- Email: [email protected]
On the relationship between self-report and facial expression of pain
O RIGINAL R EPORTS On the Relationship Between Self-Report and Facial Expression of Pain Miriam Kunz,*,† Veit Mylius,*,‡ Karsten Schepelmann,‡ and Stefan Lautenbacher† Abstract: Several investigators have reported weak or no associations between self-report and facial expression of pain, concluding that both parameters appear to be unrelated. However, studies so far have only focused on an overall association, not considering psychophysical relationships between stimulus intensities and pain responses while computing correlations. In the present study these psychophysical relationships, between stimulus intensity on the one hand and response magnitudes (of self-report and facial expression) on the other hand, were described in terms of intercept and slope. Correlation analyses were conducted between intercept and slope parameters of selfreport and facial expression of pain. Forty young, pain-free individuals were investigated for their responses to mechanically and electrically induced pain. Self-report was assessed by Visual Analog Scales. Facial expression was examined by using the Facial Action Coding System. There were significant correlations between the linear slopes of the psychophysical functions of self-report and facial expression in pressure pain. Neither the intercepts nor overall mean responses in the 2 pain-signaling systems were significantly correlated. These findings suggest that the facial expression of pain appears to mirror self-report ratings, when their increases over a range of increasing stimulus intensities are considered in parallel. Perspective: In future studies, our psycho-physically derived observation that incremental changes in facial expression during developing pain are more characteristic for individuals than static levels needs further corroboration. © 2004 by the American Pain Society Key words: Facial expression, self-report, pain, psychophysics, FACS
T
he facial expression of pain is considered to be the most prominent nonverbal pain behavior.3,23 A variety of studies have demonstrated that there are specific facial movements associated with pain and that these occur relatively consistently across a range of clinical pain conditions1,8,11,13,16 and experimental pain modalities.20 Furthermore, it has been shown that the facial expression of pain can be differentiated from other negative subjective states12 and that the magnitude of facial
Received June 9, 2004; Accepted June 10, 2004 From the *Department of Psychiatry and Psychotherapy, Philipps University Marburg, Marburg, †Department of Physiological Psychology, OttoFriedrich University Bamberg, Bamberg, and ‡Department of Neurology, Philipps University Marburg, Marburg, Germany. Supported by a research grant of the Deutsche Forschungsgemeinschaft (La 685/5). Address reprint requests to Miriam Kunz, MD, Department of Psychiatry and Psychotherapy, Philipps University Marburg, Rudolf-Bultmann Str. 8, Marburg 35033, Germany. E-mail: [email protected] 1526-5900/$30.00 © 2004 by the American Pain Society doi:10.1016/j.jpain.2004.06.002
368
expression increases across rising intensities of noxious stimulation.9 However, the vast majority of previous research did not succeed in finding a direct relationship between self-report and facial expression of pain, neither by using painful clinical tests nor by using experimental pain stimuli.2,8-12,20 Significant correlations between both parameters were reported from only 3 studies.17,21,24 All 3 studies used painful clinical tests to examine the relationship. However, there does not appear to be an intelligible explanation of why these studies yielded significant correlations, whereas the majority did not. Given that most studies could not demonstrate significant correlations between facial expression and self-report, the question arises why that is the case. Some authors argue that both response parameters tap different aspects of the multidimensional pain experience, ie, the facial expression represents the more immediate, reflexive aspects, whereas self-report ratings often tend to be processed to a higher level and be more retrospective.9 Furthermore, research so far has only focused on overall
The Journal of Pain, Vol 5, No 7 (September), 2004: pp 368-376
ORIGINAL REPORT/Kunz et al associations, ie, the association between the subjects’ self-report of painfulness to some kind of extreme stimulus and the facial expression in response to that same stimulus. From a psychophysical perspective, the stimulus-response relationship characterizes sensory and perceptual systems by relating a certain physical input to a certain behavioral output. Stimulus-response relationships evaluate pain responses over a wide range of stimulus intensities. These relationships between stimulus intensity and response intensity can be described in terms of their threshold (intercept) and their rate of increase (slope).25 Most of the previous studies used single-stimulus approaches that do not allow for calculating intercept and slope,2,8,11,20 and even when different stimulus intensities were used,9 responses were not analyzed with regard to their intercept and slope. To demonstrate the gain of information when using stimulus-response relationships, a fictitious example is given in Fig 1, in which facial responses of 2 individuals to 5 different intensities of noxious stimulation are displayed. The 2 individuals differ substantially in their capacity to encode different pain intensities in their facial expression. These differences can be described by calculating intercept and slope of the stimulus-response relationship for the facial expression (with subject 2 displaying a higher threshold for facial responses, which– once triggered for the first time– increase with a much higher rate over the remaining stimulus intensities than those of subject 1). Single-stimulus approaches fail to register such differences, because they only allow the comparison of facial responses to 1 stimulus intensity between individuals (depending on the stimulus intensity, subject 1 would be considered to be more or less facial expressive than subject 2). Consequently, single-stimulus approaches might not be sufficient to depict individual response patterns precisely enough. In this pilot study we aim to investigate whether an analysis strategy that focuses on stimulus-response relationships allows us to demonstrate significant correlations between self-report and facial expression of pain. More specifically, we aim to test whether the association between facial expression and self-report of pain appears stronger when the onset of reactions (intercept) and the increase in reactivity (slope) are considered separately (first type of analysis), instead of computing correlations by using responses averaged across all intensities (second type of analysis). In addition, we tried to answer the question of whether facial responses show a closer linkage to the unpleasantness or to the sensory intensity dimension of pain. Because the characteristics of psychophysical stimulusresponse relationships depend, among a variety of factors, on the type of stimuli applied, we used 2 types of stimuli, ie, mechanical and electrical pain stimuli. This allows for testing the degree of external validity of our results.
369
Figure 1. Fictitious example displaying scores of facial expression in response to 5 different intensities of noxious stimuli.
Materials and Methods Subjects Forty young subjects (male, N ⫽ 20; female, N ⫽ 20), mostly students of medicine and psychology between the ages of 20 and 40 years (mean age, 24.0 years; standard deviation, 3.2 years), were recruited via advertisements posted in the university buildings. None had taken any analgesic medication for at least 24 hours before the test session. Participants had been previously screened and medically examined for conditions that could affect pain perception and pain report such as diabetes, hypertension, peripheral and central neuropathy, and neuropsychological and psychiatric disorders. The study protocol was approved by the ethics committee of the medical faculty of the University of Marburg. All subjects were paid for participation and gave written informed consent.
Materials and Procedure The verbal and facial responses to noxious stimuli were assessed by using 2 types of physical stimuli, ie, mechanical and electrical. All testing was done during the hours of 3:00 PM to 6:30 PM and lasted for approximately 2 hours. The testing procedure included an examination in regard to the exclusion criteria (approximately 1 hour), application of pressure stimuli (20 minutes), a short break (10 minutes), and application of electrical stimuli (30 minutes). All subjects were seated in an armchair.
Mechanical Pain Stimuli A Fischer algometer (a force gauge fitted by a rubber disk with a surface of 1 cm2) was used to assess responses to noxious mechanical pressure.5 The algometer was slightly modified so that each stimulus onset could be recorded and could be used as a trigger signal for the video analysis. To familiarize subjects with pressure stimulation, stimuli of 2 to 3 kg were applied to the thigh before tests
370
Self-Report and Facial Expression of Pain
Facial Action Units (AUs) Selected for Further Analysis
Table 1.
