To the Editor: 

As a practicing anesthesiologist with a Ph.D. in control systems engineering, I feel compelled to comment on both the merits and hazards of closed-loop control (CLC) in the operating room, as discussed in the February 2012 article by Liu et al.  1and the accompanying editorial.2 

CLC is a mature field of engineering with well established standards of analysis, design, and reporting – points missing from Liu et al. 's article and the accompanying editorial. This absence suggests unfamiliarity with the hazards of CLC, which include the risk of introducing instability where none existed before.

My main criticism of Liu et al. 's work is that there is no evidence that the appropriate groundwork was done to ensure that the control algorithm was safe (stable) before entering the operating room. Minimizing the hazards of CLC requires a thorough stability analysis before implementation. By way of comparison, it is as important to precede CLC operational testing with computer modeling and simulation studies as it is to precede human drug trials with animal modeling and testing. Testing under extremes of “physiologically challenging conditions such as hypertension, hypotension, morbid obesity, in pediatric patients, or during major surgery such as cardiac surgery or lung transplantation,” as the authors propose for their next clinical trial (see Discussion1), should have been done by simulation before the first clinical trial.

Controller design is the key to ultimate success and acceptance of any closed-loop strategy. There are dozens of ways to design closed-loop algorithms. Of all of the available approaches, it is noteworthy that the authors decided to use the PID, or proportional, integral, derivative, approach. The details of how PID controllers work are not important for this point of discussion. The important issue is that the PID approach is the most basic and unsophisticated of algorithms–the first one learned by every CLC engineer in their first undergraduate control course–and is suitable for some simple mechanical, electrical, or hydraulic systems (e.g. , automobile cruise control), but not for complicated, time-variant, nonlinear systems with time delays (e.g. , the surgical patient).

Lastly, Liu et al. 's article lacks the necessary engineering detail to evaluate the work. By piecing together comments from the present paper and previous work,3,4it is possible, however, to determine that the controller is in fact not a PID controller. Instead, the authors used empirical methods to develop look-up tables and logic trees to mimic the proportional (P) and derivative (D) terms of a classic PID while ignoring the integral (I) term altogether. The result is that the control algorithm looks more like a crude expert system than a PID controller. An expert system of this type is what would result from a survey of anesthesiologists asked what they would do with propofol and remifentanil infusion rates given particular entropy values.

In conclusion, what the authors have done is replace the human anesthesiologist with an ideal, empirically derived, computer anesthesiologist dedicated to the one task of maintaining a set entropy value. For this menial task, the computer wins. For the advancement of closed-loop control design in the operating room, no one wins.

Studies using CLC have the potential to significantly impact our specialty and, more importantly, our patients. When considering the publication of CLC articles, the editors of ANESTHESIOLOGY must consider using a control systems engineer in the peer review process. There is precedent for recent and good systems engineering work in the anesthesia literature. I refer to the work of De Smet et al. ,5,6where the authors follow a sound engineering development strategy.

Finally, perhaps Crosby and Culley think the future bright if “a dozen operating rooms filled with patients having surgery” are managed by a “lone anesthesiologist … from a distant location.”2In my view, this is an unholy grail. Almost every paper about “closed-loop” anesthesia discusses the similarities of our environment and the cockpit. Without doubt there is one way they are exactly alike: Patients and passengers would recoil if they were to learn that their anesthesiologist or pilot would be in abstentia. 

The ideal goal of control systems engineering in the operating room is not the elimination of anesthesia staff, but the ability to prescribe an effect rather than a dose. Imagine a future where we say “set end-tidal carbon dioxide to 36, set mean arterial pressure to 80, set heart rate to 70, set pain response to 0, set awareness to 45,” and then watch as safe, stable, robust, and reliable control algorithms adjust all fluids and anesthetics at once to achieve these results. This is a bright future, indeed.

1.
Liu N, Le Guen M, Benabbes-Lambert F, Chazot T, Trillat B, Sessler DI, Fischler M: Feasibility of closed-loop titration of propofol and remifentanil guided by the spectral M-Entropy monitor. ANESTHESIOLOGY 2012; 116:286–95
2.
Crosby G, Culley DJ: Processed electroencephalogram and depth of anesthesia: Window to nowhere or into the brain? ANESTHESIOLOGY 2012; 116:235–7
3.
Liu N, Chazot T, Hamada S, Landais A, Boichut N, Dussaussoy C, Trillat B, Beydon L, Samain E, Sessler DI, Fischler M: Closed-loop coadministration of propofol and remifentanil guided by bispectral index: A randomized multicenter study. Anesth Analg 2011; 112:546–57
4.
Liu N, Chazot T, Genty A, Landais A, Restoux A, McGee K, Laloë PA, Trillat B, Barvais L, Fischler M: Titration of propofol for anesthetic induction and maintenance guided by the bispectral index: Closed-loop versus  manual control: a prospective, randomized, multicenter study. ANESTHESIOLOGY 2006; 104:686–95
5.
De Smet T, Struys MM, Neckebroek MM, Van den Hauwe K, Bonte S, Mortier EP: The accuracy and clinical feasibility of a new bayesian-based closed-loop control system for propofol administration using the bispectral index as a controlled variable. Anesth Analg 2008; 107:1200–10
6.
De Smet T, Struys MM, Greenwald S, Mortier EP, Shafer SL: Estimation of optimal modeling weights for a Bayesian-based closed-loop system for propofol administration using the bispectral index as a controlled variable: A simulation study. Anesth Analg 2007; 105:1629–38