Custos (http://www.custos.co.nz/) is a new evidence-based monitoring system designed to enhance the clinician's awareness of possible pathophysiological changes during anaesthesia. It extracts routinely collected data, weighs up the evidence of changes of several parameters over time and suggests a diagnosis when identifying concordance in changes. It aims to improve harmful event detection and decision making.1 Among Custos alerts, ‘sympathetic response’ emerges when heart rate trends upwards, pulse oximeter amplitude downwards and blood pressure increases by two standard deviations over 30 s,2 corresponding to a significant response to noxious stimulation. When heart rate responses are absent, the software produces an ‘achronotropic sympathetic response’ alert. The ‘hypovolaemia-likely’ alert is based on real-time assessment of pulse pressure variation, alerting the clinician when the increase is over 13%. The ‘hypovolaemia-possible’ alert aims at an earlier detection of hypovolaemia, and occurs when the pulse oximeter pulse amplitude trends down, heart rate trends up and blood pressure falls two standard deviations over 30 s.3
The clinical utility and validity of Custos have only started to be evaluated. Hence, we compared alerts given by the software and other independent indicators. Classical anaesthesia monitoring data collected during other studies4,5 were entered into the Custos (EBMi) software offline. They were initially collected from 28 patients undergoing intracranial neurosurgery under propofol and remifentanil anaesthesia, who had given their consent for the anonymous use of their data (Ethical Committee no. 707; University Hospital and Faculty of Medicine of the University of Liege, Liege, Belgium; Belgian number: B70720084063; internal reference: 2008/115; Chairperson Professor M. Lamy; acceptance on 29 May 2008). The Custos output files were scrutinised for the incidence of both ‘sympathetic response’ and ‘achronotropic sympathetic response’ alerts per hour of anaesthesia, as well as for ‘hypovolaemia-likely’, ‘hypovolaemia-possible’ and ‘low blood pressure’ alerts within 30 min following a validated measure of hypovolaemia, the Delta Down.6 The ‘low blood pressure alert’ uses an algorithm that incorporates the age of the patient and certain pathophysiologies – such as aortic stenosis, ischaemic heart disease or cerebrovascular disease. Of note, Custos does not repeat the same alert until 5 min has passed to avoid alarm fatigue.
According to the initial study protocol,4,5 patients were receiving target-controlled infusions (TCI) of propofol and remifentanil adjusted to keep a State Entropy (GE Healthcare Helsinki, Finland) between 40 and 60 and stable haemodynamic parameters. The anaesthesiologist in charge of managing the patient was blind to the concentration of remifentanil contained in the syringe: 25, 50 or 75 μg ml−1. The practitioner set remifentanil target concentrations on the TCI device, according to haemodynamic parameters, but without knowing the exact delivered concentration. The number of ‘sympathetic response’, ‘achronotropic sympathetic response’ and combined ‘sympathetic–achronotropic sympathetic response’ alerts per hour of anaesthesia was compared between the three groups of patients (Kruskal–Wallis and Mann–Whitney tests). The correspondence between the incidence of alerts and effective concentration of remifentanil in the syringe was assessed using the prediction probability (PK).7 The occurrence of ‘low blood pressure’, ‘hypovolaemia-likely’ and ‘hypovolaemia-possible’ alerts within 30 min following each of the three Delta Down measures performed in each patient under stable propofol–remifentanil anaesthesia was compared with a referential occurrence of hypovolaemia defined as a Delta Down value of at least 6 mmHg. Accuracy of predictions made by those alerts, as well as by all their possible combinations, was analysed using classical performance indicators and prediction probability.
