Introduction
Acute kidney injury is a common postoperative complication of paediatric cardiac surgery that is associated with increased length of stay, morbidity, and mortality. Reference Weld, Kim and Chandra1–Reference Blinder, Goldstein and Lee5 Associations between acute kidney injury and fluid overload, medication selection, intraoperative renal regional oxygen saturation, energy consumption, and vasoactive-mediated high systemic vascular resistance have previously been established. Reference Chiu, Fong and Lazzareschi6–Reference Zhang, Zhou, Wang, Guo and Zhou12 Clinically, acute kidney injury is identified by a change in serum creatinine, which often rises after the injury has begun. This delay in diagnosis may delay early recognition and intervention. Many strategies have been employed for the prevention and detection of acute kidney injury; however, there are still evolving data regarding early detection of acute kidney injury with real-time metrics in this vulnerable population.
The purpose of this study is to characterise associations between haemodynamic parameters, clinical parameters, and medical interventions for acute kidney injury using high-fidelity haemodynamic data.
Methods
Study design
This study protocol was approved by the institutional review board at Advocate Health (protocol 2039614-1). It is in concordance with the Helsinki Declaration. This study was a single-centre, retrospective study aimed to characterise the association between various clinical parameters and acute kidney injury. The main aim of this study was not necessarily the resulting model and ability to predict renal function but rather to demonstrate the relationship between the independent continuous variables and serum creatinine.
Variables of interest
The variables of interest collected included central venous pressure, heart rate, respiratory rate, arterial blood pressure (systolic, diastolic and mean), arterial saturation by pulse oximetry, renal near infrared spectroscopy, peak airway pressure, mean airway pressure, positive end expiratory pressure, body temperature, fluid balance, and vasoactive inotropic score. Patient weight and gestational age were also collected. Cardiopulmonary bypass time and aortic cross clamp time were also collected.
All the data except for vasoactive doses, cardiopulmonary bypass time, and aortic cross clamp time were collected from the T3 software. T3 software is designed to integrate multiple clinical data streams in real time. The data from all the streams can then be displayed by the software in a user-defined fashion. Additionally, T3 also estimates the venous saturation and then displays the probability of the venous saturation being under 30%, 40%, or 50% in a metric known as the index of inadequate delivery of oxygen. The T3 software collects data from the available streams at an interval of 5 s, thus offering high temporal resolution. Data are captured by T3 only postoperatively at this institution.
Central venous pressures were obtained by use of femoral lines terminating in the inferior caval vein. Line placement was confirmed by radiographs.
Renal near infrared spectroscopy values were collected. Near infrared spectroscopy values were obtained using the Casmed ForeSight Elite tissue oximeter.
Vasoactive doses were collected manually via the electronic medical record as charted. It is local practice to document every time an infusion dose has been changed and at regular intervals. Doses of all vasoactive infusions were collected for each timepoint at which the data from T3 were collected.
Fluid balance was collected manually through the electronic medical record as charted. It is local practice to update fluid balance hourly. Fluid balance for each timepoint at which T3 data were collected was collected as the fluid balance for the hour prior to that timepoint. Thus, the collected data were the absolute net fluid balance in millilitres for the preceding hour.
Certain values were calculated. Renal oxygen extraction was calculated as [(arterial saturation by pulse oximetry – renal near infrared spectroscopy)/(arterial saturation by pulse oximetry)] × 100. For example, if the arterial saturation was 80% and the renal near infrared spectroscopy value was 60%, the renal oxygen extraction would be 25%.
Renal perfusion pressure was calculated as (mean arterial pressure – central venous pressure).
Oxygen consumption in mL/min was estimated using the LaFarge equation. This estimated oxygen consumption was then used to further estimate systemic blood flow using the renal near infrared spectroscopy as the venous saturation. Systemic vascular resistance was then estimated using the collected mean arterial blood pressure and central venous pressure and the estimated systemic blood flow.
As all values will have outliers due to technical or mechanical reasons, the highest and lowest 5% were eliminated.
Acute kidney injury
Acute kidney injury was defined as per the Kidney Disease Improving Global Outcomes serum creatinine to baseline creatinine criteria. Reference Levey, Eckardt and Dorman13 Acute kidney injury was deemed present if the current creatinine to baseline creatine was greater than 1.5. For this study, the urine output criteria set was not utilised.
Serum creatinine values were typically obtained every 24 hours. The blood urea nitrogen and serum creatinine values were linked with the preceding 24 hours of haemodynamic data with the assumption that the preceding haemodynamic state impacted the resulting creatinine.
