- Email: [email protected]
Lack of Furosemide Responsiveness Predicts Acute Kidney Injury in Infants After Cardiac Surgery
Lack of Furosemide Responsiveness Predicts Acute Kidney Injury in Infants After Cardiac Surgery Aadil Kakajiwala, MBBS,* Ji Young Kim, PhD, John Z. Hughes, MPH, Andrew Costarino, MD, MSCE, John Ferguson, J. William Gaynor, MD, Susan L. Furth, MD, PhD, and Joshua J. Blinder, MD Department of Pediatrics, Division of Pediatric Nephrology, Washington University in St. Louis School of Medicine, St. Louis, Missouri; Department of Pediatrics, Division of Nephrology, and Clinical Translational Research Center, Children’s Hospital of Philadelphia; College of Medicine and Dornsife School of Public Health, Drexel University; Department of Anesthesia and Critical Care Medicine, Division of Cardiac Critical Care, and Department of Surgery, Division of Cardiothoracic Surgery, Children’s Hospital of Philadelphia; and Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania
Background. This was a retrospective study to determine whether lack of furosemide responsiveness (LFR) predicts acute kidney injury (AKI) after cardiopulmonary bypass surgery in infants. Methods. Infants (less than 1 year of age) undergoing cardiopulmonary bypass surgery, receiving routine postoperative furosemide (0.8 to 1.2 mg/kg per dose between 8 and 24 hours after surgery) were included. Urine output was measured 2 and 6 hours after furosemide dose. Lack of furosemide responsiveness was defined a priori as urine output less than 1 mL $ kgL1 $ hL1 after furosemide. Serum creatinine was corrected for fluid balance. Acute kidney injury was determined using changes in uncorrected and corrected serum creatinine. The predictive utility of LFR was assessed using receiver-operating characteristics curve analysis. Results. We analyzed 568 infants who underwent cardiopulmonary bypass. Eighty-one (14.3%) had AKI using uncorrected serum creatinine; AKI occurred in 41 (7.2%) after correcting for fluid overload. Patients with AKI had a lower response to furosemide (median urine
output 2 hours: 1.2 versus 3.4 mL $ kgL1 $ hL1, p [ 0.01; median urine output 6 hours: 1.3 versus 2.9 mL $ kgL1 $ hL1, p [ 0.01). After creatinine correction, LFR predicts AKI development (area under receiveroperating characteristics curve of 0.74 at 2 hours and 0.77 at 6 hours). After adjusting for surgical complexity using The Society of Thoracic Surgeons/European Association for Cardiothoracic Surgery mortality categories, the area under the receiver-operating characteristics curve was 0.74 at 2 hours and 0.81 at 6 hours. Patients with urine output greater than 1 mL $ kgL1 $ hL1 were unlikely to have AKI (negative predictive value, 97%). Conclusions. After correcting serum creatinine for fluid balance and adjusting for surgical complexity, LFR performs fairly at 2 hours, whereas at 6 hours, LFR is a good AKI predictor. Prospective studies are needed to validate whether diuretic responsiveness predicts AKI.
C
definitions for AKI use either urine output or serum creatinine [6–10]. In 2012, the Kidney Disease Improving Global Outcome (KDIGO) AKI criteria integrated several AKI definitions into a single methodology to define and stage AKI [6]. The KDIGO criteria have been modified for neonates and young infants (aged less than 120 days) using an absolute rise in serum creatinine to stage AKI [7]. Furosemide is a loop of Henle diuretic that enters the tubular lumen by active secretion [11]. In patients with AKI, the diuretic effect of furosemide is blunted because
ardiac surgery-associated acute kidney injury (AKI) occurs frequently and is associated with increased mortality risk, longer intensive care unit and hospital length of stay, and need for prolonged mechanical ventilation [1, 2]. Infants may be at particularly high risk of AKI after cardiac surgery, with injury occurring in as many as 64% of infants postoperatively [3–5]. Current Accepted for publication March 6, 2017. Presented at the Meeting of the Pediatric Academic Society/American Society of Pediatric Nephrology, Baltimore, MD, April 30–May 3, 2016. *This study was performed when Aadil Kakajiwala, MBBS, was at the Department of Pediatrics, Division of Nephrology, Children’s Hospital of Philadelphia. Address correspondence to Dr Kakajiwala, Department of Pediatrics, Division of Pediatric Nephrology, Washington University in St. Louis School of Medicine, Campus Box 8116, 660 Euclid Ave, St. Louis, MO 63110; email: [email protected].
Ó 2017 by The Society of Thoracic Surgeons Published by Elsevier Inc.
(Ann Thorac Surg 2017;-:-–-) Ó 2017 by The Society of Thoracic Surgeons
The Supplemental Tables can be viewed in the online version of this article [http://dx.doi.org/10.1016/j. athoracsur.2017.03.015] on http://www.annalsthoracic surgery.org.
0003-4975/$36.00 http://dx.doi.org/10.1016/j.athoracsur.2017.03.015
2
KAKAJIWALA ET AL DIURETIC RESPONSIVENESS PREDICTS AKI
Abbreviations and Acronyms AKI AUC CI IQR KDIGO LFR ROC STAT
= = = = =
acute kidney injury area under the curve confidence interval interquartile range Kidney Disease Improving Global Outcome = lack of furosemide responsiveness = receiver-operating characteristics = The Society of Thoracic Surgeons/ European Association for Cardiothoracic Surgery
of reduced tubular secretion. Hence, response to furosemide, in terms of urine output, may be a good indicator of renal tubular function. Chawla and associates [12] investigated whether diuretic responsiveness (furosemide stress test) could be used to predict AKI progression in a cohort of critically ill adults. In patients with AKI, hourly urine output at 2 hours was significantly lower than noted in patients who did not progress. Urine output 2 hours after furosemide predicted AKI progression with the receiver-operating characteristics (ROC) area under the curve (AUC) of 0.81 to 0.87. Measuring several biomarkers in addition to the furosemide stress test showed ROC AUC that was only slightly improved to 0.90 in predicting progression to stage III AKI [13]. There were no adverse effects. Given the predictive utility of the furosemide stress test in critically ill adults, there is reason to believe that a similar test would be useful in children. Several factors play a role in AKI development in patients requiring cardiopulmonary bypass [14]. For patients who have postoperative AKI, the insult’s timing is known [15]. Therefore, we undertook this study to determine whether early lack of furosemide responsiveness (LFR) after congenital heart surgery predicts AKI. We hypothesized that LFR after cardiopulmonary bypass surgery would predict AKI in a cohort of infants. In addition, we hypothesized that adjusting for surgical complexity would improve the diagnostic characteristics of LFR for AKI.
Patients and Methods Inclusion Criteria Infants (less than 1 year of age) who underwent congenital heart surgery from January 2013 to May 2015 at Children’s Hospital of Philadelphia were reviewed. Patients on extracorporeal membrane oxygenation (n ¼ 22), on renal replacement therapy (n ¼ 2) immediately after surgery, and patients not requiring cardiopulmonary bypass were excluded (n ¼ 108). The standard furosemide dose was defined as 0.8 to 1.2 mg/kg, given 8 to 24 hours (median 12.8; interquartile range [IQR]: 9.8 to 18.2) after return from the operating room. The mean dose of furosemide received was 0.98
Ann Thorac Surg 2017;-:-–-
mg/kg (SD 0.06). Patients receiving nonstandard doses or continuous infusions of furosemide (n ¼ 104) or additional diuretics between 2 and 6 hours after the initial furosemide dose (n ¼ 60) were excluded. Continuous furosemide infusions have different pharmacokinetics compared with bolus dosing, which may limit the utility of LFR (details of the high-risk excluded patients are given in Supplement Tables 1 and 2). Eleven excluded patients received early doses furosemide before 8 hours. These patients had a median age of 286 days (IQR: 224 to 345), median body surface area of 0.23 m2 (IQR: 0.20 to 0.29 m2), and median birth weight of 2.95 kg (IQR: 2.63 to 3.38 kg). Two of these patients received renal replacement therapy within 4 hours of surgery, and 3 required extracorporeal membranous oxygenation support. Of these 11 patients, 5 underwent a ventricular septal defect repair, 1 underwent pulmonary artery banding, 1 underwent atrial septal defect repair, 1 underwent arterial switch operation, 2 underwent hemiFontan procedure, and 1 had a Norwood operation. This study was approved by the Institutional Board Review with waiver of informed consent.
