Renal resistive index is associated with acute kidney injury in COVID-19 patients treated in the intensive care unit

Background Renal resistive index (RRI) is a promising tool for the assessment of acute kidney injury (AKI) in critically ill patients in general, but its role and association to AKI among patients with Coronavirus disease 2019 (COVID-19) is not known. Objective The aim of this study was to describe the pattern of RRI in relation to AKI in patients with COVID-19 treated in the intensive care unit. Methods In this observational cohort study, RRI was measured in COVID-19 patients in six intensive care units at two sites of a Swedish University Hospital. AKI was defined by the creatinine criteria in the Kidney Disease Improving Global Outcomes classification. We investigated the association between RRI and AKI diagnosis, different AKI stages and urine output. Results RRI was measured in 51 patients, of which 23 patients (45%) had AKI at the time of measurement. Median RRI in patients with AKI was 0.80 (IQR 0.71–0.85) compared to 0.72 (IQR 0.67–0.78) in patients without AKI (p = 0.004). Compared to patients without AKI, RRI was higher in patients with AKI stage 3 (median 0.83, IQR 0.71–0.85, p = 0.006) but not in patients with AKI stage 1 (median 0.76, IQR 0.71–0.83, p = 0.347) or AKI stage 2 (median 0.79, min/max 0.79/0.80, n = 2, p = 0.134). RRI was higher in patients with an ongoing AKI episode compared to patients who never developed AKI (median 0.72, IQR 0.69–0.78, p = 0.015) or patients who developed AKI but had recovered at the time of measurement (median 0.68, IQR 0.67–0.81, p = 0.021). Oliguric patients had higher RRI (median 0.84, IQR 0.83–0.85) compared to non-oliguric patients (median 0.74, IQR 0.69–0.81) (p = 0.009). After multivariable adjustment, RRI was independently associated with AKI (OR for 0.01 increments of RRI 1.22, 95% CI 1.07–1.41). Conclusions Critically ill COVID-19 patients with AKI have higher RRI compared to those without AKI, and elevated RRI may have a role in identifying severe and oliguric AKI at the bedside in these patients.

intensive care unit (ICU) or in need of mechanical ventilation [1][2][3][4], of which up to one-third have been treated with renal replacement therapy (RRT) [5,6]. COVID-19 patients who develop AKI may have a 13-fold increased risk of death compared to those who do not develop AKI [4,7,8].
Renal resistive index (RRI) is an ultrasonographic Doppler measurement of flow velocities in intraparenchymal renal arteries. Normal values are around 0.60 [9,10] with 0.70 considered the upper normal threshold in adults [11]. Elevated RRI has shown promise in early detection and prognostication of AKI in mixed ICU populations [12][13][14][15][16], and the method seems feasible within the scope of point-of-care ultrasonography (POCUS) [17].
As thromboembolism and renal microangiopathy have gained interest as possible mechanisms giving rise to AKI in COVID-19 [18,19], RRI may show to be an especially helpful tool to guide diagnosis and treatment of AKI in these patients. In a recent case-control study, reduced renal perfusion and substantially elevated RRI were described in ten COVID-19 patients with severe AKI [20]. However, RRI has not been described in larger populations of COVID-19 patients and its role in these patients remains unclear. The aim of this study was to describe the pattern of RRI in relation to AKI in patients with COVID-19 treated in the ICU. We specifically investigated if there was an association between RRI and AKI diagnosis, different AKI stages and urine output.

Study population
This was an observational cohort study conducted in six ICUs designated for COVID-19 patients (COVID-ICUs) at the Karolinska University Hospital, Stockholm, Sweden. Four ICUs were at one of two sites, and two at the other site. On specific dates, patients in each COVID-ICU were screened for participation. Inclusion criteria were infection with Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) detected by a positive reverse transcriptase-polymerase chain reaction taken from upper or lower airways, admission to a COVID-ICU, and age ≥ 18 years. Exclusion criteria were end-stage renal disease (ESRD) before ICU admission, palliative phase of intensive care, ongoing irregular cardiac rhythm or treatment with extracorporeal membrane oxygenation (ECMO).
The study complied with the Declaration of Helsinki and was approved by the Swedish Ethical Review Authority. Requirement for signed informed consent was waived. A printed information sheet was sent to each patient or next of kin with the opportunity to retrospectively withdraw participation.