Patient inclusion
Neonates with functionally univentricular hearts who underwent a Norwood operation were eligible for inclusion in this study. Reference Jacobs, Franklin and Beland14,Reference Jacobs, Franklin and Beland15 Data must have been collected and available in T3 for patients to be included in this study. T3 was implemented locally on September 1st, 2022 and a final inclusion date of March 1st, 2023 was utilised. Data were only included while patients were intubated and mechanically ventilated to allow for quantification of airway pressures. Data were available at five-second intervals for patients with T3 data. Datapoints were included in the final analyses only if there was a corresponding central venous pressure and airway pressure available. Patients requiring extracorporeal membrane oxygenation prior to the Norwood were excluded.
Statistical analyses
First, the presence of acute kidney injury was determined for each data point using the aforementioned criteria. Patients were separated into two groups: those with acute kidney injury and those without. Qualitative data between the two groups were compared using a Fisher exact test while quantitative data between the two groups were compared using a Mann–Whitney-U test.
Next, LASSO-based machine learning techniques were employed to conduct a logistic regression analysis to model serum creatinine. Thus, acute kidney injury was entered as the dependent variable and all other collected data points were entered as independent variables. Estimated systemic vascular resistance and estimated systemic blood flow were not entered into the regression as they are not easily available at the bedside.
Next, receiver operator curve analyses were conducted with acute kidney injury as the outcome of interest to evaluate the ability of variables of interests to predict acute kidney injury. The area under the curve for each analysis was calculated. Additionally, an optimal cut-off point was also identified for each of the studied variables. Optimal cut-off points were determined by identifying the point farthest away from the diagonal reference line. The sensitivity and specificity of each variable were calculated using the identified cut-off point. Variables retained in the previously conducted logistic regression had a receiver operator curve analysis done.
An acute kidney injury risk score was created using the results from the logistic regression. Variables retained in the regression model were included in the risk score and the corresponding unstandardised coefficients from the model were used as their coefficients in the risk score. This score was calculated at each time point and then compared at time points with and without acute kidney injury. A receiver operator curve analysis was then done with the acute kidney injury risk score and the binary outcome of acute kidney injury. Once again, an area under the curve, optimal cut-off point, and sensitivity and specificity were assessed.
Next, LASSO-based machine learning techniques were employed to conduct a linear regression analysis to model serum creatinine. Thus, serum creatinine was entered as the dependent variable and all other collected data except for blood urea nitrogen and serum creatinine levels were entered as independent variables. Estimated systemic vascular resistance and estimated systemic blood flow were also not entered into the regression as they are not easily available at the bedside.
All statistical analyses were done using SPSS Version 23.0 and manually input syntax. Any use of the word “significant”, “significantly”,” or “significance” will refer to statistical significance unless explicitly specified otherwise. A p-value of less than 0.05 was considered statistically significant. Missing values were left missing and not imputed.
Peri- and postoperative management
A majority of Norwood operations at our institution are done utilising a 5 mm right ventricle to pulmonary artery conduit. Only one child in the current study cohort received a Blalock–Taussig–Thomas shunt.
Patients at our institution are generally taken to the operating room where they are endotracheally intubated under general anaesthesia. A right radial arterial line or femoral arterial line is placed using a surgical cutdown method. Next, a femoral central venous line is placed using a surgical cutdown method. The neck, chest, and abdomen are then sterilely prepared. A median sternotomy is then performed with dissection until the sternum is exposed. The sternum is then divided with an oscillating saw. The patient is then heparinised and cannulated and placed on cardiopulmonary bypass, generally with an activated clotting time of over 400 ms. Once the heart is decompressed, the pulmonary arteries are controlled and the pulmonary trunk is transected. A 5 mm Gore-Tex graft is then anastomosed to the branch pulmonary arteries in end-to-side fashion.
Next, an atriotomy is performed to inspect the atrial septum and remaining atrial septum is excised.
Cardioplegia is then administered and antegrade cerebral perfusion is initiated. An aortotomy is performed on the underside of the aortic arch with a corresponding vertical incision made on the pulmonary trunk. These are then anastomosed together with pulmonary homograft used to complete this anastomosis.
Atrial pacing wires are then placed along with two sets of chest drains coursing through both pleural cavities. The skin is then closed with the sternum left open. A sterile dressing is placed.