Data Collection Subject demographic data including age, sex, and body surface area were collected. Surgical complexity was determined using The Society of Thoracic Surgeons and the European Association for Cardiothoracic Surgery congenital heart surgery mortality classification (STAT categories) [16]. Cardiopulmonary bypass time, crossclamp time, and deep hypothermic arrest time were also noted. Vasoactive-inotropic score was calculated for the first 48 hours after surgery [17] as follows: vasoactiveinotropic score ¼ dopamine dose (mg $ kg1 $ min1) þ dobutamine dose (mg $ kg1 $ min1) þ 100 epinephrine dose (mg $ kg1 $ min1) þ 10 milrinone dose (mg $ kg1 $ min1) þ 10,000 vasopressin dose (units $ kg1 $ min1) þ 100 norepinephrine dose (mg $ kg1 $ min1). Lengths of hospital and intensive care unit stay as well as mortality at 30 days and at mediumterm follow-up (2.5 years after cardiac surgery) were noted. Postoperatively, all patients had a urinary catheter, and urine output was measured continuously per standard practice. Urine output response was assessed after the first dose of furosemide. Daily fluid balance and firstmorning serum creatinine levels were collected from time of surgery through postoperative day 7. The KDIGO and neonatal KDIGO criteria were used to determine AKI using serum creatinine criteria as appropriate based on patient age [6, 7]. Reference creatinine was the lowest preoperative serum creatinine. In infants aged 120 days or less, AKI was assessed as a 0.3 mg/dL or greater creatinine increase. In patients aged more than 120 days, AKI was assessed as a 0.2 mg/dL or greater creatinine increase.
Statistical Analysis Given the low incidence of AKI in our cohort (especially of stage 2 and stage 3 AKI) and the questionable clinical
Ann Thorac Surg 2017;-:-–-
relevance of mild AKI [9], AKI stage was not utilized to assess outcomes. Patient characteristics and surgical variables in each cohort were compared using c2 and Student’s t tests. Infants undergoing more than one operation were analyzed using mixed effects modeling to account for withinsubject correlations. As the variables were not normally distributed, percentage and medians were compared. Linear mixed effects models were used for the continuous variables, and mixed effects logistic regression models were used for the binary variables. The CochranArmitage test for trend was used to assess the association of STAT category with AKI risk. Generalized linear mixed effects modeling (test for log odds of AKI developing per unit increase in urine output) was used to compare the urine output in the AKI and no AKI groups at each operation. The LFR, defined a priori as urine output less than 1 mL $ kg1 $ h1, was used to predict AKI using ROC curves. Thresholds for urine output with the highest sum of sensitivity and specificity were determined. Furthermore, the sensitivity, specificity, and negative predictive value for LFR in predicting AKI were calculated. The probabilities of AKI predicted from this model were used to create the ROC curve by plotting the sensitivity and specificity of LFR at various thresholds of urine output. We fitted the generalized mixed effects model adjusting for STAT category as an ordinal factor to adjust for surgical complexity. Serum creatinine may be diluted in fluid overload, limiting AKI detection [18]. Based on the work of Basu and colleagues [19], we corrected the creatinine for fluid balance, as follows: corrected serum creatinine ¼ measured serum creatinine [1 þ (accumulated net fluid balance / total body water)], where total body water ¼ 0.6 weight (kg) [20].
KAKAJIWALA ET AL DIURETIC RESPONSIVENESS PREDICTS AKI
3
Statistical analysis was performed using R, version 3.2.4 (R Foundation for Statistical Computing, www.R-project. org).
Results Data were collected on 676 patients; 108 patients were excluded (Fig 1). The remaining 568 patients comprised the cohort. Using uncorrected serum creatinine, 81 patients (14%) had AKI, of whom 56 (69%) were stage 1, 19 (23%) were stage 2, and 6 (7%) were stage 3 AKI. When correcting serum creatinine for fluid overload, AKI occurred in 41 patients. In the 48 patients initially classified as having AKI without creatinine correction, the median fluid balance at 24 hours was 0.03 L (IQR: 0.18 to þ0.11 L). In comparison, the 41 patients classified as having AKI using corrected serum creatinine had a median fluid balance at 24 hours of þ0.18 L (IQR: þ0.02 to þ0.40 L, p ¼ 0.001). Table 1 summarizes subject characteristics and cardiac support times. The patients who had AKI using corrected serum creatinine were younger (median age 18 days versus 89 days, p ¼ 0.006) and smaller (median body surface area 0.20 m2 versus 0.25 m2, p < 0.001). Higher STAT category was also associated with AKI (p ¼ 0.003). Cardiopulmonary bypass time (p < 0.001), cross-clamp time (p < 0.001), and deep hypothermic arrest time (p ¼ 0.01) were significantly longer in patients having AKI. The postoperative day 2 vasoactive-inotropic score was higher in the AKI group (p < 0.001). Patients who had AKI had longer stays in the intensive care unit, higher 30-day mortality, and higher overall mortality (p < 0.001 for all; Table 2).
Response to Furosemide Using uncorrected serum creatinine, 2 and 6 hours after furosemide administration, median urine output was Fig 1. Study cohort. We excluded 104 patients who received nonstandard doses of furosemide and patients who were started on a furosemide drip. An additional 60 patients who received additional doses of diuretics within 6 hours of the first furosemide dose were also excluded. (AKI ¼ acute kidney injury.)
4
KAKAJIWALA ET AL DIURETIC RESPONSIVENESS PREDICTS AKI
Ann Thorac Surg 2017;-:-–-
Table 1. Comparison of Subject Characteristics, Surgical Complexity, and Cardiac Support Times in Acute Kidney Injury and No Acute Kidney Injury Groups Uncorrected Serum Creatinine AKI (n ¼ 81)
Variables Age at surgery, days Birth weight, kg Gestational age, weeks Male Weight, kg Body surface area, m2 Maximum STAT score 1, least complex 2 3 4 5, most complex CPB time, minutes Cross-clamp time, minutes DHCA time, minutes Postoperative day 1 VIS Postoperative day 2 VIS a
17.5 2.85 38 43 3.6 0.21 5 10 26 26 14 83 43 0 5 4
No AKI (n ¼ 487)
(5–118.2) (2.52–3.36) (37–39) (53.1) (2.9–5.1) (0.19–0.27)
96 3.09 39 225 4.6 0.25
(6–152) (2.68–3.48) (38–39) (53.7) (3.3–6.2) (0.20–0.26)
(6) (12) (32) (32) (17) (66–112) (30–61) (0–35) (2.5–9.1) (0–9)
62 75 163 141 45 63 34 0 5 0
(13) (15) (33) (29) (9) (39.5–84) (22–50) (0–24) (2–7.5) (0–5)
Corrected Serum Creatinine p Value
AKI (n ¼ 41)
<0.001 1 1 0.93 0.0039 <0.001
18 2.69 37.5 23 3.2 0.20
0.005a
2 3 13 15 8 93 51 0 5 4
<0.001 <0.001 0.031 0.057 <0.001
No AKI (n ¼ 527)
(5–107) (2.25–3.190) (36.75–39) (56.1) (2.7–4.9) (0.12–0.26)
89 3.09 39 282 4.5 0.25
(5.75–150) (2.70–3.49) (37.25–39) (53.5) (3.3–6.1) (0.20–0.31)
(5) (7) (32) (36) (19) (76–119) (39–71) (0–42) (2.5–7.5) (0–8.5)
65 82 177 152 51 64 34.5 0 5 0
(12) (15) (34) (29) (10) (41–85) (22–50.75) (0–25) (2–8) (0–5)
p Value 0.006 1 1 0.87 0.005 <0.001 0.003a
<0.001 <0.001 0.01 0.66 <0.001
Cochran-Armitage test of trend.