Definitions
AKI was defined according to the Kidney Disease Improving Global Outcomes (KDIGO) classification as an increase in serum creatinine concentration (sCr) and categorized into three stages [21]. The highest sCr from ICU admission to the day of RRI measurement was compared to baseline sCr. Baseline sCr was defined as the last known value measured in a disease-free phase before admission. When no previous sCr value existed, hospital admission sCr was used. The AKI stages were defined as the following: stage 1, ≥ 1.5-to 1.9-fold increase or an absolute increase ≥ 26 μmol/l; stage 2, ≥ 2.0-to 2.9fold increase; stage 3, ≥ 3.0-fold increase or an absolute increase > 354 μmol/l or initiation of RRT. The KDIGO urine output criteria were not used since hourly urine output was not always registered in the medical records. If the sCr elevation occurred more than 7 days before the RRI measurement and its value had returned to < 1.5fold or < 26 μmol/l higher than from baseline, the patient was evaluated as having recovered from an AKI episode and was classified into the no AKI group. Oliguria at the time of RRI measurement was defined as urine output < 0.5 ml/kg ideal body weight/hour for 24 h regardless of diuretic drug administration [22]. Ideal body weight was calculated using the gender-specific Acute Respiratory Distress Syndrome Network formula [23]. Chronic kidney disease (CKD) was defined as estimated glomerular filtration rate (eGFR) < 60 mL/min/1.73 m 2 , ESRD was defined as eGFR < 15 mL/min/1.73 m 2 , and eGFR was calculated using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation [24]. For this, baseline sCr was used when classifying patients as having CKD or ESRD, and a combination of sCr and cystatin C was used when calculating eGFR at the time of RRI measurement. Comorbidities were considered present if documented in the patient's medical record or if the patient was prescribed medication for the current state. The cardiovascular disease group included patients with cardiac failure, atrial arrhythmia, prior myocardial infarction or prior cardiac surgery.

RRI measurements
All RRI measurements were performed by one of two operators (MR and OJ). Both operators had more than one year's clinical experience of the RRI method. For each site a designated ultrasound device with a curvilinear probe of 1.0-6.0 MHz was used (GE Vivid S70N, US and GE Logiq E10, US at the two sites, respectively). The patients were examined in their ICU bed in supine or prone position depending on their respiratory requirements. Both kidneys were examined, and measurements were made on both or the most accessible side since the difference in RRI values between the right and left kidney has been shown to be negligible both in healthy and critically ill patients [9,12,17]. After obtaining a complete view of the kidney, color-Doppler was applied to visualize the global organization of intrarenal blood vessels. Pulse waved Doppler at the smallest possible width between 2 and 5 mm was used to measure flow velocities in an interlobular or arcuate artery in the upper, middle and lower kidney pole. The Doppler gain was set to obtain a clear outline of flow waves with minimal background noise. The pulse waved Doppler spectrum was considered optimal when at least three consecutive similar-looking waveforms for each pole were visualized. RRI was calculated for each pole as [(peak systolic velocityend-diastolic velocity)/peak systolic velocity]. From the pole RRI values, a mean RRI was computed.

Data collection
The following clinical data were collected for each patient at the time of RRI measurement: hemodynamic parameters, vasopressor requirements, sedatives dose and ventilator settings if mechanically ventilated. Severity of illness was graded on the day of measurement using the Sequential Organ Failure Assessment (SOFA) (originally the Sepsis-related Organ Failure Assessment) score [25]. Information on comorbidities, regular and current medication and laboratory data were collected from medical records.

Statistical analysis
Patient characteristics and variables are presented using frequencies and percentages for categorical data, and medians with interquartile range (IQR) and minimum/ maximum (min/max) values for continuous data. Clinical characteristics of patients with or without AKI were compared using Fisher's exact test for dichotomous variables and Wilcoxon rank-sum test for continuous variables. Median RRI between different groups were compared using the Wilcoxon rank-sum test. To elucidate potential confounding of the association between RRI and AKI, a multivariable logistic regression model was performed including RRI as a continuous and AKI as a dichotomous variable. Variables in our dataset that were associated with AKI with a p-value < 0.2 in bivariate logistic regression were selected for a manual forward selection procedure. Variables with the strongest association with AKI were included first, and variables that no longer were associated with AKI (p-value > 0.2) during forward selection were excluded. Remaining variables constituted the final model. Odds ratios (OR) and 95% confidence intervals (CI) were calculated. Goodness-of-fit of the model was assessed using the Hosmer-Lemeshow test. For all analyses a p-value < 0.05 was considered significant. The following variables had missing data: hospital admission sCr (n = 7 [14%]) and urine output at the day of RRI measurement (n = 2 [4%]). Missing data on height (n = 1 [2%]) were substituted with the median value according to sex. Statistical analyses were performed using Stata version 15.1 (StataCorp, College Station, US).

Results
Between April 15 and May 15 in 2020, the six COVID-ICUs were screened on specific dates for each ICU when at least one of the operators was available and able to perform measurements. Out of 71 screened patients, 20 were excluded, and a total of 51 patients were analyzed (Fig. 1).

