Associations of Serum and Dialysate Potassium Concentrations With Incident Atrial Fibrillation in a Cohort Study of Older US Persons Initiating Hemodialysis for Kidney Failure

Introduction: Atrial fibrillation (AF) disproportionally affects persons on maintenance hemodialysis (HD). Associations of serum and dialysate potassium concentrations [K+] with AF incidence are poorly understood.Methods: We conducted a cohort study using Medicare claims merged with clinical data from a dialysis provider to determine whether serum-[K+] and/or dialysate-[K+] independently associated with AF incidence. Persons insured by fee-for-service Medicare aged $67 years at dialysis initiation and free from diagnosed AF prior to day 120 of dialysis were eligible. Serum-[K+] and dialysate-[K+] were assessed in 30-day intervals and patients were followed-up with for AF incidence in subsequent 30-day intervals.Results: During 2006 to 2011, 15,190 persons (mean age = 76.3 years) initiating HD had no prior AF diagnosis. Mean serum-[K+] was 4.5 mEq/l; dialysate-[K+] was 3 mEq/l in 34% and 2 mEq/l in 52% of patients. Followed-up over 21,907 person-years, 2869 persons had incident AF (incidence/100 person-years, 13.1 [95% confidence interval [CI], 12.6-13.6]). The multivariable-adjusted association of serum -[K+] with incident AF was J-shaped as follows: relative to a serum-[K+] of 4.5 mEq/l, lower serum-[K+] associated with increased AF risk, whereas confidence bands for higher serum-[K+] indicated no associ-ation. Dialysis against a dialysate-[K+] of 3 mEq/l versus 2 mEq/l independently associated with a 14% (95% CI, 5%-24%) lower incidence of AF. No effect modification between serum-[K+] and dialysate-[K+] was detected (P = 0.34).Conclusion: Lower serum-[K+] was independently associated with incident AF whereas elevated serum -[K+] was not. The findings support adoption of dialysate solutions with a dialysate-[K+] of 3 mEq/l, regardless of patients' serum-[K+], and elimination of lower dialysate-[K+] solutions from practice. Clinical trials randomizing patients to different dialysate-[K+] are warranted to establish causality.