Development and Standardization of a Furosemide Stress Test to Predict the Severity of Acute Kidney Injury

The Southern Acute Kidney Injury Network (SAKInet) [17] was formed in 2007 to collect samples from patients who developed AKI, with the goal of testing the diagnostic and prognostic accuracy of previously described and novel AKI biomarkers. For each subject, informed consent was obtained in accordance with The George Washington University Institutional Review Board-approved SAKInet protocol. We identified a subset of patients from the SAKInet cohort at the George Washington University who fulfilled the study criteria.

The protocol for cohort 2 was registered in clinicaltrials.gov. The study was carried out at the George Washington University (NCT00673244) and at the University of Chicago (NCT01275729). The respective university IRBs approved the identical protocol. Patients or their surrogates were required to sign informed consent prior to study entry. Patients were enrolled from June 2009 through December 2012. Urine sediment was assessed with the George Washington Urine Sediment Score (GW USS) as described previously [18]. Briefly, patients with a GW USS ≥2 have evidence of granular or epithelial cell casts in their urine sediment.

Inclusion criteria: (1) age greater than 18 years, admitted in an ICU; (2) AKIN stage I (6 hours of oliguria (<0.5 ml/kg/hour) or 0.3 mg/dL increase in serum creatinine or increase of 150 to 200% above baseline serum creatinine), or AKIN stage II (12 hours of oliguria (<0.5 ml/kg/hour) or increase of 200 to 300% above baseline serum creatinine); (3) indwelling bladder catheter; (4) presence of granular or epithelial cell casts on urine sediment (defined by GW USS ≥2), or fractional excretion of sodium (FeNa) >1.0%; and (5) patient deemed by the treating clinical team to be well-resuscitated.

Exclusion Criteria: (1) baseline estimated glomerular filtration rate (eGFR) <30 ml/minute/1.73m²; (2) history of renal allograft; (3) known pregnancy; (4) evidence of obstructive uropathy (for example, hydro-ureter); (5) evidence of active bleeding; (6) patients with allergy or known sensitivity to loop diuretics; (7) achievement of AKIN stage III criteria; or (8) evidence of volume depletion at the time of furosemide administration.

Patients in the SAKInet cohort who met the study criteria, and who received a furosemide dose of 1.0 mg/kg were entered into cohort 1. Replacement fluid was not protocolized in this group of subjects. Demographic and clinical data, urine sediment scores, and outcome data were abstracted from the case report forms.

Prior to FST urine was collected and scored with the GW USS [18]. Urinalysis was performed at each site (EBM-GW, JLK-UC). A pre-FST FeNa was only available if the treating team had ordered it for clinical purposes. After acquisition of informed consent, patients who were loop-diuretic naïve were given 1.0 mg/kg of intravenous furosemide. Because patients who were previously treated with loop diuretics within the previous 7 days were likely to have a blunted response over time compared to naïve patients, this group received an intravenous dose of 1.5 mg/kg (as little as 6 to 8 days of chronic loop-diuretic therapy is associated with a blunted response to furosemide due to increased distal tubular uptake of sodium in the thiazide-sensitive nephron segment)[19]. In order to minimize the risk of hypovolemia, urine output was replaced ml for ml each hour with either Ringers lactate or normal saline for six hours after the FST. The treating team could elect not to replace the volume if net volume loss was considered clinically desirable. Urine output was measured hourly for six hours and in total for 24 hours. Any and all adverse events related to furosemide were recorded including, but not limited to, tinnitus, hypokalemia, hypomagnesemia and hypotension. Patients were followed for 14 days or hospital discharge, whichever occurred first.

The primary outcome was the progression to AKIN stage III (need for RRT, increase in serum creatinine of 300% over baseline, urine output of 0.3 cc/kg/hour × 24 hours) within 14 days of FST. The secondary outcome was the composite of achieving stage AKIN III or death within 14 days of the FST.