ACTION UNIT Pressure pain AU1/2 AU4 AU6/7 AU9/10 AU12 AU17 AU25/26/27 AU45† Electrical pain AU1/2 AU4 AU6/7 AU9/10 AU12 AU14 AU17 AU25/26/27 AU45†
DESCRIPTION
PERCENT*
P VALUE
Brow raiser Brow lower Orbit tightening Levator contraction Lip corner pull Chin raise Mouth opening Blink
14.3 9.8 22.7 9.5 14.8 7.1 15.8 218.0
ns .006 ⬍.001 .008 .003 ns ns ns
Brow raiser Brow lower Orbit tightening Levator contraction Lip corner pull Dimpler Chin raise Mouth opening Blink
17.9 17.5 49.2 19.0 19.2 10.8 11.9 33.8 277.5
⬍.001 ns ⬍.001 ⬍.001 .014 ns ns .001 .010
Abbreviation: ns, not significant. *Percentage of occurrence in the entire pain-related segments. †Blinking of the eye can appear more than once in a time segment of 5 seconds.
started. Twenty stimuli varying between 1 and 5 kg were applied to the right and left forearm in a random order, which had been determined once and was used for all subjects. Pressure was increased steadily at an application rate of 1 kg per second until stimulus peak was reached and then continued at that level for another 5 seconds. The interval between stimulus application was approximately 20 to 30 seconds. Pressure application was always performed by the same experimenter, who had been trained in using the Fischer algometer.
Electrical Pain Stimuli The assessment of self-report and facial expression to electrical stimuli was part of an assessment battery for nonverbal responses, including the nociceptive flexion reflex, heart rate responses, and sympathetic skin responses. These physiological data are not reported here. Electrical stimulation was especially designed to determine the threshold of the flexion reflex (in accordance to protocol of France et al6) and the suprathreshold reflex responses. For electrical stimulation a surface electrode was attached over the sural nerve at the backside of the lower limb. The stimulus consisted of a train of 5 rectangular impulses (1-millisecond duration) at a frequency of 250 Hz (Viking IV D; Nicolet Biomedical, Madison, WI). To familiarize the subject with electrical stimulation we chose stimuli of very mild intensities at the beginning. Stimulus intervals were always 15 to 20 seconds. First, the R-III reflex threshold was determined by using an up-down staircase method.6 For that purpose, stimulation intensity was increased in 3-mA steps until the
flexion reflex (defined as changes in voltage in a time window from 80 to 150 milliseconds after stimulation) was detected for the first time and then decreased in 2-mA steps until the reflex disappeared again. After that, steps of 1 mA were used, and the procedure was continued until the reflex appeared and subsided 2 more times. Mean scores of all 3 peaks and troughs (in milliamperes) determined the reflex threshold. Only the last 2 peak trials and the last 2 trough trials were selected for further analysis (self-report, facial expression), thus resulting in 2 trials each for below and above threshold intensities. Immediately after this threshold assessment protocol, a series of 10 stimuli 5 mA above the reflex threshold were delivered. All 10 trials were used for further analysis (self-report, facial expression).
Facial Expression of Pain The face of the subject was videotaped throughout the entire session. The camera was placed in front of the subject at a distance of approximately 4 m. Before applying a stimulus, subjects were always instructed to look at a fixation point (a picture being positioned behind the camera) to ensure a frontal view of the face. Subjects were also instructed not to talk during pain induction. To mark the onset of pain stimulation on the videotape (for further analysis), we switched on a signal light concurrently. The light was placed behind the subject. The Facial Action Coding System (FACS)4 was used to assess facial expression. The FACS is based on anatomic analysis of facial muscle movements and distinguishes 44 different action units (AUs). These are the minimal units of facial activity that are anatomically separate and visually distinguishable. The intensity for each AU is rated on a 5-point scale. A FACS coder (qualified by passing an examination given by the developers of the system) identified the frequency of all 44 AUs and the intensity of 42 AUs (except for AUs 45 and 46, which did not allow for intensity coding). The Observer Video-Pro (Noldus Information Technology, Wageningen, The Netherlands) was used to facilitate FACS coding. On the basis of preceding studies,12,18,23 AUs that involve the same muscle action were combined to form new variables. AUs 6 and 7 were combined to form 1 variable, because they only represent different degrees of eye orbit tightening. AUs 9 and 10 were combined in a similar fashion (representing different degrees of levator contraction), as were AUs 25, 26, and 27 (representing varying degrees of mouth opening) and AUs 1 and 2 (representing inner and outer brow raise). Time segments of 5 seconds after stimulus had reached maximum were selected for scoring. In total, 20 segments of pressure pain (4 times series of 1 to 5 kg) plus 14 segments of electrical pain (2 stimuli just below the threshold, 2 stimuli just above the reflex threshold, and 10 stimuli 5 mA above the threshold) were analyzed. As has been done in earlier studies,1,2,7,9,16 we reduced the number of AUs to those that occurred in at least 5% of the pain-related time segments recorded. We did this separately for pressure and electrical pain. AUs that did not occur with at least that frequency were not consid-
ORIGINAL REPORT/Kunz et al
371
Table 2. Predictors and Criteria for Multiple Regression Analyses of the First and Second Type of Analyses PREDICTORS First type of analysis Regression analysis between Intercept of FACS composite score frequency Intercept of FACS composite score intensity Regression analysis between Intercept of FACS composite score frequency Intercept of FACS composite score intensity Regression analysis between Slope of FACS composite score frequency Slope of FACS composite score intensity Regression analysis between Slope of FACS composite score frequency Slope of FACS composite score intensity Second type of analysis Regression analysis between Mean composite score of FACS frequency Mean composite score of FACS intensity Regression analysis between Mean composite score of FACS frequency Mean composite score of FACS intensity
CRITERIA
Intercept of VAS sensory intensity of pain
Intercept of VAS pain unpleasantness
Slope of VAS sensory intensity of pain
Slope of VAS pain unpleasantness
Mean score of VAS sensory intensity of pain
Mean score of VAS pain unpleasantness
Abbreviations: FACS, Facial Action Coding System; VAS, Visual Analog Scale.
ered to be pain relevant in our context and were therefore not included in further analysis. With the exception of AU 14 (“dimpler”), the same AUs were activated by both types of stimulation. These AUs broken down by pain-inducing methods are listed in Table 1. To determine which AUs were significantly more frequent during pain/noxious segments than during nonpainful/nonnoxious segments, we computed Wilcoxon signed rank tests for paired data. Those AUs that were significantly more frequent during pain/noxious segments are marked in Table 1. For further analysis all AUs that occurred in 5% of the pain-related time segments were summed (separately for FACS frequency and FACS intensity and separately for each pain induction method) to form composite scores of facial expression. Composite scores that were formed out of all those AUs that occurred in 5% of the painrelated segments guaranteed normal distribution much better than composite scores that only included those AUs that were significantly more frequent during pain/ noxious segments.
Self-Report Approximately 10 seconds after each stimulus application, subjects were asked to give self-report ratings regarding the peak sensation felt. The delay of 5 to 10 seconds was introduced to guarantee an undisturbed facial expression during the time segment of relevance. Self-report was assessed by horizontal visual analog scales (VAS) to allow for the assessment of sensory intensity of pain and pain unpleasantness. The VAS for sensory intensity of pain was labeled with verbal anchors from “no pain” (0) to “extremely strong pain” (100). Pain unpleasantness was labeled with “no pain” (0) to “ex-
tremely unpleasant pain” (100). All subjects were instructed in the conceptual distinction between sensory intensity of pain and pain unpleasantness by using standard instructions.19
Statistical Analysis Psychophysical Stimulus-Response Relationships Separately for each subject, intercept and slope of the stimulus-response relationship for self-report (VAS ratings of sensory intensity and unpleasantness of pain) and of the stimulus-response relationship for facial expression (frequency and intensity of AUs) were calculated. This was done by computing linear regression analyses (with stimulus intensity always serving as the predictor). Stimulus intensities ranged from 1 to 5 kg in 5 steps for pressure pain and from below to 5 mA above reflex threshold in 3 steps for electrical pain. The goodness of fit (r2) ranged from 0.37 to 0.90 with a median of 0.67 over all individuals and parameters (FACS, VAS). Graphical inspection of the individual stimulus-response relationships showed that the type of deviation from linearity was unsystematically varying, so we started with the most parsimonious assumption, ie, the assumption of linearity.
