CLUSTERING PATIENTS WITH CHRONIC KIDNEY DISEASE BASED ON ROUTINELY MEASURED LABORATORY BIOMARKERS AND ITS ASSOCIATION WITH DISEASE PROGRESSION; FINDINGS FROM A POPULATION LEVEL FEASIBILITY STUDY FROM BRITISH COLUMBIA, CANADA

Introduction:

Current management and monitoring of patients with chronic kidney disease (CKD) are primarily based on eGFR. We investigated if clustering patients with CKD using routinely measured laboratory biomarkers is associated with disease progression.

Methods:

We used data from PROMIS, a population-level registry database for CKD patients under the specialized care of nephrologists in British Columbia, Canada. We analyzed a cohort of 40766 patients with non-dialysis dependent CKD who were registered in PROMIS between January 1, 2010 and December 31, 2019. The date of urine albumin-creatinine ratio (UACR) measurement was chosen as the anchor moment date. Patients with at least one full panel moment with measurements of estimated glomerular filtration rate (eGFR), UACR, serum albumin (ALB), bicarbonate (HCO3), calcium (CA), free hemoglobin (HgB), intact parathyroid hormone (iPTH), phosphate (PO4), and transferrin saturation (TSAT) were included. Formation of a full lab panel was based on a window period from the anchor date; ±3 months for eGFR, serum ALB, HgB, CA, PO4, HCO3 and ±6 months for TSAT and serum iPTH. Patients aged ≤18 years on moment date, had glomerulonephritis or polycystic kidney disease, had history of kidney transplant or chronic dialysis prior to the moment date or patients who received erythropoiesis-stimulating agents within 3 months prior to the moment date were excluded. Equal number of patients in each of the eGFR categories (G1 and G2 combined, G3a, G3b, G4 and G5) were included after randomly sampling 1 moment per patient where patients in G1 and G2 combined were identified first. We used Latent Profile Analysis (LPA) to identify the distinct clusters of patients. K-means clustering analysis was used as a sensitivity analysis. Patients were followed until July 8, 2024. We used Fine and Grey sub-distribution hazard model to investigate the risk of progressing to kidney failure accounting for death as a competing event.

Results:

A total of 20346 patients had 128,087 full-panel lab observations over the 10 years period. The random study sample 1 included 7715 patients, median age was 73 years, 56% were male. LPA produced 9 distinct clusters. Table 1A and 1B present patient characteristics for overall sample and by cluster. Figure 1 presents the cumulative incidence of kidney failure. Patients experienced low, moderate and high risk of kidney failure in 4, 3 and 2 clusters, respectively (Figure 1 and Table 2). Sensitivity analyses using K-means clustering produced similar results that provides confidence in our primary analyses.

Conclusions:

Clustering of patients with CKD using routinely collected laboratory biomarkers may provide valuable insight on disease progression and help in patient care. Future research involving multiple datasets from various geographical regions is necessary to validate these findings.

I have no potential conflict of interest to disclose.

I did not use generative AI and AI-assisted technologies in the writing process.