We previously showed in a multicenter, outpatient, random-order crossover trial (n=39, 11 days each arm) that the bionic pancreas (BP) reduced group mean CGM glucose (141±10 vs. 162±29 mg/dl, p<0.0001) and percent of time spent <60 mg/dl (0.6±0.6% vs. 1.9±1.7%, p<0.0001) relative to usual care (UC) with pump ± CGM (El-Khatib FH, et al. Lancet. 2017; 389:369-80). We applied a new statistical approach to these data to determine how many individuals saw benefit for each outcome. We compared the average CGM glucose and average time <60 mg/dl between the BP and UC arms for each patient using autoregressive time series models to determine significance of differences. We found that 72% of subjects had a statistically significant reduction in mean CGM glucose, 51% had significant reduction in time <60 mg/dl, 44% had a significant reduction in both, and 97% had a significant reduction in at least one outcome. For every subject with mean CGM glucose that was nominally higher in the BP arm, the mean glucose in the BP arm was <154 mg/dl (predicted A1c <7%) and the average time <60 mg/dl was less than in the UC arm (significantly less in 2/6 cases). These analyses provide a new prospective on the efficacy of the BP emphasizing the benefit to individuals rather than the group as a whole. Subjects least likely to benefit were those who had extremely tight glycemic control and minimal hypoglycemia under UC.

J. Sloane: None.

Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at http://www.diabetesjournals.org/content/license.