Self-Report and Facial Expression Two different analyses were used to compute correlations between self-report and facial expression of pain, both analyses differing in the set of predictors and criteria.
First Type of Analysis (Considering the Stimulus-
372
Self-Report and Facial Expression of Pain
Figure 3. Mean values (⫾ standard deviation) of FACS scores (frequency, intensity) in dependency of stimulus intensities in pressure pain (A) and in electrical pain (B). Figure 2. Mean values (⫾ standard deviation) of VAS ratings (intensity and unpleasantness) in dependency of stimulus intensities in pressure pain (A) and in electrical pain (B).
Response Relationship). Multiple regression analyses were computed between either the intercepts (of the stimulus-response relationships) for self-report and for facial expression or between the slopes (of the stimulusresponse relationships) for self-report and for facial expression. Second Type of Analysis (Considering the Overall Responses). In addition, an approach was used to investigate the relationship between self-report and facial expression of pain that disregarded the distinction between intercept and slope values. Therefore, mean scores of self-report and of facial expression (averaged across stimulus intensities) were used in multiple regression analysis. Predictors and criteria of both types of analyses are displayed in Table 2.
We also computed repeated analyses of variance for repeated measurement to assess the increase in self-report ratings and in facial expression dependent on the increase in stimulus intensities. Stimulus intensity served as the within-subjects factor. A second within-subjects factor, namely the type of VAS scale (intensity or unpleasantness), was considered for self-report. In case of significant results, t tests were used for pairwise comparisons. Findings were always considered to be statistically significant at ␣ ⬍ 0.05.
Results Effects of Stimulus Intensity on SelfReport and Facial Expression Self-Report The increase in self-reported pain from weak to strong stimuli (Fig 2) was significant, regarding both pressure
ORIGINAL REPORT/Kunz et al Table 3.
373
Descriptive Statistics for Intercepts and Slopes PRESSURE PAIN, MEAN (⫾SD)
VAS sensory intensity of pain VAS pain unpleasantness FACS composite score frequency FACS composite score intensity
ELECTRICAL PAIN, MEAN (⫾SD)
INTERCEPT
SLOPE
INTERCEPT
SLOPE
1.44 (⫾12.8) ⫺3.48 (⫾13.9) 0.96 (⫾1.0) ⫺0.00 (⫾0.3)
9.40 (⫾4.5) 8.62 (⫾5.0) 0.17 (⫾0.3) 0.13 (⫾0.2)
0.79 (⫾90.0) ⫺3.66 (⫾50.8) ⫺14.62 (⫾11.9) ⫺1.43 (⫾1.8)
2.38 (⫾2.9) 2.61 (⫾2.3) 0.03 (⫾0.1) 0.12 (⫾0.1)
Abbreviations: SD, standard deviation; VAS, Visual Analog Scale; FACS, Facial Action Coding System.
pain (F4,156 ⫽ 50.48; P ⬍ .001) and electrical pain (F2,78 ⫽ 9.17; P ⫽ .002). Descriptive statistics of minimum and maximum showed that pressure stimuli higher than 4 kg and all electrical stimuli were rated as being painful by all subjects (minimum for VAS intensity and unpleasantness ratings, ⬎0). This confirms that both pain induction methods included stimulus intensities that all subjects experienced as being painful. The pain ratings differed significantly between VAS unpleasantness and VAS sensory intensity of pain (F1,39 ⫽ 5.07; P ⫽ .03) for pressure pain (with higher values in VAS intensity ratings). No significant differences between the 2 VAS ratings were detected for electrical stimuli (F1,39 ⫽ 0.36; P ⫽ .55).
Facial Expression The analyses of variance for repeated measurements showed that frequency (pressure: F4,156 ⫽ 10.49; P ⫽ .001; electrical: F2,78 ⫽ 159.37; P ⬍ .001) and intensity (pressure: F4,156 ⫽ 16.51; P ⬍ .001; electrical: F2,78 ⫽ 56.38; P ⬍ .001) of facial expression increased significantly over stimulus intensities both for pressure pain (Fig 3A) and for electrical pain (Fig 3B). For pressure pain, pairwise comparisons showed no significant differences in all comparisons between stimulus intensities below 4 kg (all P values ⬎ .4). In contrast, stimulus intensities of 4 and 5 kg elicited significantly stronger facial expressions
Table 4. Multiple Correlation Coefficients (of the Multiple Regression Analyses), First Type of Analysis FACIAL EXPRESSION (FREQUENCY & INTENSITY) (PREDICTOR) Pressure pain Intercept Slope Electrical pain Intercept Slope
†F2, 39 ⫽ 3.33; P ⫽ .047.
Correlation Between Self-Report and Facial Expression First Type of Analysis The basic statistics of intercepts and slopes are given in Table 3. In regard to slope values (of the stimulus-response relationships), we found significant multiple correlation coefficients between facial expression (frequency and intensity) and pain unpleasantness as well as sensory intensity of pain, when using pressure pain (Table 4). No significant correlations between slope values of the stimulus-response relationships for facial expression and for self-report were found by using electrical stimuli. We did not find any significant correlations for pressure pain or for electrical pain between intercepts of the stimulusresponse relationship for facial expression and intercepts of the stimulus-response relationship for self-report. Accordingly, intercepts of facial expression (frequency, intensity) could not predict intercepts of self-report ratings. The correlation computation including slope parameters always led to higher correlation coefficients, although not always significant, in comparison to inter-
SELF-REPORT (CRITERION) VAS SENSORY INTENSITY OF PAIN (INTERCEPT/SLOPE)
VAS PAIN UNPLEASANTNESS (INTERCEPT/SLOPE)
.211 .414*
.134 .391†
.171 .271
.207 .290
Abbreviation: VAS, Visual Analog Scale. *F2, 39 ⫽ 3.83; P ⫽ .031.
(frequency, intensity) than stimulus intensities below 4 kg (all P values ⬍ .005). There also was a significant increase in facial expressions (frequency, intensity) from 4 to 5 kg (all P values ⬍ .005). In regard to electrical pain, all pairwise comparisons (with the exception of the comparison between frequency above and frequency 5 mA above reflex threshold) showed significant results (all P values ⬍ .005).
Table 5. Multiple Correlation Coefficients (of the Multiple Regression Analyses), Second Type of Analysis FACIAL EXPRESSION (FREQUENCY & INTENSITY) (PREDICTOR) Pressure pain Electrical pain
SELF-REPORT (CRITERION) VAS SENSORY INTENSITY OF PAIN (INTERCEPT/SLOPE)
VAS PAIN UNPLEASANTNESS (INTERCEPT/SLOPE)
.096 .225
.127 .157
Abbreviation: VAS, Visual Analog Scale.
374 cept parameters by using both pressure and electrical pain (Table 4). Correlations between pain unpleasantness and facial expression did not yield higher correlation coefficients than did correlations between sensory intensity of pain and facial expression.
Second Type of Analysis No significant multiple correlation coefficients were detected (neither for pressure pain nor for electrical pain) when responses averaged across intensities were used. Mean scores of facial expression of pain (frequency/intensity) could not significantly predict mean scores of self-report ratings (Table 5).
