Seeking the Light in Intensive Care Unit Sedation: The Optimal Sedation Strategy for Critically Ill Patients

The clinical approach to sedation in critically ill patients has changed dramatically over the last two decades, moving to a regimen of light or non-sedation associated with adequate analgesia to guarantee the patient’s comfort, active interaction with the environment and family, and early mobilization and assessment of delirium. Although deep sedation (DS) may still be necessary for certain clinical scenarios, it should be limited to strict indications, such as mechanically ventilated patients with Acute Respiratory Distress Syndrome (ARDS), status epilepticus, intracranial hypertension, or those requiring target temperature management. DS, if not indicated, is associated with prolonged duration of mechanical ventilation and ICU stay, and increased mortality. Therefore, continuous monitoring of the level of sedation, especially when associated with the raw EEG data, is important to avoid unnecessary oversedation and to convert a DS strategy to light sedation as soon as possible. The approach to the management of critically ill patients is multidimensional, so targeted sedation should be considered in the context of the ABCDEF bundle, a holistic patient approach. Sedation may interfere with early mobilization and family engagement and may have an impact on delirium assessment and risk. If adequately applied, the ABCDEF bundle allows for a patient-centered, multidimensional, and multi-professional ICU care model to be achieved, with a positive impact on appropriate sedation and patient comfort, along with other important determinants of long-term patient outcomes.


INTRODUCTION
Patients admitted to the intensive care unit (ICU) routinely experience pain, agitation and anxiety, use of invasive monitoring, and need for invasive procedures or mechanical ventilation. Appropriate analgesia and sedation are therefore essential. In mechanically ventilated patients, sedation aims to minimize oxygen consumption, keep patients comfortably connected to the ventilator, avoid recall of the critical condition and prevent psychological and physical damage.
Failure to control pain and agitation can have detrimental effects both in the short-and long-term perspective. Poorly controlled pain and agitation have been associated with worsening of the critical condition, increase in sympathetic tone, and increased rate of accidental removal of medical devices such as endotracheal tubes and vascular catheters (1)(2)(3). Moreover, uncontrolled pain during ICU stay is related to a higher incidence of chronic pain, post-traumatic stress disorder (PTSD) symptoms, and a lower health-related quality of life (HRQoL). Conversely, deep and prolonged sedation has been associated with increased duration of mechanical ventilation, delayed weaning, increased incidence of delirium, and muscle impairment leading to ICU-acquired weakness (ICU-AW), and higher hospital and 6-month mortality (1). Different studies have been conducted in the last two decades to clarify the "why" and "how" to sedate patients in ICU. The key concept is that a "calm, comfortable, and collaborative" patient (i.e., lightly sedated) allows active cognitive stimulation, earlier liberation from the endotracheal tube, active mobilization, and also improved interaction with the healthcare team and the family, all of them being important patient-centered outcomes.

THE "WHY"
The paradigm of sedation in ICU changed dramatically in the last few decades. In the 80s, the common belief was that patients should adapt to machines, leading to the large use of deep sedation (DS) and prolonged immobility. In the 90s, the paradigm changed: the machine should adapt to the patients but clinicians should avoid patients' recall and memory of critical ill, leading to the use of deep sedation and avoiding neuromuscular blocking agent (NMBA). Starting from the early twentieth century, the idea of a protocolized sedation with a daily sedation interruption (DSI) modified dramatically the sedation approach in mechanically ventilated patients. A protocolized sedation approach alone (intended as sedation delivered by team members following written and approved procedures and outlined in a protocol) is not clearly associated with a beneficial effect in terms of duration of mechanical ventilation, mortality, or ICU length of stay as highlighted by a recent Cochrane systematic review (SR) and meta-analysis (MA) review, and confirmed in a recent multicenter cluster-RCT (DESIST trial) ( Table 1) (4,5). Indeed, the association between sedation protocol and daily sedation interruption (DSI) was initially introduced by Kress et al. (6) leading to a large number of trials aiming to establish its efficacy and safety ( Table 1). Daily sedation interruption (DSI) seems to reduce time in coma, ICU and hospital length of stay (LOS), sedation time, and mechanical ventilation duration and improve the 1-year survival rates (7). Although other trials did not find the same results (8,9), subsequent systematic reviews (SRs) and meta-analysis (MA) of RCTs demonstrated a possible reduction in duration of mechanical ventilation, ICU-LOS, hospital-LOS, and also a reduction in tracheostomy rate ( Table 1). On the other hand, the main arguments against daily sedation interruption (DSI) consist in its side effects, including agitation, pain, and endotracheal tube or catheter self-removal. Although possible psychological sequelae, such as anxiety, depression, and PTSD, related to the patient's awareness of their critical situations are claimed against daily sedation interruption (DSI), different