Abstract
Background
Many prognostic models exist for the prediction of type 2 diabetes mellitus. Some of these include genetic risk factors. The objective of this review was to investigate whether including genetic variables in prediction models for type 2 diabetes mellitus is clinically useful.
Methods
Studies were included if they reported on prognostic prediction models for type 2 diabetes mellitus. Seven medical and bioengineering databases were searched using terms regarding risk prediction and diabetes combined via Boolean operators. In studies including genetic variables, the c-statistics of genetic and non-genetic models were compared.
Results
Seventy-six studies published between 2002 and 2019 were included in the review. Twenty of these (published 2008 to 2019) included genetic variables, namely deleterious alleles and specific small nuclear polymorphisms. Study samples represented the general population. When comparing genetic to non-genetic models, one study reported a statistically significantly greater c-statistic for the genetic and four for the non-genetic models. Adding genetic risk factors to a clinical model did not substantially increase the predictive accuracy of any study.
Conclusions
The use of genetic data did not show any meaningful improvements in the predictive performance for the general population compared to clinical models. Studies using genetic data were often based on small sample sizes and not externally validated, raising concerns regarding potential overfitting and lack of generalisability.
Key messages
Currently, the clinical usefulness of genetic risk scores for type 2 diabetes mellitus seems quite limited.
Genetic risk factors for type 2 diabetes mellitus and response to sulfonylurea treatment