We observed a significantly higher incidence of ‘achronotropic sympathetic response’ alerts in the 25 μg ml−1 than in the 75 μg ml−1 remifentanil group (Kruskal–Wallis H = 6.42; P = 0.04; Mann–Whitney U = 20; P = 0.02). The incidence of ‘sympathetic response’ or combined responses was not significantly different between the groups. The prediction probability of ‘achronotropic sympathetic response’ at predicting remifentanil concentration in the syringe was 0.27 [95% confidence interval (CI), 0.11 to 0.43]. ‘Sympathetic response’ and combined responses had no significant predictive ability [PK value of 0.57 (95% CI, 0.39 to 0.75) and 0.38 (95% CI, 0.18 to 0.58), respectively]. Accuracy of hypovolaemia prediction was calculated on data obtained in 27 patients (81 pairs of data, one patient with missing Delta Down values). Statistics describing the ability of ‘hypovolaemia-likely’, ‘hypovolaemia-possible’, ‘low blood pressure’ and all possible combinations of those alerts to predict a Delta Down value higher or equal to 6 mmHg adequately are reported in Table 1. ‘Hypovolaemia-likely’ alerts alone were not efficient at predicting a Delta Down value indicative of hypovolaemia, with poor sensitivity and prediction probability not significantly higher than 0.5. Contrarily, single ‘hypovolaemia-possible’ or ‘low blood pressure’ alerts showed comparable efficacy in that respect. The best sensitivity/specificity combination was provided by combined ‘hypovolaemia-likely’, ‘hypovolaemia-possible’ and ‘low blood pressure’ alerts.
Our results indicate that Custos is able to recognise patterns of change and combines pieces of evidence to suggest plausible diagnoses. Its ability to detect a sympathetic response is intuitively correct, at least with the amount of remifentanil being given. Insofar as ‘sympathetic response’ alert frequency was comparable between the groups, practitioners adequately adjusted remifentanil concentration to account for sympathetic responses when increased heart rate and blood pressure were both present. The third criteria entering into the ‘sympathetic response’ alert algorithm, namely decreased pulse amplitude, probably did not influence practitioners’ decision, because its recognition necessitates trained eye attention to blood pressure or plethysmographic tracing on the monitor. Contrarily, as demonstrated by the significantly lower ‘achronotropic sympathetic response’ alert frequency in the high remifentanil concentration group, anaesthesiologists in charge of patients were less efficient at avoiding achronotropic sympathetic responses when using low concentrations of remifentanil. They were less alerted by increases in blood pressure without tachycardia, so as to motivate a change in remifentanil concentration. Given the apparent concordance between Custos alerts and nociception–antinociception balance, studies should be specifically designed to compare it with already existing indicators of sympathetic activity (such the Surgical Plethysmographic Index) and assess its influence on antinociception management and patient outcome, while also addressing the eventual influence of nonnociception-related changes in sympathetic activity, blood pressure, heart rate and pulse amplitude on the occurrence of Custos alerts. Our results also show that Custos is able to recognise hypovolaemia as defined by the Delta Down, particularly when hypovolaemia-related alerts are combined. The prediction is not perfect, with a relatively high number of false negatives. This may be because of the imperfection of the Custos algorithms, or to the imperfect ability of Delta Down to correctly identify hypovolaemia. A prospective study using gold standard indicators of hypovolaemia is needed to clarify this point.
The main limitation of our study relies in its retrospective nature, using reanalysis of data collected during another study. Our results should serve as a basis for designing future prospective studies aiming at addressing each endpoint specifically. Custos shows promising qualities as an intra-operative decision-helping tool for anaesthesiologists. It alerts clinicians on the occurrence of intra-operative achronotropic sympathetic responses that trigger less correcting adjustments by practitioners than classical sympathetic response. Those alerts are in accordance with the antinociceptive regimen being received by the patients.
Acknowledgements relating to this article
Assistance with the study: none.
Financial support and sponsorship: this work was supported by the Department of Anaesthesia and ICM, Liege University Hospital, Liege, Belgium.
Conflicts of interest: MH is the designer of Custos software.
1. Raymer KE, Bergstrom J, Nyce JM. Anaesthesia monitor alarms: a theory-driven approach. Ergonomics
2. Connor CW, Gohil B, Harrison MJ. Triggering of systolic arterial pressure alarms using statistics-based versus threshold alarms. Anaesthesia
3. Harrison MJ. Evidence-based monitoring for the diagnosis of hypovolemia. Anesth Analg
4. Hans P, Verscheure S, Uutela K, et al. Effect of a fluid challenge on the Surgical Pleth Index during stable propofol-remifentanil anaesthesia. Acta Anaesthesiol Scand
5. Bonhomme V, Uutela K, Hans G, et al. Comparison of the surgical Pleth Index with haemodynamic variables to assess nociception-antinociception balance during general anaesthesia. Br J Anaesth
6. Deflandre E, Bonhomme V, Hans P. Delta down compared with delta pulse pressure as an indicator of volaemia during intracranial surgery. Br J Anaesth
7. Smith WD, Dutton RC, Smith NT. Measuring the performance of anesthetic depth indicators. Anesthesiology