Closing the loop: automation in anesthesiology is coming

Abstract

Anesthesiology and intensive care medicine provide fertile ground for innovation in automation, but to date we have only achieved preliminary studies in closed-loop intravenous drug administration. Anesthesiologists have yet to implement these tools on a large scale despite clear evidence that they outperform manual titration. Closed-loops continuously assess a predefined variable as input into a controller and then attempt to establish equilibrium by administering a treatment as output. The aim is to decrease the error between the closed-loop controller’s input and output. In this editorial we consider the available intravenous anesthesia closed-loop systems, try to clarify why they have not yet been implemented on a large scale, see what they offer, and propose the future steps towards automation in anesthesia.

1. Editorial

Anesthesiology and intensive care medicine provide fertile ground for innovation in automation, but to date we have only achieved preliminary studies in closed-loop intravenous drug administration. Anesthesiologists have yet to implement these tools on a large scale despite clear evidence that they outperform manual titration, [1, 2] as Spataru and colleagues show in their meta-analysis published in this issue of the Journal [3]. Closed-loops continuously assess a predefined variable as input into a controller and then attempt to establish equilibrium by administering a treatment as output [4, 5]. The aim is to decrease the error between the closed-loop controller’s input and output. A simple analogy to a single variable closed-loop is to consider it like extra pairs of eyes and hands that constantly assess and optimize a variable to reach a predefined target. Let’s consider the available intravenous anesthesia closed-loop systems, try to clarify why they have not yet been implemented on a large scale, see what they offer, and propose the future steps towards automation in anesthesia.

1.1. Closed-loop intravenous anesthesia

For more than two decades, many studies have shown the benefits of closed-loop intravenous drug administration [1, 2]. Automated propofol, norepinephrine, insulin, neuromuscular blockade, and fluid administration have all been evaluated and the results are consistent: patients that benefit from automation, instead of manually titrated drug administration, have better care [6-10]. Sparatu and colleagues analyzed 32 studies that used single variable closed-loops to administer intravenous agents and showed that time in target was consistently higher when using a closed-loop [3]. In the case of propofol and neuromuscular blockade, reversal of effect was also quicker with automated care.

1.2. Limitations to implementation

Closed-loop systems are now part of daily life – outside of medicine. They help us with mundane tasks, such as maintaining a refrigerator’s temperature, and keep us safe and comfortable during more complex tasks such as flying an airplane or driving a car. Intravenous drug administration, however, has lagged behind other innovations. Several potential reasons exist. The first concerns the importance of tightly maintaining intraoperative targets. Anesthesiologists have historically been more concerned with awareness than excessive anesthetic depth, [11] and even today some clinicians are not very concerned with burst suppression [12]. This can lead to the tendency of over titrating anesthesia and neglecting its toxic effects (e.g., hypotension, burst suppression, and potentially neuronal apoptosis). Moreover, the negative sequalae of awareness is always attributable to anesthesia whereas the toxic effects of overdose may also occur due to other factors, creating a subtle but very real treatment bias by anesthesia providers. With growing evidence that both burst suppression and low alpha power are associated with delirium, [13, 14] however, experts are starting to underline the importance of avoiding potentially toxic doses of anesthetics. A similar explanation can be found for hypotension, which has traditionally been considered a normal part of anesthesia. There is no longer such a consensus as there now exists an abundance of evidence on the association between hypotension and morbidity [15-17]. In addition to these changes in practice, other limiting factors for closed-loop systems include regulatory obligations, challenges in design, and acceptance of practitioners.

