Creatinine index and lean body mass are excellent predictors of long-term survival in haemodiafiltration patients.

BACKGROUND

No single measurement adequately defines protein-energy malnutrition. In the dialysis population, somatic protein mass, a useful marker of protein malnutrition, is estimated using the creatinine index (CI), lean body mass (LBM) or both, but the clinical usefulness of these indices remains uncertain. Moreover, calculating these indices requires formal creatinine kinetics or urine and dialysate collection. A simpler method to estimate the creatinine generation rate (G(Cr)) probably might widen its use.

METHODS

We evaluated the usefulness of creatinine-based indices for predicting mortality in a cohort of 226 French haemodiafiltration patients using the Cox proportional hazards method. We also proposed simple yet precise formulas to calculate post-dialysis creatinine (Cr(post)) concentrations and derive creatinine generation rates (G(Cr)) from readily available measures. These formulas were developed using a large database containing more than 10 000 measured Cr(post) and G(Cr) values based on formal creatinine modelling. A single set of monthly values was used to evaluate the validity of the formulas.

RESULTS

When adjusted for comorbidities, sex and Kt/V, CI and LBM/body weight (LBM/BW) were better predictors of 5 year all-cause mortality than urea-based indices [survival relative risk (RR) = 0.24, P<0.01 for CI<22 mg/kg/day; RR = 0.33, P<0.02 for LBM/BW<0.75]. When the cohort was divided according to gender, similar results were found in males, but not in females. The different formulas allowed adequate prediction of Cr(post) and G(Cr) and classification of patients with good accuracy (CI<22: sensitivity = 94%, specificity = 82%; LBW/BW<0.75: sensitivity = 89%, specificity = 90%).

CONCLUSIONS

In a haemodiafiltration population, CI and LBM are excellent predictors of long-term survival. In anuric Caucasian haemodialysis patients, CI and LBM can be estimated from biochemical and anthropometric measurements without relying on formal modelling.