(Elsevier, 2022) Galbete Jiménez, Arkaitz; Tamayo Rodríguez, Ibai; Librero, Julián; Enguita Germán, Mónica; Cambra Contin, Koldo; Ibáñez Beroiz, Berta; Ciencias de la Salud; Osasun Zientziak; Estadística, Informática y Matemáticas; Estatistika, Informatika eta Matematika
Aims: to identify all cardiovascular disease risk prediction models developed in patients with type 2 diabetes or in the general population with diabetes as a covariate updating previous studies, describing model performance and analysing both their risk of bias and their applicability. Methods: a systematic search for predictive models of cardiovascular risk was performed in PubMed. The CHARMS and PROBAST guidelines for data extraction and for the assessment of risk of bias and applicability were followed. Google Scholar citations of the selected articles were reviewed to identify studies that conducted external validations. Results: the titles of 10,556 references were extracted to ultimately identify 19 studies with models developed in a population with diabetes and 46 studies in the general population. Within models developed in a population with diabetes, only six were classified as having a low risk of bias, 17 had a favourable assessment of applicability, 11 reported complete model information, and also 11 were externally validated. Conclusions: there exists an overabundance of cardiovascular risk prediction models applicable to patients with diabetes, but many have a high risk of bias due to methodological shortcomings and independent validations are scarce. We recommend following the existing guidelines to facilitate their applicability.
