Doppler resistive index to reflect regulation of renal vascular tone during sepsis and acute kidney injury

In critically ill patients, sepsis and acute kidney injury (AKI) are very common diseases and are associated with increased hospitalization and elevated in-hospital mortality rates [1, 2]. AKI is a dynamic process that evolves from an early reversible condition to an established disease and leads to sustained renal impairment, cell death, and delayed renal recovery [3, 4]. Consequently, prompt resuscitation of the circulation and optimal perfusion pressure are the primary therapies for critically ill patients with AKI. These methods are based principally on the appropriate management of intravenous fluid replacement and vasopressor administration under strict hemodynamic monitoring [5].

To date, the pathogenesis of septic AKI is not completely known [6]. An increase in renal vascular resistance (RVR), associated with reduced renal blood flow (RBF) and, thus, potential renal ischemia has been reported [7, 8]. Conversely, some experimental studies have demonstrated that septic AKI may be associated with renal vasodilatation and increased RBF [9, 10]. Renal resistive index (RI), determined by Doppler ultrasonography, allows an estimation of RVR at the bedside. Although this parameter provides interesting information, its interpretation remains difficult in clinical practice. In patients with septic shock, the norepinephrine (NE)-induced increase in mean arterial pressure (MAP) from 65 to 75 mm Hg is associated with significant decreases in RVR and RI [11]. In contrast, some authors have reported that RI was not influenced by vasoactive drug therapy [12, 13]. Furthermore, a recent publication reported that increased RI was associated with persistent AKI in critically ill patients who require mechanical ventilation [12]. We tested the hypothesis that the relationship between renal RI and MAP may be influenced by renal injury. The primary objective of this study was to evaluate whether MAP is a determinant of renal RI in septic and critically ill patients who do or do not have AKI.

Ninety-six patients were prospectively screened during the study period. Poor-quality images from two patients made Doppler measurements for any lobar arteries impossible. Consequently, 94 patients were included in the statistical analyses (Table 1). Doppler measurements could be obtained from only one kidney in 15 patients. In this case, the right kidney has been preferentially observed (11 patients out of 15). For the 79 patients in whom bilateral Doppler could be performed, matched t test comparisons of right and left RI measurements showed no significant differences: 0.73 (95% CI 0.70 to 0.76) versus 0.73 (95% CI 0.70 to 0.75). Intra-observer variation was 3.4%, and the intraclass correlation coefficient was 0.94 (95% CI 0.92 to 0.96). Inter-observer variation was 6.2%, and the intraclass correlation coefficient was 0.92 (95% CI 0.81 to −0.97). The mean bias of inter-observer measurements was 0.00 (95% CI −0.10 to 0.09). RI was not correlated with cardiovascular risk factors and was correlated only with age in patients without AKI (ρ = 0.35; P = 0.015). RI was higher in patients presenting with an arterial partial pressure of oxygen/fraction of inspired oxygen (PaO₂/FiO₂) ratio of less than 200: 0.76 (0.72 to 0.80) versus 0.72 (0.68 to 0.78) (P = 0.02).

For patients with transient AKI, diagnosis was made at Risk, Injury, or Failure levels in 42%, 33%, and 25% of cases, respectively. For patients with persistent AKI, diagnosis was made at Risk, Injury, or Failure levels in 7%, 11%, and 82% of cases, respectively. Median renal RIs were 0.72 (0.68 to 0.75) in patients without AKI and 0.76 (0.72 to 0.80) in patients with AKI (P = 0.001) (Table 2). Median RIs were 0.75 (0.72 to 0.79) in cases of transient AKI and 0.77 (0.70 to 0.80) in cases of persistent AKI (P = 0.84).

RI did not differ between patients who received or did not receive NE - respectively, 0.73 (0.69 to 0.78) versus 0.75 (0.68 to 0.78) (P = 0.94) - and was not correlated with the NE dose (ρ = -0.06; P = 0.69). Among patients without AKI, RIs were 0.73 (0.68 to 0.76) in patients who received NE and 0.70 (0.68 to 0.75) in patients who did not receive NE (P = 0.30) and were also not correlated with NE dose (ρ = 0.20; P = 0.43).