Discussion The present study verified a substantial relationship between the individual abilities to report and facially express different levels of pressure pain. In statistical words, slopes of the stimulus-response relationships for self-report and facial expression were significantly correlated. This indicates that an increase of noxious stimulus intensity leads to a strong increase in self-report and facial expression in one individual, whereas it leads to a weak increase in both parameters in another one. Admittedly, the correlations were only statistically significant for experimental pressure pain and of moderate size. (In the case of electrical pain the correlations tended into similar directions but did not quite reach significance.) Nevertheless, in contrast to the vast majority of earlier reports, our data suggest that the verbal and facial response systems are linked. It mainly appears to be a matter of applying appropriate analysis strategies to find this linkage. Conversely to the findings regarding slope parameters, no significant correlations were obtained by using the intercepts of the stimulus-response relationships for selfreport and facial expression. The intercept can be considered to describe the activation threshold, representing the minimum level of stimulation required to elicit a specific pain response. Our results suggest that the thresholds of self-report and facial expression are independent of each other. This means that an individual who is reporting pain does not necessarily express this facially or vice versa. Prkachin and Craig22 described the facial expression of pain as a “late” signaling system, meaning that even though pain is being experienced, it might still not be facially expressed. However, our finding of no correlation between intercepts makes it likely that there are individuals who facially respond in a pain-specific manner to stimulus intensities that they do not rate as being painful. Taking together the findings on slope and intercept parameters, it appears that although the starting points in the verbal and facial response systems are unrelated, the individual degrees of reactivity in both systems during pain are related to each other. The latter only becomes obvious when applying different stimulus intensities for computation of stimulus-response functions. Mean scores of facial responses to pain (averaged
Self-Report and Facial Expression of Pain across stimulus intensities) appeared uncorrelated to mean scores of self-report ratings (averaged across stimulus intensities) when using pressure and electrical pain. Accordingly, individuals who are facially more expressive on average do not rate the stimuli as being more painful on average in comparison to those who show less facial expression of pain. As described in the introduction, disregarding intercept and slope of the stimulus-response relationship might result in missing the relationship between particular aspects of pain report and facial expression. Therefore, approaches that average responses across intensities or single stimulus approaches that do not allow for calculating intercept and slope might be misleading when investigating the association between self-report and facial expression of pain. These findings have several practical implications. First, the independence of starting points (activation thresholds) in the verbal and facial response systems suggests that the absence of a facial expression of pain cannot necessarily be interpreted as a sign that no pain is being experienced. Those engaged in pain management ought to be aware of that. Second, the correlation between slopes of the stimulus-response relationships for self-report and facial expression suggests that an increase/decrease in facial expression of pain is likely linked to an increase/decrease in subjective pain experience, thus yielding clear support for the utility of facial expression as an indicator of the subjective pain experience when intraindividual changes are considered. This might be of great importance in regard to pain assessment in individuals with limited ability in verbal communication. Observer should try to study the increase/decrease of facial responses across a variety of potential painful activities to gain more valid information on how the patient is facially encoding her/his pain experience. We only found moderate correlations (around r ⫽ 0.4) between the slopes of the stimulus-response relationships for self-report and facial responses. Among other factors, this might be due to the fact that the pain stimulus intensities, which were used for analysis, were too weak. Regarding pressure pain (Fig 3A), it becomes obvious that facial expression only started to increase with stimulus intensities from 4 kg on. Stimuli of lower intensities elicited very weak facial responses that were nonsystematically related to stimulus strength. It is quite possible that stronger stimulus intensities might have led to an even stronger correlation between slopes of the stimulus-response relationships for self-report and facial expression of pain. With electrical stimuli, we found even slightly weaker correlations between the slope values, although facial expression appeared to increase from the weakest stimulus on. A possible interpretation of this negative finding might be that pain induced by electrical current (with its sudden onset) elicits further emotional processes, such as startle reactions, that might be confounded with pain. As a consequence, facial expressions that are not entirely the result of pain but are blended with other emotions as well might occur, thus lowering the correlation between facial expression and self-report of pain.
ORIGINAL REPORT/Kunz et al
375
Comparisons between self-report and facial expression of pain induced by pressure and electrical current were a bit unbalanced because the stimuli selection (regarding number and intensity) differed substantially. Our analysis strategy was optimized in regard to pressure pain. We used this strategy a second time as a kind of first test of external validation of pain induced by electrical current, originally designed for assessing further nonverbal, physiological parameters of pain. Besides the type of analytic approach, the type of stimulus induction method, and the range of stimulus intensities, there are numerous additional factors that moderate the association between facial expression or other nonverbal pain behaviors and self-report of pain (eg, time of pain assessment,2 social modifiers,18 severity of experienced distress,24 number of nonverbal pain expressions being studied,15 pain site,15 duration of pain15). It remains to be clarified how these factors moderate the association between self-report and facial expression of pain when our analysis strategies are applied. It has often been assumed that facial expression is more closely related to the affective dimension of pain,14,17 but we could not detect a closer association to the unpleasantness dimension of pain (VAS ratings of unpleasantness). The VAS ratings of sensory intensity and unpleasantness of pain were highly interrelated (pressure pain, r ⫽ 0.76; electrical pain, r ⫽ 0.93), which probably accounts for the similar results found for sensory intensity and unpleasantness ratings. Nevertheless, using 2 scales to discriminate between the sensory intensity and affective dimensions of pain appears still advisable, given the significant differences found between VAS–sensory intensity and VAS-unpleasantness ratings when using pressure pain. The study has several limitations. First, a rather limited
number of pain stimuli levels were used to investigate the association between relationships of stimulus intensity and intensity of self-report/facial expression of pain. A less limited number of pain stimuli levels would probably ensure a more reliable and complete depiction of the stimulus response relationships. Future studies should apply more numerous pain stimuli levels when using psychophysical relationships between stimulus intensities and pain responses to compute correlations between self-report and facial expression. Second, video segments were not analyzed in a randomized order but in their order of assessment. Therefore, it is possible that the FACS codings could have been biased by expectancy effects. Third, we decided to always start with the application of pressure pain before electrical pain and did not use a random order. This was done because some individuals experience even very weak intensities of electrical stimulation as being frightening, and our procedure requires moderate intensities well above pain thresholds. This fixed order allowed us to always start with the less stressful part. In conclusion, the facial expression of pain appears to mirror self-report ratings, when their increases over a range of increasing stimulus intensities are considered in parallel (although the moderate magnitude of correlations suggests that both remain at least partially independent and under the control of third factors). In contrast, the starting points for activation of the 2 response systems appear to be independent of each other. This suggests that the facial expression of pain can be used as an estimate of the subjective pain when the reactivity over different levels of pain states or pain stimuli is analyzed. Single stimulus approaches as well as approaches averaging responses across different stimulus intensities can be misleading.
