ESPE Abstracts

Background and Aims: Hypoglycaemia is a life-threatening risk for many patients and prevention is individualised and complex. Continuous Glucose Monitoring (CGM) shows promise but current accuracy is insufficient for acute hypoglycaemia detection and data review services are complex and generic. Machine Learning is increasingly used but ignores weekly hypoglycaemia patterns and behaviour change and thus has demonstrated no real-world reduction in hypoglycaemia. Children with congenital hyperinsulinism (CHI) have recurrent hypoglycaemia and we developed HYPO-CHEAT (HYpoglycaemia-Prevention-thrOugh-Cgm-HEatmap-Assisted-Technology): a personalised technology to proactively prevent hypoglycaemia for this underserved group.

Methods: HYPO-CHEAT automatically undertakes iterative analysis of individual CGM data and generates personalised hypoglycaemia heatmaps to allow actionable visualisation of high-risk periods throughout the week. The simple to use output is accompanied by personalised targets for reflection and is specifically designed to change patient behaviours and proactively prevent hypoglycaemia (glucose <3.5mmol/l). Ten patients with CHI used a blinded CGM for four weeks to establish baseline time below range (TBR). Devices were then unblinded for four weeks, at the end of which patients and families used HYPO-CHEAT. CGM was subsequently re-blinded for four weeks.

Results: Six patients demonstrated baseline hypoglycaemia (all TBR >1.5%), while four showed negligible hypoglycaemia (mean TBR 0.2%). In those with baseline hypoglycaemia, TBR improved from a mean of 7.1% to 4.5% when devices were unblinded. Proprietary CGM software (Dexcom Clarity) failed to identify hypoglycaemia patterns in any patient. By contrast, HYPO-CHEAT identified weekly hypoglycaemia patterns in all patients. Furthermore, families all identified specific behaviours contributing to hypoglycaemia “hotspots”. After using HYPO-CHEAT, families did 70% more fingerpricks within suggested targets and reduced targeted TBR by 67%, despite the removal of real-time data through device re-blinding. By concentrating on personalised hypoglycaemia hotspots and associated behaviours, families reduced total TBR by 25% from baseline (5.4%). For those without initial hypoglycaemia and thus no use for HYPO-CHEAT, TBR increased from 0.2% at baseline to 3.2% when reblinded, strengthening the associated between HYPO-CHEAT and a reduction in TBR. Patient feedback was positive and identified the utility of HYPO-CHEAT to change fingerprick behaviour and reduce hypoglycaemia.

Conclusions: HYPO-CHEAT's automated provision of personalised, hypoglycaemia heatmaps facilitated patient reflection and behaviour change to target fingerpricks and reduce total and targeted TBR even when CGM was blinded. HYPO-CHEAT offers a highly effective and immediately available personalised technology to proactively prevent hypoglycaemia and empower self-care. As CGM is not required after the initial four-week period, burden and costs are minimised.