded.
Molecular response (MR) endpoints were defined using quantitative BCR-ABL1 transcript levels. BCR-ABL1 transcript levels of ≤0.1%, ≤0.01%, ≤0.0032%, ≤0.001% on the international scale (IS) were defined MMR, MR 4.0 , MR 4.5 , and MR 5.0 respectively. 18 MR 4.0 , MR 4.5 , and MR 5.0 are classified as DMR. 18 Achievement of sDMR (DMR maintained for at least 2 consecutive years) and early molecular response (EMR; BCR-ABL1 IS ≤ 10% at 3 and 6 months) were also documented. Survival endpoints included overall survival (OS), progression-free survival (PFS), and event-free survival (EFS). 18 OS was calculated from the date of imatinib initiation until death (of any cause, whilst on imatinib or within 60 days off imatinib treatment) or the end of treatment follow-up, whichever occurred earliest. PFS was calculated from the date of imatinib initiation until disease progression to accelerated/blast phase, transformation to acute myeloid leukemia (AML), or death (of any cause), while on imatinib or within 60 days off imatinib treatment. EFS was defined as survival with the absence of disease progression, relapse, or death (of any cause), while receiving imatinib or within 60 days off imatinib treatment.

| Statistical analysis
Continuous variables are presented as mean and standard deviation (SD), or median and interquartile range (IQR), and compared between groups using the independent two-sample t-test or the nonparametric Wilcoxon-Mann-Whitney test, respectively. Categorical variables were described using frequencies and percentages and compared between groups using Pearson's chi-squared test of independence or Fisher's exact test of independence if Cochran's rule was not met.
Survival endpoints (OS, PFS, & EFS), time to first dose modification, TTD, and TTNT were evaluated using the Kaplan-Meier method, with a log-rank test comparing between-group differences. 25 Patients without an event were censored at the date of last follow-up. A Cox proportional hazards model 26 was used to assess the independent factors associated with EFS. Data are reported as hazard ratios (HRs) with associated 95% confidence intervals (CIs). A logistics regression model was used to investigate the effect of baseline variables on achievement of EMR at 3 months, with a Wald test to assess the null hypothesis of no between-group difference. 27 Odds ratios (ORs) and associated 95% CIs are reported.
The cumulative incidences of molecular response (MMR, DMR, sDMR) and imatinib-related ADRs were modeled using the cumulative incidence competing risk method, with Gray's weighted logrank test comparing between-group differences. 28 retained in the analysis, was not used due to the possibility of introducing bias and producing estimates with higher variance. 43,44 Statistical methods for data analysis are further justified and defined in the Appendix S1, Supplementary Methods. All reported p values are two-sided, and a significance level of α = .05 was used (except for selection of variables for inclusion in multivariable regression, whereby a significance level of α = .10 was used). The statistical analyses were performed using R (version 3.3.3). 45

| Nomenclature of targets and ligands
Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guide topha rmaco logy. org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY, 46 Table 1.

| Imatinib prescribing patterns
Of the 86 patients treated with imatinib, data were collected on a total of 89 imatinib treatment courses. In the majority of treatments (n = 78, 88%), imatinib was prescribed as first-line therapy for newly diagnosed CML. Baseline demographic characteristics and distribution of disease risk scores were well-balanced in patients receiving imatinib first-line and those initiating imatinib second-line or later (Table S2). Of patients receiving imatinib as first-line treatment, 49% were initiated on the standard dose of imatinib 400 mg/day and 49% on a higher dose of 600 mg/day. In patients receiving imatinib treatment second-line or later, the majority were initiated on 400 mg/ day (64%) with only 27% initiated on 600 mg/day. Very few (n = 5, 6%) received imatinib with concomitant cancer treatments (e.g. peginterferon alfa-2a, cytarabine, or hydroxyurea).