Values are median (interquartile range) or n (%). Linear mixed effects model was used for numeric variables, and mixed effects logistic regression models were used for binary variables. AKI ¼ acute kidney injury; CPB ¼ cardiopulmonary bypass; DHCA ¼ deep hypothermic arrest; European Association for Cardiothoracic Surgery; VIS ¼ vasoactive-inotropic score.
lower in the AKI group compared with the no-AKI group (2 hours: 2.1 mL $ kg1 $ h1 versus 3.5 mL $ kg1 $ h1, p ¼ 0.006; and 6 hours: 1.9 mL $ kg1 $ h1 versus 3.0 mL $ kg1 $ h1, p ¼ 0.01). The LFR predicts AKI with an AUC of 0.64 at 2 hours (95% confidence interval [CI]: 0.56 to 0.72) and at 6 hours (95% CI: 0.55 to 0.73). Urine output less than 2.57 mL $ kg1 $ h1 at 2 hours and less than 1.96 mL $ kg1 $ h1 at 6 hours predicted AKI with maximum sensitivity (65.4% and 54.7%, respectively) and specificity (63.3% and 72.9%, respectively). When serum creatinine was corrected for fluid balance, the median urine output at 2 hours was 1.2 mL $ kg1 $ h1 (IQR: 0.7 to 2.6 mL $ kg1 $ h1) in the AKI group versus 3.4 mL $ kg1 $ h1 (IQR: 1.8 to 5.7 mL $ kg1 $ h1) in the no AKI group (p ¼ 0.01). The median urine output at 6 hours was 1.3 mL $ kg1 $ h1
STAT ¼ The Society of Thoracic Surgeons/
(IQR: 0.6 to 1.9 mL $ kg1 $ h1) in the AKI group versus 2.9 mL $ kg1 $ h1 (IQR: 1.9 to 4.1 mL $ kg1 $ h1) in the no AKI group (p ¼ 0.01; Fig 2). Figure 3 demonstrates the ROC AUC at 2 hours and 6 hours used to determine the performance of LFR after correction for fluid balance. The LFR predicts AKI with AUC 0.74 (95% CI: 0.62 to 0.86) at 2 hours and 0.77 (95% CI: 0.61 to 0.93) at 6 hours. Adjusting the ROC curve for maximum STAT showed AUC 0.74 (95% CI: 0.63 to 0.85) at 2 hours and 0.81 (95% CI: 0.69 to 0.93) at 6 hours (Fig 4). Urine output less than 1.66 mL $ kg1 $ h1 (2 hours) and less than 1.93 mL $ kg1 $ h1 (6 hours) predicted AKI with maximum sensitivity (68.2% and 80.0%, respectively) and specificity (76.5% and 73.3%, respectively). Urine output less than 1 mL $ kg1 $ h1 had a negative predictive value of 95.6% at 2 hours and 97.6% at 6 hours,
Table 2. Comparison of Subject Outcomes in Acute Kidney Injury and No Acute Kidney Injury Groups Uncorrected Serum Creatinine Comparisons
AKI (n ¼ 81)
Length of ICU stay, days Length of hospital stay, days Thirty-day mortality Overall mortality
10.5 21 9 15
(5.8–20.3) (12.3–48) (11.1) (18.5)
No AKI (n ¼ 487) 6 11 6 21
(3–11) (6–21) (1.2) (4.3)
Corrected Serum Creatinine p Value
AKI (n ¼ 41)
<0.001 <0.001 <0.001 <0.001
12.5 27.5 6 10
(8–24.3) (16–48) (14.6) (24.4)
No AKI (n ¼ 527) 6 11 9 26
(3–11) (6–22.8) (1.7) (4.9)
p Value <0.001 0.003 <0.001 <0.001
Values are median (interquartile range) or n (%). Linear mixed effects model was used for numeric variables, and mixed effects logistic regression models were used for binary variables. AKI ¼ acute kidney injury;
ICU ¼ intensive care unit.
Ann Thorac Surg 2017;-:-–-
KAKAJIWALA ET AL DIURETIC RESPONSIVENESS PREDICTS AKI
5
Fig 2. Box plots comparing the urine output (UOP) in the acute kidney injury (AKI) group and the no AKI group, determined after correcting serum creatinine for fluid balance, 2 and 6 hours after the initial furosemide dose.
indicating that patients with urine output greater than 1 mL $ kg1 $ h1 were very unlikely to have AKI.
Comment To our knowledge, this is the first study evaluating diuretic responsiveness in infants after congenital heart surgery. After correction for fluid balance, urine output less than 1.7 mL $ kg1 $ h1 at 2 hours and less than 1.9 mL $ kg1 $ h1 at 6 hours after furosemide predicts AKI development. The LFR with correction for fluid balance was a fair predictor of AKI development. After adjusting for maximum STAT category, the LFR was a better predictor of AKI development. Moreover, AKI was unlikely to develop in patients with urine output greater than 1 mL $ kg1 $ h1.
Fluid overload may dilute serum creatinine, leading to AKI underdiagnosis using creatinine-based definitions. In our study, however, we found that we potentially overestimated AKI without fluid correction. Basu and colleagues [19] studied a cohort of 92 children who underwent arterial switch operation, of whom 18 (20%) had severe AKI. Serum creatinine was corrected for net fluid balance after surgery. Several patients shifted between stages of AKI owing to this correction. There was a significant difference in the duration of postoperative ventilation, inotrope score, length of stay in the intensive care unit and postoperative hospital stay. In our cohort, serum creatinine correction showed an increased ability of LFR to predict AKI. After pediatric cardiac surgery, furosemide is administered routinely and patients undergo strict monitoring
Fig 3. Receiver-operating characteristics (ROC) curves for lack of furosemide responsiveness, determined after correcting serum creatinine for fluid balance at (A) 2 hours after furosemide dose, and (B) 6 hours after furosemide dose. The area under the ROC curve (AUC) signifies the ability of the lack of furosemide response to predict acute kidney injury (AKI) development. The AUC can range from 0.5 to 1.0. A value of 0.5 suggests the test is not useful, whereas an AUC of 1.0 suggests a “perfect test.” (UOP2 ¼ urine output 2 hours; UOP6 ¼ urine output 6 hours.)
6
KAKAJIWALA ET AL DIURETIC RESPONSIVENESS PREDICTS AKI
Ann Thorac Surg 2017;-:-–-
Fig 4. Receiver-operating characteristics curves for lack of furosemide responsiveness in determining development of acute kidney injury (AKI), after correcting serum creatinine for fluid balance at (A) 2 hours after furosemide dose, adjusted for maximum STAT category, and (B) 6 hours after furosemide dose, adjusted for maximum STAT category. (AUC ¼ area under the curve; UOP2 ¼ urine output 2 hours; UOP6 ¼ urine output 6 hours; STAT ¼ The Society of Thoracic Surgeons/European Association for Cardiothoracic Surgery.)
