Canonical Correlation Analysis (CCA) of Imatinib Treatment (ImRx) Nonadherence (NA) with Associated Patient Variables (APVs) in Chronic Myeloid Leukemia (CML) - Results from the ADAGIO Study.

BACKGROUND. ImRx for CML is a long-term treatment potentially compromised by NA. Identifying APVs may assist in reducing NA and optimizing treatment outcomes.

OBJECTIVE. To model the relationships between two NA measures and selected APVs using CCA, a multivariate analog of multiple regression to accommodate multiple criterion variables.

DESIGN AND PATIENTS. Data subset from prospective, 90d observational, open-label, multicenter study. 169 evaluable pts on ImRx for minimum 30d at enrollment [1].

MEASUREMENTS. NA at 90d vector: Basel Assessment of Adherence Scale for pts (pBAAS; 0/1 with 1=NA); and % of ImRx taken per pill count (%ImRx, subtracted from 100 to reflect NA). APVs at BL vector: age; months since CML diagnosis (mCML); months since ImRx initiation (mImRx); knowledge of CML disease, treatment, and ImRx (KCMLRx); and general health (SF-8).

RESULTS. The criterion (dependent) vector of NA indicators included pBAAS and %ImRX. The predictor (independent) vector of APVs included: age, mCML, mImRx, KCMLRx, and SF-8. Two canonical correlations were generated: 0.389 (Bartlett Chi-squared=23.564, P=0.009) and 0.170 (Bartlett Chi-squared= 3.590, P=0.464); the second correlation was deleted due to nonsignificance from zero. The canonical loadings (or structure coefficients) for the retained model were: age 0.951, mCML 0.205, mImRx 0.145, SF-8 0.016, and KCMLRx -0.367. Redundancy analysis showed that 22.1% of variance in the predictor set was explained by variables within that set.

CONCLUSIONS. The patient NA vector (composed of a binary assessment of NA per pBAAS 0/1 with 1=NA) and continuous quantification of % of ImRx not taken was related to the APV vector as follows: NA increased as patients were older, had been diagnosed with CML for a longer period of time, had been on imatinib treatment for a longer period time, and were in slightly better health at enrollment. These may be considered warning signs for NA for clinicians to consider in practice, given the long-term nature of ImRx. Importantly, better patient knowledge of disease and treatment, a clinically modifiable APV, was associated with a decrease in NA. The initial insights in patient NA with ImRX provided by these findings, though some are counterintuitive to the NA literature at large, must be studied further to better understand the dynamics of NA in the CML population.