Multivariable analysis
Bivariate and multivariable logistic regression analysis is presented in Table 2. Possible confounders eligible for inclusion in the multivariable model are all variables presented in Table 1 except those that could be influenced by   patients who had recovered from AKI earlier during the ICU course. Oliguric patients had higher RRI compared to non-oliguric patients. The association between RRI and AKI remained significant after adjustment for possible confounding.
Our results are in line with previous studies on mixed or septic ICU patients instead designed to investigate the role of RRI to predict and prognosticate AKI [12,13,16,26,27]. These studies have suggested RRI to be able to distinguish severe or persistent AKI from no or transient AKI, with optimal cut-off values for this discrimination varying from 0.69 to 0.80. The median RRI of 0.80 in patients with AKI in our population must be considered high in comparison, but may partly be due to the large proportion of patients with AKI stage 3. This is in line with the results of a recent case-control study which presented higher RRI in ten COVID-19 patients with AKI stage 3 compared to ten patients with AKI from septic shock [20]. Notably, also patients without AKI in our study had higher RRI (median 0.72) compared to non-AKI patients in ICU populations without COVID- 19 where reported values typically are lower than 0.65 [12,14,28]. It is not clear if elevated RRI in patients without AKI but infected with SARS-CoV-2 is specifically related to the infection itself, or if it reflects severity of illness as indicated by the long length of ICU stay as well as the high incidence of mechanical ventilation and vasopressor use in our population.
There is growing evidence of reduced renal microperfusion in patients with COVID-19-related AKI, both from contrast enhanced ultrasonographic measurements in patients with severe AKI [20] as well as from post-mortem findings of microvascular obstruction [29]. In addition, numerous studies have reported high rates of thrombotic complications in hospitalized COVID-19 patients in general [30][31][32], and thrombi in the renal microcirculation has been mentioned as a possible mechanism contributing to AKI in these patients [2]. Even if several renal and extrarenal factors influence the final profile of the RRI flow wave and value [33], it is conceivable that renal microcirculatory disturbances further contributed to the generally high RRI values observed in our population.
Previous studies on critically ill patients without COVID-19 have in general focused on prediction of AKI from RRI measurements performed within the first day of ICU admission [12,[14][15][16]26]. However, the ability of early RRI measurements to predict short-term AKI reversibility within 3 days recently has been challenged [34,35]. Our finding of higher RRI in COVID-19 patients with an ongoing AKI episode compared to patients who had recovered from an AKI episode earlier during the ICU course suggests that RRI values decrease with recovered renal function. This indicates that RRI also might have a role later in the ICU or hospital course, but its exact role for prediction of renal recovery or progression towards CKD while the patient still is in hospital needs to be investigated in properly designed studies. In outpatients with already established CKD, RRI ≥ 0.70 notably has shown to be predictive of both CKD progression [36,37] and mortality [38].
Our finding of elevated RRI in COVID-19-related AKI indicates that RRI may have a role in the assessment of this new and unique disease. Further, RRI has shown potential as a precocious ICU monitoring tool for detecting progression of shock states [39,40], and together with Doppler assessments of other splanchnic organs RRI can expand the bedside monitoring window for hypoperfusion in critically ill patients in general [41]. As a non-invasive and repeatable method that has been demonstrated to be fast to learn also for non-experienced sonographers [17,42], RRI should be applicable within POCUS protocols for ICU clinicians and thereby contributing as a valuable tool in the present resource scarce times of a pandemic.
Our study has several limitations. First, our study population was small and although we were still able to adjust our analysis for a number of possible confounders, the results should therefore be interpreted with caution. The small number of patients with AKI stage 1 and 2 further makes it difficult to draw conclusions on the association between RRI and AKI in these specific subgroups. Second, due to the challenges of conducting clinical research during a pandemic, RRI measurements were performed at different time points in different patients and most of the measurements were made late in the ICU course. The Karolinska University Hospital is a tertiary referral hospital, and many patients were transferred to its ICUs from other hospitals when they had already received several days of intensive care. This meant there was a delay from ICU admission to accessibility in some of the patients, and we were therefore not able to perform early measurements to investigate the ability of RRI as an early predictor of subsequent AKI development. Third, intra-and inter-observer variability for the operators were not investigated. Our group has previously shown that RRI measurements by inexperienced sonographers were reliable, accurate and precise compared to an expert after only a brief training session [17], and both operators in our study were experienced with the RRI method. Lastly, our study was affected by some of the well-known pitfalls in AKI research. The use of hospital admission sCr as baseline level in patients in whom pre-admission sCr was missing might have resulted in an underestimation of the AKI incidence. Further, using sCr decline to define recovery from an AKI episode could in patients with muscle wasting during a prolonged ICU course lead to overestimation of renal function recovery [43]. However, we used eGFR calculations based on a combination of sCr and cystatin C at the time of RRI measurement and still observed a difference in estimated renal function between patients classified with or without AKI, suggesting any such misclassification was negligible.

Conclusion
Critically ill COVID-19 patients with AKI have higher RRI compared to those without AKI, and elevated RRI may have a role in identifying severe and oliguric AKI at the bedside in these patients. The exact role of RRI as a POCUS application for AKI assessment and monitoring of ICU patients with COVID-19 should be established in further studies.