of intake and output in the intensive care unit, making furosemide responsiveness an easy to use, real time bedside AKI predictor. Assessment of AKI biomarkers in pediatrics is ideal as children often lack many comorbidities that could confound AKI studies [15]. The LFR may meet the requirements of an effective biomarker for AKI— accurate, easy to measure, noninvasive, and reproducible [21]. Early AKI detection may afford clinicians the opportunity for early intervention, limiting exposure to nephrotoxic medications, intervening with deleterious hemodynamics, and allowing closer outpatient follow-up for patients at risk of chronic renal insufficiency [22, 23]. Serum creatinine may be a late AKI indicator. Serum creatinine is affected by muscle mass and medications and may be diluted in patients who are fluid overloaded, thereby leading to underestimation of the incidence of AKI [18]. In our population, LFR between 8 and 24 hours after cardiac surgery predicted AKI substantially earlier than it was otherwise detected by serum creatinine elevation, and potentially leading to earlier intervention to mitigate the effects of AKI. Infants with urine output >1 mL $ kg1 $ h1 after furosemide are at low risk of AKI. Because AKI is associated with worse clinical outcomes, the high negative predictive value of LFR indicates that patients with adequate diuretic responsiveness are likely to do better. Hence, there is a good reason to consider diuretic responsiveness and its relationship to AKI development. Duration of cardiopulmonary bypass may partially reflect the complexity of surgical repair [24, 25]. The underlying cardiac lesions, fragile postoperative hemodynamics, cyanosis, and nonpulsatile flow during cardiopulmonary bypass may be associated with decreased perfusion states, placing the kidneys at greater risk [24]
Adjusting LFR for surgical complexity further increased its utility to predict AKI. Our study has several important limitations. (1) We excluded infants requiring additional doses of diuretics or furosemide drips during the 6-hour window. These patients are at higher risk of AKI (Supplemental Tables 1 and 2). We may be underestimating the overall AKI rates in our population, and therefore, the utility of the test. In addition, we excluded other high-risk patients (including those receiving renal replacement therapy and extracorporeal membranous oxygenation). (2) We were unable to interpret the urine output beyond 6 hours after the initial dose of furosemide as additional doses of diuretics are typically administered and urine output precision would decrease with urinary catheter removal. Studies evaluating the kinetics of furosemide suggest slower clearance of furosemide in infants as compared with that in adults [26, 27]. Many of our patients received additional doses of diuretics beyond 6 hours from the initial dose. Therefore, we were not able to assess the test beyond that timepoint. (3) Our population has a lower incidence of AKI after cardiac surgery (14.2%) as compared with other studies (40% to 60%), with a very low incidence of stage 2 and stage 3 AKI. Most previous work involving children undergoing cardiac surgery indicates that the clinical impact of AKI occurs in patients with stage 2 and stage 3, leading to potential underestimation of the predictive utility of LFR in our center. (4) Correction of the serum creatinine led to loss of 48 patients initially classified as having AKI. The fluid balance in these 48 patients was significantly more negative as compared with that in the patients with newly classified AKI. This finding suggests that the correction may have underestimated the incidence of prerenal AKI. In addition, the total body water in infants, used to correct serum creatinine, is different as compared with adults and may vary depending on body surface area
Ann Thorac Surg 2017;-:-–-
[28]. (5) Owing to the retrospective nature of the study, we were unable to assess whether patients received diuretics in the operating room or had intraoperative blood loss, although correcting serum creatinine for overall postoperative fluid balance would partially mitigate this limitation. This study suggests that the LFR is a fair AKI predictor in infants after cardiopulmonary bypass, with improved diagnostic value after correcting serum creatinine for fluid overload. After adjusting for surgical complexity, the LFR at 6 hours is a good AKI predictor in infants. Larger prospective studies are needed involving children to validate the LFR. Assessment of the pharmacokinetics of furosemide in infants with AKI will also add to the usefulness of LFR in this population.
References 1. Aydin SI, Seiden HS, Blaufox AD, et al. Acute kidney injury after surgery for congenital heart disease. Ann Thorac Surg 2012;94:1589–95. 2. Li S, Krawczeski CD, Zappitelli M, et al. Incidence, risk factors, and outcomes of acute kidney injury after pediatric cardiac surgery—a prospective multicenter study. Crit Care Med 2011;39:1493. 3. Morgan CJ, Zappitelli M, Robertson CM, et al, for the Western Canadian Complex Pediatric Therapies Follow-Up Group. Risk factors for and outcomes of acute kidney injury in neonates undergoing complex cardiac surgery. J Pediatr 2013;162:120–7. 4. Alabbas A, Campbell A, Skippen P, Human D, Matsell D, Mammen C. Epidemiology of cardiac surgery-associated acute kidney injury in neonates: a retrospective study. Pediatr Nephrol 2013;28:1127–34. 5. Blinder JJ, Goldstein SL, Lee VV, et al. Congenital heart surgery in infants: effects of acute kidney injury on outcomes. J Thorac Cardiovasc Surg 2012;143:368–74. 6. Kellum JA, Lameire N, Aspelin P, et al. Work group membership. Kidney Int 2012;2:1. 7. Jetton JG, Askenazi DJ. Update on acute kidney injury in the neonate. Curr Opin Pediatr 2012;24:191–6. 8. Mehta RL, Kellum JA, Shah SV, et al. Acute Kidney Injury Network: report of an initiative to improve outcomes in acute kidney injury. Crit Care 2007;11:R31. 9. Bellomo R, Ronco C, Kellum JA, Mehta RL, Palevsky P. Acute renal failure—definition, outcome measures, animal models, fluid therapy and information technology needs: the Second International Consensus Conference of the Acute Dialysis Quality Initiative (ADQI) group. Crit Care 2004;8:1. 10. Akcan-Arikan A, Zappitelli M, Loftis LL, Washburn KK, Jefferson LS, Goldstein SL. Modified RIFLE criteria in critically ill children with acute kidney injury. Kidney Int 2007;71:1028–35.
KAKAJIWALA ET AL DIURETIC RESPONSIVENESS PREDICTS AKI
7
11. Ponto LL, Schoenwald RD. Furosemide (frusemide) a pharmacokinetic/pharmacodynamic review (part I). Clin Pharmacokinet 1990;18:381–408. 12. Chawla LS, Davison DL, Brasha-Mitchell E, et al. Development and standardization of a furosemide stress test to predict the severity of acute kidney injury. Crit Care 2013;17: R207. 13. Koyner JL, Davison DL, Brasha-Mitchell E, et al. Furosemide stress test and biomarkers for the prediction of AKI severity. J Am Soc Nephrol 2015;26:2023–31. 14. Thiele RH, Isbell JM, Rosner MH. AKI associated with cardiac surgery. Clin J Am Soc Nephrol 2014; CJN-07830814. 15. Goldstein SL. Acute kidney injury biomarkers: renal angina and the need for a renal troponin I. BMC Med 2011;9:1. 16. O’Brien SM, Clarke DR, Jacobs JP, et al. An empirically based tool for analyzing mortality associated with congenital heart surgery. J Thorac Cardiovasc Surg 2009;138:1139–53. 17. Gaies MG, Gurney JG, Yen AH, et al. Vasoactive-inotropic score as a predictor of morbidity and mortality in infants after cardiopulmonary bypass. Pediatr Crit Care Med 2010;11:234–8. 18. Bouchard J, Mehta RL. Fluid accumulation and acute kidney injury: consequence or cause. Curr Opin Crit Care 2009;15: 509–13. 19. Basu RK, Andrews A, Krawczeski C, Manning P, Wheeler DS, Goldstein SL. Acute kidney injury based on corrected serum creatinine is associated with increased morbidity in children following the arterial switch operation. Pediatr Crit Care Med 2013;14:e218–24. 20. Liu KD, Thompson BT, Ancukiewicz M, et al. Acute kidney injury in patients with acute lung injury: impact of fluid accumulation on classification of acute kidney injury and associated outcomes. Crit Care Med 2011;39:2665. 21. Nguyen MT, Devarajan P. Biomarkers for the early detection of acute kidney injury. Pediatr Nephrol 2008;23:2151–7. 22. Fortenberry JD, Paden ML, Goldstein SL. Acute kidney injury in children: an update on diagnosis and treatment. Pediatr Clin North Am 2013;60:669–88. 23. Stafford-Smith M, Patel UD, Phillips-Bute BG, Shaw AD, Swaminathan M. Acute kidney injury and chronic kidney disease after cardiac surgery. Adv Chronic Kidney Dis 2008;15:257–77. 24. Chiravuri SD, Riegger LQ, Christensen R, et al. Factors associated with acute kidney injury or failure in children undergoing cardiopulmonary bypass: a case-controlled study. Paediatr Anaesth 2011;21:880–6. 25. Picca S, Ricci Z, Picardo S. Acute kidney injury in an infant after cardiopulmonary bypass. Semin Nephrol 2008;28:470–6. 26. Aranda JV, Lambert C, Perez J, Turmen T, Sitar DS. Metabolism and renal elimination of furosemide in the newborn infant. J Pediatr 1982;101:777–81. 27. Tuck S, Morselli P, Broquaire M, Vert P. Plasma and urinary kinetics of furosemide in newborn infants. J Pediatr 1983;103: 481–5. 28. Wells JC, Fewtrell MS, Davies PS, Williams JE, Coward WA, Cole TJ. Prediction of total body water in infants and children. Arch Dis Child 2005;90:965–71.
