Critically Ill COVID-19 Patients With Acute Kidney Injury Have Reduced Renal Blood Flow and Perfusion Despite Preserved Cardiac Function: A Case-Control Study Using Contrast-Enhanced Ultrasound

Abstract

Background: 

Acute kidney injury (AKI) is a common complication of COVID-19 critical illness but the pathophysiology is uncertain. Some evidence has indicated that a vascular aetiology may be implicated. We used contrast-enhanced ultrasound (CEUS) and echocardiography to study renal perfusion and global blood flow and compared our findings with measurements taken in a group of septic shock patients and healthy volunteers.

Methods: 

Prospective case-control study. Renal perfusion variables were assessed with CEUS; macrovascular blood flow was assessed using Doppler analysis of large renal vessels; echocardiography was used to assess right and left heart function and cardiac output.

Results: 

CEUS-derived parameters were reduced in COVID-19 associated AKI compared with healthy controls (perfusion index 3,415 vs. 548 a.u., P = 0·001; renal blood volume 7,794 vs. 3,338 a.u., P = 0·04). Renal arterial flow quantified using time averaged peak velocity was also reduced compared with healthy controls (36·6 cm/s vs. 20·9 cm/s, P = 0.004) despite cardiac index being similar between groups (2.8 L/min/m² vs. 3.7 L/min/m², P = 0.07). There were no differences in CEUS-derived or cardiac parameters between COVID-19 and septic shock patients but patients with septic shock had more heterogeneous perfusion variables.

Conclusion: 

Both large and small vessel blood flow is reduced in patients with COVID-19 associated AKI compared with healthy controls, which does not appear to be a consequence of right or left heart dysfunction. A reno-vascular pathogenesis of COVID-19 AKI seems likely.

INTRODUCTION

In December 2019 a new disease now termed COVID-19, emerged in China and subsequently became a global pandemic; the causative agent was identified as a novel coronavirus, SARS-COV-2 ⁽¹⁾. Acute kidney injury (AKI) occurs in a substantial proportion of patients who are critically ill with COVID-19. Early published literature from China reported AKI prevalence between 5% and 29% among patients admitted to intensive care ^{(2, 3)}. However, more recent reports from Europe and North America suggest AKI complicates 27% to 40% of COVID-19 intensive care unit (ICU) admissions ^{(4, 5)}.

AKI of any stage is an independent risk factor for in-hospital death in patients with COVID-19, with greater mortality associated with more severe kidney damage ^{(2, 4)}. The pathogenesis of AKI in patients with sepsis is multifactorial and complex and despite many decades of research there remain important unanswered questions ^{(6, 7)}. AKI developing as a result of COVID-19 is likely to be equally complex and due to the emergent nature of the disease there are limited data ⁽⁸⁾. Much of the limited current evidence comes from autopsies which have demonstrated the presence of viral inclusion bodies within tubular epithelial cells and acute tubular necrosis alongside endothelial injury and erythrocyte aggregation within peri-tubular capillaries ⁽⁹⁾. These findings contribute to the hypothesis advanced by some researchers that COVID-19 disease induces widespread changes in vascular endothelial cells and that resultant changes in regional blood flow may be important in the pathogenesis of COVID-19 related critical illness ⁽¹⁰⁾.

Our group is currently recruiting to an ongoing study, Microshock Renal (Clinicaltrials.gov NCT03713307) which aims to explore the pathogenesis of septic AKI and in particular the contribution of hypoperfusion in patients with septic shock. The protocol for this study has been published elsewhere and uses a relatively new technique, contrast-enhanced ultrasound (CEUS), to visualize perfusion within the renal parenchyma, alongside measures of global renal blood flow and cardiac output ⁽¹¹⁾.

The aim of the current study is to use CEUS and measures of global blood flow to investigate renal perfusion in patients with COVID-19 and established AKI. Comparison is made to a group of patients with established AKI resulting from septic shock and healthy volunteers. We hypothesized that patients with COVID-19 associated AKI would have evidence of reduced renal perfusion when compared with healthy volunteers and patients with sepsis associated AKI.

