Determinants of Doppler-based renal resistive index in patients with septic shock: impact of hemodynamic parameters, acute kidney injury and predisposing factors

Acute kidney injury (AKI) occurs in approximately 25% of sepsis in intensive care unit (ICU) [1, 2] and 50% of septic shock [2]. Septic AKI pathophysiology is complex: interactions between renal hemodynamic changes and inflammation are probably involved in both initial and established phases of AKI [3]. Overall, vascular dysfunction may play a pivotal role in these pathophysiological mechanisms [4‐6]. Thus, renal resistive index (RI) measured by Doppler ultrasonography allowing to explore noninvasively renal hemodynamics has been used to predict renal dysfunction [7‐9]. Increased RI has been shown to be associated with renal failure, especially in septic patients [7]. However, RI’s clinical use is limited by poor sensitivity and specificity. A wide overlap of RI values between patients with and without renal failure is thus observed. This may be explained by the numerous determinants that contribute to the RI [10]. RI may be determined not only by “direct” renal determinants such as renal vascular lesions impacting arteriolar resistance and compliance [11] but also by non-renal factors indirectly modifying renal hemodynamics such as mean arterial pressure (MAP) [7, 12], pulse pressure (PP) [13], heart rate (HR) [14], fluid challenge [15] or arterial partial pressures of oxygen (PaO₂) [16, 17] and carbon dioxide (PaCO₂) [17].

The relationship between MAP and RI has been demonstrated to be particularly strong and was observed repeatedly over several studies [7, 12, 18]. RI was shown to be inversely correlated with MAP, which might be consistent with physiologic renal post-glomerular vasoconstriction observed during low renal perfusion in order to maintain glomerular filtration pressure [7]. Interestingly, Dewitte et al. [12] observed that this relationship was abolished in patients with AKI. Together with animal studies showing that renal vascular lesions are at stake very early in the course of AKI [6, 19], this observation may indicate that renal vascular reactivity to MAP changes is impaired in AKI. Alternatively, vascular factors predisposing to AKI, such as diabetes and chronic arterial hypertension, may be responsible for this altered vascular response before the onset of AKI lesions [20, 21].

In an attempt to define the role of these determinants on RI in patients with septic shock, we performed RI measurements at different MAP levels over a short period of time during vasopressor syringe pump relays performed during routine care on the admission day in ICU (D1) and every 3 days until day 10 (D10).

This study aimed at exploring the determinants of RI in patients with septic shock and at studying specifically the relationship between MAP and RI in patients with and without AKI. Specifically, we put forward the hypothesis that RI changes to MAP variation were modified in patients with AKI and/or in the presence of diabetes or hypertension.

In this study on 65 patients with septic shock, analysis of RI determinants was performed using a linear mixed model. On a pathophysiological point of view, this model brings up several important observations. We first confirmed the association between high RI on D1 and subsequent AKI [7, 9] and the negative correlation between MAP and RI [7, 18]. We further extended these data by showing conversely that PP is positively correlated with RI. More importantly, the model describes how AKI and predisposing factors to AKI impact renal hemodynamic. We observed that AKI is associated with an increase in RI but did not interact with the relationship between MAP and RI. On the opposite, chronic arterial hypertension and diabetes did not modify directly RI but interacted with the MAP/RI relationship, a less steeper slope being observed when these risk factors are present. In synthesis and regarding our initial working hypothesis, the presence of predisposing factors to AKI was associated with lower RI reactivity to MAP changes, and AKI was associated with higher RI.

