Serum urea as a marker for initiation of renal replacement therapy in ICU patients with acute kidney injury

Introduction

There is no consensus on the criteria for initiation of renal replacement therapy (RRT) for acute kidney injury (AKI). Traditionally, serum urea levels are used as a surrogate marker for timing. In the past, higher levels were associated with increased hospital mortality. There is no consensus on the exact level of serum as a cutoff to start RRT, in order to impact on outcome. We want to evaluate urea cutoff criteria as described in the literature for initiation of RRT for AKI, in relation to outcome.

Methods

A retrospective study of 342 ICU patients with AKI who were started on RRT in the period 2004–2007, and who were captured in the electronic ICU database. The APACHE II score was calculated at admission to the ICU. The SOFA and nonrenal SOFA scores, blood levels of urea, creatinine, sodium and potassium were recorded at the start of RRT. Data are presented as the proportion and median (IQR).

Results

Three hundred and forty-two patients were included. The median age was 65 years (56, 73), 67.9% were male. The APACHE II score was 20 (15.26), and the nonrenal SOFA and SOFA scores were 5 (4, 8) and 8 (6, 11), respectively. Inhospital mortality was 58.2%. At the start of RRT, serum creatinine was 3.7 mg/dl (2.6, 4.9), urea was1.4 g/dl (0.9, 2.0). Intermittent RRT was the initial mode of RRT for 53.2% of patients, continuous dialysis in 22.5%, continuous venovenous hemofiltration in 14.9%, and slow low-efficiency daily dialysis in 8.5%. The traditionally used serum urea cutoff levels (1.0, 1.2, 1.4, 2, and 3 g/dl) were not associated with increased hospital death. Factors associated with inhospital death were APACHE II score (OR = 1.06, P < 0.001), continuous RRT (OR = 2.63, P < 0.001), nonrenal SOFA score at start of RRT (OR = 1.14, P < 0.001), and age (OR = 1.03, P = 0.001). Inhospital death was inversely correlated with creatinine at the start of RRT (OR = 0.63, P < 0.001). A multivariate logistic regression model demonstrated that, after correction for age, nonrenal SOFA and creatinine at the start of RRT, increasing urea levels were associated with mortality (OR = 1.50, P = 0.023), as were urea cutoff levels of 1.2 g/dl (OR = 2.28, P = 0.009) and 1.4 g/dl (OR = 2.36, P = 0.003). Higher and lower cutoff values were not associated with mortality.

Conclusion

Traditional cutoff values for initiation of RRT based on serum urea levels were not useful in assessing the prognosis. After correction for severity of illness and serum creatinine, serum urea was associated with mortality. Future models for initiation of RRT should include nonrenal indicators for outcome.