studies demonstrated that such an approach did not impact the recall of ICU stay nor increased the incidence of PTSD (10,11). Daily sedation interruption (DSI) could increase nurse overload, especially in resources limited counties where the nurse: patients ratio is frequently less than 1:1, and it should be accompanied by a light sedation (LS) approach for the rest of the day, instead of a deep sedation approach (12). Indeed, no doubts exist about the impact of deep sedation during the first 48 h of ICU and later in ICU stay, as demonstrated by Shehabi et al. in the two different studies (13,14). Deep sedation (DS) was an independent predictor of long-term mortality and time to extubation in mechanically ventilated ICU patients. Although early deep sedation and the cumulative dose of sedative agents were not associated with time to delirium after 48 h, patients with lighter sedation had a lower presence of delirium at 48 h, and also significantly more coma-and delirium-free days at 28 days. These results induced the authors to propose and test the so-called "early goal-directed sedation" concept (i.e., light sedation, LS) in a pilot RCT (15). Starting early after mechanical ventilation (MV) initiation, this approach refers to goal-directed management to target a light level of sedation and minimize benzodiazepine usage. Results showed that patients with light sedation received less benzodiazepine and propofol, had more delirium-free days and required fewer physical restraints, but mechanical ventilation (MV) duration, ICU, and hospital LOS or mortality were not modified. Starting from these contrasting results, the effects of light sedation have been examined in different subsequent RCTs summarized in SRs and MAs ( Table 1). Among the most important systematic reviews, Stephens et al. showed as the implementation of light sedation in the first 48 h of mechanical ventilation reduced mortality, mechanical ventilation duration, and ICU-LOS (16). The effects of light sedation on outcomes were also examined in the meta-analysis presented in the PADIS guidelines, showing a reduction in mechanical ventilation (MV) duration and the tracheostomy rate, with no effects on 90day mortality, the occurrence of delirium, depression, PTDS, or self-extubation (12). The heterogeneity between included RCTs and their low quality forced the guidelines panel to make use of light sedation as a conditional recommendation. At last, Aitken et al. (17) examined the effect of depth of sedation in ICU patients on outcomes that extend across the ICU stay and beyond, including 7 RCTs and 18 cohort studies. The main findings were that lighter sedation was not associated with the reduced ICU or hospital mortality. Moreover, lighter sedation was not associated with a reduced duration of mechanical ventilation, delirium occurrence, and other adverse events among the RCTs but it was among the cohort studies. Across studies, both risks of bias and heterogeneity were high. The different results found in the Aitken and PADIS MAs are difficult to explain, although the inclusion of non-RCTs in Aitken's study inevitably introduced heterogeneity in the results, the RCTs included are different, and Aitken's study included the results from the NON-SEDA trial (18) that account for 45% of the patients included in the MA. NON-SEDA trial (18) is the far    biggest RCT that investigated the effects of a non-sedation protocol compared with the light sedation, and the authors did not find any differences in terms of 90-day mortality, ventilatorfree days, ICU and hospital LOS. A possible explanation for the non-superiority of a non-sedation strategy in the RCT, as the authors stated in the discussion, could be that the depth of sedation did not differ between the groups as much as intended, especially on day 1, configuring the non-sedation approach as light sedation (mean RASS score was −2.3 on day 1 in light sedation vs. −1.3 in the non-sedation group) (18).
Although still under debate, light sedation has been suggested by the 2018 Pain, Agitation/Sedation, Delirium, Immobility, and Sleep Disruption (PADIS) guidelines (12), and reinforced by the eCASH approach (19). In particular, a protocol based, stepwise assessment for pain control and sedation management in critically ill patients should be adopted, and light sedation "should be used in all mechanically ventilated patients" (conditional recommendation, low quality of evidence) (Figure 1).
Some caveats exist in the application of light sedation. At first, its definition is not universally accepted. The PADIS guidelines define light sedation as a Richmond Agitation Sedation Scale (RASS) score between + 1 (slightly restless) and −2 (awake with eye contact to voice) or a Riker Sedation-Agitation Scale (SAS) score between 4 (calm and cooperative) and 3 (difficult to rouse and obey simple commands). The SAS does not evaluate arousal and an SAS of 3 points is the only score to assign to a sedated patient without coma (SAS = 1 or 2) (20). The RASS scale may have a positive or negative value (−5 to + 4) and it is an ordinal scale, which makes it unuseful to synthesize using an average score (21). This is particularly important since the sedation level could variate during the 24 h; for this reason, other methods for sedation assessment have been proposed, such as the Sedation Index (22), which summarizes the sedation level over 48 h (sum of the negative RASS scores (RASS −1 to −5) divided by the total number of RASS measurements performed). When Sedation Index has been used, the level of sedation was strongly related to the risk of death, delirium, and delayed time to extubation (22).
