Cargando…

Cluster Individuals Based on Phenotype and Determine the Risk for Atrial Fibrillation in the PREVEND and Framingham Heart Study Populations

BACKGROUND: Risk prediction of atrial fibrillation (AF) is of importance to improve the early diagnosis and treatment of AF. Latent class analysis takes into account the possible existence of classes of individuals each with shared risk factors, and maybe a better method of incorporating the phenoty...

Descripción completa

Detalles Bibliográficos
Autores principales: Rienstra, Michiel, Geelhoed, Bastiaan, Yin, Xiaoyan, Siland, Joylene E., Vermond, Rob A., Mulder, Bart A., Van Der Harst, Pim, Hillege, Hans L., Benjamin, Emelia J., Van Gelder, Isabelle C.
Formato: Online Artículo Texto
Lenguaje:English
Publicado: Public Library of Science 2016
Materias:
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5104331/
https://www.ncbi.nlm.nih.gov/pubmed/27832125
http://dx.doi.org/10.1371/journal.pone.0165828
Descripción
Sumario:BACKGROUND: Risk prediction of atrial fibrillation (AF) is of importance to improve the early diagnosis and treatment of AF. Latent class analysis takes into account the possible existence of classes of individuals each with shared risk factors, and maybe a better method of incorporating the phenotypic heterogeneity underlying AF. METHODS AND FINDINGS: Two prospective community-based cohort studies from Netherlands and United States were used. Prevention of Renal and Vascular End-stage Disease (PREVEND) study, started in 1997, and the Framingham Heart Study (FHS) Offspring cohort started in 1971, both with 10-years follow-up. The main objective was to determine the risk of AF using a latent class analysis, and compare the discrimination and reclassification performance with traditional regression analysis. Mean age in PREVEND was 49±13 years, 49.8% were men. During follow-up, 250(3%) individuals developed AF. We built a latent class model based on 18 risk factors. A model with 7 distinct classes (ranging from 341 to 1517 individuals) gave the optimum tradeoff between a high statistical model-likelihood and a low number of model parameters. All classes had a specific profile. The incidence of AF varied; class 1 0.0%, class 2 0.3%, class 3 7.5%, class 4 0.2%, class 5 1.3%, class 6 4.2%, class 7 21.7% (p<0.001). The discrimination (C-statistic 0.830 vs. 0.842, delta-C -0.013, p = 0.22) and reclassification (IDI -0.028, p<0.001, NRI -0.090, p = 0.049, and category-less-NRI -0.049, p = 0.495) performance of both models was comparable. The results were successfully replicated in a sample of the FHS study (n = 3162; mean age 58±9 years, 46.3% men). CONCLUSIONS: Latent class analysis to build an AF risk model is feasible. Despite the heterogeneity in number and severity of risk factors between individuals at risk for AF, latent class analysis produces distinguishable groups.