We continue to read with great interest the description of the drug titration paradox by Minto et al.1 They are, in essence, describing a classical control theory problem on how to tune a control system, with an anesthesiologist being at the center of a control-feedback loop.It turns out that we anesthesiologists, at a population level, behave like an overdamped control system2 trying to slowly identify the correct target with no a priori knowledge (fig. 1). There appears to be a natural reluctance that biases using a preconceived delivery schedule instead of dosing to clinical effect. This may be due to clinical concerns because overshooting our target potentially risks either underdosing the patient and causing awareness or overdosing the patient, leading to burst suppression and hypotension.We would argue that if we were to seek better knowledge as to patient dosing3 or if we were more aggressive in our dosing strategy, we would not see the feature. In addition, closed-loop control anesthesia may also alter the results of this paradox.4 If we implemented a differently tuned controller, we might lose the negative correlation described in the paradox (fig. 1) or even see a positive correlation.In summary, the drug titration paradox is not a property of the observed patient population or the drug model. It is a property of the anesthesia teams delivering the anesthetic. Any conclusion drawn from population data concerning dosing should be understood in this context. We otherwise agree with the assessment that big data should very cautiously approach this problem.Dr. Kennedy is a consultant for GE HealthCare (Chicago, Illinois). The other authors declare no competing interests.

Introduction Electroconvulsive Therapy may be utilized in as many as 76,000 cases annually in the US, with the majority of cases employing succinylcholine. The reported dose spans the range of 0.29 - 2.1 mg/kg, and while motor seizures only last 36 ± 6 seconds, the duration of paralysis extends to 310 ± 38 seconds. While a model of succinylcholine pharmacokinetics/pharmacodynamics exists, this has not been employed to facilitate dose selection in clinical practice. Probability ramp control was investigated for this purpose. Methods Two approaches to dose finding were implemented. The first was an optimized Up-Down Method (UDM) that utilized an initial bolus, an adjustment dose, and a decrement to decrease the adjustment when crossing the target of 95% twitch depression. The second utilized probability ramp control (PRC) comprised of an infusion sequence that stopped when 95% twitch depression was obtained, a model that mapped the times for onset and offset of blockade to a subsequent bolus, and an adjustment dose to refine this dose when needed. Two populations of 10000 randomly parameterized models were developed from published data to train and evaluate the performance. Performance was assessed with a fuzzy classifier that segmented outcomes into three sets – LOW, HIGH, and SUCCESS. A loss function was developed that determined the number of sessions required to bring all models to SUCCESS. The probability distributions were compared using the Kolmogorov-Smirnov 2 sample test, with P<0.001 considered significant. Results Optimal values for the UDM parameters BOLUS, ADJUSTMENT, and DECREMENT were 0.7834 mg/kg, 0.3334 mg/kg, and 0.4056. Optimal values for the PRC SEQUENCE were 0.2663 mg/kg/min for 3 minutes followed by 0.7028 mg/kg/min. A fourth order polynomial MODEL produced estimates of the bolus that brought 99% of models to SUCCESS on the second session, while UDM required 6 sessions to achieve 99% SUCCESS. The probability distributions were distinct with P<<0.001. Discussion PRC was able to correctly produce SUCCESS in significantly fewer sessions than UDM. Additionally, PRC is easy to implement and allows pooling of results from multiple clinicians. The performance of PRC in clinical use for ECT will require further study. Key Points Question Can probability ramp control reduce the number of ECT sessions with suboptimal succinylcholine dosing? Findings Probability ramp control found the correct dose in two sessions in 99% of simulations, compared to six sessions for the Up-Down Method. Meaning Probability ramp control is a more efficient method for finding the appropriate dose of succinylcholine for repeated sessions of ECT.

A wealth of information about respiratory system is encoded in the airflow signal. While direct measurement of airflow via spirometer with an occlusive seal is the gold standard, this may not be practical for ambulatory monitoring of patients. Advances in sensor technology have made measurement of motion of the thorax and abdomen feasible with small inexpensive devices, but estimating airflow from these time series is challenging due to the presence of complicated nonstationary oscillatory signals. To properly extract the relevant oscillatory features from thoracic and abdominal movement, a nonlinear-type time-frequency analysis tool, the synchrosqueezing transform, is employed; these features are then used to estimate the airflow by a locally stationary Gaussian process regression. It is shown that, using a dataset that contains respiratory signals under normal sleep conditions, accurate airflow out-of-sample predictions, and hence the precise estimation of an important physiological quantity, inspiration respiration ratio, can be achieved by fitting the proposed model both in the intra- and inter-subject setups. The method is also applied to a more challenging case, where subjects under general anesthesia underwent transitions from pressure support to unassisted ventilation to further demonstrate the utility of the proposed method.