A potential design that can help to not only facilitate regulatory controls, but also increase ease of implementation into clinical practice is to convert closed-loops into openloop systems. For example, the assisted fluid management (AFM) system (Edwards Lifesciences, Irvine, USA) has gained CE mark and FDA approval. This tool uses identical software to an automated fluid delivery system previously developed at the University of California Irvine, but provides the added option to suggest or refuse fluid challenges [18]. This added human intervention provides high level fluid status assessment with the added benefit of clinician control [18-20]. A final limitation to closed-loop implementation is that anesthesia consists of multiple components, including anesthetic depth, antinociception, neuromuscular blockade, mechanical ventilation, electrolyte balance, and hemodynamic optimization. Each component is important, but optimizing only one has limited impact on patient outcome. Instead, a personalized approach that integrates every component would probably have much stronger impact on outcome.

1.3. Benefits of automating care

Compliance to protocol is by far the strongest advantage of automated systems. Even in the most stringent study conditions, protocol compliance almost never reaches 100% when intravenous anesthetics are administered manually. Since closed-loops focus continuously on their input variable, time in target is higher and this translates to better compliance with the benefits of a personalized target (e.g., maintaining mean blood pressure within 10% of a patient’s baseline value). An added advantage in addition to their higher precision in care is that they simultaneously decrease practitioner workload. These tools effectively free clinicians from changing infusion speeds, which can be time consuming, repetitious, and distracting. A final strong advantage of closed-loops is that different closed-loops can be used simultaneously and maintain multiple variables within target. Several studies have evaluated such approaches with promising results [19, 21-24].

1.4. Closed-loops cannot replace human intelligence

Clearly, closed-loops today cannot replace anesthesiologists. These tools maintain targets with very high compliance while simultaneously decreasing workload due to simple repetitive tasks. They have no role, however, in choosing or modifying targets. For example, a physician must still determine the fluid strategy (e.g., more restrictive or liberal) or the exact anesthetic depth target range. Furthermore, maintaining targets is only a small part of the complex profession of anesthesiology. One key task that is irreplaceable today is that of situation awareness [25]. Only a human is capable of perceiving the key events of patient care, comprehending their meaning, anticipating future events, and communicating effectively with members of the perioperative team.

1.5. Integrating closed-loops: the future

The true potential for automation of anesthesia is not simply to have one or even multiple closed-loop systems that maintain variables in target, but to have a single multiple input multiple output (MIMO) controller. As previously stated, anesthesia consist of many components that should ideally be all maintained within therapeutic targets. Closed-loops can maintain variables with high precision thanks to near maximal, and even supra-human, compliance [10, 19]. The added advantage of a MIMO controller would be to incorporate all aspects of care and establish a logical and beneficial equilibrium. One risk of an independent system, contrarily to an interdependent system such as a MIMO controller, would be to have one component pulling others towards a poorly tolerated equilibrium. For example, when independent fluid and vasopressor systems are used concomitantly, a vasopressor controller may over treat and not consider fluid status. This could maintain target blood pressure, but at the cost of decreased cardiac output. Another limitation is that multiple independent closed-loops each require their own hardware (e.g., controller, screen, and wires), which can be particularly cumbersome at the bedside. The role of the MIMO would be to use algorithms to personalize the patient’s anesthesia while continuously checking that each component (e.g., blood pressure, cardiac output, anesthetic depth, antinociception, etc.) is optimal (Fig. 1). In doing so, every component of anesthesia would be maintained with very high adherence to predefined targets while clinicians would be free to focus on the big picture (e.g., modification of targets, communication with team members, and anticipation of key events).

Fig. 1.

Comparison of combined independent anesthesia and hemodynamic controllers with a multiple input multiple output anesthesia and hemodynamic controller. Multiple input multiple output controllers aim to offer the advantages of integrating data from all components of anesthesia and hemodynamics. This tool would theoretically achieve an equilibrium between these components and centralize automation to a single controller. This figure was designed using biorender.com. MIMO multiple input multiple output NMB neuromuscular blockade

Spataru and colleagues’ meta-analysis demonstrates once again the key benefits of closed-loops. It is necessary, today, to continue moving forward and to advance on automating all key components of intravenous anesthesia so that we can easily personalize patient care with very high compliance to protocol. Automation of anesthesia and intensive care is the future, and its implementation has already begun.