Background. This was a retrospective study to determine whether lack of furosemide responsiveness (LFR) predicts acute kidney injury (AKI) after cardiopulmonary bypass surgery in infants. Methods. Infants (less than 1 year of age) undergoing cardiopulmonary bypass surgery, receiving routine postoperative furosemide (0.8 to 1.2 mg/kg per dose between 8 and 24 hours after surgery) were included. Urine output was measured 2 and 6 hours after furosemide dose. Lack of furosemide responsiveness was defined a priori as urine output less than 1 mL $ kgL1 $ hL1 after furosemide. Serum creatinine was corrected for fluid balance. Acute kidney injury was determined using changes in uncorrected and corrected serum creatinine. The predictive utility of LFR was assessed using receiver-operating characteristics curve analysis. Results. We analyzed 568 infants who underwent cardiopulmonary bypass. Eighty-one (14.3%) had AKI using uncorrected serum creatinine; AKI occurred in 41 (7.2%) after correcting for fluid overload. Patients with AKI had a lower response to furosemide (median urine
output 2 hours: 1.2 versus 3.4 mL $ kgL1 $ hL1, p [ 0.01; median urine output 6 hours: 1.3 versus 2.9 mL $ kgL1 $ hL1, p [ 0.01). After creatinine correction, LFR predicts AKI development (area under receiveroperating characteristics curve of 0.74 at 2 hours and 0.77 at 6 hours). After adjusting for surgical complexity using The Society of Thoracic Surgeons/European Association for Cardiothoracic Surgery mortality categories, the area under the receiver-operating characteristics curve was 0.74 at 2 hours and 0.81 at 6 hours. Patients with urine output greater than 1 mL $ kgL1 $ hL1 were unlikely to have AKI (negative predictive value, 97%). Conclusions. After correcting serum creatinine for fluid balance and adjusting for surgical complexity, LFR performs fairly at 2 hours, whereas at 6 hours, LFR is a good AKI predictor. Prospective studies are needed to validate whether diuretic responsiveness predicts AKI.
C
definitions for AKI use either urine output or serum creatinine [6–10]. In 2012, the Kidney Disease Improving Global Outcome (KDIGO) AKI criteria integrated several AKI definitions into a single methodology to define and stage AKI [6]. The KDIGO criteria have been modified for neonates and young infants (aged less than 120 days) using an absolute rise in serum creatinine to stage AKI [7]. Furosemide is a loop of Henle diuretic that enters the tubular lumen by active secretion [11]. In patients with AKI, the diuretic effect of furosemide is blunted because
ardiac surgery-associated acute kidney injury (AKI) occurs frequently and is associated with increased mortality risk, longer intensive care unit and hospital length of stay, and need for prolonged mechanical ventilation [1, 2]. Infants may be at particularly high risk of AKI after cardiac surgery, with injury occurring in as many as 64% of infants postoperatively [3–5]. Current Accepted for publication March 6, 2017. Presented at the Meeting of the Pediatric Academic Society/American Society of Pediatric Nephrology, Baltimore, MD, April 30–May 3, 2016. *This study was performed when Aadil Kakajiwala, MBBS, was at the Department of Pediatrics, Division of Nephrology, Children’s Hospital of Philadelphia. Address correspondence to Dr Kakajiwala, Department of Pediatrics, Division of Pediatric Nephrology, Washington University in St. Louis School of Medicine, Campus Box 8116, 660 Euclid Ave, St. Louis, MO 63110; email: [email protected].
Ó 2017 by The Society of Thoracic Surgeons Published by Elsevier Inc.
(Ann Thorac Surg 2017;-:-–-) Ó 2017 by The Society of Thoracic Surgeons
The Supplemental Tables can be viewed in the online version of this article [http://dx.doi.org/10.1016/j. athoracsur.2017.03.015] on http://www.annalsthoracic surgery.org.
0003-4975/$36.00 http://dx.doi.org/10.1016/j.athoracsur.2017.03.015
2
KAKAJIWALA ET AL DIURETIC RESPONSIVENESS PREDICTS AKI
Abbreviations and Acronyms AKI AUC CI IQR KDIGO LFR ROC STAT
= = = = =
acute kidney injury area under the curve confidence interval interquartile range Kidney Disease Improving Global Outcome = lack of furosemide responsiveness = receiver-operating characteristics = The Society of Thoracic Surgeons/ European Association for Cardiothoracic Surgery
of reduced tubular secretion. Hence, response to furosemide, in terms of urine output, may be a good indicator of renal tubular function. Chawla and associates [12] investigated whether diuretic responsiveness (furosemide stress test) could be used to predict AKI progression in a cohort of critically ill adults. In patients with AKI, hourly urine output at 2 hours was significantly lower than noted in patients who did not progress. Urine output 2 hours after furosemide predicted AKI progression with the receiver-operating characteristics (ROC) area under the curve (AUC) of 0.81 to 0.87. Measuring several biomarkers in addition to the furosemide stress test showed ROC AUC that was only slightly improved to 0.90 in predicting progression to stage III AKI [13]. There were no adverse effects. Given the predictive utility of the furosemide stress test in critically ill adults, there is reason to believe that a similar test would be useful in children. Several factors play a role in AKI development in patients requiring cardiopulmonary bypass [14]. For patients who have postoperative AKI, the insult’s timing is known [15]. Therefore, we undertook this study to determine whether early lack of furosemide responsiveness (LFR) after congenital heart surgery predicts AKI. We hypothesized that LFR after cardiopulmonary bypass surgery would predict AKI in a cohort of infants. In addition, we hypothesized that adjusting for surgical complexity would improve the diagnostic characteristics of LFR for AKI.
Patients and Methods Inclusion Criteria Infants (less than 1 year of age) who underwent congenital heart surgery from January 2013 to May 2015 at Children’s Hospital of Philadelphia were reviewed. Patients on extracorporeal membrane oxygenation (n ¼ 22), on renal replacement therapy (n ¼ 2) immediately after surgery, and patients not requiring cardiopulmonary bypass were excluded (n ¼ 108). The standard furosemide dose was defined as 0.8 to 1.2 mg/kg, given 8 to 24 hours (median 12.8; interquartile range [IQR]: 9.8 to 18.2) after return from the operating room. The mean dose of furosemide received was 0.98
Ann Thorac Surg 2017;-:-–-
mg/kg (SD 0.06). Patients receiving nonstandard doses or continuous infusions of furosemide (n ¼ 104) or additional diuretics between 2 and 6 hours after the initial furosemide dose (n ¼ 60) were excluded. Continuous furosemide infusions have different pharmacokinetics compared with bolus dosing, which may limit the utility of LFR (details of the high-risk excluded patients are given in Supplement Tables 1 and 2). Eleven excluded patients received early doses furosemide before 8 hours. These patients had a median age of 286 days (IQR: 224 to 345), median body surface area of 0.23 m2 (IQR: 0.20 to 0.29 m2), and median birth weight of 2.95 kg (IQR: 2.63 to 3.38 kg). Two of these patients received renal replacement therapy within 4 hours of surgery, and 3 required extracorporeal membranous oxygenation support. Of these 11 patients, 5 underwent a ventricular septal defect repair, 1 underwent pulmonary artery banding, 1 underwent atrial septal defect repair, 1 underwent arterial switch operation, 2 underwent hemiFontan procedure, and 1 had a Norwood operation. This study was approved by the Institutional Board Review with waiver of informed consent.