Among patients without AKI, RI was not correlated with fluid balance on the day of inclusion (ρ = 0.21, P = 0.18) or with the cardiac index (ρ = -0.21; P = 0.37) but was significantly correlated with the MAP (ρ = -0.47; P = 0.002) (Figure 1) and the PaO₂/FiO₂ ratio (ρ = -0.33; P = 0.04). Among patients with AKI, RI was not correlated with fluid balance (ρ = -0.15, P = 0.29), the cardiac index (ρ = -0.21; P = 0.37), the MAP (ρ = 0.006; P = 0.97) (Figure 2), or the PaO₂/FiO₂ ratio (ρ = -0.23; P = 0.11). In cases of transient or persistent AKI, RI was not correlated with the MAP (ρ = 0.009, P = 0.99, and ρ = -0.0003, P = 0.99, respectively). When patients were divided into three groups according to MAP, RI was only inversely proportional to MAP in patients without AKI (P = 0.04) (Table 3). In contrast, in cases of AKI, RI remained elevated regardless of the MAP range. Logistic regression, using a univariate analysis, indicated that the factors that predicted an RI of greater than 0.75 in patients without AKI were the MAP (P = 0.02) and SAPS II (P = 0.002).

In this study, renal RI was evaluated in critically ill patients with sepsis after hemodynamic stabilization. A poor correlation between RI and MAP, age, or PaO₂/FiO₂ ratio was found in septic and critically ill patients without AKI compared with patients with AKI. Our understanding of the pathogenesis of septic AKI is limited because currently we do not have a clinically validated method to measure RBF. In patients with renal hypoperfusion, compensatory mechanisms to maintain GFR include attenuation of afferent arteriolar vasoconstriction, afferent arteriolar dilation, efferent arteriolar constriction, and neuro-/hormonal effects to increase tubular reabsorption of fluid and maintenance of cardiac output [18]. An increase in RVR, representing renal vasoconstriction and a reduced RBF leading to renal ischemia, has repeatedly been proposed to be central to the pathogenesis of septic acute renal failure [7, 8]. This paradigm has been derived from animal studies, using a variety of models, and from techniques for the measurement of RBF and for induction of sepsis without hemodynamic support [23]. However, many studies do not report decreased RBF in sepsis. Shortly after induced septic shock in ewes without significant fluid resuscitation or NE administration, Langenberg and colleagues [10] reported marked renal vasodilatation, which led to a transient increase in RBF. In parallel, sCr significantly increased after 24 hours whereas fractional excretion of sodium and fractional excretion of urea decreased in the first 48 hours (as seen in pre-renal azotemia). In human sepsis, sustained systemic vasodilatation with a high cardiac output is the dominant clinical finding during the early phase of AKI [24, 25].

Determination of increased RI on the first day of septic shock is known to have predictive value for acute renal failure in septic shock [13]. Darmon and colleagues [12] recently confirmed these results in patients who required mechanical ventilation. A Doppler renal RI value of more than 0.795 also predicted persistent AKI with good sensitivity and specificity. Our study has confirmed that RI is increased in cases of AKI, particularly in cases of persistent AKI. However, we were not able to precisely distinguish transient from persistent AKI in our selected patients. We assume that hypoperfusion (pre-renal component) and injury (renal component) often occurred at the same time in injured kidneys given that each tubular injury is a continuum from volume responsiveness to necrosis. The definition of pre-renal azotemia has conceptual or diagnostic limitations or both, and a precise renal RI cutoff point to distinguish clinical entities was difficult to determine in this study.

Alterations in renal hemodynamics may be induced by multiple factors in critically ill patients. Improving regional perfusion and inducing excessive vasoconstriction must be balanced when NE therapy is required. Indeed, intra-renal infusion of NE is used to induce AKI in animals, and NE has been shown to decrease RBF in healthy volunteers. In contrast, too low a dose of NE in patients with vasodilatory shock may result in an MAP that is below the limit of renal autoregulatory capacity. Guidelines for hemodynamic support of adult patients with sepsis advocate that vasopressors be titrated to the minimum level required to provide effective organ perfusion [26]. To prevent AKI, maintaining an MAP of at least 60 to 65 mm Hg during vasodilatory shock has also been recommended [27].

