brain research 1608 (2015) 147–156
Available online at www.sciencedirect.com
www.elsevier.com/locate/brainres
Research Report
Introduction and validation of a less painful algorithm to estimate the nociceptive flexion reflex threshold Gregor Lichtnern, Anna Golebiewski, Martin H Schneider, Falk von Dincklage Charité - Universitätsmedizin Berlin, Klinik für Anästhesiologie mit Schwerpunkt operative Intensivmedizin, Campus Charité Mitte und Campus Virchow-Klinikum, Berlin, Germany
art i cle i nfo
ab st rac t
Article history:
The nociceptive flexion reflex (NFR) is a widely used tool to investigate spinal nociception for
Accepted 25 February 2015
scientific and diagnostic purposes, but its clinical use is currently limited due to the painful
Available online 5 March 2015
measurement procedure, especially restricting its applicability for patients suffering from
Keywords:
chronic pain disorders. Here we introduce a less painful algorithm to assess the NFR
Nociceptive flexion reflex
threshold. Application of this new algorithm leads to a reduction of subjective pain ratings
NFR
by over 30% compared to the standard algorithm. We show that the reflex threshold estimates
Reflex threshold
resulting from application of the new algorithm can be used interchangeably with those of the
Pain measurement
standard algorithm after adjusting for the constant difference between the algorithms.
RIII reflex
Furthermore, we show that the new algorithm can be applied at shorter interstimulus intervals than are commonly used with the standard algorithm, since reflex threshold values remain unchanged and no habituation effects occur when reducing the interstimulus interval for the new algorithm down to 3 s. Finally we demonstrate the utility of the new algorithm to investigate the modulation of nociception through different states of attention. Taken together, the here presented new algorithm could increase the utility of the NFR for investigation of nociception in subjects who were previously not able to endure the measurement procedure, such as chronic pain patients. & 2015 Elsevier B.V. All rights reserved.
1.
Introduction
The nociceptive flexion reflex is a spinal withdrawal reflex that can be assessed by electromyographic recording of the biceps femoris muscle during electrocutaneous stimulation of the ipsilateral sural nerve (Sandrini et al., 2005; Skljarevski and Ramadan, 2002). The strong correlation of the reflex
threshold with the subjective pain threshold has made the NFR a widely used tool to investigate pain processing, pharmacological and psychological modulation of nociception as well as chronic pain conditions (Lim et al., 2011). However, the applicability of the method is currently limited to subjects who are willing to endure the repeated painful stimulations required for assessing the reflex threshold.
n Correspondence to: Charité - Universitätsmedizin Berlin, Klinik für Anästhesiologie mit Schwerpunkt operative Intensivmedizin, Campus Charité Mitte, Charitéplatz 1, 10117 Berlin, Germany. Fax: þ49 30 450 531927. E-mail address: [email protected] (G. Lichtner).
http://dx.doi.org/10.1016/j.brainres.2015.02.049 0006-8993/& 2015 Elsevier B.V. All rights reserved.
148
brain research 1608 (2015) 147–156
This poses a significant restriction for the method, especially
painful suprathreshold stimulation intensities. Therefore application of the 1:4 algorithm can be expected to induce less subjective pain than the standard 1:1 algorithm. Here we demonstrate that (i) the 1:4 algorithm leads to less painful stimuli than the standard 1:1 algorithm, while (ii) agreement of the reflex thresholds estimated by the two algorithms is sufficient to allow them to be used interchangeably. We show that (iii) the new 1:4 algorithm allows for stimulating at shorter interstimulus intervals than at which the standard algorithm is commonly applied. Finally we show that (iv) the 1:4 algorithm can be used to investigate modulation of nociception.
when applied on patients suffering from chronic pain disorders, who exhibit a higher degree of pain sensitivity and pain catastrophizing (Ruscheweyh et al., 2013, 2012; Osman et al., 2000). To increase the utility of the method, an optimisation of the measurement procedure towards a lower pain induction seems necessary. The standard procedure to assess the NFR threshold is an adaptive testing algorithm that increases the stimulation intensity when no reflex is detected and decreases it when a reflex occurs (Sandrini et al., 2005). While this standard algorithm is usually applied with variable stimulation intensity step sizes, the up and down steps are equal in size. Here we propose a less painful algorithm, which in contrast uses an up/down ratio of 1:4, meaning that upon detection of a reflex the stimulation intensity is reduced four times as much as it is increased when no reflex occurs. This stimulation paradigm inherently leads to lower average stimulation intensities compared to the standard 1:1 algorithm, as the reduction of the stimulation intensity after every reflex occurrence is four times larger. Besides this direct effect, the 1:4 algorithm utilizes the probabilistic nature of the reflex occurrence (Sandrini et al., 2005) to provoke reflex responses below the reflex threshold. Since the 1:4 algorithm reduces the stimulation intensity in larger steps than it increases it back again, its application results in repeated stimulations below the reflex threshold. Single stimuli at such subthreshold intensities are unlikely to elicit reflex responses. However, because the overall probability to elicit a reflex cumulates with repeated stimulation, the application of multiple subthreshold stimuli increases the total probability of reflex responses already below the reflex threshold. Since the stimulation intensity is reduced after every reflex occurrence, such reflex responses at stimulation intensities below the reflex threshold prevent the algorithm to reach
ns
** ***
General results
The new 1:4 algorithm led to lower average pain ratings, lower average stimulation intensities and yielded lower reflex threshold estimates compared to the standard 1:1 algorithm (Fig. 1). Population average reduction of the subjective pain ratings between the two algorithms amounted to 34% and
* ***
*** 30
***
*** ***
30
***
***
***
70 60 50 40 30 20 10
25
80 70 60 50 40 30 20
20
15
10
5
*** 1:1 algorithm 1:4 algorithm
25
Reflex threshold [mA]
Maximal subjective pain rating
90
80
Subjective pain rating
2.1.
2.2. Differences in subjective pain ratings, stimulation intensities and reflex threshold estimates between the standard 1:1 and the new 1:4 algorithm
100
***
90
Results
56 (28 male/28 female) of 60 participants completed all measurement blocks and were included in the analysis. All 4 excluded participants did not tolerate the stimulation at an interstimulus interval of 1 s and aborted the respective blocks. The remaining 56 participants had a median (IQR) age of 23 (22 25), a median height of 177 (170–184) cm and a median weight of 69 (60–80) kg.
Stimulus intensity [mA]
100
***
**
2.
20
15
10
5
10
0
0
0
Sensory deprived
Distracted
Sensory deprived
Distracted
0
Sensory deprived
Distracted
Sensory deprived
Distracted
Fig. 1 – Comparison of subjective pain ratings, stimulation intensities and reflex threshold estimates between the new 1:4 and the standard 1:1 algorithm. Shown are (a) the individual average subjective pain ratings, (b) the individual maximal subjective pain ratings, (c) the individual average applied stimulation intensities and (d) the individual average reflex threshold estimates during application of the standard 1:1 algorithm (crosses) and the new 1:4 algorithm (circles), each at interstimulus intervals of 6 s, separated between the attention state of sensory deprivation and the state of distraction. Lines represent the population means and the respective standard errors. Stars mark significant differences between the states and the algorithms (ns: not significant; *: po0.05; **: po0.01; ***: po0.001; post hoc test with Bonferroni correction).
149
Sensory deprived
Sensory deprived 4.0
10 8 6 4 2 0 -2 -4 -6 -8 -10
Between algorithm variability [mA]
mean standard deviation 95 % limit of agreement
3.5 Standard deviation 95 % Confidence band
3.0 2.5 2.0 1.5
0
5
10
15
20
0
25
Average of reflex threshold estimates (1:4 algorithm & 1:1 algorithm) [mA]
Distracted
Distracted
4.0
10 8 6 4 2 0 -2 -4 -6 -8 -10
10 20 30 40 50 60 70 80 90 100 Number of included reflex threshold estimates per subject
Between algorithm variability [mA]
Difference between reflex threshold estimates (1:4 algorithm - 1:1 algorithm) [mA]
Difference between reflex threshold estimates (1:4 algorithm - 1:1 algorithm) [mA]
brain research 1608 (2015) 147–156
3.5 3.0 2.5 2.0 1.5
0
5
10
15
20
25
Average of reflex threshold estimates (1:4 algorithm & 1:1 algorithm) [mA]
0
10 20 30 40 50 60 70 80 90 100 Number of included reflex threshold estimates per subject
Fig. 2 – Bland Altman analyses of the variability between the reflex thresholds estimated by the two algorithms. The BlandAltman plots picturing the variability between the two algorithms at interstimulus intervals of 6 s for the states of (a) sensory deprivation and (c) distraction. Each data point represents one individual, showing the individual differences between the reflex threshold estimates obtained using the two different algorithms (ordinate) over the corresponding individual average of the reflex threshold estimates obtained using the two different algorithms (abscissa). Individual reflex threshold estimates were calculated for each subject and each algorithm as the average of all values obtained during the respective measurement block. Solid lines indicate the mean differences between the algorithms and the 95% limits of agreement. Dashed lines indicate standard deviations of the differences. Figures b and d show the standard deviations of the differences between the algorithms (ordinate) obtained from multiple Bland Altman analyses performed as in a and c, but differing in the number of reflex threshold estimates per individual (abscissa) that were included into the analyses. Dashed lines indicate the standard deviations obtained from the analyses shown in (a) and (c), when all reflex threshold estimates per individual of the respective measurement block were included into the analysis. Shaded areas indicate 95% confidence bands.
37% for the state of sensory deprivation and the state of distraction, respectively. Population average reduction of the maximal subjective pain ratings between the two algorithms amounted to 14% and 17% for the state of sensory deprivation and the state of distraction, respectively. Population average reduction of stimulation intensities amounted to 28% (sensory deprivation) and 35% (distraction) and the population average reduction of the reflex threshold estimates amounted to 11% (sensory deprivation) and 24% (distraction).