Hypertension and diabetes are well-known factors for AKI, in particular through induction of chronic vascular renal disease [33]. It may be hypothesized that low RI reactivity to MAP changes indicates the presence of chronic renal vascular lesions. It would have been interesting to assess vascular disease in our patients in other vascular beds by validated techniques (such as Doppler ultrasounds of lower limbs and carotid arteries) to determine if the low reactivity to MAP changes correlates with the extent of chronic vascular damage. Unfortunately, this was not possible in our patients during the ICU stay. Confirming the lower response to MAP changes in patients with diabetes and/or hypertension outside the ICU may be interesting but unfortunately will be difficult as administering norepinephrine to modify MAP is not routine care outside the ICU. Interestingly, the interaction between MAP target and chronic hypertension regarding AKI has been highlighted in a randomized clinical trial on septic shock patients [34]. In this study, a higher MAP target (i.e., 85 mmHg) resulted in a lower risk for AKI, only in the subgroup of patients suffering from chronic hypertension. Together with this observation, our study indicates that chronic renal vascular changes may be implicated in the chain of events leading from hypertension to AKI. In addition, the correlation between PP and RI may also illustrate the relation between predisposing factors to AKI and RI. This positive correlation may be explained by the negative correlation between PP and vascular compliance [32]. In sum, a high PP may reflect a low vascular compliance which is known to be associated with a high RI [11].

We identify that recovery from severe AKI was associated with RI decrease and that the extent of this decrease was very close to the extent of the increase associated with AKI. Thus, the increase in RI associated with AKI seems to be reversible and to evolve in a time course similar to AKI itself. Importantly, this observation indicates that the increase in RI associated with AKI is likely to be related to the pathophysiology of AKI itself and not to predisposing factors. Our study was not designed to disclose whether the decrease in RI may permit to diagnose early AKI healing, i.e., before SCr decrease. This may be an interesting point to study in the future. Finally, we identified time spent in the ICU as a factor associated with higher RI. From our data set, we cannot determine whether this association relates to the progression of renal acute disease over time, effect of ICU treatments or other unknown factors.

Regarding the value of RI by itself, the model showed that despite entering several determinants associated with RI, the random effect, i.e., the variation in RI not “explained” by these determinants, was high (0.046) and even higher than the increase in RI associated with severe AKI (0.032). Thus, even when correcting for these several RI determinants, the variation in RI between patients with and without severe AKI is below the inter-individual variability of the index, which explains the low individual performance observed in most studies. Several additional RI determinants may be hypothesized: renal venous pressure, renal congestion, cardiac output, abdominal pressure, pO₂, etc. They may narrow the random effect but will result in an over-sophisticated model that is likely to perform poorly should it be used to predict AKI [35]. These data suggest that the use of the RI for clinical practice for AKI prediction is probably unrealistic, even by taking into account several RI determinants or by using a dynamic approach. Indeed, we assessed several ways to combine parameters with RI to improve its accuracy to predict AKI, such as the ratio ∆RI/∆MAP during MAP swings, but none proved effective (data not shown). Furthermore, as RI does not represent only renal vascular resistance, it can also not be used to determine the “optimal” MAP level in every patient.

Our study has several limitations. Firstly, as already highlighted, the high value of the residual random effects suggests that the participation of other determinants of the RI was not taken into account. Secondly, renal hemodynamics changes due to MAP changes may vary according to the initial MAP level, depending on whether these values remain within the auto-regulatory range [18]. The renal autoregulation plateau refers to a range of perfusion pressure values in which the renal blood flow is constant irrespective of the perfusion pressure [36, 37]. Below this plateau, the renal blood flow is directly dependent on the perfusion pressure. This phenomenon should be associated with an increase in vascular resistance, and hence potentially in RI, with increasing MAP above a threshold [37]. This may add another level of sophistication to the relationship between MAP and RI that may be extremely difficult to take into account, as the threshold for this plateau is probably highly variable between patients. Furthermore, MAP may be considered as a surrogate of renal perfusion pressure. However, this approximation does not account for the venous renal pressure. Lastly, kidney inflammation potentially linked to renal hemodynamic changes was not analyzed in our study [38].

RI was impacted by MAP, presence of severe AKI and severe AKI recovery, PP, day from ICU admission and past medical history of diabetes and/or hypertension. AKI was associated with a reversible increase in RI without significant interaction with the relationship between MAP and RI. Conversely, the presence of chronic hypertension and/or diabetes interacted with this relationship. Finally, variability of RI despite taking into account these numerous parameters remained high, suggesting that its use in clinical practice to predict AKI is probably unrealistic.