PATIENTS AND METHODS

Study design

In this observational prospective case-control study critically-ill patients with COVID-19 and Kidney Disease Improving Global Outcomes (KDIGO) stage 3 AKI were recruited ⁽¹²⁾. Renal ultrasound, including CEUS and a trans-thoracic echocardiogram, were performed at a single time point. Results were compared with a cohort of patients with septic shock and stage 3 AKI recruited as part of the ongoing Microshock Renal study. A third series of measurements obtained from healthy volunteers was used as a negative control group.

Ethical approval

Ethical approval for this study was obtained from the Yorkshire and Humber (Leeds East) Research Ethics committee (18/YH/0371). In view of the lack of capacity for consent of enrolled patients, an emergency waiver of consent was approved after ethical review and deferred patient consent for data usage was sought.

Patient selection

• COVID 19 group: Adult patients with confirmed positive polymerase chain reaction for SARS-COV-2 infection and KDIGO stage 3 AKI treated on one of the intensive care units of King's College Hospital. All patients were enrolled at a single point in time on the same day and those with the shortest duration of AKI were preferentially selected.
• Septic shock group: Adult patients with suspected or actual infection, a requirement for vasopressor therapy and a lactate >2 mmol/L after fluid resuscitation, recruited within 48 h of ICU admission as part of the ongoing Microshock Renal study. All patients, recruited to date, with analyzed data and KDIGO grade 3 AKI on the final day of study measurements (Day 4) were included in the analysis.
• Healthy controls recruited as a negative comparator group in the Microshock Renal study.

Exclusion criteria

Patients were excluded if they had a body mass index greater than 40 kg/m² due to probable poor CEUS image quality, severe chronic kidney disease defined as a baseline estimated glomerular filtration rate of 30 mL/min or less, calculated using pre-admission or admission creatinine, a previous renal transplant, ultrasonographic appearances of chronic kidney disease (bright, shrunken kidneys (<9 cm length)), severe pulmonary hypertension (>90 mm Hg), or known intolerance to SonoVue contrast agent.

Sample size

This exploratory study was undertaken rapidly in the midst of a pandemic setting. As such a formal power calculation was not made. Instead a convenience sample size was selected, based on what was achievable in the time period available.

Renal contrast-enhanced ultrasound (CEUS)

Ultrasound scans for all three cohorts were obtained by a single investigator (JW) using an Affiniti ultrasound system (Philips, UK). Training in the technique was provided by a highly experienced user (DH) and JW has acquired and analyzed over 500 images to date as part of the ongoing Microshock Renal study. Images were assessed for quality and suboptimal images, not suitable for VueBox analysis, were subsequently excluded. Both kidneys were identified and the maximal length measured. Provided both kidneys were of normal size, the kidney with the better sonographic views was selected for study. Pulsed-wave Doppler (PW) was aligned with an interlobular vessel and peak systolic velocity and end-diastolic velocity were recorded for the calculation of resistive index (RI), a measure of vascular pulsatility ⁽¹³⁾. An infusion of 2.4 mL of SonoVue (Bracco SpA, Milan, Italy), an ultrasound contrast agent was administered intravenously at a rate of 1 mL/min by a dedicated infusion pump. Ultrasound images of the entire examination were digitally recorded using a low mechanical index technique (range 0.04–0.1). Within 2 min of infusion, steady state was reliably achieved and later confirmed by off-line analysis. Contrast destruction–replenishment sequences were then undertaken every 30 s from 2 min for five cycles.

CEUS analysis

Figure 1 shows an example of grayscale ultrasound and CEUS images. Dual mode views of both grayscale and low-MI modes were analyzed offline with dedicated software (VueBox, Bracco SpA, Milan, Italy) to determine the replenishment kinetics of cortical and medullary regions of interest (ROI). For each patient one ROI was selected from the renal cortex and two other ROIs representing deeper medullary areas. Care was taken to exclude areas that contained highly vascular or avascular regions. Time-intensity curves were calculated for each ROI as illustrated in Figure 2A. Reported variables were: relative blood volume (RBV), a measure of maximal contrast intensity after replenishment; mean transit time (mTT), the time taken from destruction to 50% replenishment; wash in rate the gradient of the reperfusion slope, and the perfusion index calculated by dividing RBV by mTT. A graphical illustration of these variables is shown in Figure 2B.

Fig. 1: Dual mode view of a typical renal ultrasound imaged from the flank. Three regions of interest are selected: cotex (green), medulla ROI1 (pink), and ROI2 (yellow). The blue bounding box demonstrates the area of motion stabilization. ROI indicates regions of interest.