2.3. Agreement between reflex threshold estimates of the standard 1:1 and the new 1:4 algorithm The individual mean reflex threshold estimates were significantly correlated between the two algorithms in both
attention states at an interstimulus interval of 6 s (Pearson’s correlation 0.91 for state of sensory deprivation, 0.86 for state of distraction, po0.001 for both). BlandAltman analyses revealed a mean difference between the individual average reflex threshold estimates of the 1:1 algorithm and the 1:4 algorithm of 1.0771.97 mA (mean7SD) for the state of sensory deprivation and 1.2472.42 mA (mean7SD) for the state of distraction (Fig. 2a and c). BlandAltman analyses of differences between the reflex threshold estimates of the 1:1 and the 1:4 algorithm when including only a specific number of reflex threshold estimates per subject into the analyses showed a maximum variability of 2.97 mA (SD) for the sensory deprived state and 3.27 mA (SD) for the distracted state, both in the case when only one
150
brain research 1608 (2015) 147–156
Sensory deprivation 15 12
1s
Distracted
median interquartile range
15 12
9
9
6
6
3
3
p > 0.13
Reflex threshold [mA]
Reflex threshold [mA]
Interstimulus interval
0 1 2 3 4 5 6 7 8 9 10
0 1 2 3 4 5 6 7 8 9 10
3s
15 12 9 6 3
p > 0.66
0
15 12 9 6 3
0 1 2 3 4 5 6 7 8 9 10
15
15
12
12
9
9
6
6
3
p > 0.12
0
0 1 2 3 4 5 6 7 8 9 10
6s
p > 0.45
0
0
3
p > 0.40
p > 0.20
0
0 0 1 2 3 4 5 6 7 8 9 10 time [min]
0 1 2 3 4 5 6 7 8 9 10 time [min]
Fig. 3 – Stability of the reflex threshold estimates over time. Shown are the time courses of the population medians and the interquartile ranges of the reflex thresholds estimated using the new 1:4 algorithm at interstimulus intervals of 1 s, 3 s and 6 s, each at the attention state of sensory deprivation and distraction. Reported p-values correspond to estimated population slopes of the linear mixed effect models.
reflex threshold estimate per subject was included into the analyses (Fig. 2b and d).
2.4. Applicability of the 1:4 algorithm at different interstimulus intervals The reflex threshold estimates resulting from application of the 1:4 algorithm were stable with no systematic drift over the measurement blocks of 10 min length for all interstimulus intervals and attention states (Fig. 3). Application of the new 1:4 algorithm also did not result in significantly different pain ratings or reflex threshold estimates when comparing measurements at interstimulus intervals of 3 s and 6 s. Using the 1:4 algorithm at an interstimulus interval of 1 s resulted in partly different subjective pain ratings and reflex threshold estimates than the interstimulus intervals of 3 and 6 s (Fig. 4).
intervals of 6 s and 3 s, but not at the 1 s interval (Fig. 5a). Reflex threshold estimates were significantly reduced in the state of distraction compared to the state of sensory deprivation for all interstimulus intervals used (Fig. 5b). Population average reduction of the subjective pain ratings between the two states of attention when using the 1:4 algorithm amounted to 2%, 22% and 20% for interstimulus intervals of 1 s, 3 s and 6 s, respectively, and to 16% when using the standard 1:1 algorithm at an interstimulus interval of 6 s. Population average reduction of mean estimated reflex thresholds using the new 1:4 algorithm amounted to 10%, 20% and 21% for interstimulus intervals of 1 s, 3 s and 6 s, respectively, and to 17% when using the standard 1:1 algorithm at an interstimulus interval of 6 s.
3. 2.5. Utility of the 1:4 algorithm to investigate modulation of nociception Application of the new 1:4 algorithm resulted in significantly lower subjective pain ratings in the state of distraction compared to the state of sensory deprivation when stimulating at
Discussion
3.1. Application of the new 1:4 algorithm induces less pain compared to the standard algorithm Our results show that the 1:4 algorithm leads to lower subjective pain ratings than the standard 1:1 algorithm,
151
ns
50
** ns
ns
ns
40 30 20 10 0
1s 3s 6s
1s 3s 6s
Sensory deprived
Distracted
50
ns
*** ***
ns
*** ** **
ns
40
Reflex threshold [mA]
Subjective pain rating
ns
Maximal subjective pain rating
brain research 1608 (2015) 147–156
30 20 10 0
1s 3s 6s
1s 3s 6s
Sensory deprived
Distracted
10
ns
***
ns ns
8 6 4 2 0
1s 3s 6s
1s 3s 6s
Sensory deprived
Distracted
Fig. 4 – Effect of different interstimulus intervals on subjective pain ratings and reflex threshold estimates. Shown are (a) average subjective pain ratings, (b) maximal subjective pain ratings and (c) reflex threshold estimates at interstimulus intervals of 1 s (light grey), 3 s (dark grey) and 6 s (black) for the new 1:4 algorithm in both states of attention. Stars mark significant differences between the interstimulus intervals (ns: not significant; **: po0.01; ***: po0.001, post hoc test with Bonferroni correction). Error bars depict the standard error of the mean.
1:4 algorithm 40 35 30 25 20 15 10 5 0
***
**
1:1 algorithm ns
ns
**
12 Reflex threshold [mA]
Subjective pain rating
ns
10
1:4 algorithm
1:1 algorithm
*
***
***
*
***
***
1s
3s
6s
1s
3s
6s
8 6 4 2 0
1s
3s
6s
1s
3s
6s
Interstimulus interval
Interstimulus interval
Sensory deprived Distracted Fig. 5 – Effect of different states of attention on subjective pain ratings and reflex threshold estimates at different interstimulus intervals. Shown are (a) subjective pain ratings and (b) reflex threshold estimates in the state of sensory deprivation (black) and the state of distraction (grey) for the new 1:4 algorithm (filled bars) and the standard 1:1 algorithm (unfilled bars). Stars mark significant differences between the attention states (ns: not significant; *: po0.05; **: po0.01; ***: po0.001, post hoc test with Bonferroni correction). Error bars depict the standard error of the mean.
which can be explained by the lower stimulation intensities that are applied when using the 1:4 algorithm (Fig. 1c). One obvious reason for the lower stimulation intensities is the direct consequence of the stimulation paradigm of the new 1:4 algorithm, as decreasing the stimulation intensity by 4 mA after a reflex occurred inherently leads to lower stimulation intensities than decreasing by 1 mA. This direct effect can be expected to decrease average stimulation intensities of the new 1:4 algorithm by 1.5 mA compared to the standard 1:1 algorithm (Fig. 6). However, this direct effect explains only about half of the overall reduction of the stimulation intensities, which in total amounted to 2.7–2.8 mA when comparing the 1:4 algorithm to the 1:1 algorithm. The additional reduction when using the 1:4 algorithm can be assumed to be caused by the interaction between the 1:4 algorithm and the probabilistic nature of the reflex occurrence. The probability to elicit a reflex is a function of the applied stimulation intensity and increases gradually with increasing
stimulation intensity (Levitt, 1971; Sandrini et al., 2005). Hence there is no hard threshold above which a stimulus always leads to a reflex response and below which reflexes never occur. Instead the reflex threshold is merely the stimulation intensity at which a reflex response is elicited with a certain probability, usually 50% (Levitt, 1971). Therefore stimulations below that reflex threshold are still capable, although less probable, to elicit a reflex. The 1:4 algorithm utilizes this characteristic to provoke reflex responses already at stimulation intensities below the threshold. Because of the larger decrease of the stimulation intensity after a reflex occurs in comparison to the smaller increasing steps when no reflex is detected, every reflex occurrence is followed by multiple subthreshold stimuli. Each of these subthreshold stimulations alone has a low probability of eliciting a reflex, but the cumulated probabilities of all subthreshold stimuli together result in a higher total probability of eliciting reflex responses at stimulation intensities below the reflex threshold. Since after occurrence of such reflex responses
152
brain research 1608 (2015) 147–156
Stimulus intensity
1:1 algorithm 1:4 algorithm
1 mA
Stimulus number Fig. 6 – Stimulation paradigms of the new 1:4 and the standard 1:1 algorithm. For the standard 1:1 algorithm (crosses) the stimulation intensity is increased by 1 mA every time no reflex is detected and decreased by 1 mA after every reflex detection (arrow heads). For the new 1:4 algorithm (circles) the stimulation intensity is also increased by 1 mA every time no reflex is detected, but in contrast to the standard algorithm the stimulation intensity is decreased by 4 mA every time a reflex is detected.
as after any reflex occurrence the stimulation intensity is decreased, a higher rate of reflexes occurring below the reflex threshold leads to lower average stimulation intensities. Most parsimoniously, this effect can explain the stronger reduction of the applied stimulation intensities and the subjective pain ratings for the 1:4 algorithm in comparison to the 1:1 algorithm than what can be explained by the direct effect of the different stimulus paradigms alone. It has been proposed to reduce the induced pain for the subjects simply by using a single ascending series of stimuli and thereby reducing the number of stimuli required for the measurement to a minimum (Rhudy and France, 2011). However, such reduction of the number of stimuli also reduces the accuracy of the threshold estimation and can therefore only be applied to detect rather large differences (Levitt, 1971; Linschoten et al., 2001). In contrast, the continual estimation of the reflex threshold, as performed by our algorithm, leads to rapidly decreasing variability and thus increasing precision of the reflex threshold estimates already after 15–25 stimuli (Fig. 2b and d), as are commonly applied for the standard algorithm. Therefore the 1:4 algorithm presented here is the first to allow an estimation of the threshold with similar accuracy as the standard algorithm.