Average cardiopulmonary bypass time for this cohort was 150 minutes with an average aortic cross clamp time of 64 minutes.
Patients are brought to the cardiac ICU intubated. Balanced fluids are utilised and are generally run at 100 ml/kg/day. Boluses of balanced fluids are given as needed and generally are about 20 to 40 ml/kg in the first 12 postoperative hours. Vasoactive support generally consists of epinephrine and/or dopamine in the initial 24 hours with milrinone introduced later. Patients are generally atrially paced to maintain a heart rate of 150 beats per minute for the first 12 hours. This is done to augment cardiac output. Sedation generally consists of a dexmedetomidine infusion and a fentanyl infusion. Paralysis is utilised on a patient-to-patient basis depending on the clinical situation. Mechanical ventilation is usually achieved using a pressure responsive volume control mode with tidal volumes of 6 to 8 ml/kg. A partial pressure of carbon dioxide of 35 to 45 mmHg is targeted which generally requires a ventilatory rate of 35 to 40 breaths per minute. Intermittent boluses of both are utilised. Clinical care is titrated to either systemic oxygen delivery monitored by cerebral near infrared spectroscopy, renal near infrared spectroscopy, and measured venous saturations. Some clinicians will also utilise blood pressure goals. These goals are physician dependent. Arterial blood gases are drawn periodically and the serum lactate is also measured simultaneously. The sternum is generally closed at the bedside in the first 48 hours. Extubation generally occurs in the first 96 hours.
None of the included patients required peritoneal dialysis catheters to be placed postoperatively or required or experienced arrhythmias requiring intervention.
Results
Cohort information
A total of 27,270 datapoints from nine patients over 1,338 patient hours (55.7 days) were included in the final analysis. As per the inclusion criteria for retaining datapoints in the final analysis, central venous pressure and airway pressures must have been available at the same timepoint to be included.
The average gestational age at birth was 38 weeks, with two patients being premature. The mean 1-minute APGAR was 8 and the mean 5-minute APGAR was 9. The lowest 1-minute APGAR was 5 and the lowest 5-minute APGAR was 7. Six patients were born via caesarean section while the remainder by vaginal delivery. Average patient age at the time of the Norwood operation was 20 days. This was due to two patients first undergoing a hybrid procedure, and later the Norwood procedure closer to 2 months of life. The mean age at time of Norwood for the other seven patients was 2 days. Of the nine patients for whom data were collected, two had an identified genetic anomaly (Table 1).
Table 1. Descriptive data regarding cohort
All patients had a downtrending creatinine before going to the operating room for their Norwood.
The primary cardiac diagnosis was hypoplastic left heart syndrome in five patients, tricuspid atresia in two, and small left sided structures in the remaining two. All patients remained with functionally univentricular palliation. All but one patient had a prenatal diagnosis. Only one patient had a restrictive atrial septum.
Only one child was intubated prior to the Norwood procedure and remained intubated until the operation. This same child was the only child who required vasoactive support prior to the Norwood operation. All but the one patient who was intubated were able to feed by mouth but no nasogastric feeds were administered to any children pre-operatively. There were no pre-operative cardiorespiratory arrests.
Acute kidney injury
Of the nine patients included, two (22%) experienced acute kidney injury at some point. Collectively, acute kidney injury was present for a total of 4.3% of the patient monitoring time. Of the monitoring time during which acute kidney injury was present, acute kidney injury was stage 1 for 83% of the time and stage 2 for the remaining 17%. There was no time spent in stage 3 acute kidney injury (Table 2).
Periods of acute kidney injury were preceded by statistically significantly higher vasoactive inotropic score, higher mean airway pressure, higher systolic arterial pressure, higher mean arterial pressure, higher diastolic arterial pressure, higher central venous pressure, higher central venous pressure, higher renal perfusion pressure, lower arterial saturation, lower heart rate, higher respiratory rate, lower absolute renal near infrared spectroscopy values, higher fluid balance, lower estimated systemic blood flow, and higher estimated systemic vascular resistance. Histograms for renal oxygen extraction and renal perfusion for those with and without acute kidney injury are demonstrated in Figures 1 and 2.
Figure 1. Histograms depicting the percentage of time spent at specific renal oxygen extraction values. The x-axis is the renal oxygen extraction while the y-axis is the percent of time. The top panel is data for those with acute kidney injury while the bottom panel is for those without acute kidney injury.
