Chronic myeloid leukemia (CML) is a type of cancer formed from specific cells in the bone-marrow. CML develops in the hematopoetic system through the formation of the BCR-ABL fusion gene, which promotes the unregulated growth of CML cells via novel chimeric tyrosine kinase. This disease typically occurs in older adults (median age of onset is 65 in the US) and can manifest in one of three stages of increasing severity: chronic, accelerated, and blast. Continuous adherence (defined as the degree to which patient behavior reflects healthcare provide recommendations) to tyrosine kinase inhibitors (TKI) is crucial to properly manage CML and can add years to a patient’s life. This is evidenced by the increase in the 10-year survival rate of CML patients following TKI introduction, from 32% in the 1990s to 70% by 2010. Although consistent use of TKIs is key to achieving disease remission, studies have reported that an estimated 30% of CML patients are non-adherent.
To better understand which CML patients may be most likely to become non-adherent and when, the Bansal Group from the Division of Public Health Sciences used pharmacy claims and sociodemographic data from the Truven Health MarketScan Commercial and Medicare Supplemental Databases (MS) to identify newly diagnosed CML patients from 2007 through 2017. Proportion of days covered (PDC) was used to calculate 30-day adherence over a 12-month period for each of the 2,049 patients in the final sample and served as the longitudinal measure of adherence in the study. Because TKIs represent some of the most expensive chemotherapies available, the cost of these medications has been identified as a key driver of non-adherence. The authors incorporated a measure of this financial burden into their analysis by linking monthly medication costs with monthly income estimates were from the 2017-2018 Area Health Resource File.
Researchers used latent profile analysis (LPA) to identify distinct classes of CML patients based on fluctuations in their TKI adherence over time. The seven adherence classes derived from the LPA model (see below) were collapsed into four more clinically relevant categories of: never adherent (group 1), initially non-adherent becoming adherent (group 2), initially adherent becoming non-adherent (group 3), and stable adherent (groups 4, 5, 6, and 7). Multinomial logistic regression was then used to evaluate the relationship between membership in the broader adherence groups and demographic, economic, and clinical characteristics. Clark et al. were the first researchers to apply LPA methods to evaluate TKI adherence in CML patients. This study was published in the Journal of Oncology Pharmacy Practice.
The majority of patients (74%) fell into the stable adherent group. This group exhibited higher rates of overall medication use and lower financial burden associated with TKI prescription cost. Those in the never adherent group were the youngest (44.7 years; SD: 11.8), had the lowest monthly income ($5,128; SD: $1,231), fewest concomitant medications (4.62; SD: 3.64), and highest proportion of medication switchers (28%). Results from the multinomial logistic regression indicate that, relative to individuals in the stable adherent group, patients were significantly more likely to be classified as never adherent if they were younger, female, had an earlier year of diagnosis, had a longer duration from CML diagnosis to TKI prescription, were taking a non-imatinib TKI, and were prescribed fewer concomitant medications. Patients had a significantly greater chance of being considered initially non-adherent becoming adherent if they were younger, female, and had a longer duration from diagnosis to TKI prescription. Assignment to the initially adherent becoming non-adherent group was associated with being younger, female, having a higher monthly financial burden, having a high starting daily TKI dose, and having a lower Charlson comorbidity score.
In summary, there are multiple distinct patterns of TKI adherence in CML patients and sociodemographic and clinical characteristics may be used to predict membership in clinically meaningful adherence categories. This study supports the use of dynamic adherence measures such as LPA since relying solely on static measures such as average 12-month PDC can cause providers to miss important variation and differences within patient groups over time. Samantha Clark, the first author, concluded, "Our study provides important insight into determinants of TKI adherence in CML patients and extends current knowledge of how patient medication-taking behavior changes over time. Through the use of latent profile analysis, we were able to identify groups of non-adherent individuals who could benefit from targeted interventions and who would not have been identified using conventional measures of adherence.” Overall, this study has important implications for clinical practice and policy.
This research was supported by the National Cancer Institute and the National Institutes of Health.
Fred Hutch/UW Cancer Consortium members Jerald Radich and Aastha Bansal contributed to this work. Dr. Bansal is also a member of the Hutchinson Institute for Cancer Outcomes Research.
Clark SE, Marcum ZA, Radich JP, Bansal A. Predictors of tyrosine kinase inhibitor adherence trajectories in patients with newly diagnosed chronic myeloid leukemia. Journal of Oncology Pharmacy Practice. 2020 Nov 11: