Clinical investigation of a personalized decision support system for insulin injections in adults with Type 1 diabetes

Type 1 diabetes is a chronic condition resulting from the immune-mediated destruction of insulin-producing pancreatic beta cells. Consequently, lifelong insulin replacement therapy is required to manage the disease via multiple daily injections, most commonly using insulin pens, or continuous subcutaneous insulin infusion with an insulin pump. The recent advent of continuous glucose monitoring with glucose sensors has augmented both insulin delivery methods, transforming the standard of care for type 1 diabetes. Despite this, attaining optimal glycemic targets is still challenging for most people and carries the risk of long-term complications.In 2024, we have a variety of technological innovations that are commercially available, ranging from advanced glucose sensors to hybrid closed-loop systems to connected insulin pens. Since their introduction, insulin pens have dominated the global market and have gradually evolved over time. However, it was not until recently that smart pens and attachments began integrating continuous glucose monitoring coupled to digital platforms for combined real-time tracking. Yet, these devices still lack an adaptive decision component for unsupervised use. The concept of personalized decision support systems is an emerging avenue marked by a growing interest in addressing this unmet need for individuals with type 1 diabetes using multiple daily injections. While a limited number of systems were investigated in large clinical trials, none have demonstrated glycemic improvement to date. Nevertheless, an effective automated approach could offer value to this underserved population, given the infrequent clinical monitoring in practice despite evolving insulin needs, partly due to restricted resources.The core objectives of my thesis were to investigate the clinical outcomes and practical use of the McGill decision support system, integrating a novel optimization algorithm designed to titrate insulin injection parameters, in hopes to bridge the gap. My primary work involved conducting a 12-week randomized controlled trial in 84 adults using multiple daily injections with type 1 diabetes and suboptimal glycemic control. This trial aimed to assess the effectiveness of the McGill decision support system in improving glycemia compared to a smartphone application with a non-adaptive insulin dose calculator. The primary outcome demonstrated a statistically significant and clinically meaningful improvement in glycated hemoglobin levels (gold standard assessment of glycemic control) with the system compared to the standalone application.Notably, this trial is the first to demonstrate glycemic improvement with algorithm-guided insulin adjustments in adults on multiple daily injections. It is also the first to include a mixed methods approach, encompassing qualitative outcomes that shed light on unique patient perspectives regarding the use of this system.The second part of my thesis entailed a three-part sub-study to evaluate the practical utility of this algorithm. This was accomplished through non-inferiority comparisons of weekly (Part A) and biweekly (Part C) adjustments made by the algorithm, benchmarked against those made by various endocrinologists. A novel assessment of intra-physician variability compared each endocrinologist’s adjustments made in Part A to those made 12 weeks later (Part B), using the same dataset. The main findings revealed comparable proportions of full agreement and full disagreement in the direction of insulin dose adjustments made by the algorithm to those made by endocrinologists. Interestingly, on average, physicians only fully agreed with themselves on the direction of insulin change about two-thirds of the time. Furthermore, the same physician even occasionally disagreed with themselves, reinforcing the subjective and complex nature of human decision making. Moreover, the average absolute percentage of change made by physicians was higher than that of the algorithm, underscoring the algorithm’s conservative approach. Overall, this study highlights the algorithm’s potential utility in practice while also conceivably alleviating concerns about inadequate medical oversight. Collectively, my thesis work demonstrated the clinical effectiveness and practical utility of the McGill decision support system, paving the way for clinical translation

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.