References
8. Hadjistavropoulos T, Craig KD, Martin N, Hadjistavropoulos HD, McMurtry B: Toward a research outcome measure of pain in frail elderly in chronic care. Pain Clinic 10:71-79, 1997
1. Craig KD, Hyde SA, Patrick CJ: Genuine, suppressed and faked facial behavior during exacerbation of chronic low back pain. Pain 46:161-171, 1991 2. Craig KD, Patrick CJ: Facial expression during induced pain. J Pers Soc Psychol 48:1080-1091, 1985 3. Craig KD, Prkachin KM, Grunau RVE: The facial expression of pain, in Turk D, Melzack R (eds): Handbook of Pain Assessment. New York, NY, Guilford Press, 2001, pp 153169 4. Ekman PE, Friesen WV: Facial Action Coding System. Palo Alto, CA, Consulting Psychologists Press, 1978 5. Fischer AA: Reliability of the pressure algometer as a measure of myofascial trigger point sensitivity. Pain 28:411414, 1987 6. France CR, France JL, al’Absi M, Ring C, McIntyre D: Catastrophizing is related to pain ratings, but not nociceptive flexion reflex threshold. Pain 99:459-463, 2002 7. Hadjistavropoulos HD, Craig KD, Hadjistavropoulos T, Poole GD: Subjective judgments of deception in pain expression: Accuracy and errors. Pain 65:251-258, 1996
9. Hadjistavropoulos T, LaChapelle DL, Hadjistavropoulos HD, Green S, Asmundson GJG: Using facial expressions to assess musculoskeletal pain in older persons. Eur J Pain 6:179-187, 2002 10. Hadjistavropoulos T, LaChapelle DL, MacLeod FK, Hale C, O’Rourke N, Craig KD: Cognitive functioning and pain reactions in hospitalized elders. Pain Res Manage 3:145-151, 1998 11. Hadjistavropoulos T, LaChapelle DL, MacLeod FK, Snider B, Craig KD: Measuring movement-exacerbated pain in cognitively impaired frail elders. Clin J Pain 16:5463, 2000 12. Hale CJ, Hadjistavropoulos T: Emotional components of pain. Pain Res Manage 2:217-225, 1997 13. Hill ML, Craig KD: Detecting deception in pain expressions: The structure of genuine and deceptive facial displays. Pain 98:135-144, 2002 14. Kappesser J, Williams AC: Pain and negative emotions in
376 the face: Judgements by health care professionals. Pain 99:197-206, 2002 15. Labus JS, Keefe FJ, Jensen MP: Self-reports of pain intensity and direct observations of pain behavior: When are they correlated? Pain 102:109-124, 2003 16. LaChapelle DL, Hadjistavropoulos T, Craig KD: Pain measurement in persons with intellectual disabilities. Clin J Pain 15:13-23, 1999 17. LeResche L, Dworkin SF: Facial expressions of pain and emotions in chronic TMD patients. Pain 35:71-78, 1988 18. Patrick CJ, Craig KD, Prkachin KM: Observer judgements of acute pain: Facial action determinants. J Pers Soc Psychol 50:1291-1298, 1986 19. Price DD, McGrath PA, Rafii A, Buckingham B: The validation of visual analogue scales as ratio scale measures for chronic and experimental pain. Pain 17:45-56, 1983
Self-Report and Facial Expression of Pain 20. Prkachin KM: The consistency of facial expression of pain: A comparison across modalities. Pain 51:297-306, 1992 21. Prkachin KM, Berzins S, Mercer SR: Encoding and decoding of pain expressions: A judgement study. Pain 58:253-259, 1994 22. Prkachin KM, Craig KD: Expressing pain: The communication and interpretation of facial pain signals. J Nonv Behav19:191-205, 1994 23. Prkachin KM, Currie NA, Craig KD: Judging nonverbal expressions of pain. Can J Behav Sci 15:43-73, 1983 24. Prkachin KM, Mercer SR: Pain expression in patients with shoulder pathology: Validity, properties and relationship to sickness impact. Pain 39:257-265, 1989 25. Stevens SS: The psychophysics of sensory functions. Am Scientist 48:226-253, 1960
T
he facial expression of pain is considered to be the most prominent nonverbal pain behavior.3,23 A variety of studies have demonstrated that there are specific facial movements associated with pain and that these occur relatively consistently across a range of clinical pain conditions1,8,11,13,16 and experimental pain modalities.20 Furthermore, it has been shown that the facial expression of pain can be differentiated from other negative subjective states12 and that the magnitude of facial
Received June 9, 2004; Accepted June 10, 2004 From the *Department of Psychiatry and Psychotherapy, Philipps University Marburg, Marburg, †Department of Physiological Psychology, OttoFriedrich University Bamberg, Bamberg, and ‡Department of Neurology, Philipps University Marburg, Marburg, Germany. Supported by a research grant of the Deutsche Forschungsgemeinschaft (La 685/5). Address reprint requests to Miriam Kunz, MD, Department of Psychiatry and Psychotherapy, Philipps University Marburg, Rudolf-Bultmann Str. 8, Marburg 35033, Germany. E-mail: [email protected] 1526-5900/$30.00 © 2004 by the American Pain Society doi:10.1016/j.jpain.2004.06.002
368
expression increases across rising intensities of noxious stimulation.9 However, the vast majority of previous research did not succeed in finding a direct relationship between self-report and facial expression of pain, neither by using painful clinical tests nor by using experimental pain stimuli.2,8-12,20 Significant correlations between both parameters were reported from only 3 studies.17,21,24 All 3 studies used painful clinical tests to examine the relationship. However, there does not appear to be an intelligible explanation of why these studies yielded significant correlations, whereas the majority did not. Given that most studies could not demonstrate significant correlations between facial expression and self-report, the question arises why that is the case. Some authors argue that both response parameters tap different aspects of the multidimensional pain experience, ie, the facial expression represents the more immediate, reflexive aspects, whereas self-report ratings often tend to be processed to a higher level and be more retrospective.9 Furthermore, research so far has only focused on overall
The Journal of Pain, Vol 5, No 7 (September), 2004: pp 368-376
ORIGINAL REPORT/Kunz et al associations, ie, the association between the subjects’ self-report of painfulness to some kind of extreme stimulus and the facial expression in response to that same stimulus. From a psychophysical perspective, the stimulus-response relationship characterizes sensory and perceptual systems by relating a certain physical input to a certain behavioral output. Stimulus-response relationships evaluate pain responses over a wide range of stimulus intensities. These relationships between stimulus intensity and response intensity can be described in terms of their threshold (intercept) and their rate of increase (slope).25 Most of the previous studies used single-stimulus approaches that do not allow for calculating intercept and slope,2,8,11,20 and even when different stimulus intensities were used,9 responses were not analyzed with regard to their intercept and slope. To demonstrate the gain of information when using stimulus-response relationships, a fictitious example is given in Fig 1, in which facial responses of 2 individuals to 5 different intensities of noxious stimulation are displayed. The 2 individuals differ substantially in their capacity to encode different pain intensities in their facial expression. These differences can be described by calculating intercept and slope of the stimulus-response relationship for the facial expression (with subject 2 displaying a higher threshold for facial responses, which– once triggered for the first time– increase with a much higher rate over the remaining stimulus intensities than those of subject 1). Single-stimulus approaches fail to register such differences, because they only allow the comparison of facial responses to 1 stimulus intensity between individuals (depending on the stimulus intensity, subject 1 would be considered to be more or less facial expressive than subject 2). Consequently, single-stimulus approaches might not be sufficient to depict individual response patterns precisely enough. In this pilot study we aim to investigate whether an analysis strategy that focuses on stimulus-response relationships allows us to demonstrate significant correlations between self-report and facial expression of pain. More specifically, we aim to test whether the association between facial expression and self-report of pain appears stronger when the onset of reactions (intercept) and the increase in reactivity (slope) are considered separately (first type of analysis), instead of computing correlations by using responses averaged across all intensities (second type of analysis). In addition, we tried to answer the question of whether facial responses show a closer linkage to the unpleasantness or to the sensory intensity dimension of pain. Because the characteristics of psychophysical stimulusresponse relationships depend, among a variety of factors, on the type of stimuli applied, we used 2 types of stimuli, ie, mechanical and electrical pain stimuli. This allows for testing the degree of external validity of our results.
369
Figure 1. Fictitious example displaying scores of facial expression in response to 5 different intensities of noxious stimuli.
Materials and Methods Subjects Forty young subjects (male, N ⫽ 20; female, N ⫽ 20), mostly students of medicine and psychology between the ages of 20 and 40 years (mean age, 24.0 years; standard deviation, 3.2 years), were recruited via advertisements posted in the university buildings. None had taken any analgesic medication for at least 24 hours before the test session. Participants had been previously screened and medically examined for conditions that could affect pain perception and pain report such as diabetes, hypertension, peripheral and central neuropathy, and neuropsychological and psychiatric disorders. The study protocol was approved by the ethics committee of the medical faculty of the University of Marburg. All subjects were paid for participation and gave written informed consent.
Materials and Procedure The verbal and facial responses to noxious stimuli were assessed by using 2 types of physical stimuli, ie, mechanical and electrical. All testing was done during the hours of 3:00 PM to 6:30 PM and lasted for approximately 2 hours. The testing procedure included an examination in regard to the exclusion criteria (approximately 1 hour), application of pressure stimuli (20 minutes), a short break (10 minutes), and application of electrical stimuli (30 minutes). All subjects were seated in an armchair.
