BACKGROUND: Whilst estimates of sub-optimal adherence to oral corticosteroids in asthma range from 30 to 50%, no ideal method for measurement exists; the impact of poor adherence in severe asthma is likely to be particularly high. RESEARCH QUESTIONS: 1. What is the prevalence of suboptimal adherence detected using self-reporting and direct measures? 2. Is suboptimal adherence associated with disease activity? STUDY DESIGN AND METHODS: Data were included from individuals with severe asthma taking part in the U-BIOPRED study prescribed daily oral corticosteroids. Participants completed the MARS, a five-item questionnaire used to grade adherence on a scale from 1 to 5, and provided a urine sample for analysis of prednisolone and metabolites by liquid-chromatography mass spectrometry. RESULTS: Data from 166 participants were included in this study, mean (SD) age 54.2 (11.9) years, FEV1 65.1 (20.5) % predicted, 58% female. 37% completing the MARS reported sub-optimal adherence, and 43% with urinary corticosteroid data did not have detectable prednisolone or metabolites in their urine. Good adherence by both methods was detected in 35% participants who had both performed; adherence detection did not match between methods in 53%. Self-reported high-adherers had better asthma control and quality of life, whereas directly-measured high-adherers had lower blood eosinophils. INTERPRETATION: Low adherence is a common problem in severe asthma, whether measured directly or self-reported. We report poor agreement between the two methods suggesting some disassociation between self-assessment of medication adherence and regular oral corticosteroid use, which suggests that each approach may provide complementary information in clinical practice.