The individual in this case was a 79-year-old woman who presented with vomiting, increased stoma output and vertigo. She had a medical history of surgically resected bowel cancer, hypothyroidism, deep vein thrombosis treated with tinzaparin, and Type 2 diabetes mellitus managed with oral hypoglycaemic agents (metformin and linagliptin). Prior to admission she was independent in activities of daily living. A computerized tomography (CT) scan at presentation showed a large posterior fossa haemorrhage, which was managed with reversal of tinzaparin and blood pressure control. Metformin was stopped on admission as a result of dehydration from increased stoma output, and the woman received corrective doses of aspart as required to treat hyperglycaemia. Subsequently, the woman developed increased confusion and drowsiness, and repeat CT scan showed mass effect with hydrocephalus. She was transferred to the regional neurosurgical unit and an external ventricular drain inserted. Hyperglycaemia was initially managed with a variable rate intravenous insulin infusion with 2–4 units/h, but hyperglycaemia persisted (capillary blood glucose levels 12–16 mmol/l). Nasogastric feeding was commenced because of fluctuating consciousness [Glasgow Coma Scale 8–14] with Nutrison protein Plus 50 ml/h for 20 h (7.1 g carbohydrate/h), with a 4-h break, and this was subsequently uptitrated to 75 ml/h (10.7 g carbohydrate/h). During the woman's stay in the hospital, we were undertaking a multinational randomized controlled trial contrasting fully closed-loop insulin delivery with faster-acting insulin aspart (Fiasp; Novo Nordisk, Bagsværd, Denmark), or conventional subcutaneous insulin therapy over a period of up to 15 days of hospital stay. In the UK, the study protocol was approved by the local Research Ethics Committees (East of England Central Cambridge Ethics Committee) and the Medicines and Healthcare products Regulatory Agency (MHRA). Recruitment of this particular woman into the study was discussed with her family, and a consultee declaration form was signed by her next-of-kin. Although prognosis was guarded, there were no limits to treatment escalation at the time of recruitment into the study. The woman was randomized to closed-loop insulin delivery. Pre-study total daily insulin dose was 110 units. A subcutaneous cannula was inserted into the abdomen/arm for delivery of faster-acting insulin aspart by insulin pump (Dana R Diabecare; SOOIL Development, Seoul, Korea). A subcutaneous real-time continuous glucose monitor (Freestyle Navigator II; Abbott Diabetes Care, Alameda, CA, USA) was inserted into the abdomen/arm and calibrated according to manufacturer's instructions. The FlorenceD2W-T2 closed-loop system consisted of a model predictive control algorithm (version 0.3.70) residing on a control algorithm device (Dell Latitude 10 Tablet, Dell, UK) linked by a USB cable to the continuous glucose monitoring receiver (FreeStyle Navigator II). The tablet device communicated with the study pump via a Bluetooth wireless communication protocol. The control algorithm was initialized using the woman's weight and pre-study total daily insulin dose. The control algorithm did not receive announcements regarding timing or carbohydrate content of meals, enteral or parenteral feeds. No adjustments to target glucose were made during the study. The control algorithm adapted itself to a particular participant by updating model variables and refining participant's insulin requirements. The algorithm aimed to achieve glucose levels between 5.8 and 7.2 mmol/l, and adjusted the actual target level depending on accuracy of the model-based glucose predictions and prevailing glucose levels. Safety rules limited maximum insulin infusion and suspended insulin delivery at sensor glucose ≤4.2 mmol/l, or when sensor glucose was rapidly decreasing. If sensor glucose was unavailable for 30 min, the research team was alerted and the study pump insulin infusion rate reverted to the pre-programmed basal rate. During the study period, after 2 days of closed-loop insulin delivery, the woman developed hospital acquired pneumonia. Antibiotics, oxygen therapy and chest physiotherapy were commenced and closed-loop was continued, with notably higher insulin requirements. Despite optimal ward-based treatment, her clinical condition deteriorated and clinical focus switched to palliation; the nasogastric feed was stopped and, as insulin requirements were minimal without the feed, closed-loop and participation in the study were stopped at this point. A syringe driver was started for symptom control and she died several days later. The death was reported to the trial Data and Safety Monitoring Board as an adverse event unrelated to study devices and procedures. During the period of closed-loop insulin delivery, glucose control was safe with no glucose-related harm. The mean sensor glucose was 11.3 ± 4.3 mmol/l and percentage of time with sensor glucose between 6 and 15 mmol/l was 70.5%. Time spent in hypoglycaemia <3.9 mmol/l was 2.0% and time spent in significant hyperglycaemia >20 mmol/l was 2.6%.