As more studies report on patient preferences, techniques are needed to identify, assess and, eventually, synthesize results from a diverse set of methodologies. Data on patient preferences are valuable to decision makers in a variety of ways. Preferences for outcomes can be used to inform decision and cost-effectiveness models, while preferences for treatments can inform patient-centered outcomes research (PCOR) and patient-centered care. This project sought to identify and assess the literature reporting on the treatment preferences of adult patients with type 2 diabetes. In addition to cataloging the preference elicitation methods used, we developed and assessed a novel quality assessment checklist for preference-based studies. PubMed, EMBASE, CINAHL, and EconLit databases were searched to identify studies examining patient preferences for medications for type 2 diabetes studies published since inception of each database. The review protocol specified inclusion of studies reporting diabetes-treatment preferences among adults with type 2 diabetes, using a range of preference measurement methods. Studies were excluded if participants were not patients with type 2 diabetes and if treatments were not pharmacological therapies targeting glycemic control, or if no primary preference information was collected. Two investigators independently reviewed titles, abstracts, and articles sequentially to select studies for data abstraction based on the inclusion and exclusion criteria. Disagreements were resolved by consensus. Data on study country, year, number of respondents, preference elicitation method, number of attributes, subgroup analyses, and funding source were abstracted into standardized tables. A novel checklist (PREFS) was used to assess the data quality and validity across different types of preference studies by assessing the following: purpose of the study; respondent sampling; explanation of preference assessment methods; findings reported for total sample; and significance testing. Each item was scored, and an aggregate score was then calculated (ranging from 0 to 5). Of the 2,100 unique citations, 61 met the inclusion criteria. The studies used conjoint analysis (n = 10), time trade-off (n = 6), standard gamble (n = 2), contingent valuation (n = 1), other stated preference methods (n = 39), and revealed preferences (n = 5). Sample sizes ranged from 27 to 14,033, with an average of 562 respondents, and two-thirds included a subgroup analysis. Most studies were conducted in one country, predominantly the USA (n = 27), UK (n = 14), Canada (n = 10), and Germany (n = 7), while 14 were conducted in multiple (2-18) countries across two or more countries. There was an increase in the annual rate of studies published over time from the time of the first publication in 1985 (p = < 0.01). Most (n = 52) studies were funded by pharmaceutical or device companies, with government, academic, association, and hospital sources also funding studies. One study met all five of the PREFS criteria and 12 met four; yet four studies met none of the criteria. The average was 27. Currently, preferences reviews are limited by the mixed quality in the reporting of studies, the publication bias inherent in the literature, a lack of guidelines to conduct various methods, and the difficulty of synthesizing results from different studies. Our study is also limited by its focus on English language articles. This study provides the first systematic evaluation of the methods used in the broad existing body of research into patient preferences for type 2 diabetes medications and can serve as a primary source of information for decision makers. Future work is necessary to assess the utility of the results of reviews of preference information and to develop best-practice guidelines for the reporting of, and methods of conducting, preference studies and systematic reviews of such studies. This systematic review was registered with PROSPERO (registration number CRD42012002285).
