Though patients with chronic kidney disease (CKD) have variable risks of death, the population risk prediction tools that are currently available perform poorly.
To establish more satisfactory prediction models, Adeera Levin, MD, and colleagues from the Division of Nephrology of the Department of Medicine of the University of British Columbia undertook the Canadian Study of Prediction of Death, Dialysis and Interim Cardiovascular Events (CanPREDDICT) — a large, prospective cohort study that began in 2008 — and reported their findings at Kidney Week 2013, the American Society of Nephrology's annual meeting held November 5-10, 2013, in Atlanta, GA.
In that study, 2,544 CKD patients referred from 25 rural, urban, academic, and non-academic centers throughout Canada underwent baseline newer biomarker (NMB) tests, which included asymmetric dimethylarginine (ADMA), high sensitivity C-reactive protein (hsCRP), interleukin 6, N-terminal pro-brain natriuretic peptide (NTproBNP), troponin I, transforming growth factor beta 1 (TGFβ1), cystatin C, and fibroblast growth factor (FGF23). The mean age of the cohort was 68 years old, while the median estimated glomerular filtration rate (eGFR) was 28ml/min/1.73m2
. According to the researchers, 48% of patients had diabetes, and 15.5% died during the 3-year follow-up.
Patients were followed for 3 years at 6-month intervals and annually monitored for traditional biomarkers, such as eGFR, urine albumin creatinine ratio (uACR), hemoglobin (Hgb), and phosphate and albumin. The researchers also performed the aforementioned NMB tests after collecting blood and urine samples on schedule.
To investigate whether including NBMs improve the prediction of death risk in CKD patients to a greater extent than conventional clinical, demographic, and laboratory predictors, the researchers compared discrimination (C statistic) and classification (NRI) of proportional hazard models based on conventional markers versus a combination of conventional and NBM markers.
After adjusting for base predictors, they found that the biomarker NTproBNP was the best predictor of mortality (NRI=8.9; 95% confidence interval [CI]: 3.3-17.4), followed by hsCRP (NRI=4.5; 95% CI: 1.3-9.6) and FGF23 (NRI=3.1; 95% CI: 0.4-11.4).
While the authors found that including NBMs in risk prediction models significantly improves the precision of death prediction in CKD patients, they noted that their approach needs to be validated with further testing.
The authors disclosed support from Ortho Janssen.