The role of Kt/V and creatinine clearance on assisting optimization of serum phosphorus levels among patients on peritoneal dialysis

KEY POINTS: This is a retrospective observational multinational peritoneal dialysis study to test whether creatinine clearance could be a better marker of serum phosphorus than urea Kt/V. Creatinine clearance was not more accurate predicting serum phosphorus than urea Kt/V, but its inclusion in multivariable models added more clarity. In conclusion, using both biomarkers, instead of just one, may better assist in the optimization of serum phosphorus levels.

BACKGROUND: Hyperphosphatemia is associated with poor outcome and is still very common in peritoneal dialysis (PD) patients. Because peritoneal phosphorus clearance is closer to peritoneal creatinine clearance (CrCl) than urea clearance, we hypothesized that weekly CrCl could be a better marker of serum phosphorus in PD.

METHODS: In a retrospective observational study, data from adult PD patients were collected across five institutions in North and South America: Fresenius Medical Care Latin America, Renal Research Institute, Mount Sinai Hospital, Hospital Civil de Guadalajara, and the Brazil PD cohort. All centers analyzed routinely available laboratory data, with exclusions for missing data on serum phosphorus, CrCl, or urea Kt/V. A unified statistical protocol was used across centers. Linear mixed-effect models examined associations between longitudinal serum phosphorus levels, CrCl, and Kt/V. Adjustments were made for age, sex, and baseline phosphorus binder usage. Mixed-effects meta-analysis determined the pooled effect size of CrCl and Kt/V on serum phosphorus trajectories, adjusted for confounders.

RESULTS: There were 16,796 incident PD patients analyzed. Age, body mass index, sex, PD modality, Kt/V, and CrCl, as well as serum phosphorus, varied significantly across the different cohorts, but >70% had residual renal function. For most cohorts, both CrCl total and urea Kt/V associated negatively with serum phosphorus levels, and log-likelihood ratio tests demonstrate that models including CrCltotal have more predictive information than those including only urea Kt/V for the largest cohorts. Models including CrCltotal increase information predicting longitudinal serum phosphorus levels irrespective of baseline urea Kt/V, age, use of phosphorus binder, and sex.

CONCLUSIONS: CrCl was not more accurate in predicting serum phosphorus than urea Kt/V, but its inclusion in multivariable models predicting serum phosphorus added accuracy. In conclusion, both CrCl and Kt/V are associated with phosphorus levels, and using both biomarkers, instead of just one, may better assist in the optimization of serum phosphorus levels.