Figure 2. Histograms depicting the percentage of time spent at specific renal perfusion pressures. The x-axis is the renal perfusion pressure while the y-axis is the percent of time. The top panel is for those with acute kidney injury while the bottom panel is for those without acute kidney injury.
A LASSO logistic regression conducted to model acute kidney injury demonstrated that net fluid balance and renal oxygen extraction were independently associated with acute kidney injury. Of note, none of the pressure metrics (systolic blood pressure, mean arterial blood pressure, diastolic blood pressure, central venous pressure, or renal perfusion pressure) were retained in the regression model.
Acute kidney injury risk score
The resulting acute kidney injury risk score was (4.1 × fluid balance) + (1.9 × renal oxygen extraction).
The acute kidney injury risk score was significantly higher in acute kidney injury at 32.9 compared to 7.9 (p < 0.01). Receiver operator curve analyses demonstrated an area under the curve of 0.73 for this risk score in identifying acute kidney injury. The optimal cut-off point was found to be 52. At this cut-off, the score had 96% sensitivity, 90% specificity, 12% positive predictive value, 99% negative predictive value, and 89% accuracy.
Receiver operator curve analyses
Receiver operator curve analyses of variables to identify acute kidney injury demonstrated that fluid balance had an area under the curve of 0.55 while renal oxygen extraction had an area under the curve of 0.62. The optimal cut-off point for fluid balance was found to be 7.0 mL/hr while the optimal cut-off point for renal oxygen extraction was found to be 29%.
Regression analyses, creatinine ratio
The LASSO linear regression conducted to identify factors associated with an increase in the serum creatinine to baseline creatinine ratio identified the following factors (in order of impact): higher mean airway pressure, higher renal oxygen extraction, higher mean arterial blood pressure, higher vasoactive inotropic score, and higher fluid balance. There was also an association with renal perfusion pressure and heart rate.
Discussion
High-fidelity data from the current study demonstrate that higher net fluid balance and higher renal oxygen extraction are associated with acute kidney injury after the Norwood operation. A fluid balance greater than 7 mL/hr and renal oxygen extraction greater than 29% were found to be independently associated with acute kidney injury as defined by the Kidney Disease Improving Global Outcomes serum creatinine criteria.
Additional associated factors were higher mean airway pressure, higher renal oxygen extraction, higher renal perfusion pressure, and higher vasoactive inotropic score.
In a clinical context, it appears that increasing creatinine to baseline creatinine ratio and acute kidney injury are correlated with lower cardiac output and higher systemic vascular resistance states. Although this correlation may be intuitive, this study also demonstrated that a higher vasoactive inotropic score and higher renal perfusion pressure were associated with increasing creatinine to baseline creatinine ratio, contrary to commonly held clinical belief. There was a correlation between renal perfusion pressure, vasoactive inotropic score, and systemic vascular resistance indicating that use of vasoconstrictor therapies to increase renal perfusion pressure is likely not beneficial and may in fact be harmful.
This is admittedly not a widely accepted clinical management phenomenon but has been demonstrated previously in adult studies. One in particular suggested a strategy of increased vasopressor support and decreased fluid resuscitation was associated with greater risk of acute kidney injury in 281,010 adults who underwent abdominal surgery across 26 separate centres. Reference Chiu, Fong and Lazzareschi6
Conceptually, this is understandable given that organs require oxygen and systemic oxygen delivery is a product of oxygen content and cardiac output. Oxygen content further breaks down into the bound and dissolved components that are mediated by haemoglobin, saturation of oxygen, and the partial pressure of oxygen. Cardiac output is characterised by the Fick principle and becomes the quotient of oxygen consumption and the arteriovenous oxygen content difference. This implies that cardiac output is inversely proportional to the arteriovenous oxygen content difference. Therefore, oxygen delivery is objectively not characterised by blood pressure.
