“…A total of 116 RCTs were selected by title or abstract by the first author. After reading the text, 38 articles were excluded because of follow up less than 2 months [5–7], pharmacokinetic/dynamic studies [8–12], too few patients [13–19], only hospitalized participants [20,21], an extension or a substudy of an RCT with no relevance to the topic of the review [22,23], being a substudy of an included RCT [24–26], oral agents added to insulin therapy [27,28], no RCT after all [29–31], only describing within‐group and not between‐group outcomes [32], pooled data of two trials [33], underpowered [34], comparing different time schedules [35], comparing patient‐ or physician‐driven regimens [36–38] and cost‐effectiveness analysis [39], comparison with troglitazone [40,41]. One trial did not include hypoglycaemia as an outcome measurement [42].…”