| Imatinib dose modifications and treatment discontinuations
Imatinib dose modifications were commonly observed in this realworld cohort. Within the first 12 months of imatinib treatment, 58% (95% CI, 46% to 67%) of patients required an imatinib dose modification (of any type), 44% (95% CI, 33% to 54%) a dose reduction or temporary interruption of imatinib treatment, and 34% (95% CI, 23% to 44%) a dose escalation ( Figure 1). Of patients requiring an imatinib dose reduction or temporary treatment interruption, 63% were receiving imatinib doses of 600 mg/day or greater at first dose change, whilst 35% were receiving 400 mg/day. Adverse events were the most common reason reported for imatinib dose reductions and treatment interruptions (86% of all dose reductions/interruptions). Of those requiring an imatinib dose escalation, 58% were initiated on a standard imatinib dose of 400 mg/day and 35% initiated on 600 mg/day. The most common reported reason(s) for dose escalation was poor response (54% of all dose escalations), followed by good tolerability (43%), and relapse or disease progression (8%).
Nine patients also discontinued imatinib treatment due to attainment of sDMR, of which two patients then relapsed within a median follow-up period of 9 months (IQR 6 to 22) post discontinuation.
In univariable regression analysis, the following baseline variables were associated with a higher risk of occurrence of grade ≥ 3 ADRs with imatinib treatment: imatinib starting dose, total body weight, CCI score, concomitant use of a medicine with the potential for imatinib drug-drug interactions, treatment with another antineoplastic agent for CML, and family history of cardiovascular disease (p < .10;  The cumulative incidence of MMR among evaluable patients (n = 73) was 58% (95% CI, 46% to 69%) by 2 years, whilst the cumulative incidence of DMR (n = 72 evaluable) was 42% (95% CI, 30%
In univariable regression analysis, the following baseline variables were associated with MMR achievement; ELTS score, Sokal score, BCR-ABL1 transcript type and concomitant use of medicines with potential for imatinib drug-drug interactions (

| Considering clinical trial exclusion criteria
Overall, 48 patients treated with imatinib (56%) would have been excluded from the DASISION and ENESTnd trials due to serious F I G U R E 2 Cumulative incidence of imatinib-related adverse drug reactions (ADRs) Cumulative incidence of imatinib-related ADRs by 3 years (95% confidence intervals) calculated using the cumulative incidence competing risk method.

Patients who would have been ineligible for both DASISION and
ENESTnd clinical trials (based on the exclusion criteria) were signifi-

F I G U R E 3
Cumulative incidence of major molecular response (MMR), deep molecular response (DMR) and sustained DMR (sDMR) in patients treated with imatinib. Cumulative incidence of molecular response at certain time points are presented with their associated 95% confidence intervals. Cumulative incidence was calculated using the cumulative incidence competing risk method.   Figure 5). Furthermore, the ineligible cohort had a significantly higher risk of occurrence of imatinib-related grade ≥ 3 F I G U R E 5 Cumulative incidence of (A) major molecular response (MMR), (B) deep molecular response (DMR) and (C) sustained DMR (sDMR) in patients treated with imatinib, by likely eligibility for the ENESTnd 37 Figure 6). Patients considered ineligible for clinical trial inclusion were also more likely to experience recurrent imatinib-related ADRs ( were applied to patients in our study (Figure 7).

