Comparison of Acute Kidney Disease Staging by Estimated Glomerular Filtration Rate and by Serum Creatinine Ratio in Patients With Dialysis-requiring Acute Kidney Injury: A National Cohort Study From Taiwan

Abstract

Rationale & objective: The comparative performance of 2 staging systems for acute kidney disease (AKD), based on estimated glomerular filtration rate (AKD_eGFR) or serum creatinine ratio (AKD_sCr), remains uncertain. Our objective is to assess the predictive ability of these staging systems concerning outcomes.

Study design: Population-based retrospective observational cohort study.

Setting & participants: In total, 40,849 hospitalized patients with acute kidney injury requiring dialysis between July 1, 2015 and June 30, 2022, in the Taiwan National Health Insurance Research Database.

Tests compared: AKD stages were defined according to the Acute Dialysis Quality Initiative 16 Workgroup (AKD_sCr) and 2021 Kidney Disease: Improving Global Outcomes Consensus (AKD_eGFR). Cox proportional hazard models were constructed to examine associations between AKD stages and outcomes.

Outcomes: Mortality and sustained kidney recovery.

Results: The AKD_eGFR staging system predicted both outcomes better than the AKD_sCr staging system. Hazard ratios and 95% confidence intervals for mortality across increasing AKD_eGFR stages were 1.29 (1.19-1.41), 1.78 (1.65-1.92), 2.68 (2.49-2.87), 3.11 (2.84-3.42), and 6.34 (6.07-6.63). Conversely, within the AKD_sCr staging system, hazard ratios and 95% confidence intervals for mortality across ascending stages were 0.84 (0.75-0.94), 0.90 (0.77-1.04), 0.93 (0.75-1.16), and 4.32 (4.18-4.47). Subgroup and sensitivity analyses yielded consistent results.

Limitations: Retrospective analysis based on a claims database.

Conclusions: We conclude that the AKD_eGFR staging system performs better than the AKD_sCr staging system in patients with acute kidney injury requiring dialysis regarding all-cause mortality and kidney recovery. Future investigations are warranted to evaluate the comparative efficacy in patients with acute kidney injury not requiring dialysis.

Keywords: Acute kidney disease; acute kidney injury; dialysis; mortality; renal recovery; staging.

Figure 1

The associations between AKD stages and outcomes in the AKD_eGFR and the AKD_sCr staging systems. The adjusted hazard ratios for mortality (A) and sustained renal recovery (B) across AKD stages in the AKD_eGFR staging system (upper portion) and the AKD_sCr staging system (lower portion) were shown. AKD stage 0 served as the reference. Dots and error bars represented the point estimates and the 95% confidence intervals of the hazard ratios. The covariates adjusted for were listed in Supplementary Tables S3 and S4.

Figure 2

Subgroup analysis of mortality risk in the AKD_eGFR staging systems. The adjusted hazard ratios and 95% confidence intervals of different AKD_eGFR stages for mortality in the multivariable Cox proportional hazard model according to prespecified subgroups were shown. The adjusted hazard ratio and 95% confidence interval in the original 40,849 cohort were also shown. Dots and error bars represented the point estimates and the 95% confidence intervals of the hazard ratios. CKD was defined as a baseline eGFR < 60 mL/min/1.73 m². Two types of P values for interaction were provided. An interaction product variable, combining the covariate and AKD stage, was added to the multivariable model to test for interaction. The overall P value (P_overall) was estimated by comparing the multivariable models with and without the interaction product variable using a likelihood-ratio test. The P value for interaction (P_int) at each AKD stage was determined by the significance of the interaction product variable in the model. P < 0.003 was marked with an asterisk (∗) and claimed as statistically significant based on the Bonferroni correction, considering 2 outcomes and 9 exposures. Abbreviations: CCI, Charlson comorbidity index; CKD, chronic kidney disease.

Figure 3

Subgroup analysis of mortality risk in the AKD_sCr staging systems. The adjusted hazard ratios (represented by dots) and 95% confidence intervals (represented by error bars) for sustained renal recovery were displayed for the original cohort of 40,849 patients and in prespecified subgroups. The definitions of CKD, P_int, P_overall, and statistical significance were the same as those in Figure 2. Abbreviations: CCI, Charlson comorbidity index; CKD, chronic kidney disease.