Fig. 2: A, Time–intensity replenishment curves after bubble destruction. Cortical signal (green) is typically less than medullary signal and more uniform. B, Illustration of variables derived from a time–intensity curve following ultrasound contrast destruction and reperfusion. Replenishment kinetic variables measured include Wash-in rate (WiR) = maximal slope; mean transit time (mTT) = time to 50% echo power; relative blood volume (RBV) = maximum power – minimum power).

Measurement of renal arterial flow

The renal artery was identified on the CEUS image and a measurement of the diameter taken using a magnified view as illustrated in Figure 3A. After alignment with the renal artery, PW Doppler signal was recorded and the envelope manually traced to provide time-averaged peak velocity (TAPV) as shown in Figure 3B. This data was used to calculate volumetric renal blood flow (RBF) using the following equation:$$\begin{array}{l}RBF=(\pi r^2)(TAPV*Ti)*HR*2\\ (r=renalarteryradius;Ti=TAPVduration,HR=heartrate)\end{array}$$

Fig. 3: A, Magnified view of the renal artery and measurements made using callipers in contrast mode. B, Pulsed wave Doppler signal of the renal artery from the same location as (A).

Doppler signal from the intralobular artery was used to calculate the Resistive Index as follows:$$RI=\frac{PeakSystolicVelocity-EndDiastolicVelo\cot y}{PeakSystolicVelocity}$$

Transthoracic echocardiography

Transthoracic echocardiography measurements were made contemporaneously with renal ultrasound by the same investigator, accredited to the British Society of Echocardiography. Assessments of biventricular and valvular function were made to ensure severe cardiac disease did not confound measurements of renal perfusion. Cardiac output was calculated using pulsed-wave Doppler of the left ventricular outflow tract ⁽¹⁴⁾. Right ventricular systolic pressure was estimated using tricuspid valve regurgitation velocity and estimates of right atrial pressure from inferior vena cava (IVC) variability. Right ventricular function was assessed using tricuspid annular plane systolic excursion (TAPSE) and right ventricular systolic velocity (RVs’) from Tissue Doppler Imaging.

Outcome measures

The primary outcome measure was a difference in CEUS-derived replenishment kinetic variables between patients with COVID-19 and healthy controls. Secondary outcomes were differences in CEUS-derived replenishment kinetic variables between patients with COVID-19 and septic shock and differences in cardiac output, renal blood flow, and resistive index between the three groups.

Statistical analysis

Distribution of data was assessed using D’Agostino-Pearson omnibus normality test. Normally distributed data is reported as mean ± SD and non-normally distributed data as median. Differences between three or more groups were assessed using Kruskal–Waliis test with Dunn correction for multiple comparisons. Comparison between two groups was conducted using unpaired t test, Mann– Whitney U test, or Fisher exact test depending on the nature of the data. Statistical analysis was conducted using Prism v. 6.0 (Graph Pad). P values of <0.05 were taken to indicate significant differences.

RESULTS

Patient characteristics

Characteristics of patients at the time of study enrolment are shown in Table 1. Twelve patients with COVID-19 were recruited but image quality was poor in two patients who were excluded from analysis. Patients with COVID-19 had been in the ICU longer than patients with septic shock and had a higher burden of organ dysfunction but appeared similar in other respects, including baseline renal function. All patients were receiving renal replacement therapy in the form of continuous veno-venous hemodiafiltration (CVVHDF) at the time that study measurements were obtained and there were no significant differences in the duration of CVVHDF between the two patient groups. Patients in the septic shock group showed some signs of recovery of renal function with a higher urine output in the preceding 24 h than COVID-19 patients (706 ± 716 mL vs. 155 ± 199 mL, P 0.01). Septic patients also had a more negative fluid balance for the 24 h preceding study measurements (−597 ± 750 mL vs. 1,493 ± 1,583 mL, P 0.0003) but not for overall fluid balance over the length of ICU stay (5,703 ± 5,001 mL vs. 8,927 ± 4,875 mL, P 0.12). Baseline demographic data for the healthy control group is provided in the supplementary material, https://links.lww.com/SHK/B138.