Introduction: Drug-induced sleep endoscopy (DISE) employs a drug such as propofol to allow observation of airway collapse to predict success in surgical interventions for obstructive sleep apnea. A number of protocols have been published for the administration of propofol during DISE, but these protocols do not make use of pharmacokinetic modeling to estimate the propofol concentration at airway collapse and the probability distribution for this quantity. We hypothesized that this information could be used to improve the delivery of propofol during DISE. Methods: Deidentified data from 404 patients sedated with Probability Ramp Control was analyzed. The estimated effect site concentration at the time of airway collapse (ECac) was fit to a gamma distribution. Correlation between ECac and age/weight were performed using Spearman ρ. This probability distribution was utilized to assess the ability of a published protocol to produce airway collapse in a reasonable duration in a representative cohort of patients. Results: The parameters of the gamma distribution were a=8.3 (95% confidence interval, 7.2–9.5), b=0.58 (95% confidence interval, 0.51–0.67). The mean ECac was 4.8±1.7 µg/mL; 95% of ECac fell in the range 2.4–8.5 µg/mL. ECac was uncorrelated with age (ρ=−0.06, P=0.27) and weight (ρ=−0.11, P=0.03). The previously published protocol displayed variability across the cohort and failed to achieve effect site concentrations sufficient for the least sensitive patients in our cohort. Conclusions: Effect site propofol concentrations at the time of airway collapse during DISE span a wide range with a skewed distribution. The lack of correlation with either age or weight suggests that the pharmacokinetic model is sufficient for describing the effect of age and weight on propofol dosing. The impact of these factors on propofol dosing during DISE is examined, and workable solutions to this clinical challenge are proposed.

Infusion systems are complicated electromechanical systems that are used to deliver anesthetic drugs with moderate precision. Four types of systems are describedgravity feed, in-line piston, peristaltic, and syringe. These systems are subject to a number of failure modesocclusion, disconnection, siphoning, infiltration, and air bubbles. The relative advantages of the various systems and some of the monitoring capabilities are discussed. A brief example of the use of an infusion system during anesthetic induction is presented. With understanding of the functioning of these systems, users may develop greater comfort.

Purpose of reviewSedation for nonoperating room procedures is experiencing a considerable increase in demand. Respiratory compromise is one of the most common adverse events seen in sedation. Capnography is a modality that has been widely adopted in this area, but may not be well suited to the special demands of nonoperating room sedation. This review is an assessment of new technologies that may improve outcomes beyond those achievable with capnography.Recent findingsNew devices for detecting the onset of apnea and for assessing respiratory depression have emerged which have advantages over conventional capnography for detecting apnea without excessive false positive and false negative rates. In addition, monitors that assess respiratory drive have become available, and these may prove useful in regulating depth of sedation.SummaryNo single monitor is ideal for all settings. During brief endoscopic sedation, detection of apnea is paramount, while during longer procedures, avoiding excessive respiratory depression is more critical. The clinician must choose the appropriate monitor based on an understanding of the challenges of the particular environment.

PURPOSE OF REVIEW:Provide a practical update on drug-induced sleep endoscopy (DISE) for anesthesia providers, which can also serve as a reference for those preparing to establish a DISE program.RECENT FINDINGS:New developments in surgical approaches to OSA and the growing global incidence of the condition have stimulated increased interest and demand for drug-induced sleep endoscopy. New techniques include transoral robotic surgery and hypoglossal nerve stimulation. Recent DISE literature has sought to address numerous debates including relevance of DISE findings to those during physiologic sleep and the most appropriate depth and type of sedation for DISE. Propofol and dexmedetomidine have supplanted midazolam as the drugs of choice for DISE. Techniques based on pharmacokinetic models of propofol are superior to empiric dosing with regard to risk of respiratory compromise and the reliability of dexmedetomidine to achieve adequate conditions for a complete DISE exam is questionable.SUMMARY:The role of DISE in surgical evaluation and planning for treatment of OSA continues to develop. Numerous questions as to the optimal anesthetic approach remain unanswered. Multicenter studies that employ a standardized approach using EEG assessment, pharmacokinetic-pharmacodynamic modelling, and objectively defined clinical endpoints will be helpful. There may be benefit to undertaking DISE studies in non-OSA patients.