3.2. Reflex threshold estimates of the 1:4 and the 1:1 algorithm can be used interchangeably The Bland Altman analyses performed here revealed variabilities between the new 1:4 algorithm and the standard 1:1 algorithm of 1.98–2.97 mA for the sensory deprived state and 2.42–3.27 mA for the distracted state, depending on how many individual reflex threshold estimates were averaged as the reflex threshold estimate. These variabilities between the algorithms are well below the variability between two measurement sessions performed with the same standard 1:1
algorithm, which we found in a previous study to amount to 4.44 mA (Dincklage et al., 2009). Other studies that investigated the test-retest variability between two measurements using the same method to estimate the NFR threshold showed results in a similar (Micalos et al., 2009; Biurrun Manresa et al., 2011, 2014) or even higher magnitude (Lewis et al., 2012). Our algorithm continually calculates reflex threshold estimates after each stimulation, in contrast to the commonly used algorithm, which provides only a single threshold estimate (Rhudy and France, 2011; Willer, 1977). However, the continual algorithm can be used equivalently to estimate a single threshold value by averaging the estimated thresholds, resulting in low variability after 15–25 stimuli (Fig. 2b and d), which is in the range of the number of applied stimuli of the commonly used single-threshold algorithm (Sandrini et al., 2005; Skljarevski and Ramadan, 2002). This commonly used stimulation paradigm applies one ascending series of stimuli with a rather large step size (e.g. 4 mA) until a reflex is detected, then stimulates in a descending series with a smaller step size (e.g. 2 mA) until no reflex is elicited, which is followed by two more ascending/ descending series of again smaller steps (e.g. 1 mA). The NFR threshold is then defined as the average stimulation intensity of the last two peaks and troughs, i.e. of the two ascending/ descending series with the smallest step size. As the step size of those is the same for the up and down steps, this algorithm estimates the reflex threshold in equal measure to the 1:1 algorithm used in this study, and likewise equal to all algorithms that estimate the reflex threshold as the stimulation intensity at which 50% of stimuli elicit reflexes. Frequently, in studies investigating the nociceptive flexion reflex, constant suprathreshold stimulations are applied to examine changes in reflex magnitude. In such studies the individual reflex threshold is estimated once and the subjects are then stimulated with intensities well above their individual reflex threshold. When the 1:4 algorithm is used to provide threshold values for suprathreshold stimulation it is necessary to keep in mind that even when the 1:4 algorithm and the 1:1 algorithm can be used interchangeably for assessing a NFR threshold, this does not mean that the threshold values are identical, but only that both algorithms can be used equivalently to investigate any research question: Since the average individual reflex threshold estimates are lower for the new 1:4 algorithm compared to the 1:1 algorithm it is necessary to correct the threshold values measured with the 1:4 algorithm by adding the mean difference of 1.16 mA (Bland and Altman, 1986) when directly comparing threshold values between the algorithms or using the threshold of the 1:4 algorithm as a basis to calculate stimulus intensities for suprathreshold stimuli. Accordingly, it has to be kept in mind that the threshold gained through the 1:4 algorithm represents the stimulus intensity of 50% probability of reflex occurrence only in the context of the 1:4 algorithm, which utilizes conditional probability to reduce the necessary stimulus intensity. Without the underlying conditional probability context of the 1:4 algorithm, stimulation at the threshold of the 1:4 algorithm would result in a lower probability of reflex occurrence, as is reflected in the lower threshold value for the 1:4 algorithm in comparison to the 1:1 algorithm. After adding the mean difference between
brain research 1608 (2015) 147–156
the 1:1 and the 1:4 algorithm of 1.16 mA to the threshold gained from the 1:4 algorithm, this transformed threshold value would produce a 50% reflex occurrence rate in a context without conditional probability. Because threshold estimates of both algorithms can be interconverted by simply adding or subtracting the mean difference between the methods, properties of the 1:1 algorithm, such as correlation with pain threshold and stability of the threshold estimates, are equally applicable to the 1:4 algorithm. In conclusion, the variability between the 1:1 and the 1:4 algorithm is lower than the variability between two measurement sessions with the same algorithm that were held under the most identical conditions and which can be regarded as the underlying intra-individual variability of the reflex itself. Therefore, the differences between the two algorithms are within the accuracy of the standard 1:1 algorithm and according to Bland and Altman we can assume that, after adjusting for the constant difference between the methods, the threshold estimates of the 1:4 algorithm can be used interchangeably with those of the 1:1 algorithm (Bland and Altman, 1986), but with the benefit of less induced pain in patients.
3.3. The 1:4 algorithm can be applied at interstimulus intervals down to 3 s The nociceptive flexion reflex threshold is routinely measured with interstimulus intervals of 5–20 s (Lim et al., 2011), as shorter interstimulus intervals might lead to unwanted habituation (Dimitrijević et al., 1972; Von Dincklage et al., 2013). However, shorter interstimulus intervals would be desirable, as the duration of the unpleasant measurement procedure is reduced if the aim is to assess a singular reflex threshold estimate and the time resolution of the measurement is increased if the aim is to continually assess the reflex threshold over a time interval. Our data suggests that the 1:4 algorithm can be applied at interstimulus intervals down to 3 s as the estimated reflex thresholds and subjective pain ratings do not significantly differ from those values recorded for longer interstimulus intervals. However, the maximal subjective pain ratings are significantly higher when using the 1:4 algorithm at an ISI of 3 s (Fig. 4b), which might decrease tolerability of the measurement. Further reduction of the interstimulus interval to 1 s seems not recommendable, as several subjects did not tolerate stimulation at that interstimulus interval. In summary, it might be possible to use the 1:4 algorithm at shorter ISIs than the 1:1 algorithm, resulting in a shorter measurement duration. If applied at the same ISI as the 1:1 algorithm, the number of stimuli, and therefore measurement time, of the 1:4 algorithm required to estimate the reflex threshold with similar accuracy is comparable to that of the 1:1 algorithm (Fig. 2b and d). The absence of discernable habituation effects even at short interstimulus intervals might be explained by the randomly varied interval time of 720% for each stimulus (Dimitrijević et al., 1972). Moreover, due to the nature of the 1:4 algorithm reflexes are less often provoked than with the 1:1 algorithm. This results in a greater average time interval between evoked reflexes, which might further explain the absence of habituation for the 1:4 algorithm.
153
3.4. The 1:4 algorithm can be used to investigate modulation of pain and nociception The nociceptive flexion reflex has been used to investigate spinal pain processing and analgesic pharmacological and psychological modulation of nociception. Pain modulation due to attention and distraction has raised increasing interest (Moont et al., 2010; Roy et al., 2011; Villemure and Bushnell, 2002) and the pain sensation as well as the NFR threshold were found to be influenced by the state of attention of the subjects (Ruscheweyh et al., 2011; Edwards et al., 2007; Willer et al., 1979). Indeed we found that reflex threshold estimates and subjective pain ratings of subjects who were distracted from the stimulation were significantly reduced compared to the estimates and ratings of the same subjects, when they were sensory deprived and thus focussed on the stimulations. The relative reductions of the reflex threshold estimates and the subjective pain ratings between the two states of attention when applying the 1:4 algorithms were of comparable size to those when using the standard 1:1 algorithm. The relative reduction of the pain ratings was in line with previous findings (Edwards et al., 2007), while the relative reduction of the reflex threshold was even more profound. The comparable reduction of reflex threshold estimates and pain ratings between the states for the standard 1:1 and the new 1:4 algorithm demonstrates that the 1:4 algorithm can be used to investigate the effects of modulation on nociception.
3.5.
Limitations
The subjects included in our study were mostly young, healthy students. However, even though the reflex threshold estimates for different age groups or morbid states might differ from the sample presented here, the applicability of this method to estimate the threshold remains valid, since its requirements are independent of differences in population parameters. The sole requirement of the method is an increasing probability of eliciting a reflex with increasing stimulation intensity, which is given independently of age or state of health. Therefore all reflex threshold measurement methods that are based on adaptive testing procedures have been successfully used with all age groups as well as chronic illness patients (Lim et al., 2011; Neziri et al., 2010; Rhudy and France, 2011; Von Dincklage et al., 2012) and it can be safely assumed that the 1:4 algorithm can be used for all population groups.
3.6.
Conclusion
We have demonstrated that measurement of the NFR threshold using the 1:4 algorithm presented here induces less pain than the standard 1:1 algorithm and that the threshold estimates of both algorithms can be used interchangeably after adjusting for the mean difference between both algorithms. Moreover, the threshold estimation using the 1:4 algorithm is sensitive enough to investigate modulation of pain processing and nociception. The new 1:4 algorithm as a method for assessing the NFR threshold in a less painful manner should increase the utility of the NFR as a tool to
154
brain research 1608 (2015) 147–156
investigate nociception in subjects who were previously not able to endure the painful standard measurement procedure, such as chronic pain patients.
4.
Experimental procedure
4.1.
Participants
The study was approved by the local ethics committee of the Charité - Universitätsmedizin Berlin. 60 healthy volunteers (30 male/30 female) were recruited and included in the study after written formal consent. Participants were required to abstain from caffeine within 6 h, from alcohol and strenuous exercise within 24 h and from analgesic drugs within 48 h before the measurements. Exclusion criteria were acute or chronic pain disorders or neuropathy. Participants received 40€ as compensation for their efforts.
4.2.
Experimental setup
The study was performed in a quiet air-conditioned room. Participants were seated in a comfortable chair with hip flexion angle of 1201, knee flexion of 1601 and ankle flexion of 1201 using a foot support. Electrocutaneous stimulation was performed via surface electrodes below and behind the left lateral malleolus to stimulate the sural nerve. Each stimulus consisted of five rectangular pulses of 1 ms each at 200 Hz (DS5; Digitimer Ltd, Welwyn Garden City, Hertfordshire, UK) (Tørring et al., 1981). The occurrence of the nociceptive flexion reflex was assessed by an electromyogram of the biceps femoris muscle, measured via surface electrodes placed over the muscle. The electromyographic signal was band-pass filtered between 2 Hz and 10 kHz, amplified by a factor of 10000 (Neuropac Four Mini; Nihon Kohden, Tokyo, Japan) and digitized at a sampling rate of 20 kHz (Power1401 mk II; CED Ltd, Cambridge, England). Signals were rectified and analyzed using customized scripts for Signal 3.10 (CED Ltd, Cambridge, England). Occurrence of a reflex was defined as an intervalpeak-z-score (baseline corrected peak amplitude in units of noise standard deviation) (Rhudy and France, 2007) above 6.085 between 90 and 180 ms after stimulation. Baseline and noise standard deviation were assessed from the interval of -90 to -10 ms before the stimulation. During the measurements participants were asked to verbally rate each stimulus on a numerical rating scale (NRS) from 0 to 100, in which 0 equals no sensation, 1 9 equal painless sensations and 10 100 equal painful sensations with 100 as the strongest pain imaginable. A rating of 10 was thus implicitly defined as the lowest painful rating. Prior to the measurement sequence participants were familiarized to the stimulation and to rating the pain using the NRS by a sequence of 20 stimuli in the NRS range of 0 60. To minimize predictability and reduce habituation of the reflex, the interstimulus times (1 s, 3 s and 6 s) were randomly varied by 720% for each stimulus (Dimitrijević et al., 1972).
4.3.
Experimental session
A measurement session consisted of twelve blocks, between which the stimulation algorithm (standard 1:1 algorithm vs. new 1:4 algorithm), the participants’ state of attention (sensory deprived vs. distracted) and the interstimulus interval (1 s vs. 3 s vs. 6 s) were varied. The order of the 12 measurement blocks was randomized for each participant. Blocks with interstimulus intervals of 1 s, 3 s and 6 s consisted of 600, 200 and 100 stimuli, respectively, equalling a duration of about 10 min per block. For measurements according to the standard 1:1 algorithm, the stimulation intensity was first increased in steps of 4 mA until the first reflex was detected and then decreased in steps of 1 mA until no reflex was detected. Continuing from this intensity, the intensity was increased by 1 mA every time no reflex was detected and decreased by 1 mA every time a reflex was detected (Willer, 1977; Dincklage et al., 2009). The new 1:4 algorithm was identical to the 1:1 algorithm except that the stimulation intensity was decreased by 4 mA every time a reflex was detected (Fig. 6). Reflex thresholds for both stimulation algorithms were continually estimated after each stimulus as the stimulus intensity associated with a 50%probability of reflex occurrence using a logistic regression of the last 7 stimulireflex pairs (Dincklage et al., 2009). The states of attention were either a state of sensory deprivation, in which the participants wore earmuffs and were instructed to keep their eyes closed and sit still, or a state of distraction from the stimulation by the task of throwing a ball back and forth between their hands. The interstimulus intervals at which consecutive stimuli were applied were 1 s, 3 s and 6 s.