Table 1 - Demographic and baseline patient characteristics at time of study enrolment

	COVID-19 (n = 10)	Septic shock (n = 13)	P value
Age	60 ± 7	56 ± 18	0.43
Gender	7 male (70%)	10 male (77%)	1.0
Comorbid (DM)	4 (40%)	4 (30%)	0.68
Comorbid (BP)	7 (70%)	6 (46%)	0.40
Comorbid (CVS)	3 (30%)	0 (0%)	0.17
Comorbid (Resp)	2 (20%)	2 (15%)	0.41
Comorbid (any)	10 (100%)	11 (84%)	0.48
Documented CKD stage 3	1 (10%)	3 (23%)	0.60
Baseline creatinine (μmol /L)	89 ± 5	74 ± 6	0.08
Baseline GFR (mL/min)	73 ± 4	76 ± 4	0.58
Days since symptoms	23 ± 8	NR	NA
ICU day	10 (7–16)	4 (4)	<0.0001
Duration of RRT	6 (2–9)	4 (3–4)	0.10
SOFA score	16 ± 2	14 ± 4	0.02
HR (beats/min)	88 ± 11	86 ± 16	0.72
MAP (mm Hg)	80 (74–88)	76 (69–83)	0.25
CVP (mm Hg)	13.1 ± 4.9	12.3 ± 4.5	0.68
Norepinephrine dose (mcg/kg/min)	0 (0–0.07)	0.02 (0–0.2)	0.47
Other inotropes	0 (0%)	4 (30%)	0.10
Lactate (mmol/L)	1.0 (0.9–1.4)	1.6 (1.1–1.9)	0.07
WCC × 109 /L	17.8 (11.9–26.3)	16.9 (10.5–26.7)	0.72
CRP (mg/L)	207 ± 136	164 ± 107	0.38

Comorbidities: CVS indicates ischemic heart disease, chronic heart failure, chronic dysrhythmia; BP, hypertension; DM, diabetes mellitus type 1 or 2; Respiratory: asthma, chronic airways disease; any: all recorded comorbidity including those listed above.CKD, chronic kidney disease; CRP, C-reactive protein; CVP, central venous pressure; GFR, glomerular filtration rate; HR, heart rate; MAP, mean arterial pressure; RRT, renal replacement therapy; SOFA, Sequential Organ Failure Assessment; WCC, white cell count.Values expressed as mean ± SD or median (IQR).Differences between groups assessed using unpaired t test, Mann–Whitney test, or Fisher exact test.

Renal artery blood flow

Figure 4 shows measurements obtained from the renal artery. Both TAPV and RBF were reduced in patients from both groups compared with healthy controls, but there was no difference in measurements between the two patient groups. RI, measured in the interlobular artery, was significantly higher in both groups compared with controls implying a degree of downstream flow occlusion. RI was significantly higher in patients with COVID-19 than those with septic shock.

Fig. 4: Renal blood flow variables in critically ill patients presenting with COVID-19 and septic shock compared with healthy controls (HC). TAPV indicates time averaged peak velocity.

Echocardiographic measurements

Echocardiographic variables are presented in Table 2. Cardiac index was increased in both COVID-19 and septic shock when compared with controls (3.7 ± 0.8 L/min/m² vs. 3.5 ± 1.0 L/min/m² vs. 2.8 ± 0.6 L/min/m²P 0.04). Right heart function was normal and equal between groups when assessed by TAPSE and RVs.

Table 2 - Echocardiographic-derived parameters for left and right heart function.

	COVID 19 (n = 10)	Septic shock (n = 13)	Healthy controls (n = 12)	P value
Cardiac output (L/min)	7.2 ± 1.8	6.5 ± 1.7	5.0 ± 1.3	0.009: HC vs. COVID
Cardiac index (L/min/m2)	3.7 ± 0.8	3.5 ± 1.0	2.8 ± 0.6	0.04: HC vs. COVID
RVSP (mm Hg)	48.1 ± 27.9	36.7 ± 19.3	24.4 ± 7.1	0.02: HC vs. COVID
TAPSE (mm)	2.4 (1.8–2.7)	2.1 (1.8–2.6)	2.6 (2.5–2.8)	0.2
RVs’ (mm/s)	19.8 (9.6–22.1)	18.9 (14.3–23.0)	13.6 (12.4–14.4)	0.08

RVs’ indicates right ventricular systolic velocity (normal >9.5 mm/s); RVSP, right ventricular systolic pressure; TAPSE, tricuspid annulus plane systolic excursion (normal >1.7 mm).