Data Collection Subject demographic data including age, sex, and body surface area were collected. Surgical complexity was determined using The Society of Thoracic Surgeons and the European Association for Cardiothoracic Surgery congenital heart surgery mortality classification (STAT categories) [16]. Cardiopulmonary bypass time, crossclamp time, and deep hypothermic arrest time were also noted. Vasoactive-inotropic score was calculated for the first 48 hours after surgery [17] as follows: vasoactiveinotropic score ¼ dopamine dose (mg $ kg1 $ min1) þ dobutamine dose (mg $ kg1 $ min1) þ 100 epinephrine dose (mg $ kg1 $ min1) þ 10 milrinone dose (mg $ kg1 $ min1) þ 10,000 vasopressin dose (units $ kg1 $ min1) þ 100 norepinephrine dose (mg $ kg1 $ min1). Lengths of hospital and intensive care unit stay as well as mortality at 30 days and at mediumterm follow-up (2.5 years after cardiac surgery) were noted. Postoperatively, all patients had a urinary catheter, and urine output was measured continuously per standard practice. Urine output response was assessed after the first dose of furosemide. Daily fluid balance and firstmorning serum creatinine levels were collected from time of surgery through postoperative day 7. The KDIGO and neonatal KDIGO criteria were used to determine AKI using serum creatinine criteria as appropriate based on patient age [6, 7]. Reference creatinine was the lowest preoperative serum creatinine. In infants aged 120 days or less, AKI was assessed as a 0.3 mg/dL or greater creatinine increase. In patients aged more than 120 days, AKI was assessed as a 0.2 mg/dL or greater creatinine increase.
Statistical Analysis Given the low incidence of AKI in our cohort (especially of stage 2 and stage 3 AKI) and the questionable clinical
Ann Thorac Surg 2017;-:-–-
relevance of mild AKI [9], AKI stage was not utilized to assess outcomes. Patient characteristics and surgical variables in each cohort were compared using c2 and Student’s t tests. Infants undergoing more than one operation were analyzed using mixed effects modeling to account for withinsubject correlations. As the variables were not normally distributed, percentage and medians were compared. Linear mixed effects models were used for the continuous variables, and mixed effects logistic regression models were used for the binary variables. The CochranArmitage test for trend was used to assess the association of STAT category with AKI risk. Generalized linear mixed effects modeling (test for log odds of AKI developing per unit increase in urine output) was used to compare the urine output in the AKI and no AKI groups at each operation. The LFR, defined a priori as urine output less than 1 mL $ kg1 $ h1, was used to predict AKI using ROC curves. Thresholds for urine output with the highest sum of sensitivity and specificity were determined. Furthermore, the sensitivity, specificity, and negative predictive value for LFR in predicting AKI were calculated. The probabilities of AKI predicted from this model were used to create the ROC curve by plotting the sensitivity and specificity of LFR at various thresholds of urine output. We fitted the generalized mixed effects model adjusting for STAT category as an ordinal factor to adjust for surgical complexity. Serum creatinine may be diluted in fluid overload, limiting AKI detection [18]. Based on the work of Basu and colleagues [19], we corrected the creatinine for fluid balance, as follows: corrected serum creatinine ¼ measured serum creatinine [1 þ (accumulated net fluid balance / total body water)], where total body water ¼ 0.6 weight (kg) [20].
KAKAJIWALA ET AL DIURETIC RESPONSIVENESS PREDICTS AKI
3
Statistical analysis was performed using R, version 3.2.4 (R Foundation for Statistical Computing, www.R-project. org).
Results Data were collected on 676 patients; 108 patients were excluded (Fig 1). The remaining 568 patients comprised the cohort. Using uncorrected serum creatinine, 81 patients (14%) had AKI, of whom 56 (69%) were stage 1, 19 (23%) were stage 2, and 6 (7%) were stage 3 AKI. When correcting serum creatinine for fluid overload, AKI occurred in 41 patients. In the 48 patients initially classified as having AKI without creatinine correction, the median fluid balance at 24 hours was 0.03 L (IQR: 0.18 to þ0.11 L). In comparison, the 41 patients classified as having AKI using corrected serum creatinine had a median fluid balance at 24 hours of þ0.18 L (IQR: þ0.02 to þ0.40 L, p ¼ 0.001). Table 1 summarizes subject characteristics and cardiac support times. The patients who had AKI using corrected serum creatinine were younger (median age 18 days versus 89 days, p ¼ 0.006) and smaller (median body surface area 0.20 m2 versus 0.25 m2, p < 0.001). Higher STAT category was also associated with AKI (p ¼ 0.003). Cardiopulmonary bypass time (p < 0.001), cross-clamp time (p < 0.001), and deep hypothermic arrest time (p ¼ 0.01) were significantly longer in patients having AKI. The postoperative day 2 vasoactive-inotropic score was higher in the AKI group (p < 0.001). Patients who had AKI had longer stays in the intensive care unit, higher 30-day mortality, and higher overall mortality (p < 0.001 for all; Table 2).
Response to Furosemide Using uncorrected serum creatinine, 2 and 6 hours after furosemide administration, median urine output was Fig 1. Study cohort. We excluded 104 patients who received nonstandard doses of furosemide and patients who were started on a furosemide drip. An additional 60 patients who received additional doses of diuretics within 6 hours of the first furosemide dose were also excluded. (AKI ¼ acute kidney injury.)
4
KAKAJIWALA ET AL DIURETIC RESPONSIVENESS PREDICTS AKI
Ann Thorac Surg 2017;-:-–-
Table 1. Comparison of Subject Characteristics, Surgical Complexity, and Cardiac Support Times in Acute Kidney Injury and No Acute Kidney Injury Groups Uncorrected Serum Creatinine AKI (n ¼ 81)
Variables Age at surgery, days Birth weight, kg Gestational age, weeks Male Weight, kg Body surface area, m2 Maximum STAT score 1, least complex 2 3 4 5, most complex CPB time, minutes Cross-clamp time, minutes DHCA time, minutes Postoperative day 1 VIS Postoperative day 2 VIS a
17.5 2.85 38 43 3.6 0.21 5 10 26 26 14 83 43 0 5 4
No AKI (n ¼ 487)
(5–118.2) (2.52–3.36) (37–39) (53.1) (2.9–5.1) (0.19–0.27)
96 3.09 39 225 4.6 0.25
(6–152) (2.68–3.48) (38–39) (53.7) (3.3–6.2) (0.20–0.26)
(6) (12) (32) (32) (17) (66–112) (30–61) (0–35) (2.5–9.1) (0–9)
62 75 163 141 45 63 34 0 5 0
(13) (15) (33) (29) (9) (39.5–84) (22–50) (0–24) (2–7.5) (0–5)
Corrected Serum Creatinine p Value
AKI (n ¼ 41)
<0.001 1 1 0.93 0.0039 <0.001
18 2.69 37.5 23 3.2 0.20
0.005a
2 3 13 15 8 93 51 0 5 4
<0.001 <0.001 0.031 0.057 <0.001
No AKI (n ¼ 527)
(5–107) (2.25–3.190) (36.75–39) (56.1) (2.7–4.9) (0.12–0.26)
89 3.09 39 282 4.5 0.25
(5.75–150) (2.70–3.49) (37.25–39) (53.5) (3.3–6.1) (0.20–0.31)
(5) (7) (32) (36) (19) (76–119) (39–71) (0–42) (2.5–7.5) (0–8.5)
65 82 177 152 51 64 34.5 0 5 0
(12) (15) (34) (29) (10) (41–85) (22–50.75) (0–25) (2–8) (0–5)
p Value 0.006 1 1 0.87 0.005 <0.001 0.003a
<0.001 <0.001 0.01 0.66 <0.001
Cochran-Armitage test of trend.