Drug infusions are an increasingly common way of delivering anesthetic agents. With inhaled anesthetic agents, the concentration in the vessel-rich group associated with adequate anesthesia spans a narrow range, and these concentrations can be readily estimated from end-tidal concentrations. Intravenous agents are far more variable in the effective concentration, and monitoring this concentration at the point of care is not likely to be available any time soon. Manufacturers of infusion systems have recognized the need for limits on the infusion rates to avoid errors in programming, and libraries of such limits are promulgated based on the judgment of individuals who may have experience in delivering these drugs … or not. Berman1 presents a methodology for establishing limits based on the documented actions of anesthesiologists participating in the Multicenter Perioperative Outcomes Group database. Presumably, anesthesiologists are rational actors, and their actions should reflect safe practice. Large databases collected from electronic medical records of multiple institutions reduce the chances of institutional bias. How can we argue against this? Applying this methodology to driving speeds seems fairly easy: data from the GPS navigation company TomTom5 suggest that the average speed on Interstate 15 in Utah and Nevada is 77.67 miles per hour, while on the Washington DC beltway, it is 46 miles per hour. But before we develop a statistical model of highway speeds and change the signs, recall that the infusion rate is not analogous to car speed; rather, it is the gas pedal of the pharmacokinetic car. While it is possible that there is a relationship between fuel consumption and speed in a single automobile model under ideal conditions (Volkswagen has done some pioneering work in this area), applying this approach with corporate average fuel efficiency would not work well for regulating speeds on the hills of Interstate 15 or in stop-and-go traffic on the beltway. What if we could estimate the concentrations of drugs based on their administration? Well, we can. It is called target-controlled infusion (TCI) and will eventually be available to US practitioners.2,3 With TCI, we do not really worry about infusion rates; we control plasma or effect site concentrations. Even if we do not have access to a TCI pump, there are programs such as Rugloop6 and TIVA Trainer7 that can give us these numbers. If you want to go to the pharmacokinetic speed shop, MATLAB models8 are publicly available. Using this approach, my standard induction consists of boluses of 1000 µg/kg of propofol and 1.5 µg/kg of remifentanil over 1 minute, followed by infusions of 330 µg/kg/min of propofol and 0.3 µg/kg/min until I have sufficient information to specify the maintenance infusion rates. An example is depicted in the Figure. While this approach might seem unusual to an American practitioner, this is a fairly common strategy for TCI: set the pumps to high targets for induction, then turn down the targets for maintenance. Egan and Shafer4 referred to this as “surfing the wave,” although their example of an initial hand-delivered bolus is more analogous to tow-in surfing.9Figure.: Pharmacokinetic simulation of delivering propofol 1000 µg/kg and remifentanil 1.5 µg/kg over 1 min, followed by propofol 330 µg/kg/min and remifentanil 0.3 µg/kg/min for 48 s, followed by propofol 60 µg/kg/min and remifentanil 0.18 µg/kg/min for the remaining time using the Cortínez and Kim models (see OpenTCI.org) for a 72-year-old, 90-kg, 170-cm male.In my practice of total intravenous anesthesia, I attempt to meticulously document the infusion rates; with institutional review board permission, I reviewed 57 consecutive records from cases performed in first 2 months of 2017 and found that 35% (118 of 333) of recorded propofol rates and 43% (140 of 323) of recorded remifentanil rates exceeded the 90% upper limits reported by Berman.1 Should I be concerned about my dosing patterns? The propofol rates that exceeded the 90% cutoff comprised about 5% of the total case time, and only 0.5% of the time after induction. The impact of these brief periods when the “pedal is to the metal” have a trivial impact on the average effect site concentration of propofol; they are employed to get the effect site concentration to a higher target quickly, without overshoot. While my approach may seem complex and cumbersome, it has the advantage of allowing me to observe the transition from consciousness to unconsciousness with moderate precision. My goal is not to establish an effect site concentration that works in 50% or 90% of patients; I am trying to gauge the pharmacodynamic response for the patient in front of me. If effect site concentrations are analogous to driving speeds, pharmacodynamic responses are more analogous to following distances; on the beltway, I do not care as much about whether I am going the speed limit as whether I am too close to the car ahead of me to stop. I certainly would not try to explain away a fender bender by pointing to the dashboard indicator of my fuel efficiency. Could large databases provide insight into response probabilities, allowing us to “crowdsource pharmacokinetics?”10 Only if we control our drugs to individualized effect site targets, rather than using the same population-based infusion rates for everyone. In other words, “don’t drive like my brother!” DISCLOSURES Name: Jeff E. Mandel, MD, MS. Contribution: This author wrote the manuscript. This manuscript was handled by: Maxime Cannesson, MD, PhD.