Pressure metrics have been used as surrogate markers of oxygen delivery because clinical pressure monitoring has been more readily available than blood gas analysis for over 200 years. In fact, pressure is the product of flow and resistance. Thus, changes in systemic blood pressure may be due to an increase in either systemic blood flow or systemic vascular resistance. Unless, systemic vascular resistance can be quantified, a change in blood pressure cannot be assumed to be secondary to an increase in systemic blood flow. In fact, systemic blood pressure measurements may go down despite maintenance of, or even increase in, systemic blood flow, if there is a decrease in systemic vascular resistance. Reference Loomba16–Reference Sheikholeslami, Dyson and Villarreal18 Of these two, only systemic blood flow contains oxygen while systemic vascular resistance does not. Hence, an increase in systemic blood flow contributes to oxygen delivery while increasing systemic vascular resistance exclusively does not. This has been demonstrated by the current study and several previous studies. Reference Bronicki, Sebastin and Savorgnan19–Reference Loomba, Flores and Bronicki21
Increasing renal perfusion pressure does not result in increasing renal oxygen delivery. In fact, increasing renal perfusion pressure by increasing renal vascular resistance only may lead to a decrease in renal blood flow and thus renal oxygen delivery. The fraction of renal plasma flow that is filtered across the glomerulus is known as the filtration fraction and is the quotient of glomerular filtration rate and renal plasma flow. The kidney itself maintains glomerular filtration rate by two main mechanisms: the myogenic mechanism and the tubuloglomerular feedback mechanism. In the myogenic mechanism, increased flow in the renal afferent leads to increased stretch which when detected leads to a feedback loop that then constricts the afferent vessels. Reference Fage, Asfar, Radermacher and Demiselle22 In the tubuloglomerular feedback mechanism, chloride ion concentration is by the macula densa. If the chloride ion concentration is sensed to be low, then the afferent vessels are dilated. If the chloride ion concentration is sensed to be high, then the afferent vessels are constricted. Reference Fage, Asfar, Radermacher and Demiselle22,Reference Wilcox23 Neither of the two intrinsic renal feedback mechanisms (myogenic and tubuloglomerular) for maintaining the glomerular filtration rate directly include a pressure phenomenon.
Pressure does come into play when the Starling equation is considered which equates glomerular filtration rate to a function of the following variables: the filtration coefficient of the glomerulus, glomerular capillary hydrostatic pressure, hydrostatic pressure of fluid inside Bowman’s capsule, the oncotic pressure in the glomerular capillary blood, and the oncotic pressure of the fluid inside Bowman’s capsule. While systemic blood pressure does impact glomerular hydrostatic capillary pressure, there are several mechanisms which regulate this, and the glomerular filtration rate is most sensitive to pressures within the glomerulus itself. Furthermore, regulation of resistance within the afferent renal vessels is a complex process and is not solely dependent on regional vascular resistance (similar to systemic vascular resistance).
Thus, renal physiology itself demonstrates the shortcomings of maintaining renal function by titrating blood pressure. While there is a critical threshold of renal artery pressure below which renal blood flow and glomerular filtration rate become dependent on mean arterial pressure, this is quite low and this can be noted by closely following oxygen indices such as renal oxygen extraction. This is in concordance with the current study’s findings of the absence of an association between renal perfusion pressure or mean arterial blood pressure with acute kidney injury. This is further supported by this study’s finding that renal oxygen extraction ratio is associated with acute kidney injury.
Medications such as norepinephrine, epinephrine, and vasopressin have been demonstrated to be associated with reductions in renal blood flow in healthy humans by increase in afferent and efferent glomerular vascular resistance. Reference Smythe, Nickel and Bradley24–Reference Sato, Matsuzawa and Eguchi30 Glomerular filtration rate is maintained early with these changes as the resulting vasoconstriction is greater in the efferent vessels, although ultimately glomerular filtration rate may decrease. Reference Hellebrekers, Liard, Laborde, Greene and Cowley31–Reference Barclay, Cooke and Kenney33 Studies have demonstrated these findings even in the setting of an increase in systemic blood pressure. Reference Shenasky, Gillenwater, Graham and Wooster34 Although not reproduced in human studies, there are animal studies that have demonstrated that epinephrine may improve renal blood flow or leave it unchanged.
Findings regarding dopamine have been more variable with some noting an increase in renal blood flow and others a decrease in renal blood flow. Reference Sato, Matsuzawa and Eguchi30,Reference Lauschke, Teichgräber, Frei and Eckardt35–Reference Drieghe, Manoharan and Heyndrickx39 The variable effects of dopamine seem, in part, related to underlying clinical setting. Reference Drieghe, Manoharan and Heyndrickx39 Even some studies demonstrating improved renal blood flow associated with dopamine still failed to find a change in renal oxygen economy or function. Reference Day, Phu and Mai40 Clinical studies have also failed to find consistent improvement in renal function associated with dopamine. Reference Kellum and Janine41–Reference Prins, Plötz, Uiterwaal and van Vught45
The current study’s findings that a higher vasoactive inotropic score was associated with acute kidney injury are consistent with earlier published findings, Reference Weld, Kim and Chandra1 although it must be acknowledged that the current study can only characterise associations and not firm causality. It is worth noting that a majority of increases in vasoactive doses tended to precede the development of acute kidney injury, not follow it.