4.4.
Data analysis and statistics
To investigate the differences between the subjective pain ratings for the two algorithms, the different states of attention and the different interstimulus intervals, the data from each of the 12 measurement blocks was averaged for each individual and analysed using a 2 2 3 (algorithms states of attention interstimulus intervals) repeated measurements analysis of variance (RM-ANOVA). Testing for statistical significance of specific differences between conditions that were used in the RM-ANOVA was performed by post hoc tests. Reported p values have been adjusted by Bonferroni correction to account for multiple comparisons. Differences between the stimulus intensities and the reflex threshold estimates for the two algorithms, the different states of attention and the different interstimulus intervals were investigated in the same way. To investigate the interchangeability of the two algorithms, a Bland Altman analysis was performed (Bland and Altman, 1986). For this analysis the variability between the two algorithms was calculated as the mean and the standard deviation of the differences between the individual reflex threshold estimates of the 1:1 algorithm compared to the reflex thresholds estimated in the same individual using the 1:4 algorithm, separated by the state of attention. For the primary analysis the individual differences between the algorithms were calculated as the differences between the
brain research 1608 (2015) 147–156
averages of all values of the respective measurement blocks. To also evaluate the variability between the two algorithms for lower numbers of reflex threshold estimates per subject than all values of the respective measurement blocks, we randomly picked fixed numbers of reflex threshold estimates from each subject, averaged them per subject and conducted a Bland Altman analysis with the resulting averaged reflex threshold estimates. This procedure was performed for every number of reflex threshold estimates per subject repeatedly through bootstrap resampling to provide accurate estimates of the variability between the algorithms for every number of reflex threshold estimates per subject. To analyse whether the reflex threshold estimated using the new 1:4 algorithm was stable over the time course of the measurements or whether the algorithm induced a systematic drift of the reflex threshold estimates over time, linear mixedeffects models were calculated to estimate mean population fits for the time courses of the reflex threshold estimates for each interstimulus interval and each attention state. All calculations and statistical analyses were performed using MATLAB R2013a (The MathWorks, Natick, USA) and SPSS Statistics 21 (IBM, Armonk, USA). Mixed effect models were computed using R 3.0.0 (R Core Team, R Foundation for Statistical Computing, Vienna, Austria).
Funding and declaration of interests The study was funded through the Internal Research Funding of the Department of Anesthesiology and Intensive Care Medicine, Charité - Universitätsmedizin Berlin. Gregor Lichtner is in sideline student employment by Dolosys GmbH, a spin-off of Charité - Universitätsmedizin Berlin developing reflex measurement devices. No conflict of interests or financial interests regarding the study exist.
r e f e r e n c e s
Biurrun Manresa, J.A., Fritsche, R., Vuilleumier, P.H., Oehler, C., Mørch, C.D., Arendt-Nielsen, L., Andersen, O.K., Curatolo, M., 2014. Is the conditioned pain modulation paradigm reliable? A test-retest assessment using the nociceptive withdrawal reflex. PLoS ONE 9, e100241, http://dx.doi.org/10.1371/journal. pone.0100241. Biurrun Manresa, J.A., Neziri, A.Y., Curatolo, M., Arendt-Nielsen, L., Andersen, O.K., 2011. Test–retest reliability of the nociceptive withdrawal reflex and electrical pain thresholds after single and repeated stimulation in patients with chronic low back pain. Eur. J. Appl. Physiol. 111, 83–92, http://dx.doi. org/10.1007/s00421-010-1634-0. Bland, J.M., Altman, D.G., 1986. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1, 307–310. Dimitrijevic´, M.R., Faganel, J., Gregoric´, M., Nathan, P.W., Trontelj, J.K., 1972. Habituation: effects of regular and stochastic stimulation. J. Neurol. Neurosurg. Psychiatry 35, 234–242, http: //dx.doi.org/10.1136/jnnp.35.2.234.
155
Dincklage, F. von, Hackbarth, M., Schneider, M., Baars, J.H., Rehberg, B., 2009. Introduction of a continual RIII reflex threshold tracking algorithm. Brain Res. 1260, 24–29, http://dx. doi.org/10.1016/j.brainres.2009.01.001. Edwards, L., Ring, C., France, C.R., al’ Absi, M., McIntyre, D., Carroll, D., Martin, U., 2007. Nociceptive flexion reflex thresholds and pain during rest and computer game play in patients with hypertension and individuals at risk for hypertension. Biol. Psychol. 76, 72–82, http://dx.doi.org/ 10.1016/j.biopsycho.2007.06.005. Levitt, H., 1971. Transformed up down methods in psychoacoustics. J. Acoust. Soc. Am. 49, 467–477. Lewis, G.N., Rice, D.A., Jourdain, K., McNair, P.J., 2012. Influence of stimulation location and posture on the reliability and comfort of the nociceptive flexion reflex. Pain Res. Manag. J. Can. Pain Soc. 17, 110–114. Lim, E.C.W., Sterling, M., Stone, A., Vicenzino, B., 2011. Central hyperexcitability as measured with nociceptive flexor reflex threshold in chronic musculoskeletal pain: a systematic review. Pain 152, 1811–1820, http://dx.doi.org/10.1016/j. pain.2011.03.033. Linschoten, M.R., Harvey, L.O., Eller, P.M., Jafek, B.W., 2001. Fast and accurate measurement of taste and smell thresholds using a maximum-likelihood adaptive staircase procedure. Percept. Psychophys. 63, 1330–1347, http://dx.doi.org/10.3758/ BF03194546. Micalos, P.S., Drinkwater, E.J., Cannon, J., Arendt-Nielsen, L., Marino, F.E., 2009. Reliability of the nociceptive flexor reflex (RIII) threshold and association with pain threshold. Eur. J. Appl. Physiol. 105, 55–62, http://dx.doi.org/10.1007/s00421-0080872-x. Moont, R., Pud, D., Sprecher, E., Sharvit, G., Yarnitsky, D., 2010. “Pain inhibits pain” mechanisms: is pain modulation simply due to distraction?. Pain 150, 113–120, http://dx.doi.org/ 10.1016/j.pain.2010.04.009. Neziri, A.Y., Andersen, O.K., Petersen-Felix, S., Radanov, B., Dickenson, A.H., Scaramozzino, P., Arendt-Nielsen, L., Curatolo, M., 2010. The nociceptive withdrawal reflex: normative values of thresholds and reflex receptive fields. Eur. J. Pain 14, 134–141, http://dx.doi.org/10.1016/j. ejpain.2009.04.010. Osman, A., Barrios, F.X., Gutierrez, P.M., Kopper, B.A., Merrifield, T., Grittmann, L., 2000. The pain catastrophizing scale: further psychometric evaluation with adult samples. J. Behav. Med. 23, 351–365, http://dx.doi.org/10.1023/A:1005548801037. Rhudy, J.L., France, C.R., 2007. Defining the nociceptive flexion reflex (NFR) threshold in human participants: a comparison of different scoring criteria. Pain 128, 244–253, http://dx.doi.org/ 10.1016/j.pain.2006.09.024. Rhudy, J.L., France, C.R., 2011. Reliability and validity of a brief method to assess nociceptive flexion reflex (NFR) threshold. J. Pain 12, 782–791, http://dx.doi.org/10.1016/j. jpain.2011.01.002. Roy, M., Lebuis, A., Peretz, I., Rainville, P., 2011. The modulation of pain by attention and emotion: a dissociation of perceptual and spinal nociceptive processes. Eur. J. Pain 15 (e1–641), e10, http://dx.doi.org/10.1016/j.ejpain.2010.11.013 641. Ruscheweyh, R., Albers, C., Kreusch, A., Sommer, J., Marziniak, M., 2013. The effect of catastrophizing self-statements on pain perception and the nociceptive flexor reflex (RIII reflex). J. Pain 29, 725–732, http://dx.doi.org/10.1097/AJP.0b013e318272ec0c. Ruscheweyh, R., Kreusch, A., Albers, C., Sommer, J., Marziniak, M., 2011. The effect of distraction strategies on pain perception and the nociceptive flexor reflex (RIII reflex). Pain 152, 2662–2671, http://dx.doi.org/10.1016/j.pain.2011.08.016. Ruscheweyh, R., Verneuer, B., Dany, K., Marziniak, M., Wolowski, A., C¸olak-Ekici, R., Schulte, T.L., Bullmann, V., Grewe, S., Gralow, I., Evers, S., Knecht, S., 2012. Validation of the pain
156
brain research 1608 (2015) 147–156
sensitivity questionnaire in chronic pain patients. Pain 153, 1210–1218, http://dx.doi.org/10.1016/j.pain.2012.02.025. Sandrini, G., Serrao, M., Rossi, P., Romaniello, A., Cruccu, G., Willer, J.C., 2005. The lower limb flexion reflex in humans. Prog. Neurobiol. 77, 353–395, http://dx.doi.org/10.1016/j. pneurobio.2005.11.003. Skljarevski, V., Ramadan, N.M., 2002. The nociceptive flexion reflex in humans – review article. Pain 96, 3–8, http://dx.doi. org/10.1016/S0304-3959(02)00018-0. Tørring, J., Pedersen, E., Klemar, B., 1981. Standardisation of the electrical elicitation of the human flexor reflex. J. Neurol. Neurosurg. Psychiatry 44, 129–132. Villemure, C., Bushnell, M.C., 2002. Cognitive modulation of pain: how do attention and emotion influence pain processing?. Pain 95, 195–199, http://dx.doi.org/10.1016/S0304-3959(02) 00007-6. Von Dincklage, F., Correll, C., Schneider, M.H.N., Rehberg, B., 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, http: //dx.doi.org/10.1111/j.1365-2044.2012.07187.x. Von Dincklage, F., Olbrich, H., Baars, J.H., Rehberg, B., 2013. Habituation of the nociceptive flexion reflex is dependent on inter-stimulus interval and stimulus intensity. J. Clin. Neurosci. 20, 848–850, http://dx.doi.org/10.1016/j. jocn.2012.07.013. Willer, J.C., 1977. Comparative study of perceived pain and nociceptive flexion reflex in man. Pain 3, 69–80, http://dx.doi. org/10.1016/0304-3959(77)90036-7. Willer, J.C., Boureau, F., Albe-Fessard, D., 1979. Supraspinal influences on nociceptive flexion reflex and pain sensation in man. Brain Res. 179, 61–68, http://dx.doi.org/10.1016/00068993(79)90489-X.
