Abstract Aim To compare the efficacy of the closed‐loop Diabeloop for highly unstable diabetes (DBLHU) system with the open‐loop predictive low glucose suspend (PLGS) system in patients with highly unstable type 1 diabetes (T1D) who experience acute metabolic events. Methods DBLHU‐WP10 was an interventional, controlled, randomized, open‐label study that comprised two cycles of N‐of‐1 trials (2‐of‐1 trials). Each trial consisted of two crossover 4‐week periods of treatment with either DBLHU or PLGS in randomized order. The primary outcome was the percentage of time spent in the 70‐180 mg/dL glucose range (time in range [TIR]). Results Five out of seven randomized patients completed the aggregated 2‐of‐1 trials. TIR was significantly higher with DBLHU (73.3% ± 1.7%) compared with PLGS (43.5% ± 1.7%; P < .0001). The percentage of time below 70 mg/dL was significantly lower with DBLHU (0.9% ± 0.4%) versus PLGS (3.7% ± 0.4%; P < .0001). DBLHU was also significantly superior to PLGS in reducing hyperglycaemic excursions and improving almost all other secondary outcomes, including glucose variability and satisfaction score. No adverse event could be related to the experimental treatment. Conclusions DBLHU was superior to PLGS in improving the metabolic control of patients with highly unstable T1D who require an islet or pancreas transplant but who either have a contraindication or refuse to consent.