We assessed the distribution of demographic and clinical variables. Differences between proportions of patients with certain characteristics were assessed with the chi-square, Fisher exact, Student t, and Mann-Whitney tests as appropriate. The primary analysis was to assess the urine output response to the FST, which was determined by assessing the area under the curve (AUC) receiver operating characteristics (ROC) comparing the primary endpoint of progression to AKIN stage III and the secondary endpoint of death/AKIN III within 14 days of the FST. Multivariable logistic regression was used to create three models. Model 1 is a clinical model using the Acute Physiology and Chronic Health Evaluation (APACHE II) score, baseline urinary flow rate (UFR), baseline eGFR, and AKIN stage II at study entry. Model 2 has all univariate variables with a difference <0.10 entered as covariates. Model 3 is multivariate, backward elimination, logistic regression. All means are reported + standard error (SE) unless otherwise specified. Statistical analysis was performed using SPSS 18.0 (Chicago, Ill). Methodology used to calculate FeNa, APACHE II score[20], and cardiovascular (CV) SOFA[21] score and eGFR[22] is shown in Additional file 1.

The FST was well-tolerated with no episodes of hypotension or any other adverse event deemed attributable to the test. We assessed the UFR in response to furosemide. The maximum UFR was within the first 2 to 3 hours (Table 2, Figure 2). We compared the UFR in response to FST between those patients that progressed and did not progress to AKIN stage III (Table 2, Figure 2). For each hourly interval, progressors had a lower UFR response compared to non-progressors (P <0.001). We also compared the UFR of FST between subjects who were furosemide naïve versus those that were not; there was no difference between these (Additional file 1 - Table S5).

We tested various combinations of the UO intervals to assess which had the best discriminative capacity (Table 3). We found that the sum of the first 2 hours of UO after the FST had the highest AUC to predict the primary outcome (0.87 in both cohort 1 and cohort 2). We also assessed the sensitivity and specificity of various 2-hour urine volumes to predict the primary and secondary outcomes (Table 4). The 2-hour UO of 200 ml or less had the best sensitivity and specificity to predict the primary outcome.

In this pilot study, we have demonstrated that the FST is feasible and well tolerated in critically ill patients with AKI. Furosemide administration can be associated with vasodilation and hypotension, but we did not observe any of these complications during our study. We took careful measures to decrease this potential adverse event by ensuring that the patients were deemed clinically well-resuscitated, and when appropriate received isovolemic replacement of UO with isotonic fluids. This may in part explain why we did not observe any adverse events.

The performance of the FST to predict the primary outcome was robust and consistent in both cohorts, with a range in ROC AUC of 0.82 to 0.87 (Table 3). Importantly, in comparison, the capacity of baseline UFR to predict the primary outcome had an ROC AUC of 0.71 to 0.76 (Additional file 1 - Table S2). This finding supports the concept that the FST offers important clinical information not captured by baseline UFR alone. In our study, the performance of the FST was comparable or exceeded the performance of several AKI biomarkers in predicting AKI progression [6, 8, 9]. We found the first 2-hour interval had the best predictive capacity (0.87), and this interval corresponds with the maximum UFR in response to the FST. When we assessed specific UO cutoffs we found that the 2-hour UO of 200 cc offered the best combination of sensitivity and specificity (87.1% and 84.1%, respectively). Because acute tubular necrosis causes intratubular obstruction and back leak, we were unsure whether the standard UO kinetics in response to furosemide would be similar to those seen in patients without renal disease [23, 24]. Previous investigators have shown that in patients without AKI, the maximum diuretic effect of furosemide occurs within the first three hours [25]. We showed similar kinetics in our study (Figure 2). Patients who progressed to AKIN III compared to those that did not progress were similar in age, CV SOFA score, APACHE II score, and baseline eGFR, and had a similar incidence of comorbidity (Table 1). Nephrotoxic exposure and clinical phenotype was also similar in progressors and non-progressors (Table 1). Not surprisingly, the progressor group tended to have more patients with AKIN stage II, a lower baseline UFR, and a higher mean cast score prior to FST. In multivariable analyses, UO response to FST was still statistically associated with progression to AKIN stage III, even when these variables were placed into the model (Additional file 1 - Table S4).

