Use of consensus clustering to identify distinct subtypes of chronic kidney disease and associated mortality risk

Chronic kidney disease (CKD) is a complex condition with diverse etiology and outcomes. Utilizing a data-driven clustering approach holds promise in identifying distinct CKD subgroups associated with specific risk profiles for death.

Unsupervised consensus clustering was utilized to classify chronic kidney disease (CKD) into subtypes based on 45 baseline characteristics in a cohort of 6,526 participants from the US National Health and Nutrition Examination Survey (NHANES) spanning the years 1999-2000 to 2017-2018.We examined the associations between CKD subgroups and clinical endpoints related to mortality, including all-cause mortality, cardiovascular disease mortality, cancer mortality, and mortality due to other causes.

A total of 6,526 individuals with CKD were classified into four clusters at baseline. Cluster 1 (n = 508) comprised patients with relatively favorable levels of cardiac and kidney function markers, lower prevalence of cancer and higher prevalence of obesity, lower medication usage, and younger age. Cluster 4 (n = 2,029) comprised patients with the worst cardiac and kidney function markers. The characteristics of cluster 2 (n = 1,439) and 3 (n = 2,550) fell in between these two clusters. From cluster 1 to cluster 4, we observed a gradual increase in the hazard ratios of all-cause mortality, cardiovascular disease mortality, and mortality due to other causes. Additionally, further sensitivity analysis revealed patient heterogeneity among predefined subgroups with similar baseline kidney function and mortality risks.

Consensus clustering integrated baseline clinical and laboratory measures, revealing distinct CKD subgroups with markedly different risks of death, suggesting that further examination of patient subgroups could advance precision medicine.