Deruddre and colleagues [11] suggested that measuring renal RI could help optimize renal perfusion in critically ill patients. In that study, a significant decrease in renal RI was observed in 11 patients with septic shock when MAP was increased from 65 to 75 mm Hg by NE titration. However, measurements of renal RI remain difficult to interpret in clinical practice. Our findings confirm previous results, showing that RI is dependent on factors apart from NE infusion [12, 13]. In patients with acute lung injury, low oxygenation levels have been reported to increase renal RI [28]. In our study, RI was inversely correlated with the PaO₂/FiO₂ ratio in patients without AKI. Recently, Darmon and colleagues [12] also found a significant negative correlation between MAP and RI in critically ill patients receiving mechanical ventilation (see Online Supplementary Material). However, in that study, MAP was not statistically different between patients without AKI and patients with transient or persistent AKI, despite an increasing RI.

Our study has reported that RI is inversely correlated with MAP in septic critically ill patients without AKI but not in patients with AKI. This may have been because of unresponsive renal vasoconstriction or renal vascular obstruction caused by a sustained kidney injury, despite hemodynamic support, which led to increased RI. On the other hand, in septic critically ill patients without AKI, an inverse correlation might reflect increased renal perfusion pressure with associated reduced activation of RVR, reflected by RI, leading to pressure-dependent renal perfusion. Sepsis or sedation could explain this modified renal vasomotor response in our patients. However, this association seems to be relatively poor. Indeed, the correlation coefficient that we found between RI and MAP means that most of the changes in RI lie somewhere other than MAP. Consequently, renal circulatory response to sepsis estimated by Doppler ultrasonography cannot be reliably predicted simply from changes in systemic hemodynamics. This finding is consistent with a recent study showing that the development of septic AKI often seems to be accompanied by an uncoupling between systemic vascular resistance and RVR [29].

The following points need to be considered when assessing the clinical relevance of our study. First, this study was purely observational as it was difficult to perform a blinded study. Second, the present study included a single-center population of patients who had developed vasodilatory shock. Patients exhibiting an important cardiogenic component of septic shock, defined by the need for epinephrine or dobutamine or both, were deliberately excluded. Third, the pressure of the renal interstitium was impossible to measure in clinical practice, and renal Doppler measurements do not reflect microcirculation vessel tone. In addition, measurement of GFR and renal oxygenation by continuous renal vein thermodilution was not conducted, as we were unable to use invasive hemodynamic monitoring in this observational study. No patient included in this study presented clinical intra-abdominal hypertension, but this was not assessed by intravesical measurement. Other confounding factors not identified in this study also may have influenced our results. Finally, the subjective part of renal ultrasound and the precision of measurements, assessed by the inter-observer variation in comparison with the Doppler measurement differences, are the main limitations when determining renal RI.

Although future AKI studies need to determine the impact of early recognition of pressure-dependent renal perfusion on renal recovery, the timing of volume replacement to facilitate early restoration of kidney perfusion and to optimize vasopressor infusion dose can be an important determinant in preventing the progression of AKI and facilitating recovery. Estimating renal circulatory response to sepsis by Doppler ultrasonography can be hazardous because many factors influence RI value. Increased RI may be associated with low renal perfusion pressure or low oxygenation levels in patients without AKI but may also be associated with sustained renal injury. Thus, its value in septic critically ill patients in the ICU should be interpreted carefully before concluding that there is renal injury or estimating optimal MAP.

A poor correlation between renal RI and MAP found only in septic and critically ill patients without AKI suggests that determinants of RI are numerous. Consequently, renal circulatory response to sepsis estimated by Doppler ultrasonography cannot reliably be predicted simply from changes in systemic hemodynamics. The value of a single RI measurement at ICU admission to determine optimal MAP remains uncertain.