Data-driven metabolic subtypes predict future adverse events in individuals with type 1 diabetes

Abstract

Data-driven multivariable metabolic subtypes predicted the divergence of complication burden across multiple clinical endpoints simultaneously. In particular, men with the metabolic syndrome and women with high HDL-cholesterol should be recognised as important subgroups in interventional studies and public health guidelines on type 1 diabetes.

Commentary: Data-driven subgrouping in epidemiology and medicine

“…Groop et al 3 and suggested some correspondence between the independently specified data-driven clusters even though the data used were fundamentally different, i.e., genetic loci 2 versus clinical and biomarker data at the time of diabetes diagnosis. 3 The examples of SOM analyses demonstrated by Mäkinen and co-workers, 5,6,8 as well as the recent other clustering applications in diabetes; 2,3 suggest alluring potential for data-driven subgroup analyses in epidemiology and medicine. These types of approaches are likely to inform on the fundamental genetic and metabolic variation defining the complexity of polygenic diseases.…”

“…1 Subgrouping would also be pertinent for scientific applications, for example, finding genetically and metabolically distinct individuals and patients, 2,3 complex disease subtyping 4 and disentangling subgroup specific risk factors and risk assessment. 5,6 In all the abovementioned applications the statistical approach on subgrouping shares similar characteristics. The amount of data available is huge, e.g., due to increasing use of electronic health care registries and large collections of multiple omics data combined with extensive clinical information in large biobanks.…”

“…In fact, Mäkinen et al have recently also presented a reassuring application of SOM analysis in the area of type 1 diabetes. 6 In this new work they revisited their own data and SOM analyses they published earlier in 2008. 5 In their new paper 6 they analysed new clinical endpoints and looked at the predictive characteristics of the earlier derived 5 six data-driven metabolic subtypes after seven years of additional follow-up.…”

“…They successfully replicated an epidemiological multivariable model, based on the SOM algorithm, across two time periods by using pre-defined data-driven baseline subtypes with new prospective data. 6 This study focused on epidemiological aspect of data-driven subtyping and thus the authors strongly cautioned translational implications. Notably, regardless of statistical methods, prediction of vascular endpoints at an individual level from general biochemical data is not possible.…”

“…7 Nevertheless, there is considerable overall value in applying the SOM algorithm in largescale population cohorts to reveal subgroups that may not be covered by traditional approaches or accurately addressed by current treatment guidelines. Mäkinen et al 6 demonstrated this for patients with type 1 diabetes: baseline metabolic subtypes, derived from an array of quantitative biomarkers, contained a wealth of diagnostic and prognostic information of potential value for public health interventions, e.g., informing on a subset of women with high high-density lipoprotein cholesterol who would still be at high risk for cardiovascular endpoints.…”