Available online at www.sciencedirect.com
www.elsevier.com/locate/brainres
Research Report
Introduction and validation of a less painful algorithm to estimate the nociceptive flexion reflex threshold Gregor Lichtnern, Anna Golebiewski, Martin H Schneider, Falk von Dincklage Charité - Universitätsmedizin Berlin, Klinik für Anästhesiologie mit Schwerpunkt operative Intensivmedizin, Campus Charité Mitte und Campus Virchow-Klinikum, Berlin, Germany
art i cle i nfo
ab st rac t
Article history:
The nociceptive flexion reflex (NFR) is a widely used tool to investigate spinal nociception for
Accepted 25 February 2015
scientific and diagnostic purposes, but its clinical use is currently limited due to the painful
Available online 5 March 2015
measurement procedure, especially restricting its applicability for patients suffering from
Keywords:
chronic pain disorders. Here we introduce a less painful algorithm to assess the NFR
Nociceptive flexion reflex
threshold. Application of this new algorithm leads to a reduction of subjective pain ratings
NFR
by over 30% compared to the standard algorithm. We show that the reflex threshold estimates
Reflex threshold
resulting from application of the new algorithm can be used interchangeably with those of the
Pain measurement
standard algorithm after adjusting for the constant difference between the algorithms.
RIII reflex
Furthermore, we show that the new algorithm can be applied at shorter interstimulus intervals than are commonly used with the standard algorithm, since reflex threshold values remain unchanged and no habituation effects occur when reducing the interstimulus interval for the new algorithm down to 3 s. Finally we demonstrate the utility of the new algorithm to investigate the modulation of nociception through different states of attention. Taken together, the here presented new algorithm could increase the utility of the NFR for investigation of nociception in subjects who were previously not able to endure the measurement procedure, such as chronic pain patients. & 2015 Elsevier B.V. All rights reserved.
1.
Introduction
The nociceptive flexion reflex is a spinal withdrawal reflex that can be assessed by electromyographic recording of the biceps femoris muscle during electrocutaneous stimulation of the ipsilateral sural nerve (Sandrini et al., 2005; Skljarevski and Ramadan, 2002). The strong correlation of the reflex
threshold with the subjective pain threshold has made the NFR a widely used tool to investigate pain processing, pharmacological and psychological modulation of nociception as well as chronic pain conditions (Lim et al., 2011). However, the applicability of the method is currently limited to subjects who are willing to endure the repeated painful stimulations required for assessing the reflex threshold.
n Correspondence to: Charité - Universitätsmedizin Berlin, Klinik für Anästhesiologie mit Schwerpunkt operative Intensivmedizin, Campus Charité Mitte, Charitéplatz 1, 10117 Berlin, Germany. Fax: þ49 30 450 531927. E-mail address: [email protected] (G. Lichtner).
http://dx.doi.org/10.1016/j.brainres.2015.02.049 0006-8993/& 2015 Elsevier B.V. All rights reserved.
148
brain research 1608 (2015) 147–156
This poses a significant restriction for the method, especially
painful suprathreshold stimulation intensities. Therefore application of the 1:4 algorithm can be expected to induce less subjective pain than the standard 1:1 algorithm. Here we demonstrate that (i) the 1:4 algorithm leads to less painful stimuli than the standard 1:1 algorithm, while (ii) agreement of the reflex thresholds estimated by the two algorithms is sufficient to allow them to be used interchangeably. We show that (iii) the new 1:4 algorithm allows for stimulating at shorter interstimulus intervals than at which the standard algorithm is commonly applied. Finally we show that (iv) the 1:4 algorithm can be used to investigate modulation of nociception.
when applied on patients suffering from chronic pain disorders, who exhibit a higher degree of pain sensitivity and pain catastrophizing (Ruscheweyh et al., 2013, 2012; Osman et al., 2000). To increase the utility of the method, an optimisation of the measurement procedure towards a lower pain induction seems necessary. The standard procedure to assess the NFR threshold is an adaptive testing algorithm that increases the stimulation intensity when no reflex is detected and decreases it when a reflex occurs (Sandrini et al., 2005). While this standard algorithm is usually applied with variable stimulation intensity step sizes, the up and down steps are equal in size. Here we propose a less painful algorithm, which in contrast uses an up/down ratio of 1:4, meaning that upon detection of a reflex the stimulation intensity is reduced four times as much as it is increased when no reflex occurs. This stimulation paradigm inherently leads to lower average stimulation intensities compared to the standard 1:1 algorithm, as the reduction of the stimulation intensity after every reflex occurrence is four times larger. Besides this direct effect, the 1:4 algorithm utilizes the probabilistic nature of the reflex occurrence (Sandrini et al., 2005) to provoke reflex responses below the reflex threshold. Since the 1:4 algorithm reduces the stimulation intensity in larger steps than it increases it back again, its application results in repeated stimulations below the reflex threshold. Single stimuli at such subthreshold intensities are unlikely to elicit reflex responses. However, because the overall probability to elicit a reflex cumulates with repeated stimulation, the application of multiple subthreshold stimuli increases the total probability of reflex responses already below the reflex threshold. Since the stimulation intensity is reduced after every reflex occurrence, such reflex responses at stimulation intensities below the reflex threshold prevent the algorithm to reach
ns
** ***
General results
The new 1:4 algorithm led to lower average pain ratings, lower average stimulation intensities and yielded lower reflex threshold estimates compared to the standard 1:1 algorithm (Fig. 1). Population average reduction of the subjective pain ratings between the two algorithms amounted to 34% and
* ***
*** 30
***
*** ***
30
***
***
***
70 60 50 40 30 20 10
25
80 70 60 50 40 30 20
20
15
10
5
*** 1:1 algorithm 1:4 algorithm
25
Reflex threshold [mA]
Maximal subjective pain rating
90
80
Subjective pain rating
2.1.
2.2. Differences in subjective pain ratings, stimulation intensities and reflex threshold estimates between the standard 1:1 and the new 1:4 algorithm
100
***
90
Results
56 (28 male/28 female) of 60 participants completed all measurement blocks and were included in the analysis. All 4 excluded participants did not tolerate the stimulation at an interstimulus interval of 1 s and aborted the respective blocks. The remaining 56 participants had a median (IQR) age of 23 (22 25), a median height of 177 (170–184) cm and a median weight of 69 (60–80) kg.
Stimulus intensity [mA]
100
***
**
2.
20
15
10
5
10
0
0
0
Sensory deprived
Distracted
Sensory deprived
Distracted
0
Sensory deprived
Distracted
Sensory deprived
Distracted
Fig. 1 – Comparison of subjective pain ratings, stimulation intensities and reflex threshold estimates between the new 1:4 and the standard 1:1 algorithm. Shown are (a) the individual average subjective pain ratings, (b) the individual maximal subjective pain ratings, (c) the individual average applied stimulation intensities and (d) the individual average reflex threshold estimates during application of the standard 1:1 algorithm (crosses) and the new 1:4 algorithm (circles), each at interstimulus intervals of 6 s, separated between the attention state of sensory deprivation and the state of distraction. Lines represent the population means and the respective standard errors. Stars mark significant differences between the states and the algorithms (ns: not significant; *: po0.05; **: po0.01; ***: po0.001; post hoc test with Bonferroni correction).
149
Sensory deprived
Sensory deprived 4.0
10 8 6 4 2 0 -2 -4 -6 -8 -10
Between algorithm variability [mA]
mean standard deviation 95 % limit of agreement
3.5 Standard deviation 95 % Confidence band
3.0 2.5 2.0 1.5
0
5
10
15
20
0
25
Average of reflex threshold estimates (1:4 algorithm & 1:1 algorithm) [mA]
Distracted
Distracted
4.0
10 8 6 4 2 0 -2 -4 -6 -8 -10
10 20 30 40 50 60 70 80 90 100 Number of included reflex threshold estimates per subject
Between algorithm variability [mA]
Difference between reflex threshold estimates (1:4 algorithm - 1:1 algorithm) [mA]
Difference between reflex threshold estimates (1:4 algorithm - 1:1 algorithm) [mA]
brain research 1608 (2015) 147–156
3.5 3.0 2.5 2.0 1.5
0
5
10
15
20
25
Average of reflex threshold estimates (1:4 algorithm & 1:1 algorithm) [mA]
0
10 20 30 40 50 60 70 80 90 100 Number of included reflex threshold estimates per subject
Fig. 2 – Bland Altman analyses of the variability between the reflex thresholds estimated by the two algorithms. The BlandAltman plots picturing the variability between the two algorithms at interstimulus intervals of 6 s for the states of (a) sensory deprivation and (c) distraction. Each data point represents one individual, showing the individual differences between the reflex threshold estimates obtained using the two different algorithms (ordinate) over the corresponding individual average of the reflex threshold estimates obtained using the two different algorithms (abscissa). Individual reflex threshold estimates were calculated for each subject and each algorithm as the average of all values obtained during the respective measurement block. Solid lines indicate the mean differences between the algorithms and the 95% limits of agreement. Dashed lines indicate standard deviations of the differences. Figures b and d show the standard deviations of the differences between the algorithms (ordinate) obtained from multiple Bland Altman analyses performed as in a and c, but differing in the number of reflex threshold estimates per individual (abscissa) that were included into the analyses. Dashed lines indicate the standard deviations obtained from the analyses shown in (a) and (c), when all reflex threshold estimates per individual of the respective measurement block were included into the analysis. Shaded areas indicate 95% confidence bands.
37% for the state of sensory deprivation and the state of distraction, respectively. Population average reduction of the maximal subjective pain ratings between the two algorithms amounted to 14% and 17% for the state of sensory deprivation and the state of distraction, respectively. Population average reduction of stimulation intensities amounted to 28% (sensory deprivation) and 35% (distraction) and the population average reduction of the reflex threshold estimates amounted to 11% (sensory deprivation) and 24% (distraction).
2.3. Agreement between reflex threshold estimates of the standard 1:1 and the new 1:4 algorithm The individual mean reflex threshold estimates were significantly correlated between the two algorithms in both
attention states at an interstimulus interval of 6 s (Pearson’s correlation 0.91 for state of sensory deprivation, 0.86 for state of distraction, po0.001 for both). BlandAltman analyses revealed a mean difference between the individual average reflex threshold estimates of the 1:1 algorithm and the 1:4 algorithm of 1.0771.97 mA (mean7SD) for the state of sensory deprivation and 1.2472.42 mA (mean7SD) for the state of distraction (Fig. 2a and c). BlandAltman analyses of differences between the reflex threshold estimates of the 1:1 and the 1:4 algorithm when including only a specific number of reflex threshold estimates per subject into the analyses showed a maximum variability of 2.97 mA (SD) for the sensory deprived state and 3.27 mA (SD) for the distracted state, both in the case when only one
150
brain research 1608 (2015) 147–156
Sensory deprivation 15 12
1s
Distracted
median interquartile range
15 12
9
9
6
6
3
3
p > 0.13
Reflex threshold [mA]
Reflex threshold [mA]
Interstimulus interval
0 1 2 3 4 5 6 7 8 9 10
0 1 2 3 4 5 6 7 8 9 10
3s
15 12 9 6 3
p > 0.66
0
15 12 9 6 3
0 1 2 3 4 5 6 7 8 9 10
15
15
12
12
9
9
6
6
3
p > 0.12
0
0 1 2 3 4 5 6 7 8 9 10
6s
p > 0.45
0
0
3
p > 0.40
p > 0.20
0
0 0 1 2 3 4 5 6 7 8 9 10 time [min]
0 1 2 3 4 5 6 7 8 9 10 time [min]
Fig. 3 – Stability of the reflex threshold estimates over time. Shown are the time courses of the population medians and the interquartile ranges of the reflex thresholds estimated using the new 1:4 algorithm at interstimulus intervals of 1 s, 3 s and 6 s, each at the attention state of sensory deprivation and distraction. Reported p-values correspond to estimated population slopes of the linear mixed effect models.
reflex threshold estimate per subject was included into the analyses (Fig. 2b and d).
2.4. Applicability of the 1:4 algorithm at different interstimulus intervals The reflex threshold estimates resulting from application of the 1:4 algorithm were stable with no systematic drift over the measurement blocks of 10 min length for all interstimulus intervals and attention states (Fig. 3). Application of the new 1:4 algorithm also did not result in significantly different pain ratings or reflex threshold estimates when comparing measurements at interstimulus intervals of 3 s and 6 s. Using the 1:4 algorithm at an interstimulus interval of 1 s resulted in partly different subjective pain ratings and reflex threshold estimates than the interstimulus intervals of 3 and 6 s (Fig. 4).
intervals of 6 s and 3 s, but not at the 1 s interval (Fig. 5a). Reflex threshold estimates were significantly reduced in the state of distraction compared to the state of sensory deprivation for all interstimulus intervals used (Fig. 5b). Population average reduction of the subjective pain ratings between the two states of attention when using the 1:4 algorithm amounted to 2%, 22% and 20% for interstimulus intervals of 1 s, 3 s and 6 s, respectively, and to 16% when using the standard 1:1 algorithm at an interstimulus interval of 6 s. Population average reduction of mean estimated reflex thresholds using the new 1:4 algorithm amounted to 10%, 20% and 21% for interstimulus intervals of 1 s, 3 s and 6 s, respectively, and to 17% when using the standard 1:1 algorithm at an interstimulus interval of 6 s.
3. 2.5. Utility of the 1:4 algorithm to investigate modulation of nociception Application of the new 1:4 algorithm resulted in significantly lower subjective pain ratings in the state of distraction compared to the state of sensory deprivation when stimulating at
Discussion
3.1. Application of the new 1:4 algorithm induces less pain compared to the standard algorithm Our results show that the 1:4 algorithm leads to lower subjective pain ratings than the standard 1:1 algorithm,
151
ns
50
** ns
ns
ns
40 30 20 10 0
1s 3s 6s
1s 3s 6s
Sensory deprived
Distracted
50
ns
*** ***
ns
*** ** **
ns
40
Reflex threshold [mA]
Subjective pain rating
ns
Maximal subjective pain rating
brain research 1608 (2015) 147–156
30 20 10 0
1s 3s 6s
1s 3s 6s
Sensory deprived
Distracted
10
ns
***
ns ns
8 6 4 2 0
1s 3s 6s
1s 3s 6s
Sensory deprived
Distracted
Fig. 4 – Effect of different interstimulus intervals on subjective pain ratings and reflex threshold estimates. Shown are (a) average subjective pain ratings, (b) maximal subjective pain ratings and (c) reflex threshold estimates at interstimulus intervals of 1 s (light grey), 3 s (dark grey) and 6 s (black) for the new 1:4 algorithm in both states of attention. Stars mark significant differences between the interstimulus intervals (ns: not significant; **: po0.01; ***: po0.001, post hoc test with Bonferroni correction). Error bars depict the standard error of the mean.
1:4 algorithm 40 35 30 25 20 15 10 5 0
***
**
1:1 algorithm ns
ns
**
12 Reflex threshold [mA]
Subjective pain rating
ns
10
1:4 algorithm
1:1 algorithm
*
***
***
*
***
***
1s
3s
6s
1s
3s
6s
8 6 4 2 0
1s
3s
6s
1s
3s
6s
Interstimulus interval
Interstimulus interval
Sensory deprived Distracted Fig. 5 – Effect of different states of attention on subjective pain ratings and reflex threshold estimates at different interstimulus intervals. Shown are (a) subjective pain ratings and (b) reflex threshold estimates in the state of sensory deprivation (black) and the state of distraction (grey) for the new 1:4 algorithm (filled bars) and the standard 1:1 algorithm (unfilled bars). Stars mark significant differences between the attention states (ns: not significant; *: po0.05; **: po0.01; ***: po0.001, post hoc test with Bonferroni correction). Error bars depict the standard error of the mean.
which can be explained by the lower stimulation intensities that are applied when using the 1:4 algorithm (Fig. 1c). One obvious reason for the lower stimulation intensities is the direct consequence of the stimulation paradigm of the new 1:4 algorithm, as decreasing the stimulation intensity by 4 mA after a reflex occurred inherently leads to lower stimulation intensities than decreasing by 1 mA. This direct effect can be expected to decrease average stimulation intensities of the new 1:4 algorithm by 1.5 mA compared to the standard 1:1 algorithm (Fig. 6). However, this direct effect explains only about half of the overall reduction of the stimulation intensities, which in total amounted to 2.7–2.8 mA when comparing the 1:4 algorithm to the 1:1 algorithm. The additional reduction when using the 1:4 algorithm can be assumed to be caused by the interaction between the 1:4 algorithm and the probabilistic nature of the reflex occurrence. The probability to elicit a reflex is a function of the applied stimulation intensity and increases gradually with increasing
stimulation intensity (Levitt, 1971; Sandrini et al., 2005). Hence there is no hard threshold above which a stimulus always leads to a reflex response and below which reflexes never occur. Instead the reflex threshold is merely the stimulation intensity at which a reflex response is elicited with a certain probability, usually 50% (Levitt, 1971). Therefore stimulations below that reflex threshold are still capable, although less probable, to elicit a reflex. The 1:4 algorithm utilizes this characteristic to provoke reflex responses already at stimulation intensities below the threshold. Because of the larger decrease of the stimulation intensity after a reflex occurs in comparison to the smaller increasing steps when no reflex is detected, every reflex occurrence is followed by multiple subthreshold stimuli. Each of these subthreshold stimulations alone has a low probability of eliciting a reflex, but the cumulated probabilities of all subthreshold stimuli together result in a higher total probability of eliciting reflex responses at stimulation intensities below the reflex threshold. Since after occurrence of such reflex responses
152
brain research 1608 (2015) 147–156
Stimulus intensity
1:1 algorithm 1:4 algorithm
1 mA
Stimulus number Fig. 6 – Stimulation paradigms of the new 1:4 and the standard 1:1 algorithm. For the standard 1:1 algorithm (crosses) the stimulation intensity is increased by 1 mA every time no reflex is detected and decreased by 1 mA after every reflex detection (arrow heads). For the new 1:4 algorithm (circles) the stimulation intensity is also increased by 1 mA every time no reflex is detected, but in contrast to the standard algorithm the stimulation intensity is decreased by 4 mA every time a reflex is detected.
as after any reflex occurrence the stimulation intensity is decreased, a higher rate of reflexes occurring below the reflex threshold leads to lower average stimulation intensities. Most parsimoniously, this effect can explain the stronger reduction of the applied stimulation intensities and the subjective pain ratings for the 1:4 algorithm in comparison to the 1:1 algorithm than what can be explained by the direct effect of the different stimulus paradigms alone. It has been proposed to reduce the induced pain for the subjects simply by using a single ascending series of stimuli and thereby reducing the number of stimuli required for the measurement to a minimum (Rhudy and France, 2011). However, such reduction of the number of stimuli also reduces the accuracy of the threshold estimation and can therefore only be applied to detect rather large differences (Levitt, 1971; Linschoten et al., 2001). In contrast, the continual estimation of the reflex threshold, as performed by our algorithm, leads to rapidly decreasing variability and thus increasing precision of the reflex threshold estimates already after 15–25 stimuli (Fig. 2b and d), as are commonly applied for the standard algorithm. Therefore the 1:4 algorithm presented here is the first to allow an estimation of the threshold with similar accuracy as the standard algorithm.