Clinically, there are few acute kidney injury studies that include both renal perfusion pressure and renal oxygen delivery metrics. Those that do include these parameters have demonstrated that renal oxygen delivery has more prognostic ability than renal perfusion pressure. Reference Weld, Kim and Chandra1,Reference Choi, Kim and Chin8 This is supported by the findings of this current study as well. Some studies that demonstrate an association between higher renal perfusion pressure and improved clinical outcomes do not include metrics for renal oxygen delivery. If renal vascular resistance remains relatively constant in a patient cohort and pressure goes up, this implies that blood flow and subsequently oxygen delivery have improved. Presumably, if renal oxygen delivery metrics were included, the relative importance of each would be able to be quantified as it has with the current study. In the current study, there was a positive correlation of renal perfusion pressure with creatinine ratio, contrary to widely held clinical belief. This may be due to a pharmacologic means of increasing renal perfusion pressure which had other direct or indirect effects on overall or renal haemodynamics. Reference Hendon, Kane and Golem46,Reference Loomba and Flores47 Future studies should include both to help further characterise the utility of both the metrics.
A recent study by Fragasso and colleagues evaluated data from 419 postoperative paediatric cardiac surgical patients after biventricular repair. Using machine learning techniques using multiple timepoints for each patient, they try to develop a predictive model for acute kidney injury. Several laboratory markers, clinical variables, and haemodynamic variables were included in this study, and the importance of systolic, mean, and diastolic arterial blood pressures was found to be quite low in the resulting models. Reference Fragasso, Raggi, Passaro, Tardella, Lasinio and Ricci48
The findings from the current study highlight important clinical implications. First, clinical monitoring should be based on oximetric indices and pressure-based indices, not simply pressure-based indices alone. These indices can include venous saturations measured by blood gas analysis or near infrared spectroscopy. Reference Loomba, Rausa and Sheikholeslami49–Reference Dabal, Rhodes, Borasino, Law, Robert and Alten51 The benefit of oximetric indices has been previously described. Reference Joffe, Al Aklabi and Bhattacharya2,Reference Dorum, Ozkan, Cetinkaya and Koksal9 Subsequently, clinical intervention should consider oximetric indices along with other parameters to ensure adequate oxygen delivery. Additionally, vasoactive medications may not provide the benefit in renal function that is often anecdotally expected.
The current study utilises high-fidelity haemodynamic monitoring data, allowing for granular analyses of many datapoints of various parameters. As the analyses are done at frequent timepoints, this results in a large sample size with adequate statistical power. Additionally, the regression techniques utilised here are particularly well suited for the data. Similar analyses to model cardiac arrest in paediatric patients have utilised similar methods as well. Reference Bose, Verigan and Hanson52,Reference Bose, Verigan and Hanson53 There are few acute kidney injury studies that have done this to date. It has been demonstrated that patients undergoing the Norwood procedure are at higher risk of acute kidney injury. The methodology utilises machine learning techniques to characterise the associations present in these data. While these are strengths of this study, there are also certainly limitations. First, this is a single-centre retrospective study and study specific practices may impact the development of acute kidney injury and the natural history of acute kidney injury within the institution. The retrospective nature of the data also makes it difficult to comment on causality although the high temporal resolution does allow some insight into causality. Additionally, since the focus was primarily on physiologic parameters, additional medications were not captured to quantify their effect. These include agents such as acetaminophen, dexmedetomidine, fenoldopam, aminophylline, furosemide, bumetanide, and chlorothiazide, among others, which have previously been described as possibly having an association with acute kidney injury. Reference Nater, Wong, Ikeda, Heenan, Loomba and Penk7,Reference Van den Eynde, Cloet and Van Lerberghe10,Reference Loomba, Villarreal and Dhargalkar54–Reference Loomba, Uppuluri and Chandra57 The study design is really only able to identify associations and not causation. Additionally, while the sample size is the number of data points not the number of patients, this is still a study with a low number of patients. Findings must be interpreted with this taken into consideration.
Conclusion
Among patients with univentricular physiology undergoing the Norwood procedure, increased risk of acute kidney injury. Renal perfusion pressure is not significantly associated with acute kidney injury.
Table 2. Comparison between variables in those with and without acute kidney injury
Acknowledgements
None.
Financial support
None.
Competing interests
None.
Ethical standard
None.
