Evaluating beneficial drug effects in a non-interventional setting: a review of effectiveness studies based on Swedish Prescribed Drug Register data

Abstract

Aims: To describe and assess current effectiveness studies published up to 2014 using Swedish Prescribed Drug Register (SPDR) data. Methods: Study characteristics were extracted. Each study was assessed concerning the clinical relevance of the research question, the risk of bias according to a structured checklist, and as to whether its findings contributed to new knowledge. The biases encountered and ways of handling these were retrieved. Results: A total of 24 effectiveness studies were included in the review, the majority on cardiovascular or psychiatric disease (n = 17; 71%). The articles linked data from four (interquartile range: three to four) registers, and were published in 21 different journals with an impact factor ranging from 1.58 to 51.66. All articles had a clinically relevant research question. According to the systematic quality assessments, the overall risk of bias was low in one (4%), moderate in eight (33%) and high in 15 (62%) studies. Overall, two (8%) studies were assessed as contributing to new knowledge. Frequently occurring problems were selection bias making the comparison groups incomparable, treatment bias with suboptimal handling of drug exposure and an intention-to-treat approach, and assessment bias including immortal time bias. Good examples of how to handle bias problems included propensity score matching and sensitivity analyses. Conclusion: Although this review illustrates that effectiveness studies based on dispensed drug register data can contribute to new evidence of intended effects of drug treatment in clinical practice, the expectations of such data to provide valuable information need to be tempered due to methodological issues.