Urine Biomarkers and Perioperative Acute Kidney Injury: The Impact of Preoperative Estimated GFR

Abstract

Background: The interaction between baseline kidney function and the performance of biomarkers of acute kidney injury (AKI) on the development of AKI is unclear.

Study design: Post hoc analysis of prospective cohort study.

Setting & participants: The 1,219 TRIBE-AKI Consortium adult cardiac surgery cohort participants.

Predictor: Unadjusted postoperative urinary biomarkers of AKI measured within 6 hours of surgery.

Outcome: AKI was defined as AKI Network stage 1 (any AKI) or higher, as well as a doubling of serum creatinine level from the preoperative value or the need for post-operative dialysis (severe AKI).

Measurements: Stratified analyses by preoperative estimated glomerular filtration rate (eGFR) ≤ 60 versus > 60mL/min/1.73m(2).

Results: 180 (42%) patients with preoperative eGFRs≤60mL/min/1.73m(2) developed clinical AKI compared with 246 (31%) of those with eGFRs>60mL/min/1.73m(2) (P<0.001). For log2-transformed biomarker concentrations, there was a significant interaction between any AKI and baseline eGFR for interleukin 18 (P=0.007) and borderline significance for liver-type fatty acid binding protein (P=0.06). For all biomarkers, the adjusted relative risk (RR) point estimates for the risk for any AKI were higher in those with elevated baseline eGFRs compared with those with eGFRs≤60mL/min/1.73m(2). However, the difference in magnitude of these risks was low (adjusted RRs were 1.04 [95% CI, 0.99-1.09] and 1.11 [95% CI, 1.07-1.15] for those with preoperative eGFRs≤60mL/min/1.73m(2) and those with higher eGFRs, respectively). Although no biomarker displayed an interaction for baseline eGFR and severe AKI, log2-transformed interleukin 18 and kidney injury molecule 1 had significant adjusted RRs for severe AKI in those with and without baseline eGFRs≤60mL/min/1.73m(2).

Limitations: Limited numbers of patients with severe AKI and post-operative dialysis.

Conclusions: The association between early postoperative AKI urinary biomarkers and AKI is modified by preoperative eGFR. The degree of this modification and its impact on the biomarker-AKI association is small across biomarkers. Our findings suggest that distinct biomarker cutoffs for those with and without a preoperative eGFR≤60mL/min/1.73m(2) is not necessary.

Keywords: Urine biomarkers; acute kidney injury (AKI); acute renal failure (ARF); cardiac surgery; effect modification; estimated glomerular filtration rate (eGFR); interleukin 18 (IL-18); liver-type fatty acid binding protein (L-FABP); perioperative AKI; prognosis; surgical complication.

Figure 1

TRIBE-AKI subjects included in the analyses

Figure 2. Peri-operative biomarker distributions stratified along baseline eGFR and AKI status

The bar plots represent the median (horizontal line) and 25–75% IQR(boxes) for those subjects with (blue) and without (green) postoperative AKI. In those with a preoperative eGFR ≤ 60 ml/min/1.73m², pre-operative L-FABP and albumin were higher in those who went on to develop AKI, while only urinary albumin levels were elevated pre-operatively in those with an eGFR > 60 ml/min/1.73m². At the majority of post-operative timepoints biomarkers were elevated in those with AKI regardless of baseline eGFR. (* p <0.05). CysC, cystatin C.

Figure 3. Log-transformed (base 2) Continuous biomarkers and the Logit of Any AKI

The plots demonstrate the presence or absence of an interaction between continuous biomarkers, AKI and baseline eGFR (dichotomized at 60 ml/min/1.73m²). The red line represents patients with eGFR ≤60 and blue line represents patients with eGFR > 60. The plots demonstrate the Interquartile range of values for the individual biomarkers at this early postoperative timepoint (grey bars). The plots that cross or nearly intersect are those that demonstrate a significant (or trend to significant) interaction (IL-18: p=0.007; L-FABP: p=0.06; cystatin C: p=0.1). Importantly for those biomarkers which displayed an interaction (or trend) that this interaction takes place at values on the extreme high end of the spectrum well above what we saw in our cohort.