The second caveat is that when light sedation is not integrated into a bundle of patient-centered outcomes, could not be as useful. The eCASH-early Comfort using Analgesia, minimal Sedatives and maximal Humane care-as well as the ABCDEF-R bundle (where R = respiratory-drive-control has been proposed recently) (12,23), suggest the integration of light sedation into a bundle to avoid pain, anxiety, agitation, delirium, and immobility, to reduce the post-intensive care syndrome (PICS) incidence. These targets could be achieved by treating pain first and by accompanying light sedation to communications aids, noise reduction to promote sleep, early mobilization, delirium monitoring, and family involvement.
Occasionally, deep sedation may be required. In the case of patients with ARDS, historically deep sedation along with NMBA continuous infusion has been considered the best clinical practice (24). Following recent evidence (25), practical clinical guidelines (26) suggest avoiding routine use of NMBA infusion in adults with ARDS before optimizing mechanical ventilation and assessing ARDS severity. Moreover, the authors suggest against using an NMBA infusion (Suggestion, low certainty of evidence) in patients who tolerate ventilation using a lighter sedation strategy; whenever an NMBA is required to facilitate lung-protective ventilation, it should be used intermittently (Suggestion, low certainty of evidence) (26).

THE "HOW"
When the "why" of sedation in ICU has been explored, clinicians want to know "how" to lightly sedate the patients. On this topic, there is no doubt that benzodiazepines should be avoided to minimize the risk of delirium (27)(28)(29)(30)(31)(32)(33). As a result, dexmedetomidine and propofol remain the possible choices.
Dexmedetomidine is a highly selective α2 adrenergic receptor antagonist that produces dose-dependent sedation with no respiratory depression and modest hemodynamic effects (34). Dexmedetomidine may promote sleep via more physiological pathways in comparison with GABAergic sedatives (i.e., benzodiazepines, propofol), favoring the N3 (or slow wave sleep) stage (35). Propofol is a gammaaminobutyric acid (GABA) receptor agonist, by binding to the β subunit of the postsynaptic GABA-A receptor, it induces postsynaptic membrane hyperpolarization and inhibits neuronal depolarization.
Pain, Agitation/Sedation, Delirium, Immobility, and Sleep Disruption guidelines recommend the use of propofol or dexmedetomidine over the benzodiazepine for sedation in critically ill, mechanically ventilated adults, as a conditional recommendation with low quality of evidence (12). Different RCTs have investigated the effects of dexmedetomidine as sedative agents, only one of them compared dexmedetomidine with propofol (29). The SPICE III trial is the most recent and biggest RCT on this topic, including approximately 8 times as many patients as the other trials (36). It compared the use of dexmedetomidine with usual care (propofol or midazolam or a combination of the two) for sedation therapy in mechanically ventilated patients. This open-label, randomized trial enrolled 3,904 patients, with mortality at 90 days as the primary outcome. Secondary outcomes included mortality, cognitive function, and HRQoL assessed at 180 days. Results from the study revealed no difference in 90-and 180-day mortality, nor in cognitive function and HRQoL (17).
The explanation for such unexpected results should be found in the high proportion of patients (60% on day 1 and 50% on day 2) who required deep sedation (not allowed in all the other trials), which may have independently influenced outcomes, as discovered previously by the same authors (14,22). Moreover, more than 70% of patients in the dexmedetomidine arm received propofol or midazolam to reach goal sedation in the first 48 h following randomization, and more than 50% of the dexmedetomidine cohort continued to receive propofol at the study day 10 (with only 30% of patients requiring deep sedation). The inadequacy of dexmedetomidine to maintain light sedation has been reported also in MIDEX (29) and PRODEX (29) trials, with, respectively, 43.8 and 72.5% of patients requiring rescue sedation in the dexmedetomidine arm. Another interesting result of the SPICE III trial is the higher rate of adverse events in the dexmedetomidine cohort compared with the usual care cohort (9.6 vs. 1.8%, respectively, P < 0.0001), although lower than in the other trials.
