You can respond to this article at http://www.anaesthesiacorrespondence.com
Utility of Nociceptive Flexion Reflex Threshold, Bispectral Index, Composite Variability Index and Noxious Stimulation Response Index as measures for nociception during general anaesthesia
Article first published online: 18 MAY 2012
Anaesthesia © 2012 The Association of Anaesthetists of Great Britain and Ireland
Volume 67, Issue 8, pages 899–905, August 2012
How to Cite
von Dincklage, F., Correll, C., Schneider, M. H. N., Rehberg, B. and Baars, J. H. (2012), Utility of Nociceptive Flexion Reflex Threshold, Bispectral Index, Composite Variability Index and Noxious Stimulation Response Index as measures for nociception during general anaesthesia. Anaesthesia, 67: 899–905. doi: 10.1111/j.1365-2044.2012.07187.x
- Issue published online: 9 JUL 2012
- Article first published online: 18 MAY 2012
- Accepted: 29 March 2012
Movement and haemodynamic responses to noxious stimuli during general anaesthesia are regarded as signs of nociception. We compared the Nociceptive Flexion Reflex Threshold (NFRT), Bispectral Index (BIS), Composite Variability Index (CVI), Noxious Stimulation Response Index (NSRI) and the calculated propofol/remifentanil effect-compartment concentrations (Ce) as predictors for such responses in 50 female subjects at laryngeal mask airway insertion and skin incision. The following prediction probabilities (PK-values) were obtained at laryngeal mask airway insertion and skin incision, respectively. For movement responses: NFRT = 0.77 and 0.72; p = 0.0001 and 0.004, respectively; BIS = 0.41 and 0.56, p = 0.29 and 0.5, respectively; CVI = 0.48 and 0.57, p = 0.76 and 0.88, respectively; NSRI = 0.49 and 0.76, p = 0.92 and 0.0001, respectively; propofol-Ce = 0.35 and 0.66, p = 0.04 and 0.03, respectively; remifentanil-Ce = 0.68 and 0.72, p = 0.01 and 0.003, respectively. For heart rate responses: NFRT = 0.68 and 0.75, p = 0.04 and 0.01, respectively; BIS = 0.37 and 0.59, p = 0.15 and 0.41, respectively; CVI = 0.41 and 0.44, p = 0.39 and 0.37, respectively; NSRI = 0.48 and 0.53, p = 0.84 and 0.78, respectively; propofol-Ce = 0.42 and 0.56, p = 0.39 and 0.53, respectively; remifentanil-Ce = 0.58 and 0.54, p = 0.35 and 0.73, respectively. We conclude that the NFRT best predicts movement and heart rate responses to noxious stimuli. Effect-compartment concentrations and NSRI also predict movement (but not heart rate) responses satisfactorily.
Although the cerebral effect of hypnotics may be monitored using processed electroencephalography, no comparable monitor exists to measure nociception during general anaesthesia. In clinical practice, opioids are administered according to haemodynamic and movement responses to noxious stimuli .
The correlation of such responses with the hypnotic depth as measured by the bispectral index (BIS) is relatively poor [2, 3]. The composite variability index (CVI), which is derived from the BIS, may provide a better correlation and whereas promising results are presented in abstract form [4, 5], no full research paper has yet been published.
The nociceptive flexion reflex (NFR) is a polysynaptic spinal withdrawal reflex, which can be assessed by electromyography of the biceps femoris muscle during the application of electrocutaneous stimuli to the ipsilateral sural nerve. The stimulus intensity required to elicit the reflex at its threshold can be used as an objective measure of the individual nociceptive threshold [6, 7]. We have demonstrated before that the Nociceptive Flexion Reflex Threshold (NFRT) can be used to predict movement responses following noxious electrical stimulation under different anaesthetic regimens [2, 3, 8].
The noxious stimulation response index (NSRI) is derived from drug effect-site concentrations, calculating their combined effect based on a hierarchical interaction model [9, 10]. The model has been sophisticated further to adjust for stimuli of different strength and has been calibrated to allow its application as a monitoring instrument for pain .
The purpose of this study was to investigate the utility of these different measures of nociception during general anaesthesia, using their ability to predict movement and heart rate responses following laryngeal mask airway (LMA) insertion and skin incision.