The electrophysiology suite is a foreign location to many anesthesiologists. The initial experience was with shorter procedures under conscious sedation, and the value of greater tailoring of the sedation/anesthesia by anesthesiologists was not perceived until practice patterns had already been established. Although better control of ventilation with general anesthesia may be expected, suppression of arrhythmias, blunting of the hemodynamic adaptation to induced arrhythmias, and interference by muscle relaxants with identification of the phrenic nerve may be seen. We review a range of electrophysiology procedures and discuss anesthetic approaches that balance patient safety and favorable outcomes.

Procedural sedation is commonly employed in endoscopic procedures, and increasingly uses propofol. The use of propofol is commonly restricted to anesthesia providers, and this may increase the cost of care. Administration of propofol requires a special set of skills to deal with the variability of patient response and the consequences of improper dosing. This has stoked interest in the use of automated systems to reduce manpower costs associated with propofol. This article examines why propofol poses challenges for human control, and how various automated systems have been used to address these challenges. We examine target-controlled infusions, patient-controlled sedation, the SEDASYS System, and optimized ramp induction. The article emphasizes on how the various approaches deal with the range of variability in propofol response. No single system is capable of dealing with all patients without some human supervision and intervention.

BACKGROUND:Anesthesia and sedation are associated with paradoxical breathing. Respiratory inductance plethysmography (RIP) permits measurement of respiratory motion in clinical settings not conducive to spirometry, but correlation of RIP volume changes and spirometer flow in the time domain is degraded by the development of paradoxical breathing. The Hilbert-Huang transform (HHT) is a nonlinear signal analysis method that permits the instantaneous magnitude and phase of nonstationary signals to be estimated in the frequency domain. We hypothesized that these frequency domain estimates would provide higher correlation between RIP and spirometer signals than time domain signals during the transition between normal and paradoxical breathing. METHODS:From 51 patients undergoing sevoflurane anesthesia for minor procedures, a 5-minute epoch containing transitions between pressure support ventilation and spontaneous ventilation was selected for analysis. Pearson correlation for models based on HHT magnitude and phase was compared with models based on time domain signals. Bland-Altman analysis was performed to assess deviation from linearity in the models. RESULTS:For the 51 patients analyzed, the modulation of tidal volume over the epoch ranged from 30% to 215% of epoch mean. The coefficient of determination for time domain analysis was 0.62 ± 0.2 compared with 0.93 ± 0.07 for the HHT model incorporating phase. This improvement of 0.31 (99% confidence interval, 0.24-0.37) was significant (P < 0.0001). No trend was observed in prediction residuals. CONCLUSIONS:Under conditions of changing ventilation, HHT-derived magnitude and phase measures provide higher correlation with spirometry than those obtained with traditional time domain methods.

Introduction: During advanced bronchoscopic procedures, adequate sedation is required to obtain patient cooperation. Sedation with conventional agents such as midazolam and fentanyl carries a risk of respiratory depression. Dexmedetomidine and ketamine are associated with minimal respiratory depression. We hypothesized that the combination of these agents would reduce the requirement for fentanyl and midazolam and decrease respiratory depression with similar patient and proceduralist satisfaction. Methods: 50 patients undergoing flexible bronchoscopy and curvilinear-probe endobronchial ultrasound were randomly assigned to receive dexmedetomidine / ketamine or placebo. Supplemental fentanyl and midazolam could be requested by the proceduralist. Ventilation was assessed by respiratory inductance plethysmography (RIP), arterial saturation by pulse oximetry, and satisfaction by the Likert scale. Patients, proceduralists, and anesthesiologists were blinded to allocation. Midazolam and fentanyl plasma concentrations were estimated from dosing history using validated pharmacokinetic models, relative minute ventilation derived from RIP, and desaturation by cumulative time below 90%. Results: Groups were similar. Patients in the experimental group achieved lower mean fentanyl and midazolam levels. Satisfaction scores and time below oxygen saturation of 90% were not different between groups. The relative decrease below baseline minute volume was greater in the control group 0.736 (0.566 0.848) compared to experimental 0.764 (0.592 0.891), P<0.0001. Conclusions: The addition of dexmedetomidine and ketamine to midazolam/fentanyl sedation reduced the requirement for these agents, did not alter satisfaction or oxygen desaturation, but did reduce the magnitude of decrease in minute ventilation. These effects may be greater in an open label study.

