Continuous glucose monitor metrics from five studies identify participants at risk for type 1 diabetes development

Abstract

Aims/hypothesis: We aimed to assess whether continuous glucose monitor (CGM) metrics can accurately predict stage 3 type 1 diabetes diagnosis in those with islet autoantibodies (AAb). Methods: Baseline CGM data were collected from participants with ≥1 positive AAb type from five studies: ASK (n=79), BDR (n=22), DAISY (n=18), DIPP (n=8) and TrialNet Pathway to Prevention (n=91). Median follow-up time was 2.6 years (quartiles: 1.5 to 3.6 years). A participant characteristics-only model, a CGM metrics-only model and a full model combining characteristics and CGM metrics were compared. Results: The full model achieved a numerically higher performance predictor estimate (C statistic=0.74; 95% CI 0.66, 0.81) for predicting stage 3 type 1 diabetes diagnosis compared with the characteristics-only model (C statistic=0.69; 95% CI 0.60, 0.77) and the CGM-only model (C statistic=0.68; 95% CI 0.61, 0.75). Greater percentage of time >7.8 mmol/l (p<0.001), HbA1c (p=0.02), having a first-degree relative with type 1 diabetes (p=0.02) and testing positive for IA-2 AAb (p<0.001) were associated with greater risk of type 1 diabetes diagnosis. Additionally, being male (p=0.06) and having a negative GAD AAb (p=0.09) were selected but not found to be significant. Participants classified as having low (n=79), medium (n=98) or high (n=41) risk of stage 3 type 1 diabetes diagnosis using the full model had a probability of developing symptomatic disease by 2 years of 5%, 13% and 48%, respectively. Conclusions/interpretation: CGM metrics can help predict disease progression and classify an individual’s risk of type 1 diabetes diagnosis in conjunction with other factors. CGM can also be used to better assess the risk of type 1 diabetes progression and define eligibility for potential prevention trials.