Do diabetes remission definitions make sense? evidence from clinical records and epidemiological studies