Application of machine learning to assess interindividual variability in rapid-acting insulin responses following subcutaneous injection in people with type 1 diabetes.

Introduction Circulating insulin concentrations mediate vascular-inflammatory and prothrombotic factors. However, whether interindividual differences in circulating insulin levels are associated with different inflammatory and prothrombotic profiles in type 1 diabetes (T1D) is unknown. We applied an unsupervised, machine-learning approach to assess whether interindividual differences in rapid-acting insulin levels associate with parameters of vascular health in T1D patients. Methods We reanalysed baseline pre-treatment meal-tolerance test data from two randomised control trials in which 32 patients consumed a mixed-macronutrient meal and self-administered a single dose of rapid-acting insulin individualised by carbohydrate-counting. Postprandial serum insulin, tumour necrosis factor alpha (TNFα), plasma fibrinogen, human tissue factor (HTF activity) and plasminogen-activator inhibitor-1 (PAI-1) were measured. Two-step clustering categorised individuals based on shared clinical characteristics. For analyses, insulin pharmacokinetic summary statistics were normalised, allowing standardised intra-individual comparisons. Results Despite standardisation of insulin dose, individuals exhibited marked interpersonal variability in peak insulin concentrations (48.63%), time to peak (64.95%), and insulin incremental area under the curve (60.34%). Two clusters were computed: cluster 1 (n=14) representing increased serum insulin concentrations; cluster 2 (n=18) representing reduced serum insulin concentrations (cluster 1: 389.50±177.10 vs. cluster 2: 164.29±;41.91 pmol/L.IU.hr-1; P<0.001). Cluster 2 was characterised by increased fibrinogen, PAI-1, TNFα levels, higher HTF activity, higher HbA1c, BMI, and lower eGDR (increased insulin resistance), older age, and longer diabetes duration (P