CEUS-derived renal perfusion variables

CEUS-derived perfusion parameters for the renal cortex and medulla are shown in Figures 5 and 6. All measures of cortical perfusion were reduced compared with healthy controls in both COVID-19 and septic shock patients. While the reduction in perfusion appeared relatively homogenous in the COVID-19 group there appeared to be more heterogeneity within the septic shock group. Perfusion changes in the renal medulla appeared to be less marked. Patients in the septic shock group did not exhibit any reduction in medullary perfusion compared with healthy controls. By contrast, patients with COVID-19 showed a reduction in medullary mean transit time and perfusion index.

Fig. 5: CEUS parameters for renal cortical perfusion in critically ill patients presenting with COVID-19 and septic shock compared with healthy controls (HC). CEUS indicates contrast-enhanced ultrasound; MTT, mean transit time; PI, perfusion index; RBV, renal blood volume; WiR, wash in rate. Differences between groups assessed using Kruskal–Wallis test with Dunn correction for multiple comparisons.

Fig. 6: CEUS parameters for renal medullary perfusion in critically ill patients presenting with COVID-19 and septic shock compared with healthy controls (HC). CEUS indicates contrast-enhanced ultrasound; MTT, mean transit time; PI, perfusion index; RBV, renal blood volume; WiR, wash in rate. Differences between groups assessed using Kruskal–Wallis test with Dunn correction for multiple comparisons.

Long-term outcomes

Mortality in the COVID cohort was 60% versus 26% for patients with septic shock. Of the surviving patients, the median time to recover renal function for COVID-19 was 54 (19–65) days versus 7 (6–15) days for septic shock. The median ICU stay for survivors with COVID-19 was 52 (49–56) days versus septic shock 14 (11–46) days. Non-survivors died after a median duration of 16 (11–35) days for COVID-19 versus 26 (9–31 days) for septic shock.

DISCUSSION

In this case-control study, we have demonstrated a significant reduction in both micro- and macrovascular renal perfusion in COVID-19 and septic shock associated AKI. These differences exist despite preserved biventricular function and cardiac index.

Prerenal azotaemia, left heart failure with hypoperfusion and right heart failure resulting in cardiorenal syndrome have been suggested as potential aetiological factors in the development of AKI in COVID-19 (8). Due to the evolving nature of the pandemic complete outcome data are not available at the time of writing but it appears that COVID-19 related AKI is more frequent, of increased severity, and slower to recover than other causes of AKI in critically ill patients, such as sepsis ⁽¹⁵⁾. COVID-19 related AKI may also have a pathophysiology distinct from that seen in sepsis.

Post-mortem evidence has suggested that patients dying of COVID-19 can manifest an organ-specific “endotheliitis” with viral particles present in both renal endothelium and tubular epithelium. Microvascular obstruction from erythrocyte deposition and fibrin thrombi result in ischemic glomeruli and frank necrosis, although no overt vasculitic process was evident ⁽⁹⁾. By contrast, septic AKI is widely described as a largely functional phenomenon, with a relative paucity of histological change ⁽¹⁶⁾. Of the pathology that has been demonstrated, leukocyte infiltration, acute tubular injury, and apoptosis are features, with an absence of intravascular thrombosis ⁽¹⁷⁾. Given the suggestion that the pathogenesis of COVID-19 AKI was, at least in part, driven by changes in the renal vasculature we designed the current study to assess if significant alterations in renal perfusion were present in this patient group.

Accurate in vivo measurement of renal perfusion is challenging in patients with septic shock but the relatively novel technique of CEUS has the potential to cast new light on this area. Large animal studies utilizing CEUS have been conflicting however; a porcine model demonstrated a reduction in cortical perfusion while an ovine model showed a variety of responses, some animals having increased and others reduced perfusion ^{(18, 19)}. Pilot studies in critically ill patients have shown CEUS lends itself favorably to the real-time study of renal perfusion at the bedside, being safe, feasible, and reproducible in both cardiac surgery and septic shock ^{(20, 21)}. One relatively small study examined cortical perfusion in septic shock and demonstrated an early reduction in perfusion-based variables, albeit with significant interindividual variation ⁽²²⁾. Another study has used CEUS to examine renal perfusion in COVID-19 associated AKI. This case series has demonstrated the safety and feasibility of the method and has described renal perfusion abnormalities in critically ill patients ⁽²³⁾. The current study uses a similar methodology to that previously described, modeling replenishment kinetics to provide a relative measure of microcirculatory blood flow, while conventional Doppler ultrasound enables the study of renal macrovascular variables, such as resistive index and time averaged peak velocity in the renal artery.

