Background
Insulin resistance (IR) is a physiological state in which glucose disposal is impaired and accompanied by a compensatory hyperinsulinaemia, and is a primary risk factor for the development of type 2 diabetes and non-alcoholic fatty liver disease. The euglycaemic–hyperinsulinaemic clamp (EIC) is the reference standard for the measurement of whole-body insulin sensitivity but is laborious and expensive to perform. In a study from Stanford University, multiple Olink® Target 96 panels were used to assess the value of plasma proteomic profiling and to develop signatures correlating with the M value derived from the EIC. Using two independent cohorts (RISC and ULSAM), 828 proteins were measured in the fasting plasma of 966 participants and protein profiles were evaluated in the context of regular clinical variables and M values measured by EIC.
Outcome
After adjusting for multiple confounders (such as BMI), a total of 271 and 241 proteins were significantly associated with IR M values, for RISC and ULSAM respectively. Despite the fact that the characteristics of the two cohorts were significantly different (in terms of age for example), around 50% of the markers identified in each cohort replicated between them, in both directions. Machine learning (LASSO) analysis identified several proteins common to both cohorts that have well-established links to IR (e.g. IGFBP2, LEP), as well as more novel markers (e.g. RTN4R, ADGRG1 and INHBC).
Analysis of protein signatures based on up to 67 proteins substantially improved M value estimations when added to routine clinical variables. Furthermore, a protein-only model performed as well as models that combined clinical variables and proteins together in most cases, suggesting that plasma protein signatures for IR do not only provide incremental improvement of predictive clinical variables, but could even replace them. Further analysis suggested that models composed of smaller subsets of proteins could be almost as effective as the full signature, especially for models that were applied across both cohorts. The authors concluded that plasma proteomic profiling has the potential to improve individual assessments of insulin sensitivity and provides opportunities to improve the identification of individuals at risk of insulin resistance-related adverse health consequences.