A subsequent Bayesian secondary analysis of the SPICE III trial (36) found that dexmedetomidine exhibited a high probability of reduced 90-day mortality in older patients (more than 65 years) and a high probability of increased 90-day mortality in the younger patients of non-operative status.
An important topic is the cost of sedation. Park and Jeong (37) carried out a cost-minimization analysis based on MIDEX and PRODEX studies, demonstrating that patients on a mechanical ventilator and treated with midazolam and propofol had medical costs (related drug costs, ICU care costs, and costs of treating adverse events), respectively, 17.6 and 15.2% more than patients treated with dexmedetomidine. A secondary analysis of the SEDCOM trial (27) (an RCT comparing dexmedetomidine and midazolam for sedation in critically ill patients) (38) showed a median total intensive care unit cost savings of $9,679 using dexmedetomidine in patients ventilated for more than 24 h. The primary cost drivers were reduced costs of ICU stay and reduced costs of mechanical ventilation. Aggarwal et al. (39) conducted a cost-minimization analysis for short sedation in ICU, demonstrating that dexmedetomidine was associated with significant cost savings of ∼$6,000 compared with propofol and midazolam in mechanically ventilated, adult patients undergoing short-term sedation (<24 h). Although the limited generalizability of cost data outside the US, the costsaving was mainly related to the reduced ICU length of stay and the required monitoring.
An important area of research involves mechanically ventilated septic patients. In many basic and translational studies, dexmedetomidine has demonstrated anti-inflammatory and bacterial killing properties superior to those of gamma-aminobutyric acid (GABA) agonists, thereby, reducing the incidence of subsequent infections and 28-day mortality in patients with sepsis (28,40). Although indirectly correlated, another secondary analysis of the SPICE III trial (41) was carried out in patients with septic shock. On multivariableadjusted analysis, dexmedetomidine appeared to be associated with lower vasopressor requirements to maintain the target MAP.
Despite the immunomodulatory and anti-inflammatory properties of dexmedetomidine, in a recent multicenter, double-blind, randomized, controlled trial that enrolled adult mechanically ventilated patients with sepsis, the use of dexmedetomidine for light sedation did not demonstrate more days of life without acute brain dysfunction (coma or delirium) compared with propofol. In addition, no differences were found in ventilator-free days at 28 days, death at 90 days, or global cognition at 6 months (42).

NEUROMONITORING TO GUIDE SEDATION
As mentioned earlier, light sedation aims to obtain the minimum depth of sedation required to keep the patient free from agitation and anxiety, avoiding oversedation. Traditionally, the most widespread method to monitor the depth of sedation is clinical, using Richmond Agitation Sedation Scale (RASS) or Riker Sedation-Agitation Scale (SAS). However, clinical scales require awake patients and are not useful to monitor deep sedation (43). Therefore, other methods have been developed in the recent times to evaluate sedation depth through processedand raw-EEG signals. Processed-EEG signal (Bispectral Index, PSI, Entropy) has been proved, both in OR and ICU, to be related to the depth of sedation measured with clinical scales (43,44) and to reduce sedative dosages (45). Low values of a processed-EEG index (corresponding to unnecessary deep sedation, burst-suppression or isoelectric EEG) are associated with a higher incidence of delirium and mortality (46). Moreover, processed-EEG monitoring systems can also show the raw EEG traces and spectral quantitative array, allowing clinicians to identify specific electroencephalographic signatures of sedative drugs in addition to the general benefits of raw EEG monitoring (47).
Concerning the impact of objective measurement of the level of sedation on outcomes, the literature is scarce on high-quality studies. A recent systematic review and metaanalysis (48) included four RCTs and found no benefits of BIS monitoring on the clinical outcomes or resource utilization. A possible explanation of this conflicting and insufficient evidence may rely on the intrinsic limitations of the numerical dimensionless scales of alertness/unconsciousness derived from the electroencephalographic signal (49-51) rather than the electroencephalogram itself which presents a strong neurobiological background supporting its use (47,52) which should be promoted by the validated educational programs (53).

CONCLUSION
The approach to the management of critically ill patients is multidimensional, and targeted sedation should be considered in the context of the ABCDEF bundle. No doubt exists about the need to limit deep sedation to restricted clinical circumstances, while the light sedation (or non-sedation) approach could not explicit its beneficial effects when taken alone and not in the context of an ABCDEF bundle approach. When light sedation is applied, propofol or dexmedetomidine should be used, carefully evaluating their possible contraindications. Moreover, whenever clinical evaluation of the level of sedation is not feasible, a processed and raw EEG signal could be useful to avoid the detrimental effects of over-sedation (Figure 1).