In the current issue, Lovich et al.1 present work demonstrating the impact of reduced dead-volume in infusion systems during changes in carrier flow. Although this may seem obvious after careful consideration, it has been my experience that most of us do not exercise careful consideration when setting up infusion systems; we are driven by convenience and are captives to purchasing agents who only consider cost. Without data to drive our decisions, it is hard to know when we should follow the path of least resistance and when we should dig in and insist on properly configured systems. This article provides us with these data. So how do we use them? Stopcocks are ubiquitous, and are often used to combine multiple infusions, but they can add significant dead-volume to infusion systems. Multiport infusion devices that eliminate dead-volume are new items that must be justified. The impact of added dead-volume is never good, but it depends on the magnitude of carrier flow variations and the concentration of drugs administered. Gravity-fed carriers can exhibit a range of flows as the height of IV bag relative to the vein changes or as blood pressure cuffs go up above the IV.2 Large-bore catheters are more susceptible to changes in venous resistance than small-bore catheters.3 Placing a carrier in the sweeper position (the last stopcock port) increases the potential for variations in flow to affect drug delivery. The consequences of clearing the dead-volume depend on what is being administered. If we are providing analgosedation with remifentanil 100 μg/mL at 0.1 μg/kg/min to a 100-kg patient with a 2-mL dead-volume, clearing the dead-volume yields a 200-μg bolus, which could lead to a serious complication. Conversely, delivering no analgesic during the 20 minutes required for refilling the dead-volume may cause problems, too. As Lovich et al. have shown, the consequences of dead-volume can be predicted by understanding the characteristics of the infusion system. For reasons that elude me, I am periodically called upon to teach residents about the relative efficiency of Mapleson circuits in spontaneous versus positive pressure ventilation. I have not used a Mapleson circuit to deliver volatile agents in more than 25 years (although I use a Waters’ canister with some regularitya), but it is a rare day in which I do not use an infusion system that is affected by the issues raised by this article. Infusion systems are the infrastructure of modern anesthesia, not celebrated by politicians on the Fourth of July, but essential nonetheless, and worthy of understanding. DISCLOSURES Name: Jeff E. Mandel, MD, MS. Contribution: This author helped write the manuscript. Attestation: Jeff E. Mandel approved the final manuscript. This manuscript was handled by: Maxime Cannesson, MD, PhD.

BACKGROUND: Accurate monitoring of respiratory rate may be useful for the early detection of patient deterioration. Monitoring of respiratory rate in the operating room under general anesthesia by spirometry is technically straightforward and demonstrates high fidelity. Accurate measurement of the respiratory rate of an unattended patient outside the operating room is fraught with challenges. Monitors such as capnometry and thoracic impedance pneumography have significant drawbacks. Respiratory acoustic monitoring (RRa (TM)) is a new technology for respiratory rate monitoring, which has been demonstrated to provide accurate respiratory rates in patients recovering from anesthesia, but the performance of this RRa-enabled monitor under conditions of major respiratory rate variation has not been evaluated.METHODS: We enrolled 53 patients undergoing urologic procedures in the operating room under general anesthesia with a laryngeal mask airway, spontaneous ventilation, and no muscle relaxation in an observational study. Respiratory signals (RRa and in-circuit pneumotachograph) were stored for later analysis. Artifacts were excluded based on visual inspection of the raw respiratory waveforms. Instantaneous respiratory rates were obtained from the pneumotachograph signal using the Hilbert-Huang Transform. Instantaneous rate estimates (IREs) were compared with RRa by 3 methods. First, the mean delay between IREs and RRa was determined. Second, precision was obtained by Bland-Altman analysis for repeated measures. Third, for all disparities in rates exceeding 4 breaths per minute (bpm), the probability of persistent error was determined as a function of time, with 95% confidence intervals estimated by bootstrap analysis.RESULTS: Data were collected from 53 patients. Three patients were excluded due to missing data. There were no adverse events related to RRa monitoring. RRa demonstrated a median delay of 45 seconds (interquartile range 20 seconds) to detect a 1- bpm change in IREs. Bland-Altman revealed 95% limits of agreement of-2.1 to 2.2 bpm across the range of 7 to 48 bpm. Disparities in respiratory rate >4 bpm between the 2 methods did not persist beyond 160 seconds, and 90% of these differences resolved within 33 seconds (95% confidence interval 23-48 seconds).CONCLUSIONS: The data demonstrate that, under conditions of general anesthesia with a laryngeal mask airway and spontaneous ventilation, the RRa rapidly detects changes in respiratory rate, demonstrates minimal bias, and when errors in rate occur, these do not persist. The utility of this monitoring technology in detecting rate changes in unattended patients will require further study.