After approval of the local ethics committee and written informed consent, we recruited 50 female patients undergoing breast surgery. Inclusion criteria were a propofol-based anaesthesia regimen and use of a LMA. The single exclusion criterion was pre-operative treatment with analgesic substances, either chronic or during the last 24 h before surgery.
Peri-operative care was conducted according to clinic standards and independently of the study. This included premedication which was administered approximately 30–60 min before induction of anaesthesia, using midazolam at doses of 0.5–1 mg.kg−1. Before induction of anaesthesia, intravenous access via a forearm vein and standard monitoring including non-invasive blood pressure, electrocardiography and pulse oximetry were established (Intellivue MP70 patient monitor, Philips, Eindhoven, Netherlands). The monitor data were continually recorded for study purposes using the software Trendface (ixellence GmbH, Wildau, Germany). After application and testing of the electrodes for BIS, CVI and NFRT, the continual recording of these variables was started.
Anaesthesia was induced via target controlled infusion pumps (Injectomat TIVA Agilia, Fresenius Kabi, Bad Homburg, Germany) using propofol (bolus injection followed by continuous infusion to maintain calculated effect-site concentrations of 3–5 μg.ml−1, using the model of Schnider et al. ) followed by remifentanil (bolus injection followed by continuous infusion to maintain calculated effect-site concentrations of 0–4 ng.ml−1, using the model of Minto et al. ) after the loss of consciousness. Effect-site concentrations were recorded throughout the study for calculation of the NSRI. Just before and during induction of anaesthesia a tight-fitting facemask was applied and end-tidal carbon dioxide concentrations were kept stable through manual assisted ventilation. After at least 3 min to allow for the equilibration of remifentanil a LMA (LMA Unique, LMA, San Diego, USA) was inserted. Laryngeal mask airway sizes 3, 4 and 5 were used according to body weight. If the patient moved any body part during insertion or for 60 s after, this was counted as a ‘response’. Heart rate responses were calculated as the difference between the maximum heart rate during the 60 s before LMA insertion and the maximum heart rate until 60 s after insertion, with an increase > 5 min−1 counted as a ‘response’. After confirming adequate ventilation via the LMA, patients were positioned for surgery followed by disinfection, sterile covering and subsequently skin incision. Movement and heart rate responses were calculated as for LMA insertion.
The BIS was recorded via bilateral frontal electroencephalogram electrodes using the BIS VISTA bilateral monitoring system (Covidien, Boulder, USA). The CVI was calculated by Covidien staff from datasets recorded using the BIS VISTA bilateral monitoring system.
To elicit the NFR of the left biceps femoris muscle, stimuli were repeatedly applied to the left sural nerve at its retromalleolar pathway via surface electrodes. Stimuli were applied automatically at randomised intervals of 8–12 s to avoid habituation, with each stimulus consisting of a volley of five rectangular electrical pulses of 1 ms duration each, at 200 Hz (2 × DS5 Stimulator, Digitimer Ltd, Hertfordshire, UK). The reflex response was recorded over the left biceps femoris muscle using an electromyogram. The recorded signals were amplified and filtered (g.BSamp Biosignal Amplifier, g.tec, Schiedlberg, Austria), digitised at a sampling rate of 5 kHz (Mikro 1401 mk II, CED Ltd, Cambridge, England), rectified and analysed using the software Signal 3.10 (CED Ltd, Cambridge, England).
To determine the threshold of the NFR an automated threshold tracking system was used . This system varies the stimulus intensity applied to the sural nerve according to an up-down-staircase algorithm with a variable step length to estimate the stimulus intensity associated with a 50%-probability of reflex occurrence, which is defined as the reflex threshold . Reflex occurrence was defined as an interval peak z score > 10.32 in the poststimulation interval of 90 ms to 150 ms .
Calculation of the NSRI was performed using the formulae of Luginbuehl et al. . The NSRI is based on a hierarchical interaction model in which an incoming noxious stimulus is supposed to be first attenuated dose-dependently by the opioid, modelled as a fractional Emax model with a fixed slope of one. The probability that the attenuated stimulus would lead to a response or not is then dependent on hypnotic dose, modelled as a positive Emax model with a variable slope. To adjust for stimuli of differing strength, the stimuli are normalised to the stimulus of ‘shaking and shouting’ at the patient by calculating a relative stimulus strength by dividing the concentration of the hypnotic at 50% stimulus-tolerance with the concentration of the hypnotic at 50% tolerance of ‘shaking and shouting’. Finally the NSRI is normalised to a scale from 0 to 100 and calibrated so that a value of 50 corresponds with a 50% tolerance of the stimulus and a value of 20 corresponds with a 90% tolerance of the stimulus.