Values are median (interquartile range) or n (%). Linear mixed effects model was used for numeric variables, and mixed effects logistic regression models were used for binary variables. AKI ¼ acute kidney injury; CPB ¼ cardiopulmonary bypass; DHCA ¼ deep hypothermic arrest; European Association for Cardiothoracic Surgery; VIS ¼ vasoactive-inotropic score.
lower in the AKI group compared with the no-AKI group (2 hours: 2.1 mL $ kg1 $ h1 versus 3.5 mL $ kg1 $ h1, p ¼ 0.006; and 6 hours: 1.9 mL $ kg1 $ h1 versus 3.0 mL $ kg1 $ h1, p ¼ 0.01). The LFR predicts AKI with an AUC of 0.64 at 2 hours (95% confidence interval [CI]: 0.56 to 0.72) and at 6 hours (95% CI: 0.55 to 0.73). Urine output less than 2.57 mL $ kg1 $ h1 at 2 hours and less than 1.96 mL $ kg1 $ h1 at 6 hours predicted AKI with maximum sensitivity (65.4% and 54.7%, respectively) and specificity (63.3% and 72.9%, respectively). When serum creatinine was corrected for fluid balance, the median urine output at 2 hours was 1.2 mL $ kg1 $ h1 (IQR: 0.7 to 2.6 mL $ kg1 $ h1) in the AKI group versus 3.4 mL $ kg1 $ h1 (IQR: 1.8 to 5.7 mL $ kg1 $ h1) in the no AKI group (p ¼ 0.01). The median urine output at 6 hours was 1.3 mL $ kg1 $ h1
STAT ¼ The Society of Thoracic Surgeons/
(IQR: 0.6 to 1.9 mL $ kg1 $ h1) in the AKI group versus 2.9 mL $ kg1 $ h1 (IQR: 1.9 to 4.1 mL $ kg1 $ h1) in the no AKI group (p ¼ 0.01; Fig 2). Figure 3 demonstrates the ROC AUC at 2 hours and 6 hours used to determine the performance of LFR after correction for fluid balance. The LFR predicts AKI with AUC 0.74 (95% CI: 0.62 to 0.86) at 2 hours and 0.77 (95% CI: 0.61 to 0.93) at 6 hours. Adjusting the ROC curve for maximum STAT showed AUC 0.74 (95% CI: 0.63 to 0.85) at 2 hours and 0.81 (95% CI: 0.69 to 0.93) at 6 hours (Fig 4). Urine output less than 1.66 mL $ kg1 $ h1 (2 hours) and less than 1.93 mL $ kg1 $ h1 (6 hours) predicted AKI with maximum sensitivity (68.2% and 80.0%, respectively) and specificity (76.5% and 73.3%, respectively). Urine output less than 1 mL $ kg1 $ h1 had a negative predictive value of 95.6% at 2 hours and 97.6% at 6 hours,
Table 2. Comparison of Subject Outcomes in Acute Kidney Injury and No Acute Kidney Injury Groups Uncorrected Serum Creatinine Comparisons
AKI (n ¼ 81)
Length of ICU stay, days Length of hospital stay, days Thirty-day mortality Overall mortality
10.5 21 9 15
(5.8–20.3) (12.3–48) (11.1) (18.5)
No AKI (n ¼ 487) 6 11 6 21
(3–11) (6–21) (1.2) (4.3)
Corrected Serum Creatinine p Value
AKI (n ¼ 41)
<0.001 <0.001 <0.001 <0.001
12.5 27.5 6 10
(8–24.3) (16–48) (14.6) (24.4)
No AKI (n ¼ 527) 6 11 9 26
(3–11) (6–22.8) (1.7) (4.9)
p Value <0.001 0.003 <0.001 <0.001
Values are median (interquartile range) or n (%). Linear mixed effects model was used for numeric variables, and mixed effects logistic regression models were used for binary variables. AKI ¼ acute kidney injury;
ICU ¼ intensive care unit.
Ann Thorac Surg 2017;-:-–-
KAKAJIWALA ET AL DIURETIC RESPONSIVENESS PREDICTS AKI
5
Fig 2. Box plots comparing the urine output (UOP) in the acute kidney injury (AKI) group and the no AKI group, determined after correcting serum creatinine for fluid balance, 2 and 6 hours after the initial furosemide dose.
indicating that patients with urine output greater than 1 mL $ kg1 $ h1 were very unlikely to have AKI.
Comment To our knowledge, this is the first study evaluating diuretic responsiveness in infants after congenital heart surgery. After correction for fluid balance, urine output less than 1.7 mL $ kg1 $ h1 at 2 hours and less than 1.9 mL $ kg1 $ h1 at 6 hours after furosemide predicts AKI development. The LFR with correction for fluid balance was a fair predictor of AKI development. After adjusting for maximum STAT category, the LFR was a better predictor of AKI development. Moreover, AKI was unlikely to develop in patients with urine output greater than 1 mL $ kg1 $ h1.
Fluid overload may dilute serum creatinine, leading to AKI underdiagnosis using creatinine-based definitions. In our study, however, we found that we potentially overestimated AKI without fluid correction. Basu and colleagues [19] studied a cohort of 92 children who underwent arterial switch operation, of whom 18 (20%) had severe AKI. Serum creatinine was corrected for net fluid balance after surgery. Several patients shifted between stages of AKI owing to this correction. There was a significant difference in the duration of postoperative ventilation, inotrope score, length of stay in the intensive care unit and postoperative hospital stay. In our cohort, serum creatinine correction showed an increased ability of LFR to predict AKI. After pediatric cardiac surgery, furosemide is administered routinely and patients undergo strict monitoring
Fig 3. Receiver-operating characteristics (ROC) curves for lack of furosemide responsiveness, determined after correcting serum creatinine for fluid balance at (A) 2 hours after furosemide dose, and (B) 6 hours after furosemide dose. The area under the ROC curve (AUC) signifies the ability of the lack of furosemide response to predict acute kidney injury (AKI) development. The AUC can range from 0.5 to 1.0. A value of 0.5 suggests the test is not useful, whereas an AUC of 1.0 suggests a “perfect test.” (UOP2 ¼ urine output 2 hours; UOP6 ¼ urine output 6 hours.)
6
KAKAJIWALA ET AL DIURETIC RESPONSIVENESS PREDICTS AKI
Ann Thorac Surg 2017;-:-–-
Fig 4. Receiver-operating characteristics curves for lack of furosemide responsiveness in determining development of acute kidney injury (AKI), after correcting serum creatinine for fluid balance at (A) 2 hours after furosemide dose, adjusted for maximum STAT category, and (B) 6 hours after furosemide dose, adjusted for maximum STAT category. (AUC ¼ area under the curve; UOP2 ¼ urine output 2 hours; UOP6 ¼ urine output 6 hours; STAT ¼ The Society of Thoracic Surgeons/European Association for Cardiothoracic Surgery.)
