Introduction
Sepsis is a life-threatening clinical condition associated with severe infection [
1]. Sepsis-associated brain dysfunction (SABD) is considered as cerebral dysfunction that accompanies sepsis in the absence of central nervous system infection and other possible causes of brain dysfunction (i.e., structural central nervous system lesions or drug overdose) [
2]. SABD is probably the most frequent sepsis-related organ dysfunction, affecting up to 70% of patients with sepsis and frequently occurring early, often before any other organ involvement [
3]. The development of SABD has been associated with higher mortality, lower quality of life in survivors and long-term neurological sequelae [
4]. Unfortunately, owing to the unspecific spectrum of symptoms, SABD remains a diagnosis of exclusion, usually contextualized in daily clinical practice according to the patient’s history.
The pathophysiology of SABD is still unclear and probably multifactorial, involving diffuse neuroinflammation, excitotoxicity and cerebral ischaemia [
5]. In the absence of vascular occlusion, cerebral ischaemia in patients with sepsis may be the result of reduced cerebral blood flow (CBF) secondary to hypotension [
6,
7]. Moreover, alterations in cerebral microcirculation may occur even with normal systemic haemodynamics and could contribute to tissue hypoxia [
8,
9]. As such, an alteration in cerebral autoregulation (CAR) (i.e., the homeostatic mechanism that protects the brain tissue from the potentially damaging effects of hypo- and hyperperfusion) could be considered as the main determinant of cerebral ischaemia during sepsis because small cerebral arterioles, which are an essential part of the microcirculation and of CAR, could become unable to actively regulate their calibre and maintain CBF constant over a wide range of external stimuli, including cerebral perfusion pressure (CPP) or carbon dioxide [
10]. Some authors have suggested that CAR may be altered in patients with sepsis. However, previous studies involved small cohorts of patients, used different CAR assessment methods and did not consider some potential confounders, such as severity of disease, use of vasopressors, age or extracerebral organ dysfunction, in the association of impaired CAR with brain dysfunction [
11‐
15]. The aim of this study was therefore to evaluate the association of altered CAR with the occurrence of SABD, as well as to identify clinical factors associated with altered CAR in patients with sepsis.
Discussion
In this large, multicentre study, CAR was altered in half of patients with sepsis and was independently associated with the occurrence of SABD, although not with ICU survival. Lower MAP, a history of CKD, and fungal infection were associated with altered CAR. This study represents the largest cohort of patients affected by sepsis in whom CAR has been tested and its relationship with SABD analysed considering potential confounders. The present findings thus support the concept that cerebral hypoxia could contribute to the development of SABD.
A recent study identified several risk factors for the occurrence of SABD (i.e., acute renal failure, dysglycaemia, hypercapnia, hypernatremia and
Staphylococcus aureus infection), but no systemic hemodynamic variables were identified [
19]. Two previous studies evaluating static CAR (i.e., changes in CBF following the manipulation of ABP under steady state using vasopressors) reported preserved CAR during sepsis [
14,
20], whereas another reported impaired CAR in patients with septic shock [
15]. In contrast, several studies assessing dynamic CAR (i.e., changes in CBF in response to sudden spontaneous or induced [head-up tilt, Valsalva manoeuvre or standing up]) fluctuations of ABP showed that impaired CAR was more frequently observed in patients with sepsis than in those without it [
12] or in patients with sepsis with SABD when than in those without brain dysfunction [
11,
13]. It has previously been suggested that dynamic CAR may be more easily impaired than static CAR in different diseases because of the different control mechanisms involved [
21]. One potential explanation would be the dysregulation of the autonomic nervous system or of the baroreceptor reflex arc, which preferentially influences the dynamic pathway, that is typically observed in sepsis [
22]. The poor temporal resolution of static CAR assessment could not adequately describe the changes in CBF that occur during sepsis, where significant hemodynamic fluctuations, particularly in the early stage of the disease, are reported.
In our study, SABD was related to well-known risk factors for brain dysfunction, such as the presence of vascular disease (i.e., increased risk of cerebrovascular injury) and the use of mechanical ventilation (i.e., increased risk of delirium) [
23]. Our findings also suggest that impairment of dynamic CAR has a significant influence on the occurrence of SABD. Given the increased levels of energetic substrates required by complex brain activity, even a slight reduction in CBF, as has been observed in patients with sepsis [
24], may impair superior cognitive functions. Alteration in CAR together with microcirculatory alterations could explain local hypoperfusion in absence of severe systemic hypotension. Several potential causes of impaired CAR during sepsis have been identified, including nitric oxide accumulation, blood-brain barrier breakdown due to neuro-inflammation, and impaired microcirculation [
9,
25,
26]. This is the first study, to the best of our knowledge, where the association of impaired CAR with the occurrence of SABD has been evaluated in a large cohort with correction for several confounders, and the present results strongly support the concept that cerebral hypoxia could contribute to the development of SABD.