The results of the current study demonstrate significantly reduced cortical replenishment kinetics between both patients with COVID-19 and sepsis associated AKI compared with healthy controls. A reduction in variables reflective of blood volume and blood flow was observed in both groups, but with greater variability in patients with sepsis associated AKI. This heterogeneity may be a result of sub-phenotypes of septic AKI, where a variety of changes occur in response to a septic insult and is the subject of ongoing study ^{(11, 24)}. Additionally, the greater urine output observed in the septic group raises the possibility that some of these patients may have been recovering renal function at the time CEUS measurements were performed. By comparison patients with COVID-19 associated AKI demonstrated relatively uniform changes with a reduction in all renal perfusion parameters. This could support the evolving hypothesis that suggests COVID-19 AKI is a mainly vascular phenomenon driven by endotheliopathy and microthrombi-based occlusion of small blood vessels ⁽⁹⁾. If flow through the capillary bed of the kidney is reduced by obstruction from erythrocyte and fibrin deposition, vascular resistance would increase. This hypothesis is supported by the significant increase in resistive index observed in the current study. This microvascular obstruction is likely to result in congestion within the larger renal vessels and this reduction in flow is demonstrable in the measurements taken from the renal and interlobular arteries.

Another potential explanation for the reduction in renal blood flow would be a similar process to the mechanisms postulated for septic AKI. Oxidative stress and inflammation may induce an adaptive response by tubular cells to reduce energy consumption, downregulate metabolism, and undergo cell cycle arrest ⁽²⁵⁾. If the primary insult in COVID-19 associated AKI was a reduction in demand, due to cellular hibernation, one might expect a reduction in supply as a secondary phenomenon and this could manifest as a reduction in blood flow.

The expectedly elevated cardiac indices found in both patient groups suggest that observed reduction in renal blood flow is not due to cardiac insufficiency. Right ventricular systolic pressure was elevated in both the septic and COVID-19 patients; however, this finding is probably physiological and explained by the fact that all patients were mechanically ventilated, with relatively high intrathoracic pressures. It is feasible that the increase in right ventricular systolic pressure was indicative of overall venous hypertension and renal venous congestion, but we found no evidence of right heart failure and relatively normal IVC variability and central venous pressures. We therefore feel it is unlikely that venous hypertension contributed substantially to the reduced arterial forward flow seen in the current study.

Our study has several limitations. First, like all small case-control series conducted at a single center the results require confirmation by other investigators and in larger numbers of patients. The challenging nature of conducting clinical research during a pandemic meant that we were unable to recruit a larger cohort. We only assessed patients with severe AKI receiving renal replacement therapy and exploration of these parameters in patients with COVID-19 with earlier stages of AKI or normal renal function will be important to establish the temporal relationship between vascular changes and AKI onset. A longitudinal study with additional timepoints would provide further information about the development and resolution of these changes and enhance understanding of underlying pathophysiology. Lastly, in an attempt to develop a novel technique to quantify renal blood flow with ultrasound we have used CEUS to define the arterial diameter; however, the calculation assumes the predominate intra-arterial flow type is laminar, which may not be the case and has led to an overestimate of this variable.

To our knowledge the findings presented here are the first in vivo measurements of both renal macrovascular and microvascular blood flow obtained from critically ill patients with COVID-19 associated AKI. They provide some objective evidence for the hypothesis that COVID-19 AKI has a predominantly vascular pathogenesis and that the source of this pathology may be found within the renal microcirculation. Given the small size of the study cohorts these results should be defined as hypothesis generating and further research into this area is required.

CONCLUSION

In this case-control study we have demonstrated that both renal macro and microvascular flow is significantly reduced in patients with COVID-19 related AKI and is independent of changes in cardiac output or right ventricular function. A renal microvascular pathogenesis for COVID-19 AKI is a plausible hypothesis and should be confirmed by further, larger studies.