| DISCUSS ION
This study shows a high rate of molecular response and good longterm survival with imatinib treatment for people with CML in realworld clinical practice. Survival outcomes with imatinib-treatment in this study (3- Minor differences in survival outcomes could be explained by differences in definitions and censoring between studies. Notably, major and deep molecular response rates observed in this study (3-year MMR and DMR rates of 63% and 42%, respectively) are higher than those previously reported in controlled clinical trials. [49][50][51]53,54 In the ENESTnd and DASISION trials, 3-year cumulative MMR rates of imatinib-treated patients were 53% and 55%, respectively, whilst 3-year cumulative DMR rates were 26% and 14%, respectively. 49,50 Similarly, the rate of MMR at 3-years in the TOPS trial was 52% for the imatinib 400 mg/day arm and 50% for the imatinib 800 mg/day arm, with DMR achieved in 13% of patients in both the 400 and 800 mg/day treatment groups. 51 Our findings are also consistent with a study of 208 patients treated with firstline imatinib outside clinical trials which reported estimated 7-year MMR and DMR rates of 70% and 52%, respectively. 55 Achievement of sDMR has been associated with significant improvements in long-term survival outcomes with imatinib, 3,53,56 and is considered the gateway to obtaining TFR. 57 Although elective discontinuation of imatinib due to attainment of sDMR was only considered in a small sample of patients in this real-world cohort, we observed a similar rate of remission after imatinib discontinuation to previous reports (12-month TFR rate between 41% to 68%). [58][59][60][61][62][63][64][65] This raises the possibility that, at least in some patients, CML may be cured with imatinib treatment. Imatinib discontinuation has a large economic impact, with cost analysis of the Euro-SKI (European A major finding of this study is the high incidence of ADRs resulting in imatinib discontinuation, with an 18-month probability of 15% and 3-year probability of 25%. This is notably higher than previous Although plasma concentration data were not available for this cohort, we conducted a follow-up study using physiologically based pharmacokinetic (PBPK) modeling and simulation to predict the imatinib plasma concentration-time profile of patients included in this real-world study. 83 Notably, the PBPK model showed significant correlations between predicted steady-state imatinib exposure and clinical outcomes (achievement of EMR, and the occurrence of grade ≥ 3 imatinib-related ADRs). 83 In multivariable regression analyses, lower total body weight was predictive of higher rates of occurrence of imatinib-related grade ≥ 3 ADRs. This is supported in a study by Shin 76,86,87 This study adds to the evidence that geographic ancestry is an important covariate in the inter-individual variability of imatinib treatment outcomes. 88   other studies without acknowledging differences in study design (including lack of randomization in this real-world study, and differential monitoring of outcomes and management of events) and definitions of outcome measures.
Despite these limitations, the similarity in survival results and adverse event profiles with those of controlled clinical trials provides a level of confidence in the data, with differences observed likely to reflect true differences between real-world and protocol-driven practices. Importantly, characteristics of the patients included in this study were consistent with expectations of a CML patient receiving care in the Australian oncology setting. 126 Conversely, patients included in the IRIS, ENESTnd, and DASISION controlled trials were younger than expected in a real-world setting (median age of 50 years in IRIS, 1 46 years in ENESTnd, 37 and 49 years in DASISION). 38 There are several strengths in the methodology of this study, including use of the CTCAE and Naranjo algorithm to classify ADRs, re-abstraction of data by a second investigator in 30% of randomly selected patients, use of the cumulative incidence competing risk method to evaluate molecular response rates and ADR incidences, and use of a multiple imputation method in cases of missing data. Importantly, this real-world data on 89 imatinib treatment courses represents a total of 421 patient years of experience with imatinib treatment in CML. As a real-world study, this data has presented new and important insights into prescribing practices and clinical outcomes of patients receiving imatinib treatment with complex comorbidities and on multiple medicines, without the potential selection bias present in controlled clinical trials.
In summary, this study found that imatinib induces fast and deep molecular responses that translate to good survival outcomes in a real-world setting. A higher incidence of imatinib-related ADRs were observed in this real-world cohort, compared to controlled clinical trials. Baseline evaluation of concomitant medicine use and preexisting comorbidities, together with consideration of biological and clinical factors, can help identify patients with an excellent prognosis and those who may require careful monitoring and/or intervention. Early high-doses of imatinib, followed by rapid individualized dose-adaptation to good tolerability can be a strategy to achieve a balance between efficacy and tolerability.

AUTH O R CO NTR I B UTI O N S
All authors conceived the study. J.A collected, analyzed, and interpreted the data, and wrote the manuscript. A.G, A.M, and N.W.D also contributed to the interpretation of the data and revised the manuscript. All authors approved the final manuscript.

This work was supported by the Peter Coates Postgraduate
Scholarship in Ethnopharmacology provided by GlaxoSmithKline.
The authors thank pharmacy students from the University of Sydney (Maddison Mansfield, Maisah Joarder, Samantha Brown, and Zahraa Falfaly) for their assistance with data collection, the Hematology departments at Concord Repatriation General Hospital and Royal North Shore Hospital for facilitating access to data, and Dejana Munjiza for her assistance with R programming.

This work was supported by the Peter Coates Postgraduate
Scholarship in Ethnopharmacology provided by GlaxoSmithKline.

D I SCLOS U R E
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available because of privacy or ethical restrictions.

E TH I C S A PPROVA L A N D PATI E NT CO N S E NT S TAT E M E N T
The research included in this study was approved by the Sydney