3.2. Reflex threshold estimates of the 1:4 and the 1:1 algorithm can be used interchangeably The Bland Altman analyses performed here revealed variabilities between the new 1:4 algorithm and the standard 1:1 algorithm of 1.98–2.97 mA for the sensory deprived state and 2.42–3.27 mA for the distracted state, depending on how many individual reflex threshold estimates were averaged as the reflex threshold estimate. These variabilities between the algorithms are well below the variability between two measurement sessions performed with the same standard 1:1
algorithm, which we found in a previous study to amount to 4.44 mA (Dincklage et al., 2009). Other studies that investigated the test-retest variability between two measurements using the same method to estimate the NFR threshold showed results in a similar (Micalos et al., 2009; Biurrun Manresa et al., 2011, 2014) or even higher magnitude (Lewis et al., 2012). Our algorithm continually calculates reflex threshold estimates after each stimulation, in contrast to the commonly used algorithm, which provides only a single threshold estimate (Rhudy and France, 2011; Willer, 1977). However, the continual algorithm can be used equivalently to estimate a single threshold value by averaging the estimated thresholds, resulting in low variability after 15–25 stimuli (Fig. 2b and d), which is in the range of the number of applied stimuli of the commonly used single-threshold algorithm (Sandrini et al., 2005; Skljarevski and Ramadan, 2002). This commonly used stimulation paradigm applies one ascending series of stimuli with a rather large step size (e.g. 4 mA) until a reflex is detected, then stimulates in a descending series with a smaller step size (e.g. 2 mA) until no reflex is elicited, which is followed by two more ascending/ descending series of again smaller steps (e.g. 1 mA). The NFR threshold is then defined as the average stimulation intensity of the last two peaks and troughs, i.e. of the two ascending/ descending series with the smallest step size. As the step size of those is the same for the up and down steps, this algorithm estimates the reflex threshold in equal measure to the 1:1 algorithm used in this study, and likewise equal to all algorithms that estimate the reflex threshold as the stimulation intensity at which 50% of stimuli elicit reflexes. Frequently, in studies investigating the nociceptive flexion reflex, constant suprathreshold stimulations are applied to examine changes in reflex magnitude. In such studies the individual reflex threshold is estimated once and the subjects are then stimulated with intensities well above their individual reflex threshold. When the 1:4 algorithm is used to provide threshold values for suprathreshold stimulation it is necessary to keep in mind that even when the 1:4 algorithm and the 1:1 algorithm can be used interchangeably for assessing a NFR threshold, this does not mean that the threshold values are identical, but only that both algorithms can be used equivalently to investigate any research question: Since the average individual reflex threshold estimates are lower for the new 1:4 algorithm compared to the 1:1 algorithm it is necessary to correct the threshold values measured with the 1:4 algorithm by adding the mean difference of 1.16 mA (Bland and Altman, 1986) when directly comparing threshold values between the algorithms or using the threshold of the 1:4 algorithm as a basis to calculate stimulus intensities for suprathreshold stimuli. Accordingly, it has to be kept in mind that the threshold gained through the 1:4 algorithm represents the stimulus intensity of 50% probability of reflex occurrence only in the context of the 1:4 algorithm, which utilizes conditional probability to reduce the necessary stimulus intensity. Without the underlying conditional probability context of the 1:4 algorithm, stimulation at the threshold of the 1:4 algorithm would result in a lower probability of reflex occurrence, as is reflected in the lower threshold value for the 1:4 algorithm in comparison to the 1:1 algorithm. After adding the mean difference between
brain research 1608 (2015) 147–156
the 1:1 and the 1:4 algorithm of 1.16 mA to the threshold gained from the 1:4 algorithm, this transformed threshold value would produce a 50% reflex occurrence rate in a context without conditional probability. Because threshold estimates of both algorithms can be interconverted by simply adding or subtracting the mean difference between the methods, properties of the 1:1 algorithm, such as correlation with pain threshold and stability of the threshold estimates, are equally applicable to the 1:4 algorithm. In conclusion, the variability between the 1:1 and the 1:4 algorithm is lower than the variability between two measurement sessions with the same algorithm that were held under the most identical conditions and which can be regarded as the underlying intra-individual variability of the reflex itself. Therefore, the differences between the two algorithms are within the accuracy of the standard 1:1 algorithm and according to Bland and Altman we can assume that, after adjusting for the constant difference between the methods, the threshold estimates of the 1:4 algorithm can be used interchangeably with those of the 1:1 algorithm (Bland and Altman, 1986), but with the benefit of less induced pain in patients.
3.3. The 1:4 algorithm can be applied at interstimulus intervals down to 3 s The nociceptive flexion reflex threshold is routinely measured with interstimulus intervals of 5–20 s (Lim et al., 2011), as shorter interstimulus intervals might lead to unwanted habituation (Dimitrijević et al., 1972; Von Dincklage et al., 2013). However, shorter interstimulus intervals would be desirable, as the duration of the unpleasant measurement procedure is reduced if the aim is to assess a singular reflex threshold estimate and the time resolution of the measurement is increased if the aim is to continually assess the reflex threshold over a time interval. Our data suggests that the 1:4 algorithm can be applied at interstimulus intervals down to 3 s as the estimated reflex thresholds and subjective pain ratings do not significantly differ from those values recorded for longer interstimulus intervals. However, the maximal subjective pain ratings are significantly higher when using the 1:4 algorithm at an ISI of 3 s (Fig. 4b), which might decrease tolerability of the measurement. Further reduction of the interstimulus interval to 1 s seems not recommendable, as several subjects did not tolerate stimulation at that interstimulus interval. In summary, it might be possible to use the 1:4 algorithm at shorter ISIs than the 1:1 algorithm, resulting in a shorter measurement duration. If applied at the same ISI as the 1:1 algorithm, the number of stimuli, and therefore measurement time, of the 1:4 algorithm required to estimate the reflex threshold with similar accuracy is comparable to that of the 1:1 algorithm (Fig. 2b and d). The absence of discernable habituation effects even at short interstimulus intervals might be explained by the randomly varied interval time of 720% for each stimulus (Dimitrijević et al., 1972). Moreover, due to the nature of the 1:4 algorithm reflexes are less often provoked than with the 1:1 algorithm. This results in a greater average time interval between evoked reflexes, which might further explain the absence of habituation for the 1:4 algorithm.
153
3.4. The 1:4 algorithm can be used to investigate modulation of pain and nociception The nociceptive flexion reflex has been used to investigate spinal pain processing and analgesic pharmacological and psychological modulation of nociception. Pain modulation due to attention and distraction has raised increasing interest (Moont et al., 2010; Roy et al., 2011; Villemure and Bushnell, 2002) and the pain sensation as well as the NFR threshold were found to be influenced by the state of attention of the subjects (Ruscheweyh et al., 2011; Edwards et al., 2007; Willer et al., 1979). Indeed we found that reflex threshold estimates and subjective pain ratings of subjects who were distracted from the stimulation were significantly reduced compared to the estimates and ratings of the same subjects, when they were sensory deprived and thus focussed on the stimulations. The relative reductions of the reflex threshold estimates and the subjective pain ratings between the two states of attention when applying the 1:4 algorithms were of comparable size to those when using the standard 1:1 algorithm. The relative reduction of the pain ratings was in line with previous findings (Edwards et al., 2007), while the relative reduction of the reflex threshold was even more profound. The comparable reduction of reflex threshold estimates and pain ratings between the states for the standard 1:1 and the new 1:4 algorithm demonstrates that the 1:4 algorithm can be used to investigate the effects of modulation on nociception.
3.5.
Limitations
The subjects included in our study were mostly young, healthy students. However, even though the reflex threshold estimates for different age groups or morbid states might differ from the sample presented here, the applicability of this method to estimate the threshold remains valid, since its requirements are independent of differences in population parameters. The sole requirement of the method is an increasing probability of eliciting a reflex with increasing stimulation intensity, which is given independently of age or state of health. Therefore all reflex threshold measurement methods that are based on adaptive testing procedures have been successfully used with all age groups as well as chronic illness patients (Lim et al., 2011; Neziri et al., 2010; Rhudy and France, 2011; Von Dincklage et al., 2012) and it can be safely assumed that the 1:4 algorithm can be used for all population groups.
3.6.
Conclusion
We have demonstrated that measurement of the NFR threshold using the 1:4 algorithm presented here induces less pain than the standard 1:1 algorithm and that the threshold estimates of both algorithms can be used interchangeably after adjusting for the mean difference between both algorithms. Moreover, the threshold estimation using the 1:4 algorithm is sensitive enough to investigate modulation of pain processing and nociception. The new 1:4 algorithm as a method for assessing the NFR threshold in a less painful manner should increase the utility of the NFR as a tool to
154
brain research 1608 (2015) 147–156
investigate nociception in subjects who were previously not able to endure the painful standard measurement procedure, such as chronic pain patients.
4.
Experimental procedure
4.1.
Participants
The study was approved by the local ethics committee of the Charité - Universitätsmedizin Berlin. 60 healthy volunteers (30 male/30 female) were recruited and included in the study after written formal consent. Participants were required to abstain from caffeine within 6 h, from alcohol and strenuous exercise within 24 h and from analgesic drugs within 48 h before the measurements. Exclusion criteria were acute or chronic pain disorders or neuropathy. Participants received 40€ as compensation for their efforts.
4.2.
Experimental setup
The study was performed in a quiet air-conditioned room. Participants were seated in a comfortable chair with hip flexion angle of 1201, knee flexion of 1601 and ankle flexion of 1201 using a foot support. Electrocutaneous stimulation was performed via surface electrodes below and behind the left lateral malleolus to stimulate the sural nerve. Each stimulus consisted of five rectangular pulses of 1 ms each at 200 Hz (DS5; Digitimer Ltd, Welwyn Garden City, Hertfordshire, UK) (Tørring et al., 1981). The occurrence of the nociceptive flexion reflex was assessed by an electromyogram of the biceps femoris muscle, measured via surface electrodes placed over the muscle. The electromyographic signal was band-pass filtered between 2 Hz and 10 kHz, amplified by a factor of 10000 (Neuropac Four Mini; Nihon Kohden, Tokyo, Japan) and digitized at a sampling rate of 20 kHz (Power1401 mk II; CED Ltd, Cambridge, England). Signals were rectified and analyzed using customized scripts for Signal 3.10 (CED Ltd, Cambridge, England). Occurrence of a reflex was defined as an intervalpeak-z-score (baseline corrected peak amplitude in units of noise standard deviation) (Rhudy and France, 2007) above 6.085 between 90 and 180 ms after stimulation. Baseline and noise standard deviation were assessed from the interval of -90 to -10 ms before the stimulation. During the measurements participants were asked to verbally rate each stimulus on a numerical rating scale (NRS) from 0 to 100, in which 0 equals no sensation, 1 9 equal painless sensations and 10 100 equal painful sensations with 100 as the strongest pain imaginable. A rating of 10 was thus implicitly defined as the lowest painful rating. Prior to the measurement sequence participants were familiarized to the stimulation and to rating the pain using the NRS by a sequence of 20 stimuli in the NRS range of 0 60. To minimize predictability and reduce habituation of the reflex, the interstimulus times (1 s, 3 s and 6 s) were randomly varied by 720% for each stimulus (Dimitrijević et al., 1972).
4.3.
Experimental session
A measurement session consisted of twelve blocks, between which the stimulation algorithm (standard 1:1 algorithm vs. new 1:4 algorithm), the participants’ state of attention (sensory deprived vs. distracted) and the interstimulus interval (1 s vs. 3 s vs. 6 s) were varied. The order of the 12 measurement blocks was randomized for each participant. Blocks with interstimulus intervals of 1 s, 3 s and 6 s consisted of 600, 200 and 100 stimuli, respectively, equalling a duration of about 10 min per block. For measurements according to the standard 1:1 algorithm, the stimulation intensity was first increased in steps of 4 mA until the first reflex was detected and then decreased in steps of 1 mA until no reflex was detected. Continuing from this intensity, the intensity was increased by 1 mA every time no reflex was detected and decreased by 1 mA every time a reflex was detected (Willer, 1977; Dincklage et al., 2009). The new 1:4 algorithm was identical to the 1:1 algorithm except that the stimulation intensity was decreased by 4 mA every time a reflex was detected (Fig. 6). Reflex thresholds for both stimulation algorithms were continually estimated after each stimulus as the stimulus intensity associated with a 50%probability of reflex occurrence using a logistic regression of the last 7 stimulireflex pairs (Dincklage et al., 2009). The states of attention were either a state of sensory deprivation, in which the participants wore earmuffs and were instructed to keep their eyes closed and sit still, or a state of distraction from the stimulation by the task of throwing a ball back and forth between their hands. The interstimulus intervals at which consecutive stimuli were applied were 1 s, 3 s and 6 s.
