BACKGROUND: In clinical trials, patients have expressed greater satisfaction with inhaled human insulin (EXUBERA, Pfizer) than with injectable insulin. No studies to date have attempted to quantify the strength of preferences for these alternative routes of administration. OBJECTIVE: To elicit health state preference values from people with diabetes mellitus for treatment with inhaled human insulin compared with injectable insulin. STUDY DESIGN: A patient preference study. METHODS: Written descriptions were developed for five clinical scenarios: two for type 1 diabetes and three for type 2 diabetes. Each scenario required adjustment or initiation of insulin treatment because of poor glycaemic control. Two alternative insulin regimens were described for each scenario: injectable-only or inhaled human insulin to replace or reduce the number of daily injections. Equal efficacy was assumed within each of these scenario pairs.A total of 344 UK adults (66% male), 132 (mean age 49 years) with type 1 diabetes and 212 (mean age 63 years) with type 2 diabetes, rated scenario pairs corresponding to their own type of diabetes and rated their own health by time trade-off (TTO), by correspondence with EQ-5D health descriptions and on the EQ-5D visual analogue scale. Respondents stated their preference for, or indifference between, the injection-only or inhalation variant comprising each scenario pair. TTO utilities and EQ-5D utilities by UK community tariff were compared within each scenario pair, for the total sample rating, each scenario pair, and by subgroups of stated preference for each variant. RESULTS: A majority, ranging from 63% to 81% across the scenarios, preferred inhalation. Mean differences in TTO scores were 0.074, 0.076, 0.088, 0.053 and 0.043 for the five scenarios, respectively (p < 0.005 for all). Mean EQ-5D differences were 0.043, 0.029, 0.037, 0.020 and 0.021 for the five scenarios, respectively (p < 0.05 for scenarios 1 and 3), driven mainly by differences on the pain/discomfort dimension of the EQ-5D. Differences in favour of inhalation among those preferring inhalation, were greater than differences in favour of injections among those preferring injections. Mean self-rated health was similar between respondents with type 1 and type 2 diabetes, at 0.83 (TTO) and 0.75 (EQ-5D). The TTO was more sensitive than EQ-5D. Self-rated health by EQ-5D compared closely with reported values from the UK Prospective Diabetes Study (UKPDS). CONCLUSIONS: This study highlights the utility differences that people with diabetes perceive between the prospect of inhaled and injected routes of insulin administration, even under the assumption of no difference in efficacy. These differences are magnified when the comparison in utility scores is between the majority who prefer the inhaled route and the minority who prefer the injectable route.