of intake and output in the intensive care unit, making furosemide responsiveness an easy to use, real time bedside AKI predictor. Assessment of AKI biomarkers in pediatrics is ideal as children often lack many comorbidities that could confound AKI studies [15]. The LFR may meet the requirements of an effective biomarker for AKI— accurate, easy to measure, noninvasive, and reproducible [21]. Early AKI detection may afford clinicians the opportunity for early intervention, limiting exposure to nephrotoxic medications, intervening with deleterious hemodynamics, and allowing closer outpatient follow-up for patients at risk of chronic renal insufficiency [22, 23]. Serum creatinine may be a late AKI indicator. Serum creatinine is affected by muscle mass and medications and may be diluted in patients who are fluid overloaded, thereby leading to underestimation of the incidence of AKI [18]. In our population, LFR between 8 and 24 hours after cardiac surgery predicted AKI substantially earlier than it was otherwise detected by serum creatinine elevation, and potentially leading to earlier intervention to mitigate the effects of AKI. Infants with urine output >1 mL $ kg1 $ h1 after furosemide are at low risk of AKI. Because AKI is associated with worse clinical outcomes, the high negative predictive value of LFR indicates that patients with adequate diuretic responsiveness are likely to do better. Hence, there is a good reason to consider diuretic responsiveness and its relationship to AKI development. Duration of cardiopulmonary bypass may partially reflect the complexity of surgical repair [24, 25]. The underlying cardiac lesions, fragile postoperative hemodynamics, cyanosis, and nonpulsatile flow during cardiopulmonary bypass may be associated with decreased perfusion states, placing the kidneys at greater risk [24]
Adjusting LFR for surgical complexity further increased its utility to predict AKI. Our study has several important limitations. (1) We excluded infants requiring additional doses of diuretics or furosemide drips during the 6-hour window. These patients are at higher risk of AKI (Supplemental Tables 1 and 2). We may be underestimating the overall AKI rates in our population, and therefore, the utility of the test. In addition, we excluded other high-risk patients (including those receiving renal replacement therapy and extracorporeal membranous oxygenation). (2) We were unable to interpret the urine output beyond 6 hours after the initial dose of furosemide as additional doses of diuretics are typically administered and urine output precision would decrease with urinary catheter removal. Studies evaluating the kinetics of furosemide suggest slower clearance of furosemide in infants as compared with that in adults [26, 27]. Many of our patients received additional doses of diuretics beyond 6 hours from the initial dose. Therefore, we were not able to assess the test beyond that timepoint. (3) Our population has a lower incidence of AKI after cardiac surgery (14.2%) as compared with other studies (40% to 60%), with a very low incidence of stage 2 and stage 3 AKI. Most previous work involving children undergoing cardiac surgery indicates that the clinical impact of AKI occurs in patients with stage 2 and stage 3, leading to potential underestimation of the predictive utility of LFR in our center. (4) Correction of the serum creatinine led to loss of 48 patients initially classified as having AKI. The fluid balance in these 48 patients was significantly more negative as compared with that in the patients with newly classified AKI. This finding suggests that the correction may have underestimated the incidence of prerenal AKI. In addition, the total body water in infants, used to correct serum creatinine, is different as compared with adults and may vary depending on body surface area
Ann Thorac Surg 2017;-:-–-
[28]. (5) Owing to the retrospective nature of the study, we were unable to assess whether patients received diuretics in the operating room or had intraoperative blood loss, although correcting serum creatinine for overall postoperative fluid balance would partially mitigate this limitation. This study suggests that the LFR is a fair AKI predictor in infants after cardiopulmonary bypass, with improved diagnostic value after correcting serum creatinine for fluid overload. After adjusting for surgical complexity, the LFR at 6 hours is a good AKI predictor in infants. Larger prospective studies are needed involving children to validate the LFR. Assessment of the pharmacokinetics of furosemide in infants with AKI will also add to the usefulness of LFR in this population.
References 1. Aydin SI, Seiden HS, Blaufox AD, et al. Acute kidney injury after surgery for congenital heart disease. Ann Thorac Surg 2012;94:1589–95. 2. Li S, Krawczeski CD, Zappitelli M, et al. Incidence, risk factors, and outcomes of acute kidney injury after pediatric cardiac surgery—a prospective multicenter study. Crit Care Med 2011;39:1493. 3. Morgan CJ, Zappitelli M, Robertson CM, et al, for the Western Canadian Complex Pediatric Therapies Follow-Up Group. Risk factors for and outcomes of acute kidney injury in neonates undergoing complex cardiac surgery. J Pediatr 2013;162:120–7. 4. Alabbas A, Campbell A, Skippen P, Human D, Matsell D, Mammen C. Epidemiology of cardiac surgery-associated acute kidney injury in neonates: a retrospective study. Pediatr Nephrol 2013;28:1127–34. 5. Blinder JJ, Goldstein SL, Lee VV, et al. Congenital heart surgery in infants: effects of acute kidney injury on outcomes. J Thorac Cardiovasc Surg 2012;143:368–74. 6. Kellum JA, Lameire N, Aspelin P, et al. Work group membership. Kidney Int 2012;2:1. 7. Jetton JG, Askenazi DJ. Update on acute kidney injury in the neonate. Curr Opin Pediatr 2012;24:191–6. 8. Mehta RL, Kellum JA, Shah SV, et al. Acute Kidney Injury Network: report of an initiative to improve outcomes in acute kidney injury. Crit Care 2007;11:R31. 9. Bellomo R, Ronco C, Kellum JA, Mehta RL, Palevsky P. Acute renal failure—definition, outcome measures, animal models, fluid therapy and information technology needs: the Second International Consensus Conference of the Acute Dialysis Quality Initiative (ADQI) group. Crit Care 2004;8:1. 10. Akcan-Arikan A, Zappitelli M, Loftis LL, Washburn KK, Jefferson LS, Goldstein SL. Modified RIFLE criteria in critically ill children with acute kidney injury. Kidney Int 2007;71:1028–35.
KAKAJIWALA ET AL DIURETIC RESPONSIVENESS PREDICTS AKI
7
11. Ponto LL, Schoenwald RD. Furosemide (frusemide) a pharmacokinetic/pharmacodynamic review (part I). Clin Pharmacokinet 1990;18:381–408. 12. Chawla LS, Davison DL, Brasha-Mitchell E, et al. Development and standardization of a furosemide stress test to predict the severity of acute kidney injury. Crit Care 2013;17: R207. 13. Koyner JL, Davison DL, Brasha-Mitchell E, et al. Furosemide stress test and biomarkers for the prediction of AKI severity. J Am Soc Nephrol 2015;26:2023–31. 14. Thiele RH, Isbell JM, Rosner MH. AKI associated with cardiac surgery. Clin J Am Soc Nephrol 2014; CJN-07830814. 15. Goldstein SL. Acute kidney injury biomarkers: renal angina and the need for a renal troponin I. BMC Med 2011;9:1. 16. O’Brien SM, Clarke DR, Jacobs JP, et al. An empirically based tool for analyzing mortality associated with congenital heart surgery. J Thorac Cardiovasc Surg 2009;138:1139–53. 17. Gaies MG, Gurney JG, Yen AH, et al. Vasoactive-inotropic score as a predictor of morbidity and mortality in infants after cardiopulmonary bypass. Pediatr Crit Care Med 2010;11:234–8. 18. Bouchard J, Mehta RL. Fluid accumulation and acute kidney injury: consequence or cause. Curr Opin Crit Care 2009;15: 509–13. 19. Basu RK, Andrews A, Krawczeski C, Manning P, Wheeler DS, Goldstein SL. Acute kidney injury based on corrected serum creatinine is associated with increased morbidity in children following the arterial switch operation. Pediatr Crit Care Med 2013;14:e218–24. 20. Liu KD, Thompson BT, Ancukiewicz M, et al. Acute kidney injury in patients with acute lung injury: impact of fluid accumulation on classification of acute kidney injury and associated outcomes. Crit Care Med 2011;39:2665. 21. Nguyen MT, Devarajan P. Biomarkers for the early detection of acute kidney injury. Pediatr Nephrol 2008;23:2151–7. 22. Fortenberry JD, Paden ML, Goldstein SL. Acute kidney injury in children: an update on diagnosis and treatment. Pediatr Clin North Am 2013;60:669–88. 23. Stafford-Smith M, Patel UD, Phillips-Bute BG, Shaw AD, Swaminathan M. Acute kidney injury and chronic kidney disease after cardiac surgery. Adv Chronic Kidney Dis 2008;15:257–77. 24. Chiravuri SD, Riegger LQ, Christensen R, et al. Factors associated with acute kidney injury or failure in children undergoing cardiopulmonary bypass: a case-controlled study. Paediatr Anaesth 2011;21:880–6. 25. Picca S, Ricci Z, Picardo S. Acute kidney injury in an infant after cardiopulmonary bypass. Semin Nephrol 2008;28:470–6. 26. Aranda JV, Lambert C, Perez J, Turmen T, Sitar DS. Metabolism and renal elimination of furosemide in the newborn infant. J Pediatr 1982;101:777–81. 27. Tuck S, Morselli P, Broquaire M, Vert P. Plasma and urinary kinetics of furosemide in newborn infants. J Pediatr 1983;103: 481–5. 28. Wells JC, Fewtrell MS, Davies PS, Williams JE, Coward WA, Cole TJ. Prediction of total body water in infants and children. Arch Dis Child 2005;90:965–71.