Renal and cerebral vasculature share multiple characteristics, such as high blood flow rates, high pressure load and effective autoregulation of local flow. Chronic renal disease negatively affects microvascular cerebral function, favouring endothelial dysfunction, chronic inflammation, accumulation of urea or of vasoactive species, and alterations in sympathetic nervous system-driven vascular resistance [
27]. CAR has been found to be altered in paediatric patients with chronic renal dysfunction or in adult critically ill patients with acute kidney injury requiring renal replacement therapy [
28,
29]. The association between CKD and altered CAR in our study supports the hypothesis of kidney-brain crosstalk and negative effects of renal dysfunction on cerebral vasculature. The association between fungal infection and altered CAR is not easily explained. Specific effects of non-cerebral fungal infections on brain endothelial and microvascular function are unknown. Fungal infections are highly invasive, especially in cases of abdominal or respiratory infection [
30]; however, only ten patients in our cohort had fungal infections, and the possibility that a subclinical fungal involvement of the cerebral parenchyma would be the main reason for impaired CAR in this cohort remains unlikely.
Although a lower baseline MAP was a significant predictor of impaired CAR, we did not specifically assess the lower limit (LL) of CAR integrity, which may significantly vary in healthy subjects and can be affected by different chronic conditions; that is, in patients with chronic hypertension, the LL will be shifted towards higher values than in normotensive patients, and CBF could become “pressure-dependent” for lower values of MAP than in physiologic conditions [
31]. In one experimental study on rats, LL was also right-shifted, which would translate into a higher risk of cerebral hypoperfusion for MAP values considered as “normal” in clinical practice [
32]. Whether higher blood pressures should be targeted in selected patients with sepsis to avoid cerebral hypoperfusion remains to be further evaluated. Interestingly, although several studies reported a potential role of hypercapnia in the CAR impairment during sepsis [
15,
20], no correlation was found between PaCO
2 and autoregulation in our cohort. Thus, in our cohort CAR is impaired, regardless of the PaCO
2 level. In a recent meta-analysis, there was no clear association between PaCO
2 and/or CO
2 reactivity with impaired CAR during sepsis [
33]. As such, future studies prospectively evaluating the effects of changes in CO
2 and CAR function are warranted.
In contrast to the previous study by Bindra et al. [
34], we observed no association between altered CAR and mortality. However, their study differed substantially with regard to patient characteristics (i.e., 28 patients in septic shock, all sedated and on mechanical ventilation), method of CAR assessment (i.e., use of near-infrared spectroscopy and not TCD to evaluate CAR) and timing of outcome assessment (i.e., 3 months) compared with our cohort. Furthermore, brain dysfunction in patients with sepsis is not a direct cause of death in the early phase, but it is associated with long-term cognitive alterations, which were not assessed in this study.
Our study has several strengths. To date, it represents the largest cohort of patients with sepsis in whom CAR has been tested. We included patients affected by sepsis of different severities; risk of bias in selection and treatment is limited by the multicentre nature of the study. We assessed CAR using Mxa, a validated method which enables CAR assessment using spontaneous fluctuations in MAP, thus limiting potentially stressful procedural and/or pharmacological interventions. Such fluctuations depend on specific neuronal pathways and occur in healthy or critically ill subjects at very low frequency or as a physiologic response to variations in intra-thoracic pressures [
35]. Mxa has been validated in patients with traumatic brain injury [
36] and is already used in patients with sepsis [
13].
The present study also has several limitations. First, TCD technique only evaluates CBF in large intracranial arteries, thus impairment of microcirculation potentially leading to local ischaemia could not be specifically analysed. Second, Mxa is a mathematical simplification of a complex biological phenomenon and only investigates dynamic components of CAR. Static and dynamic CAR have been reported to be affected differently in healthy volunteers and patients with sepsis [
20]. Moreover, Mxa is only one method used to assess CAR, and different findings might have been obtained if other methods had been used [
37]. The threshold traditionally applied to define altered autoregulation is Mxa > 0.3; however, the absolute value of such thresholds has been questioned, and Mxa should be considered a continuous index that reflects a wide spectrum of severity of CAR impairment. Third, the diagnosis of SABD was based on Glasgow Coma Scale score and some signs of delirium and was not standardized on specific scales or scores, and no additional findings from brain monitoring were used. Nevertheless, we lacked widely accepted criteria to define septic encephalopathy [
2]. The Confusion Assessment Method for the ICU has been validated for diagnosis of delirium, but its benefit in SABD diagnosis has been questioned [
38]. In particular, the spectrum of clinical presentation of SABD is much wider than delirium alone and includes behavioural and personality changes and even post-traumatic stress disorder symptoms, so that unstructured assessment may be a better screen for this condition [
39]. Also, none of the available neuromonitoring tools (i.e., electroencephalography, biomarkers or neuroimaging) are specific for SABD [
40]. Moreover, our clinical definition was very similar to the definition used in other recent studies dealing with SABD [
19]. Fourth, sedation generates different effects in the autonomic nervous system and CAR [
41]. However, only half of the patients received sedatives and/or analgesics with different classes of drugs, so specific subgroup analyses would have been limited. Fifth, we assessed CAR only once in the early phase of sepsis, whereas this might improve over time in the first days following sepsis diagnosis [
13]. However, in this study, only CAR assessed on day 1 correlated with the occurrence of SABD thereafter, suggesting that early alterations have the largest influence on the development of brain dysfunction in these patients.