Part of the Steamroller and Not Part of the Road: Better Blood Pressure Management Through Automation

In a survey published >10 years ago, Lienhart et al1 found that intraoperative hypotension was one of the most frequent factors associated with death related to anesthesia. As we know, hypotension is common during the intraoperative period and has been shown to occur most predominantly during anesthesia induction.2 Cumulative hypotension in the operating room has been associated with an increased risk of stroke,3 acute kidney injury,4 myocardial injury,5–7 and overall mortality.8 Even short periods of hypotension, however, may be sufficient to increase morbidity,7,9 and 2 large studies have recently confirmed that even mild hypotension still increases the overall mortality risk in adult patients undergoing noncardiac surgery.10,11 Hypotension may be even more frequent during spinal anesthesia, occurring in up to 60% of patients12,13 undergoing cesarean deliveries, which may affect the mother (nausea, vomiting, and, in extreme cases, cardiovascular collapse) and the fetus (fetal acidosis and hypoxemia). Avoiding hypotension is therefore a priority for anesthesiologists. While negative outcomes attributable to hypotension during anesthesia are well demonstrated, the evidence surrounding intraoperative hypertension and/or significant variability of blood pressure during surgery remains less clear. First, intraoperative hypertension has been demonstrated in at least one study to be noncontributory toward postoperative mortality.11 Moreover, it has been shown in at least in one study that lesser, not greater, blood pressure variability has been associated with higher mortality.14 The authors speculated that this finding could reflect an association with undesirable effects of autonomic dysfunction and result from limited reserve capacity in the cardiovascular system rather than be causative in itself. Thus, while there are only limited data at present, there is at least some evidence that intraoperative hypertension and blood pressure variability are not as dangerous as intraoperative hypotension. At present, then, the evidence supports rapid treatment of hypotensive episodes, even of those of short duration. Treatment should generally correspond to the presumed etiology, whether vasodilatory, hypovolemic, or cardiac. In general and during spinal anesthesia, the vast majority of hypotension is attributable to the vasodilatory and antisympathetic effects of the anesthetics themselves (with hypovolemia being next most common and cardiac causes, fortunately, being the least common of the 3). First-line treatment of anesthetic-induced hypotension, therefore, is almost always via vasopressors like phenylephrine or mixed agents like ephedrine. COMPUTER-ASSISTED VASOPRESSOR ADMINISTRATION This issue of Anesthesia& Analgesia contains 2 articles on closed-loop computer-controlled vasopressor administration for tight blood pressure control. The first, by Ribeiro Marques et al,15 describes the use of a novel closed-loop system using phenylephrine to treat normovolemic vasodilation induced by sodium nitroprusside in a swine model. They compared the performance of the system to physicians managing the same cases in a cross-over trial and hypothesized that the system would maintain blood pressure at a predefined target level more precisely and with less use of vasopressors. In their results, they report that the system did maintain tighter control than the physicians, but the small sample size limited the power of the study and the statistical significance of this difference. Nevertheless, the authors rightly note that the system prevented hypotension as effectively as the physicians and with less overshoot. Further, the physicians in the study had only one focus—blood pressure control—without any of the other tasks, distractions, and responsibilities of normal anesthesia care. The second study is by Ngan Kee et al16 and compared the performance of 2 different closed-loop approaches (boluses versus infusion) for the administration of phenylephrine in 214 healthy patients undergoing cesarean deliveries under spinal anesthesia. They demonstrated that when using closed-loop control guided by intermittent noninvasive blood pressure cuff monitoring at 1-minute intervals, the use of intermittent boluses resulted in more precise control of blood pressure and less total drug administration as compared with continuous infusion. The significance of this study is 2-fold: (1) closed-loop vasopressor administration can be achieved using intermittent noninvasive blood pressure monitoring; and (2) computer-directed bolus dosing may have added benefits in this setting when compared to a continuous infusion. Moreover, this is the largest clinical cohort in which computer-assisted blood pressure management has been reported to date. Both of these clinical studies utilizing unique computer-guided vasopressor systems, in addition to other such systems reported in the last few years (including one of our own), highlight the rapidly growing interest in how forms of basic automation may improve and standardize patient care.17–20 THE CASE FOR COMPUTER ASSISTANCE Steward Brand, the author and futurist, famously stated that “Once a new technology rolls over you, if you’re not part of the steamroller you’re part of the road.” The health care system is arguably being changed by technology faster now than at any point in history. As much as health care providers may be loath to admit it, we appear to be sitting at a juncture point between an antiquated system—one where medicine is practiced almost uniquely by every individual providing care—and a new one that is demanding ever more evidence of outcomes, proof of value, and standardization across patients and providers. The “art” of medicine needs to be redefined. To be certain, there is still room for individual insights and customization of care by both providers and patients themselves, but to simply not have any standard, pathway, or care plan for common diseases and conditions is no longer justifiable. We only need to look to concepts like enhanced recovery after surgery, the perioperative surgical home, and increasingly the larger health system conglomerates—the consistent common feature is standardization of care. Closed-loop control of infusions has been reported in many publications from as far back as the 1970s.21–23 They have been used for many clinical targets and incorporated many control types, from proportional-integral-derivative to reinforcement learning to fuzzy logic.24–26 The first vasopressor controller was likewise described in the 1970s, and research has continued into the last decade with both vasodilators and vasoconstrictors.26–28 The one consistent finding reported across time, studies, and control targets is that the computer systems are able to maintain the clinical targets more tightly than human operators.28–32 Closed-loop systems like those described in this issue are still predominantly a research tool today. However, we expect that in the coming years this will begin to change, possibly rapidly. The regulatory approval process for such devices will take time and significant effort from thought leaders and industry. This being said, we suspect that automated closed-loop systems will eventually be widespread—as common in our clinical spaces as they are presently in the rest of our lives. These systems, once deployed in critical care areas, will allow providers to set and maintain targets while reducing the need for manual adjustment. Moreover, they will allow increased focus on other crucial aspects of patient care, including establishing better communication and rapport with patients and surgeons. Standardized protocols and closed-loop systems are just 2 of the components that will help to decrease the variability related to the current implementation of health care. A third component that has not yet been applied to anesthesia is artificial intelligence. IBM Watson (IBM, Armonk, NY), for example, has already been utilized in some medical centers to aid physicians in disease diagnosis and determining potential treatment options.33,34 Tools like these are not subject to biases or cognitive errors and may provide insights into both individuals and populations concerning diagnoses and management plans. At present, artificial intelligence is being applied to all manner of problem solving in health care, from reading radiological studies to analyzing pathology specimens, and more. One might envision these same systems soon being applied to real-time intraoperative data in order to help manage hemodynamics, predict critical events before they occur, or provide advice for interventions that the anesthesia team may have overlooked. One need only look at the evolution of our cell phones and how we use them to interact with social networks, find reviews and recommendations for nearby restaurants and services, and even to provide us directions in real-time to see that technology is undeniably changing our world. The steamroller of health care reform and advancing technology is on us; whether and how we embrace it in the coming months and years will determine whether we are part of the steamroller or part of the proverbial road. DISCLOSURES Name: Alexandre Joosten, MD. Contribution: This author helped draft and approve the final manuscript. Conflicts of Interest: Dr Joosten is a consultant for Edwards Lifesciences. Name: Joseph Rinehart, MD. Contribution: This author helped draft and approve the final manuscript. Conflicts of Interest: Dr Rinehart is the owner and director of Sironis, Inc and a consultant for Edwards Lifesciences. This manuscript was handled by: Maxime Cannesson, MD, PhD.