Mediation of serum albumin in the association of serum potassium with mortality in Chinese dialysis patients: a prospective cohort study

BACKGROUND: The clinical importance of hypokalemia is likely underrecognized in Chinese dialysis patients, and whether its clinical effect was mediated by serum albumin is not fully elucidated. This study aimed to explore the association between serum potassium and mortality in dialysis patients of a Chinese nationwide multicenter cohort, taking albumin as a consideration. METHODS: This was a prospective nation-wide multicenter cohort study. Restricted cubic splines were used to test the linearity of serum potassium and relationships with all-cause (AC) and cardiovascular (CV) mortality and a subsequent two-line piecewise linear model was fitted to approach the nadir. A mediation analysis was performed to examine relations of albumin to potassium and mortalities. RESULTS: A total of 10,027 patients were included, of whom 6605 were peritoneal dialysis and 3422 were hemodialysis patients. In the overall population, the mean age was 51.7 ± 14.8 years, 55.3%(5546/10,027) were male, and the median dialysis vintage was 13.60 (4.70, 39.70) months. Baseline serum potassium was 4.30 ± 0.88 mmol/L. After a median follow-up period of 26.87 (14.77, 41.50) months, a U-shape was found between potassium and mortality, and a marked increase in risk at lower potassium but a moderate elevation in risk at higher potassium were observed. The nadir for AC mortality risk was estimated from piecewise linear models to be a potassium concentration of 4.0 mmol/L. Interestingly, the significance of the association between potassium and mortality was attenuated when albumin was introduced into the extended adjusted model. A subsequent significant mediation by albumin for potassium and AC and CV mortalities were found (P < 0.001 for both), indicating that hypokalemia led to higher mortality mediated by low serum albumin, which was a surrogate of poor nutritional status and inflammation. CONCLUSIONS: Associations between potassium and mortalities were U-shaped in the overall population. The nadir for AC mortality risk was at a potassium of 4.0 mmol/L. Serum albumin mediated the association between potassium and AC and CV mortalities.