The effect-site concentrations of propofol and remifentanil for calculation of the NSRI were recorded from the TCI infusion pumps. The NSRI was calculated from these data using the parameter values from Luginbuehl et al.  and Bouillon et al. .
The sample size was calculated based on parameter variability data from previous studies. Based on a probability of positive responses to the stimuli of at least 25%, an alpha-error of 0.05 and a power of 0.8, we calculated a minimum sample size of 40 patients. To adjust for possible drop-outs and patients whose heart rate could not be monitored with sufficient accuracy, we included 50 patients in total.
Analysis was performed with six predictors (NFRT, BIS, CVI, NSRI, propofol Ce, remifentanil Ce) each for two responses (movement responses, heart rate responses) to each of two stimuli (LMA insertion, skin incision). To quantify the accuracy of the predictors in determining the correct response, the prediction probability was calculated for each predictor-response pair. The prediction probability is a measure of the accuracy of a monitoring device, where a value of 1 stands for a 100% correct differentiation between positive and negative responses, whereas a value of 0.5 represents only a 50% chance of a correct differentiation. For dichotomic responses, as in this study, the prediction probability equals the area under the curve of the sensitivity-specificity-curve (receiver operating characteristic). The estimation of prediction probability-values was performed using the spreadsheet macro PKMACRO by Smith et al. . Standard errors of the estimates were computed by the jackknife method.
The study was performed in 50 female patients undergoing breast surgery, with a median (IQR [range]) population age of 54 (64–43 [75–30]) years , a height of 164 (171–160 [185–152]) cm, a weight of 64 (73–57 [98–47]) kg and a body mass index of 23 (27–21 [32–17]) kg.m−2. The mean drug concentrations and predictor values at loss of consciousness, LMA insertion and skin incision are shown in Table 1. Observation of movement responses was successful in all 50 included patients and positive movement responses occurred in 26 patients after LMA insertion and in 18 patients after skin incision. Heart rate responses of the heart rate could not be monitored in all patients due to technical problems with the data extraction from the monitor (n = 3), pre-operative treatment with beta blockers (n = 5) and atrial fibrillation (n = 1). From the remaining 41 patients, increases in heart rate occurred in 15 patients after LMA insertion and in 10 patients after skin incision.
|Loss of consciousness||LMAinsertion||Skin incision|
|NFRT; mA||10.7 (10.4)||31.5 (28.8)||42.9 (33.9)|
|BIS||62.8 (14.5)||42.8 (12.4)||26.7 (8.8)|
|CVI||3.7 (1.3)||2.8 (1.6)||0.8 (0.5)|
|NSRI||85.3 (5.1)||38.8 (14.1)||66.5 (20.1)|
|Propofol Ce; μg.ml−1||3.7 (0.7)||3.7 (0.7)||3.5 (0.7)|
|Remifentanil Ce; ng.ml−1||0.0 (0.0)||2.3 (0.8)||0.9 (0.8)|
Figures 1–6 show the original data points for each predictor, separated according to the respective stimuli and responses. The difference between the mean values for positive and negative responses was largest for NFRT (Fig. 1), to an extent for movement response to skin incision for NSRI (Fig. 4), and also for movement responses with skin incision and LMA insertion for Ce propofol and remifentanil (Figs 5 and 6).
