REVIEW Case-crossover study design in pharmacoepidemiology: systematic review and recommendations Giulia P. Consiglio 1 , Andrea M. Burden 1 , Malcolm Maclure 2 , Lisa McCarthy 1,3 and Suzanne M. Cadarette 1 * 1 Leslie Dan Faculty of Pharmacy, University of Toronto, Toronto Ontario, Canada 2 Department of Anesthesiology, Pharmacology and Therapeutics, University of British Columbia, Vancouver, British Columbia, Canada 3 Women’s College Research Institute, Toronto Ontario, Canada ABSTRACT Purpose The purpose of this study is to systematically identify and review articles that use the case-crossover study design in the area of pharmacoepidemiology. Methods A systematic search of MEDLINE® (Ovid Technologies, New York City, NY, USA), EMBASE® (Elsevier Inc., Philadelphia, PA, USA), and Web of Science® (Thomson Reuters, New York City, NY, USA) was completed to identify all English language articles that applied the case-crossover study design in the area of pharmacoepidemiology. The number of reviews, methodological contributions, and empirical pharmacoepidemiologic applications were summarized by publication year. Empirical applications were retrieved, and methodological details (outcome, exposure, exposure windows, sensitivity analysis, statistical reporting) were tabulated and compared to methodological recommendations based on exposure characteristics, exposure windows, and discordant pairs data display. Results Of 836 unique articles identified, 99 pharmacoepidemiologic studies were eligible: 20 methodological contributions, 9 review papers, and 70 empirical applications. Only three empirical applications in the area of pharmacoepidemiology were published before 2000. Since 2000, the number of empirical pharmacoepidemiologic applications published annually has generally increased over time, to before a high of 15 published in 2011. The design was mainly applied to examine drug safety (96%), and most applications investigated: psychotropic (24%) and analgesic (17%) exposure drug classes; and considered hospitalization (23%) and cardiovascular/cerebrovascular (21%) events. Only 31% of applications displayed sufficient data to enable readers to confirm odds ratios presented. Conclusions Use of the case-crossover design in pharmacoepidemiology has increased rapidly in the last decade. As the application of the case-crossover design continues to increase, it is important to develop standards of practice, especially for display of data. Copyright © 2013 John Wiley & Sons, Ltd. key words—case-crossover; case-only; pharmacoepidemiology; review literature as topic Received 18 December 2012; Revised 20 June 2013; Accepted 29 July 2013 INTRODUCTION Two main study designs dominate when identifying risk factors for disease: the cohort study and case-control study. 1–3 Self-controlled, case-only designs are increas- ingly recognized as complementary, and in some cases, advantageous over cohort and case-control designs. 1–7 In particular, case-only designs are able to control for time-invariant factors not typically recorded in healthcare databases, such as long-term frailty, stable disease severity, regular over-the-counter drug and supplement use, physical activity, diet, smoking, and alcohol consumption. The case-crossover study design is a case-only design that is ideal to control for time- invariant confounding when the exposure effects are transient (one-time or short-term) and the outcome’s onset is abrupt and clearly defined. 4,5,8 This design compares exposure status immediately before the out- come (hazard period) with exposure in designated con- trol period(s) earlier in time. An important assumption of the case-crossover design is that transient exposures have stable prevalence over time, thus there is equal opportunity to be exposed and unexposed during the hazard and control periods. If time trends in the expo- sure exist, such as when a new drug enters the market, the case-crossover design will result in biased esti- mates. 8,9 We sought to systematically review the use of the case-crossover design in pharmacoepidemiology and examine how applied case-crossover studies align with current methodological recommendations. 4–6,8 *Correspondence to: S. M. Cadarette, Leslie Dan Faculty of Pharmacy, University of Toronto, Toronto, Ontario, Canada. E-mail: s.cadarette@utoronto.ca Copyright © 2013 John Wiley & Sons, Ltd. pharmacoepidemiology and drug safety 2013; 22: 1146–1153 Published online 13 September 2013 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/pds.3508