OBJECTIVES: Non-adherence to prescription medication is a serious problem. Accurate assessment of adherence is important for improving patients’ health outcomes and informing decisions made by providers and payers. This study examined how adherence was defined, measured, and reported in the scientific literature. METHODS: A systematic review of studies reporting on adherence to prescription medication for chronic diseases was conducted. Embase® and MEDLINE® databases were searched. In addition, bibliographies of identified review papers were assessed for inclusion. Mean overall adherence was estimated using a random effects model. RESULTS: The review included 266 studies that met predefined inclusion criteria. The definition of non-adherence varied across studies; 73.6% of all studies defined non-adherence as missed/skipped doses only, while 24.3% defined it as discontinuation of therapy only, and 2.1% defined it as either. Adherence was recorded using different tools including claims data (55.2%), patient self-reports (30.5%), pill counts (12.8%), and laboratory tests (1.5%). Furthermore, included studies used various metrics to assess determinants of adherence, and some studies used more than one measure. The most common metrics were odds ratios (65.6%), regression coefficients (13.0%), hazard ratios (11.2%), and relative risks (5.4%). The meta-analysis showed that the mean overall adherence was highest in studies that defined non-adherence as missed/skipped doses only (51.7% of all participants). Mean adherence in studies defining non-adherence as discontinuation of therapy only and in studies defining it as missed/skipped doses or discontinuation of therapy was 45.4% and 45.0%, respectively. CONCLUSIONS: Inconsistencies in the adherence literature pose a challenge to the interpretation, usefulness, and synthesis of adherence data. There is a clear need for methodological standardisation in adherence research and assessment.