Mechanical Pain Stimuli A Fischer algometer (a force gauge fitted by a rubber disk with a surface of 1 cm2) was used to assess responses to noxious mechanical pressure.5 The algometer was slightly modified so that each stimulus onset could be recorded and could be used as a trigger signal for the video analysis. To familiarize subjects with pressure stimulation, stimuli of 2 to 3 kg were applied to the thigh before tests
370
Self-Report and Facial Expression of Pain
Facial Action Units (AUs) Selected for Further Analysis
Table 1.
ACTION UNIT Pressure pain AU1/2 AU4 AU6/7 AU9/10 AU12 AU17 AU25/26/27 AU45† Electrical pain AU1/2 AU4 AU6/7 AU9/10 AU12 AU14 AU17 AU25/26/27 AU45†
DESCRIPTION
PERCENT*
P VALUE
Brow raiser Brow lower Orbit tightening Levator contraction Lip corner pull Chin raise Mouth opening Blink
14.3 9.8 22.7 9.5 14.8 7.1 15.8 218.0
ns .006 ⬍.001 .008 .003 ns ns ns
Brow raiser Brow lower Orbit tightening Levator contraction Lip corner pull Dimpler Chin raise Mouth opening Blink
17.9 17.5 49.2 19.0 19.2 10.8 11.9 33.8 277.5
⬍.001 ns ⬍.001 ⬍.001 .014 ns ns .001 .010
Abbreviation: ns, not significant. *Percentage of occurrence in the entire pain-related segments. †Blinking of the eye can appear more than once in a time segment of 5 seconds.
started. Twenty stimuli varying between 1 and 5 kg were applied to the right and left forearm in a random order, which had been determined once and was used for all subjects. Pressure was increased steadily at an application rate of 1 kg per second until stimulus peak was reached and then continued at that level for another 5 seconds. The interval between stimulus application was approximately 20 to 30 seconds. Pressure application was always performed by the same experimenter, who had been trained in using the Fischer algometer.
Electrical Pain Stimuli The assessment of self-report and facial expression to electrical stimuli was part of an assessment battery for nonverbal responses, including the nociceptive flexion reflex, heart rate responses, and sympathetic skin responses. These physiological data are not reported here. Electrical stimulation was especially designed to determine the threshold of the flexion reflex (in accordance to protocol of France et al6) and the suprathreshold reflex responses. For electrical stimulation a surface electrode was attached over the sural nerve at the backside of the lower limb. The stimulus consisted of a train of 5 rectangular impulses (1-millisecond duration) at a frequency of 250 Hz (Viking IV D; Nicolet Biomedical, Madison, WI). To familiarize the subject with electrical stimulation we chose stimuli of very mild intensities at the beginning. Stimulus intervals were always 15 to 20 seconds. First, the R-III reflex threshold was determined by using an up-down staircase method.6 For that purpose, stimulation intensity was increased in 3-mA steps until the
flexion reflex (defined as changes in voltage in a time window from 80 to 150 milliseconds after stimulation) was detected for the first time and then decreased in 2-mA steps until the reflex disappeared again. After that, steps of 1 mA were used, and the procedure was continued until the reflex appeared and subsided 2 more times. Mean scores of all 3 peaks and troughs (in milliamperes) determined the reflex threshold. Only the last 2 peak trials and the last 2 trough trials were selected for further analysis (self-report, facial expression), thus resulting in 2 trials each for below and above threshold intensities. Immediately after this threshold assessment protocol, a series of 10 stimuli 5 mA above the reflex threshold were delivered. All 10 trials were used for further analysis (self-report, facial expression).
Facial Expression of Pain The face of the subject was videotaped throughout the entire session. The camera was placed in front of the subject at a distance of approximately 4 m. Before applying a stimulus, subjects were always instructed to look at a fixation point (a picture being positioned behind the camera) to ensure a frontal view of the face. Subjects were also instructed not to talk during pain induction. To mark the onset of pain stimulation on the videotape (for further analysis), we switched on a signal light concurrently. The light was placed behind the subject. The Facial Action Coding System (FACS)4 was used to assess facial expression. The FACS is based on anatomic analysis of facial muscle movements and distinguishes 44 different action units (AUs). These are the minimal units of facial activity that are anatomically separate and visually distinguishable. The intensity for each AU is rated on a 5-point scale. A FACS coder (qualified by passing an examination given by the developers of the system) identified the frequency of all 44 AUs and the intensity of 42 AUs (except for AUs 45 and 46, which did not allow for intensity coding). The Observer Video-Pro (Noldus Information Technology, Wageningen, The Netherlands) was used to facilitate FACS coding. On the basis of preceding studies,12,18,23 AUs that involve the same muscle action were combined to form new variables. AUs 6 and 7 were combined to form 1 variable, because they only represent different degrees of eye orbit tightening. AUs 9 and 10 were combined in a similar fashion (representing different degrees of levator contraction), as were AUs 25, 26, and 27 (representing varying degrees of mouth opening) and AUs 1 and 2 (representing inner and outer brow raise). Time segments of 5 seconds after stimulus had reached maximum were selected for scoring. In total, 20 segments of pressure pain (4 times series of 1 to 5 kg) plus 14 segments of electrical pain (2 stimuli just below the threshold, 2 stimuli just above the reflex threshold, and 10 stimuli 5 mA above the threshold) were analyzed. As has been done in earlier studies,1,2,7,9,16 we reduced the number of AUs to those that occurred in at least 5% of the pain-related time segments recorded. We did this separately for pressure and electrical pain. AUs that did not occur with at least that frequency were not consid-
ORIGINAL REPORT/Kunz et al
371
Table 2. Predictors and Criteria for Multiple Regression Analyses of the First and Second Type of Analyses PREDICTORS First type of analysis Regression analysis between Intercept of FACS composite score frequency Intercept of FACS composite score intensity Regression analysis between Intercept of FACS composite score frequency Intercept of FACS composite score intensity Regression analysis between Slope of FACS composite score frequency Slope of FACS composite score intensity Regression analysis between Slope of FACS composite score frequency Slope of FACS composite score intensity Second type of analysis Regression analysis between Mean composite score of FACS frequency Mean composite score of FACS intensity Regression analysis between Mean composite score of FACS frequency Mean composite score of FACS intensity
CRITERIA
Intercept of VAS sensory intensity of pain
Intercept of VAS pain unpleasantness
Slope of VAS sensory intensity of pain
Slope of VAS pain unpleasantness
Mean score of VAS sensory intensity of pain
Mean score of VAS pain unpleasantness
Abbreviations: FACS, Facial Action Coding System; VAS, Visual Analog Scale.
ered to be pain relevant in our context and were therefore not included in further analysis. With the exception of AU 14 (“dimpler”), the same AUs were activated by both types of stimulation. These AUs broken down by pain-inducing methods are listed in Table 1. To determine which AUs were significantly more frequent during pain/noxious segments than during nonpainful/nonnoxious segments, we computed Wilcoxon signed rank tests for paired data. Those AUs that were significantly more frequent during pain/noxious segments are marked in Table 1. For further analysis all AUs that occurred in 5% of the pain-related time segments were summed (separately for FACS frequency and FACS intensity and separately for each pain induction method) to form composite scores of facial expression. Composite scores that were formed out of all those AUs that occurred in 5% of the painrelated segments guaranteed normal distribution much better than composite scores that only included those AUs that were significantly more frequent during pain/ noxious segments.
