Is the MARS questionnaire a reliable measure of medication adherence in childhood asthma?

Objective: To assess the reliability of the Medication Adherence Report Scale (MARS-5) for assessing adherence in clinical practice and research. Methods: Prospective cohort study following electronically measured inhaled corticosteroids (ICS) adherence for 1year in 2-13-year-old children with persistent asthma. The relationship between electronically measured adherence and MARS-5 scores (ranging from 5 to 25) was assessed by Spearman's rank correlation coefficient. A ROC (receiver operating characteristic) curve was performed testing MARS-5 against electronically measured adherence. Sensitivity, specificity, positive and negative likelihood ratios of the closest MARS-5 cut-off values to the top left-hand corner of the ROC curve were calculated. Results: High MARS scores were obtained (median 24, interquartile range 22-24). Despite a statistically significant correlation between MARS-5 and electronically assessed adherence (Spearman's rho = 0.47; p < 0.0001), there was considerable variation of adherence rates at every MARS-5 score. The area under the ROC curve was 0.7188. A MARS-5 score 23 had the best predictive ability for electronically assessed adherence, but positive and negative likelihood ratios were too small to be useful (1.65 and 0.27, respectively). Conclusions: Self-report using MARS-5 is too inaccurate to be a useful measure of adherence in children with asthma, both in clinical practice and in research.