SMS Reminders to Improve Medication Adherence of Aromatase Inhibitors

This study is a prospective randomised controlled parallel trial of SMS reminder versus standard care to investigate whether SMS reminder improves adherence to oral aromatase inhibitor (AI) therapy amongst women with breast cancer at 1-year follow-up. All subjects will be randomised to receive either SMS reminder or standard care in a 1:1 ratio. Balanced permuted block randomisation will be implemented with random varying of block sizes of 4 or 6. The sample size is estimated based on the primary endpoint of medication adherence (as measured via SMAQ) at 1-year. Assuming that the proportion of medication adherence in the intervention and control groups are 80% and 60% respectively, then based on a 5% significance level and a power of 90%, a minimum sample of 240 subjects (i.e. 120 per group) will be required. Further assuming an attrition rate of 10%, the anticipated trial size is 280 overall. The assessment of medication adherence at 1-year (as measured via SMAQ) between the SMS reminder and control groups will be made using chi-square test. Adjustment for baseline covariate and potential confounders will be made using the logistic regression where appropriate. Other measures of adherence such as the 1-year MPR, PDC and pill count will be compared between treatment arms via the t- test, with adjustment for confounders made via the analysis of covariance (ANCOVA). Natural log transformation will be implemented on the estrone and androstenedione measures to normalise the data. The assessment of secondary outcomes involving the log transformed estrone and androstenedione levels will be made using the analysis of covariance (ANCOVA) to adjust for the respective baseline levels and other potential confounders. Comparison of estradiol levels (defined as \<18.4 versus ≥18.4) between the two arms will be made using chi-square test. The logistic regression analysis will be implemented to adjust for baseline covariate and other potential confounders where appropriate. The assessment of knowledge, attitude, behavior, as well as barrier and facilitating factors between the intervention and control groups will be made using chi-square test, with treatment effect quantified based on odds ratio estimate and its 95% CI. Further adjustment for baseline scores and other potential confounders will be made using the logistic regression where appropriate. Additionally, the investigators will compare the 4 different measures of adherence in a separate paper, to determine which of these provide a more meaningful or reliable information in the local context based on sensitivity analysis. All analyses will be performed according to intention-to-treat using STATA, assuming a two-sided test at the 5% level of significance.