Self-Report Approximately 10 seconds after each stimulus application, subjects were asked to give self-report ratings regarding the peak sensation felt. The delay of 5 to 10 seconds was introduced to guarantee an undisturbed facial expression during the time segment of relevance. Self-report was assessed by horizontal visual analog scales (VAS) to allow for the assessment of sensory intensity of pain and pain unpleasantness. The VAS for sensory intensity of pain was labeled with verbal anchors from “no pain” (0) to “extremely strong pain” (100). Pain unpleasantness was labeled with “no pain” (0) to “ex-
tremely unpleasant pain” (100). All subjects were instructed in the conceptual distinction between sensory intensity of pain and pain unpleasantness by using standard instructions.19
Statistical Analysis Psychophysical Stimulus-Response Relationships Separately for each subject, intercept and slope of the stimulus-response relationship for self-report (VAS ratings of sensory intensity and unpleasantness of pain) and of the stimulus-response relationship for facial expression (frequency and intensity of AUs) were calculated. This was done by computing linear regression analyses (with stimulus intensity always serving as the predictor). Stimulus intensities ranged from 1 to 5 kg in 5 steps for pressure pain and from below to 5 mA above reflex threshold in 3 steps for electrical pain. The goodness of fit (r2) ranged from 0.37 to 0.90 with a median of 0.67 over all individuals and parameters (FACS, VAS). Graphical inspection of the individual stimulus-response relationships showed that the type of deviation from linearity was unsystematically varying, so we started with the most parsimonious assumption, ie, the assumption of linearity.
Self-Report and Facial Expression Two different analyses were used to compute correlations between self-report and facial expression of pain, both analyses differing in the set of predictors and criteria.
First Type of Analysis (Considering the Stimulus-
372
Self-Report and Facial Expression of Pain
Figure 3. Mean values (⫾ standard deviation) of FACS scores (frequency, intensity) in dependency of stimulus intensities in pressure pain (A) and in electrical pain (B). Figure 2. Mean values (⫾ standard deviation) of VAS ratings (intensity and unpleasantness) in dependency of stimulus intensities in pressure pain (A) and in electrical pain (B).
Response Relationship). Multiple regression analyses were computed between either the intercepts (of the stimulus-response relationships) for self-report and for facial expression or between the slopes (of the stimulusresponse relationships) for self-report and for facial expression. Second Type of Analysis (Considering the Overall Responses). In addition, an approach was used to investigate the relationship between self-report and facial expression of pain that disregarded the distinction between intercept and slope values. Therefore, mean scores of self-report and of facial expression (averaged across stimulus intensities) were used in multiple regression analysis. Predictors and criteria of both types of analyses are displayed in Table 2.
We also computed repeated analyses of variance for repeated measurement to assess the increase in self-report ratings and in facial expression dependent on the increase in stimulus intensities. Stimulus intensity served as the within-subjects factor. A second within-subjects factor, namely the type of VAS scale (intensity or unpleasantness), was considered for self-report. In case of significant results, t tests were used for pairwise comparisons. Findings were always considered to be statistically significant at ␣ ⬍ 0.05.
Results Effects of Stimulus Intensity on SelfReport and Facial Expression Self-Report The increase in self-reported pain from weak to strong stimuli (Fig 2) was significant, regarding both pressure
ORIGINAL REPORT/Kunz et al Table 3.
373
Descriptive Statistics for Intercepts and Slopes PRESSURE PAIN, MEAN (⫾SD)
VAS sensory intensity of pain VAS pain unpleasantness FACS composite score frequency FACS composite score intensity
ELECTRICAL PAIN, MEAN (⫾SD)
INTERCEPT
SLOPE
INTERCEPT
SLOPE
1.44 (⫾12.8) ⫺3.48 (⫾13.9) 0.96 (⫾1.0) ⫺0.00 (⫾0.3)
9.40 (⫾4.5) 8.62 (⫾5.0) 0.17 (⫾0.3) 0.13 (⫾0.2)
0.79 (⫾90.0) ⫺3.66 (⫾50.8) ⫺14.62 (⫾11.9) ⫺1.43 (⫾1.8)
2.38 (⫾2.9) 2.61 (⫾2.3) 0.03 (⫾0.1) 0.12 (⫾0.1)
Abbreviations: SD, standard deviation; VAS, Visual Analog Scale; FACS, Facial Action Coding System.
pain (F4,156 ⫽ 50.48; P ⬍ .001) and electrical pain (F2,78 ⫽ 9.17; P ⫽ .002). Descriptive statistics of minimum and maximum showed that pressure stimuli higher than 4 kg and all electrical stimuli were rated as being painful by all subjects (minimum for VAS intensity and unpleasantness ratings, ⬎0). This confirms that both pain induction methods included stimulus intensities that all subjects experienced as being painful. The pain ratings differed significantly between VAS unpleasantness and VAS sensory intensity of pain (F1,39 ⫽ 5.07; P ⫽ .03) for pressure pain (with higher values in VAS intensity ratings). No significant differences between the 2 VAS ratings were detected for electrical stimuli (F1,39 ⫽ 0.36; P ⫽ .55).
Facial Expression The analyses of variance for repeated measurements showed that frequency (pressure: F4,156 ⫽ 10.49; P ⫽ .001; electrical: F2,78 ⫽ 159.37; P ⬍ .001) and intensity (pressure: F4,156 ⫽ 16.51; P ⬍ .001; electrical: F2,78 ⫽ 56.38; P ⬍ .001) of facial expression increased significantly over stimulus intensities both for pressure pain (Fig 3A) and for electrical pain (Fig 3B). For pressure pain, pairwise comparisons showed no significant differences in all comparisons between stimulus intensities below 4 kg (all P values ⬎ .4). In contrast, stimulus intensities of 4 and 5 kg elicited significantly stronger facial expressions
Table 4. Multiple Correlation Coefficients (of the Multiple Regression Analyses), First Type of Analysis FACIAL EXPRESSION (FREQUENCY & INTENSITY) (PREDICTOR) Pressure pain Intercept Slope Electrical pain Intercept Slope
†F2, 39 ⫽ 3.33; P ⫽ .047.
Correlation Between Self-Report and Facial Expression First Type of Analysis The basic statistics of intercepts and slopes are given in Table 3. In regard to slope values (of the stimulus-response relationships), we found significant multiple correlation coefficients between facial expression (frequency and intensity) and pain unpleasantness as well as sensory intensity of pain, when using pressure pain (Table 4). No significant correlations between slope values of the stimulus-response relationships for facial expression and for self-report were found by using electrical stimuli. We did not find any significant correlations for pressure pain or for electrical pain between intercepts of the stimulusresponse relationship for facial expression and intercepts of the stimulus-response relationship for self-report. Accordingly, intercepts of facial expression (frequency, intensity) could not predict intercepts of self-report ratings. The correlation computation including slope parameters always led to higher correlation coefficients, although not always significant, in comparison to inter-
SELF-REPORT (CRITERION) VAS SENSORY INTENSITY OF PAIN (INTERCEPT/SLOPE)
VAS PAIN UNPLEASANTNESS (INTERCEPT/SLOPE)
.211 .414*
.134 .391†
.171 .271
.207 .290
Abbreviation: VAS, Visual Analog Scale. *F2, 39 ⫽ 3.83; P ⫽ .031.
(frequency, intensity) than stimulus intensities below 4 kg (all P values ⬍ .005). There also was a significant increase in facial expressions (frequency, intensity) from 4 to 5 kg (all P values ⬍ .005). In regard to electrical pain, all pairwise comparisons (with the exception of the comparison between frequency above and frequency 5 mA above reflex threshold) showed significant results (all P values ⬍ .005).
Table 5. Multiple Correlation Coefficients (of the Multiple Regression Analyses), Second Type of Analysis FACIAL EXPRESSION (FREQUENCY & INTENSITY) (PREDICTOR) Pressure pain Electrical pain
SELF-REPORT (CRITERION) VAS SENSORY INTENSITY OF PAIN (INTERCEPT/SLOPE)
VAS PAIN UNPLEASANTNESS (INTERCEPT/SLOPE)
.096 .225
.127 .157
Abbreviation: VAS, Visual Analog Scale.
374 cept parameters by using both pressure and electrical pain (Table 4). Correlations between pain unpleasantness and facial expression did not yield higher correlation coefficients than did correlations between sensory intensity of pain and facial expression.