4.4.
Data analysis and statistics
To investigate the differences between the subjective pain ratings for the two algorithms, the different states of attention and the different interstimulus intervals, the data from each of the 12 measurement blocks was averaged for each individual and analysed using a 2 2 3 (algorithms states of attention interstimulus intervals) repeated measurements analysis of variance (RM-ANOVA). Testing for statistical significance of specific differences between conditions that were used in the RM-ANOVA was performed by post hoc tests. Reported p values have been adjusted by Bonferroni correction to account for multiple comparisons. Differences between the stimulus intensities and the reflex threshold estimates for the two algorithms, the different states of attention and the different interstimulus intervals were investigated in the same way. To investigate the interchangeability of the two algorithms, a Bland Altman analysis was performed (Bland and Altman, 1986). For this analysis the variability between the two algorithms was calculated as the mean and the standard deviation of the differences between the individual reflex threshold estimates of the 1:1 algorithm compared to the reflex thresholds estimated in the same individual using the 1:4 algorithm, separated by the state of attention. For the primary analysis the individual differences between the algorithms were calculated as the differences between the
brain research 1608 (2015) 147–156
averages of all values of the respective measurement blocks. To also evaluate the variability between the two algorithms for lower numbers of reflex threshold estimates per subject than all values of the respective measurement blocks, we randomly picked fixed numbers of reflex threshold estimates from each subject, averaged them per subject and conducted a Bland Altman analysis with the resulting averaged reflex threshold estimates. This procedure was performed for every number of reflex threshold estimates per subject repeatedly through bootstrap resampling to provide accurate estimates of the variability between the algorithms for every number of reflex threshold estimates per subject. To analyse whether the reflex threshold estimated using the new 1:4 algorithm was stable over the time course of the measurements or whether the algorithm induced a systematic drift of the reflex threshold estimates over time, linear mixedeffects models were calculated to estimate mean population fits for the time courses of the reflex threshold estimates for each interstimulus interval and each attention state. All calculations and statistical analyses were performed using MATLAB R2013a (The MathWorks, Natick, USA) and SPSS Statistics 21 (IBM, Armonk, USA). Mixed effect models were computed using R 3.0.0 (R Core Team, R Foundation for Statistical Computing, Vienna, Austria).
Funding and declaration of interests The study was funded through the Internal Research Funding of the Department of Anesthesiology and Intensive Care Medicine, Charité - Universitätsmedizin Berlin. Gregor Lichtner is in sideline student employment by Dolosys GmbH, a spin-off of Charité - Universitätsmedizin Berlin developing reflex measurement devices. No conflict of interests or financial interests regarding the study exist.
r e f e r e n c e s
Biurrun Manresa, J.A., Fritsche, R., Vuilleumier, P.H., Oehler, C., Mørch, C.D., Arendt-Nielsen, L., Andersen, O.K., Curatolo, M., 2014. Is the conditioned pain modulation paradigm reliable? A test-retest assessment using the nociceptive withdrawal reflex. PLoS ONE 9, e100241, http://dx.doi.org/10.1371/journal. pone.0100241. Biurrun Manresa, J.A., Neziri, A.Y., Curatolo, M., Arendt-Nielsen, L., Andersen, O.K., 2011. Test–retest reliability of the nociceptive withdrawal reflex and electrical pain thresholds after single and repeated stimulation in patients with chronic low back pain. Eur. J. Appl. Physiol. 111, 83–92, http://dx.doi. org/10.1007/s00421-010-1634-0. Bland, J.M., Altman, D.G., 1986. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1, 307–310. Dimitrijevic´, M.R., Faganel, J., Gregoric´, M., Nathan, P.W., Trontelj, J.K., 1972. Habituation: effects of regular and stochastic stimulation. J. Neurol. Neurosurg. Psychiatry 35, 234–242, http: //dx.doi.org/10.1136/jnnp.35.2.234.
155
Dincklage, F. von, Hackbarth, M., Schneider, M., Baars, J.H., Rehberg, B., 2009. Introduction of a continual RIII reflex threshold tracking algorithm. Brain Res. 1260, 24–29, http://dx. doi.org/10.1016/j.brainres.2009.01.001. Edwards, L., Ring, C., France, C.R., al’ Absi, M., McIntyre, D., Carroll, D., Martin, U., 2007. Nociceptive flexion reflex thresholds and pain during rest and computer game play in patients with hypertension and individuals at risk for hypertension. Biol. Psychol. 76, 72–82, http://dx.doi.org/ 10.1016/j.biopsycho.2007.06.005. Levitt, H., 1971. Transformed up down methods in psychoacoustics. J. Acoust. Soc. Am. 49, 467–477. Lewis, G.N., Rice, D.A., Jourdain, K., McNair, P.J., 2012. Influence of stimulation location and posture on the reliability and comfort of the nociceptive flexion reflex. Pain Res. Manag. J. Can. Pain Soc. 17, 110–114. Lim, E.C.W., Sterling, M., Stone, A., Vicenzino, B., 2011. Central hyperexcitability as measured with nociceptive flexor reflex threshold in chronic musculoskeletal pain: a systematic review. Pain 152, 1811–1820, http://dx.doi.org/10.1016/j. pain.2011.03.033. Linschoten, M.R., Harvey, L.O., Eller, P.M., Jafek, B.W., 2001. Fast and accurate measurement of taste and smell thresholds using a maximum-likelihood adaptive staircase procedure. Percept. Psychophys. 63, 1330–1347, http://dx.doi.org/10.3758/ BF03194546. Micalos, P.S., Drinkwater, E.J., Cannon, J., Arendt-Nielsen, L., Marino, F.E., 2009. Reliability of the nociceptive flexor reflex (RIII) threshold and association with pain threshold. Eur. J. Appl. Physiol. 105, 55–62, http://dx.doi.org/10.1007/s00421-0080872-x. Moont, R., Pud, D., Sprecher, E., Sharvit, G., Yarnitsky, D., 2010. “Pain inhibits pain” mechanisms: is pain modulation simply due to distraction?. Pain 150, 113–120, http://dx.doi.org/ 10.1016/j.pain.2010.04.009. Neziri, A.Y., Andersen, O.K., Petersen-Felix, S., Radanov, B., Dickenson, A.H., Scaramozzino, P., Arendt-Nielsen, L., Curatolo, M., 2010. The nociceptive withdrawal reflex: normative values of thresholds and reflex receptive fields. Eur. J. Pain 14, 134–141, http://dx.doi.org/10.1016/j. ejpain.2009.04.010. Osman, A., Barrios, F.X., Gutierrez, P.M., Kopper, B.A., Merrifield, T., Grittmann, L., 2000. The pain catastrophizing scale: further psychometric evaluation with adult samples. J. Behav. Med. 23, 351–365, http://dx.doi.org/10.1023/A:1005548801037. Rhudy, J.L., France, C.R., 2007. Defining the nociceptive flexion reflex (NFR) threshold in human participants: a comparison of different scoring criteria. Pain 128, 244–253, http://dx.doi.org/ 10.1016/j.pain.2006.09.024. Rhudy, J.L., France, C.R., 2011. Reliability and validity of a brief method to assess nociceptive flexion reflex (NFR) threshold. J. Pain 12, 782–791, http://dx.doi.org/10.1016/j. jpain.2011.01.002. Roy, M., Lebuis, A., Peretz, I., Rainville, P., 2011. The modulation of pain by attention and emotion: a dissociation of perceptual and spinal nociceptive processes. Eur. J. Pain 15 (e1–641), e10, http://dx.doi.org/10.1016/j.ejpain.2010.11.013 641. Ruscheweyh, R., Albers, C., Kreusch, A., Sommer, J., Marziniak, M., 2013. The effect of catastrophizing self-statements on pain perception and the nociceptive flexor reflex (RIII reflex). J. Pain 29, 725–732, http://dx.doi.org/10.1097/AJP.0b013e318272ec0c. Ruscheweyh, R., Kreusch, A., Albers, C., Sommer, J., Marziniak, M., 2011. The effect of distraction strategies on pain perception and the nociceptive flexor reflex (RIII reflex). Pain 152, 2662–2671, http://dx.doi.org/10.1016/j.pain.2011.08.016. Ruscheweyh, R., Verneuer, B., Dany, K., Marziniak, M., Wolowski, A., C¸olak-Ekici, R., Schulte, T.L., Bullmann, V., Grewe, S., Gralow, I., Evers, S., Knecht, S., 2012. Validation of the pain
156
brain research 1608 (2015) 147–156
sensitivity questionnaire in chronic pain patients. Pain 153, 1210–1218, http://dx.doi.org/10.1016/j.pain.2012.02.025. Sandrini, G., Serrao, M., Rossi, P., Romaniello, A., Cruccu, G., Willer, J.C., 2005. The lower limb flexion reflex in humans. Prog. Neurobiol. 77, 353–395, http://dx.doi.org/10.1016/j. pneurobio.2005.11.003. Skljarevski, V., Ramadan, N.M., 2002. The nociceptive flexion reflex in humans – review article. Pain 96, 3–8, http://dx.doi. org/10.1016/S0304-3959(02)00018-0. Tørring, J., Pedersen, E., Klemar, B., 1981. Standardisation of the electrical elicitation of the human flexor reflex. J. Neurol. Neurosurg. Psychiatry 44, 129–132. Villemure, C., Bushnell, M.C., 2002. Cognitive modulation of pain: how do attention and emotion influence pain processing?. Pain 95, 195–199, http://dx.doi.org/10.1016/S0304-3959(02) 00007-6. Von Dincklage, F., Correll, C., Schneider, M.H.N., Rehberg, B., 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, http: //dx.doi.org/10.1111/j.1365-2044.2012.07187.x. Von Dincklage, F., Olbrich, H., Baars, J.H., Rehberg, B., 2013. Habituation of the nociceptive flexion reflex is dependent on inter-stimulus interval and stimulus intensity. J. Clin. Neurosci. 20, 848–850, http://dx.doi.org/10.1016/j. jocn.2012.07.013. Willer, J.C., 1977. Comparative study of perceived pain and nociceptive flexion reflex in man. Pain 3, 69–80, http://dx.doi. org/10.1016/0304-3959(77)90036-7. Willer, J.C., Boureau, F., Albe-Fessard, D., 1979. Supraspinal influences on nociceptive flexion reflex and pain sensation in man. Brain Res. 179, 61–68, http://dx.doi.org/10.1016/00068993(79)90489-X.