The prediction probabilities of each predictor for each stimulus and each response are shown in Table 2, confirming the statistical significance of the values suggested in the figures.
|Movement responses||Heart rate responses|
|LMA||Skin incision||LMA||Skin incision|
|NFRT||0.77 (0.07)||p = 0.0001||0.72 (0.08)||p = 0.004||0.68 (0.09)||p = 0.04||0.75 (0.10)||p = 0.01|
|BIS||0.41 (0.08)||p = 0.29||0.56 (0.09)||p = 0.50||0.37 (0.09)||p = 0.15||0.59 (0.11)||p = 0.41|
|CVI||0.46 (0.13)||p = 0.76||0.48 (0.15)||p = 0.88||0.36 (0.16)||p = 0.39||0.31 (0.21)||p = 0.37|
|NSRI||0.49 (0.09)||p = 0.92||0.76 (0.07)||p = 0.0001||0.48 (0.10)||p = 0.84||0.53 (0.11)||p = 0.78|
|Propofol Ce||0.35 (0.07)||p = 0.04||0.66 (0.07)||p = 0.03||0.42 (0.09)||p = 0.39||0.56 (0.10)||p = 0.53|
|Remifentanil Ce||0.68 (0.07)||p = 0.01||0.72 (0.07)||P = 0.003||0.58 (0.09)||p = 0.35||0.54 (0.10)||p = 0.73|
Our results demonstrate that movement responses following skin incision can be predicted using NFRT, NSRI or drug effect-site concentrations. However, movement responses following LMA insertion could only be predicted using NFRT or drug effect-site concentrations.
Figure 5 seems to suggest a paradoxical effect of propofol on movement responses with LMA insertion: the higher the propofol concentration the more the movement response (Fig. 5). As NSRI is calculated on the assumption that propofol decreases movement with increasing dose, the paradoxical association found here may be artificially contributing to a poorer performance of NSRI in our study.
The paradoxical effect for propofol can be explained as an artifact that occurs at population level caused by the manner of drug dosing. In this study, propofol was administered in increasing doses until the loss of consciousness. Therefore, patients who had a high drug tolerance received higher doses of propofol. These patients with a high drug tolerance can also be expected to require a higher dose of remifentanil to suppress reactions to the noxious stimuli. However, as remifentanil dosing was not adjusted according to drug tolerance, the patients with a high drug tolerance were relatively underdosed with remifentanil and therefore showed a higher rate of positive reactions to the noxious stimuli. This yields an apparent association between higher doses of propofol and more movement responses.
The inability of BIS and CVI to predict responses in this study despite their rather good performance in other studies [2–5] can be explained by the study design. As we did not strictly control anaesthetic doses through the study protocol, we assessed the devices in a situation that mirrors normal clinical practice with normal dosing variability and predictor variability. Measures like BIS and CVI seem not to be able to distinguish between positive and negative responses in such a clinical setting, as they are able during the simplified task in experimental settings with a reduced inter-individual variability.
Our finding with relation to the NFRT is consistent with some other results [2, 3, 8]. However, ours is the first demonstration of its applicability in a clinical setting, investigating responses to skin incision as well as correlation with haemodynamic responses. The NFRT has disadvantages in that the procedure required to site the electrodes and apparatus to record the reflex is rather complex and therefore takes time and experience. Furthermore, the recording of the reflex can be influenced by the use of neuromuscular blocking drugs, electromagnetic noise, anatomical problems due to the location on the nerves, or neurological problems that disrupt the reflex. It is possible that the NFRT represents a useful tool for research, rather than clinical measurement.
By contrast, the drug concentration measures seem to provide a more promising approach for clinical use. For the prediction of movement responses following skin incision the NSRI achieved a comparable level of accuracy to the NFRT. However, further investigations are needed to improve the underlying models and interaction parameters for responses other than movement. These models will also need to take into account of concomitant drug use.
Competing interests and acknowledgements
The project was supported by Deutsche Forschungsgemeinschaft, Bonn, Germany; grants BA2868/2-1 and BA2868/3-1. Dr von Dincklage is participant in the ‘Friedrich C. Luft’ Clinical Scientist Pilot Program funded by Volkswagen Foundation and Charité Foundation. We are very grateful to PD Dr Martin Luginbuehl (Bern, Switzerland) for many helpful discussions and assistance with NSRI calculation.
- 5Lower Composite Variability Index (CVI) was associated with better clinical global impression scores. European Journal of Anaesthesiology 2009; 26 (Supplement 45): 30., , , , .
- 9Defining depth of anesthesia. In: Schuettler J, Schwilden H, eds. Modern Anesthetics. Berlin: Springer, 2008: 209–23., .
- 10Hypnotic and opioid anesthetic drug interactions on the cns, focus on response surface modeling. In: Schuettler J, Schwilden H, eds. Modern Anesthetics. Berlin: Springer, 2008: 471–87..