THE ANALYTICAL FRAMEWORK OF CLINICAL TRIALS EVALUATING CLINICAL OUTCOMES OF ARTIFICIAL INTELLIGENCE-BASED DIGITAL HEALTH INTERVENTIONS FOR METABOLIC DISORDERS: A SYSTEMATIC LITERATURE REVIEW
Author(s)
Filip Stanicic, PhD (c), Dimitrije Grbic, PhD (c), Vlad Zah, PhD;
ZRx Outcomes Research, Inc., Mississauga, ON, Canada
ZRx Outcomes Research, Inc., Mississauga, ON, Canada
OBJECTIVES: This systematic literature review (SLR) aimed to provide a methodological guidance for trials investigating clinical outcomes of artificial intelligence-based digital health interventions (AI-DHI) for metabolic disorders.
METHODS: The SLR was in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis guidelines. Literature was searched in PubMed, Embase, and as hand-search (Google Scholar, reference lists). Population included patients with metabolic disorders using AI-DHI involved in clinical trial investigating clinical outcomes. The National Institute for Health and Care Excellence checklist was used for quality appraisal.
RESULTS: Evidence synthesis included 24 studies overall, 16 for diabetes mellitus (DM) and 8 for obesity. Regarding DM studies, 8 included type 1, 5 included type 2, 2 studies included both types, and 1 included gestational DM. The most common designs were controlled, parallel-group, multiple-arm trials (66.7%). The most frequent comparator was the usual care. Studies were mostly conducted only in the US (37.5%), while 12.5% were multicentre from multiple countries. Most studies were open-label (67.5%), probably because of the intervention type, and performed power analysis before the recruitment (70.8%). However, some publication contained no information on blinding (29.2%) and power analysis (12.5%). Dropout rates (overall and per study arm) should be <20%, which was the case in 75.0% of included trials (12.5% did not report and 12.5% had higher dropout). Outcomes that pointed out the value of AI-DHI in DM studies were glucose level measures (68.8%) and HbA1c rates (62.5%). Body weight (87.5%) and body mass index (62.5%) were used in obesity trials. Regression models were mostly used tests for assessing statistical significance (41.7%).
CONCLUSIONS: The study provided an overview of trials exploring the clinical value of AI-DHIs in metabolic disorders. The findings will help investigators to appropriately design future studies based on the current research practice and ensure the evidence quality.
METHODS: The SLR was in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis guidelines. Literature was searched in PubMed, Embase, and as hand-search (Google Scholar, reference lists). Population included patients with metabolic disorders using AI-DHI involved in clinical trial investigating clinical outcomes. The National Institute for Health and Care Excellence checklist was used for quality appraisal.
RESULTS: Evidence synthesis included 24 studies overall, 16 for diabetes mellitus (DM) and 8 for obesity. Regarding DM studies, 8 included type 1, 5 included type 2, 2 studies included both types, and 1 included gestational DM. The most common designs were controlled, parallel-group, multiple-arm trials (66.7%). The most frequent comparator was the usual care. Studies were mostly conducted only in the US (37.5%), while 12.5% were multicentre from multiple countries. Most studies were open-label (67.5%), probably because of the intervention type, and performed power analysis before the recruitment (70.8%). However, some publication contained no information on blinding (29.2%) and power analysis (12.5%). Dropout rates (overall and per study arm) should be <20%, which was the case in 75.0% of included trials (12.5% did not report and 12.5% had higher dropout). Outcomes that pointed out the value of AI-DHI in DM studies were glucose level measures (68.8%) and HbA1c rates (62.5%). Body weight (87.5%) and body mass index (62.5%) were used in obesity trials. Regression models were mostly used tests for assessing statistical significance (41.7%).
CONCLUSIONS: The study provided an overview of trials exploring the clinical value of AI-DHIs in metabolic disorders. The findings will help investigators to appropriately design future studies based on the current research practice and ensure the evidence quality.
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
MT9
Topic
Medical Technologies
Topic Subcategory
Digital Health
Disease
SDC: Diabetes/Endocrine/Metabolic Disorders (including obesity)