Using the Difference Between Estimated Glomerular Filtration Rate by Cystatin C and Creatinine to Improve Mortality Risk Prediction in Elderly Patients With CKD in the HUNT Study.

RATIONALE &

OBJECTIVE

The discrepancy in estimated glomerular filtration rate by cystatin C (eGFRcys) versus creatinine (eGFRcr) has been used as a surrogate for sarcopenia. We studied whether this eGFRcys - eGFRcr difference could improve prediction of kidney failure versus death, which is important in the management of patients with chronic kidney disease (CKD). We hypothesized that it improved the death prediction but not that of kidney failure.

STUDY

DESIGN

A five-year cohort study to assess prognostic accuracy.

SETTING &

PARTICIPANTS

The population included 1,146 participants with creatinine (cr) and cystatin C (cys) measurements from the population-based the Nord-Trøndelag Health Study, Norway. Those aged ≤ 65 years or with estimated glomerular filtration rate (eGFR)cr ≥ 45 mL/min/1.73m2 were excluded.

EXPOSURES: The mortality risk equation in patients with CKD (MREK) includes age, sex, eGFRcr, albuminuria, smoking status, history of stroke, diabetes, and heart failure. The kidney failure risk equation includes age, sex, eGFRcr, and albuminuria.

OUTCOMES: Kidney failure or death at 5 years.

ANALYTICAL APPROACH: The performances of MREK and kidney failure risk equation with and without eGFR_diff (= eGFRcys - eGFRcr) were compared: calibration (likelihood ratio, Akaike information criterion, Brier score), discrimination (C-statistics), reclassification (net reclassification improvement and integrated discrimination improvement).

RESULTS

The mean ± SD age was 80 ± 7 years, 42% were men, the mean eGFRcr was 36 ± 8 and eGFR_diff was 1.04 ± 12 mL/min/1.73m2; 42 participants (4%) reached kidney failure and 444 (39%) died. C-statistics (95% CI) for MREK improved with eGFR_diff from 70.1% (66.7-73.4) to 73.0% (69.8-76.1) (P = 0.003). The proportion of participants correctly reclassified also improved (net reclassification improvement 14% [10%-17%]), and the separation between the average predicted risk for participants who died versus not (integrated discrimination improvement +0.03 [ 0.02-0.04]).

LIMITATIONS: Untested generalizability in other populations.

CONCLUSIONS

Including eGFR_diff into the kidney failure risk and mortality risk equations significantly improved mortality risk prediction, but not kidney failure, in patients with CKD. Serum creatinine level is influenced by many non-GFR determinants, including sarcopenia, and the discrepancy of creatinine versus cystatin C could be helpful in predialytic decision-making.