The concept of using furosemide to evaluate AKI is not entirely new. In 1973, Baek and colleagues [26] assessed 15 patients who did not have clinically apparent AKI at that time, subjected them to a furosemide challenge, and then evaluated the patients' free water clearance (C_H2O). They found that C_H2O near zero and a poor response to furosemide signaled that 'acute renal failure was imminent'. In this modest sized study, the dose of furosemide was not standardized and the study did not report if the patients had early stage AKI or any evidence of AKI at all. Nonetheless, our findings confirm the findings of that original report. Moreover, clinicians regularly give patients with oliguric AKI a furosemide challenge. However, there has not been a standardized approach with fluid replacement, early assessment, and appropriate clinical cutoffs to guide care.

In this study, we have used the FST as a functional test to predict progressive AKI. Urine biomarkers have been used previously to predict worsening AKI. The predictive value of the FST compares favorably with other recent biomarker studies. Hall and colleagues determined the ability of urine neutrophil gelatinase associated lipocalin (NGAL), kidney injury molecule-1 (KIM-1) and IL-18 to predict worsening AKI (unadjusted AUC values were 0.71, 0.64 and 0.63 respectively) [27]. The TRIBE-AKI consortium found unadjusted AUC values of 0.63 for IL-18, and 0.58 and 0.74 for urinary and plasma NGAL, respectively [6]. Koyner and colleagues in a separate study found that π- glutathione S-transferase (GST) predicted progression to stage III AKI with an AUC of 0.86 [28]. We recently found that urinary angiotensinogen predicts worsening AKI with an AUC of 0.70 [17].

Although the findings in this study show good performance metrics for the FST, the use of the FST in patients who are not appropriately resuscitated can be potentially deleterious. We cannot overemphasize the point that patients need to be euvolemic before undertaking any type of furosemide challenge, and that volume replacement is mandatory in patients who are not obviously volume overloaded, as the mean UO in response to the challenge was over 1.3 L in 6 hours. In addition, the FST should be conducted in an appropriate clinical setting where UO, heart rate, and blood pressure can be monitored frequently.

The study has several limitations. Since this is a pilot study, larger more comprehensive studies of the FST are warranted in order to fully understand the advantages and disadvantages of this dynamic functional test. Although previous clinical trials of furosemide use in AKI have not shown any beneficial effect, we cannot be sure that the FST did not impact the natural history of AKI and therefore affect its predictive performance [29, 30]. Some investigators have suggested that furosemide is protective in AKI because its administration may decrease tubular oxygen consumption, in which case its early administration in AKI would be protective [31]. Clinical trials using furosemide early in the course of AKI are underway and may help determine whether furosemide has a role in the treatment of AKI [32]. In addition, we did not specifically study patients with acute decompensated heart failure, nephrotic syndrome, or other patient populations with diuretic resistance. As such, we cannot be certain that the FST will perform similarly in those patient populations.

Similar to the approach used in acute coronary syndrome before the advances in thrombolytic therapy, the clinical syndrome of angina (that is, chest pain and dyspnea) was followed by a biomarker assessment (that is, creatine phosphosphokinase-MB) to further risk-assess the patient. For those patients who were confirmed with a biomarker, stress testing (for example, dobutamine or treadmill stress test) was used to confirm the presence of severe coronary artery disease. We believe that a similar process can begin in patients with AKI. Patients with renal angina [33] can be further assessed with AKI biomarkers. For those patients in whom AKI biomarkers confirm AKI, FST could be used to assess the severity and prognosis of AKI. Because indiscriminate use of loop diuretics can be harmful, appropriate resuscitation prior to FST is mandatory, and the FST should not be used as a primary screening diagnostic. We suggest that future studies to test this hypothesis be conducted.