Second Type of Analysis No significant multiple correlation coefficients were detected (neither for pressure pain nor for electrical pain) when responses averaged across intensities were used. Mean scores of facial expression of pain (frequency/intensity) could not significantly predict mean scores of self-report ratings (Table 5).
Discussion The present study verified a substantial relationship between the individual abilities to report and facially express different levels of pressure pain. In statistical words, slopes of the stimulus-response relationships for self-report and facial expression were significantly correlated. This indicates that an increase of noxious stimulus intensity leads to a strong increase in self-report and facial expression in one individual, whereas it leads to a weak increase in both parameters in another one. Admittedly, the correlations were only statistically significant for experimental pressure pain and of moderate size. (In the case of electrical pain the correlations tended into similar directions but did not quite reach significance.) Nevertheless, in contrast to the vast majority of earlier reports, our data suggest that the verbal and facial response systems are linked. It mainly appears to be a matter of applying appropriate analysis strategies to find this linkage. Conversely to the findings regarding slope parameters, no significant correlations were obtained by using the intercepts of the stimulus-response relationships for selfreport and facial expression. The intercept can be considered to describe the activation threshold, representing the minimum level of stimulation required to elicit a specific pain response. Our results suggest that the thresholds of self-report and facial expression are independent of each other. This means that an individual who is reporting pain does not necessarily express this facially or vice versa. Prkachin and Craig22 described the facial expression of pain as a “late” signaling system, meaning that even though pain is being experienced, it might still not be facially expressed. However, our finding of no correlation between intercepts makes it likely that there are individuals who facially respond in a pain-specific manner to stimulus intensities that they do not rate as being painful. Taking together the findings on slope and intercept parameters, it appears that although the starting points in the verbal and facial response systems are unrelated, the individual degrees of reactivity in both systems during pain are related to each other. The latter only becomes obvious when applying different stimulus intensities for computation of stimulus-response functions. Mean scores of facial responses to pain (averaged
Self-Report and Facial Expression of Pain across stimulus intensities) appeared uncorrelated to mean scores of self-report ratings (averaged across stimulus intensities) when using pressure and electrical pain. Accordingly, individuals who are facially more expressive on average do not rate the stimuli as being more painful on average in comparison to those who show less facial expression of pain. As described in the introduction, disregarding intercept and slope of the stimulus-response relationship might result in missing the relationship between particular aspects of pain report and facial expression. Therefore, approaches that average responses across intensities or single stimulus approaches that do not allow for calculating intercept and slope might be misleading when investigating the association between self-report and facial expression of pain. These findings have several practical implications. First, the independence of starting points (activation thresholds) in the verbal and facial response systems suggests that the absence of a facial expression of pain cannot necessarily be interpreted as a sign that no pain is being experienced. Those engaged in pain management ought to be aware of that. Second, the correlation between slopes of the stimulus-response relationships for self-report and facial expression suggests that an increase/decrease in facial expression of pain is likely linked to an increase/decrease in subjective pain experience, thus yielding clear support for the utility of facial expression as an indicator of the subjective pain experience when intraindividual changes are considered. This might be of great importance in regard to pain assessment in individuals with limited ability in verbal communication. Observer should try to study the increase/decrease of facial responses across a variety of potential painful activities to gain more valid information on how the patient is facially encoding her/his pain experience. We only found moderate correlations (around r ⫽ 0.4) between the slopes of the stimulus-response relationships for self-report and facial responses. Among other factors, this might be due to the fact that the pain stimulus intensities, which were used for analysis, were too weak. Regarding pressure pain (Fig 3A), it becomes obvious that facial expression only started to increase with stimulus intensities from 4 kg on. Stimuli of lower intensities elicited very weak facial responses that were nonsystematically related to stimulus strength. It is quite possible that stronger stimulus intensities might have led to an even stronger correlation between slopes of the stimulus-response relationships for self-report and facial expression of pain. With electrical stimuli, we found even slightly weaker correlations between the slope values, although facial expression appeared to increase from the weakest stimulus on. A possible interpretation of this negative finding might be that pain induced by electrical current (with its sudden onset) elicits further emotional processes, such as startle reactions, that might be confounded with pain. As a consequence, facial expressions that are not entirely the result of pain but are blended with other emotions as well might occur, thus lowering the correlation between facial expression and self-report of pain.
ORIGINAL REPORT/Kunz et al
375
Comparisons between self-report and facial expression of pain induced by pressure and electrical current were a bit unbalanced because the stimuli selection (regarding number and intensity) differed substantially. Our analysis strategy was optimized in regard to pressure pain. We used this strategy a second time as a kind of first test of external validation of pain induced by electrical current, originally designed for assessing further nonverbal, physiological parameters of pain. Besides the type of analytic approach, the type of stimulus induction method, and the range of stimulus intensities, there are numerous additional factors that moderate the association between facial expression or other nonverbal pain behaviors and self-report of pain (eg, time of pain assessment,2 social modifiers,18 severity of experienced distress,24 number of nonverbal pain expressions being studied,15 pain site,15 duration of pain15). It remains to be clarified how these factors moderate the association between self-report and facial expression of pain when our analysis strategies are applied. It has often been assumed that facial expression is more closely related to the affective dimension of pain,14,17 but we could not detect a closer association to the unpleasantness dimension of pain (VAS ratings of unpleasantness). The VAS ratings of sensory intensity and unpleasantness of pain were highly interrelated (pressure pain, r ⫽ 0.76; electrical pain, r ⫽ 0.93), which probably accounts for the similar results found for sensory intensity and unpleasantness ratings. Nevertheless, using 2 scales to discriminate between the sensory intensity and affective dimensions of pain appears still advisable, given the significant differences found between VAS–sensory intensity and VAS-unpleasantness ratings when using pressure pain. The study has several limitations. First, a rather limited
number of pain stimuli levels were used to investigate the association between relationships of stimulus intensity and intensity of self-report/facial expression of pain. A less limited number of pain stimuli levels would probably ensure a more reliable and complete depiction of the stimulus response relationships. Future studies should apply more numerous pain stimuli levels when using psychophysical relationships between stimulus intensities and pain responses to compute correlations between self-report and facial expression. Second, video segments were not analyzed in a randomized order but in their order of assessment. Therefore, it is possible that the FACS codings could have been biased by expectancy effects. Third, we decided to always start with the application of pressure pain before electrical pain and did not use a random order. This was done because some individuals experience even very weak intensities of electrical stimulation as being frightening, and our procedure requires moderate intensities well above pain thresholds. This fixed order allowed us to always start with the less stressful part. In conclusion, the facial expression of pain appears to mirror self-report ratings, when their increases over a range of increasing stimulus intensities are considered in parallel (although the moderate magnitude of correlations suggests that both remain at least partially independent and under the control of third factors). In contrast, the starting points for activation of the 2 response systems appear to be independent of each other. This suggests that the facial expression of pain can be used as an estimate of the subjective pain when the reactivity over different levels of pain states or pain stimuli is analyzed. Single stimulus approaches as well as approaches averaging responses across different stimulus intensities can be misleading.
References
8. Hadjistavropoulos T, Craig KD, Martin N, Hadjistavropoulos HD, McMurtry B: Toward a research outcome measure of pain in frail elderly in chronic care. Pain Clinic 10:71-79, 1997
1. Craig KD, Hyde SA, Patrick CJ: Genuine, suppressed and faked facial behavior during exacerbation of chronic low back pain. Pain 46:161-171, 1991 2. Craig KD, Patrick CJ: Facial expression during induced pain. J Pers Soc Psychol 48:1080-1091, 1985 3. Craig KD, Prkachin KM, Grunau RVE: The facial expression of pain, in Turk D, Melzack R (eds): Handbook of Pain Assessment. New York, NY, Guilford Press, 2001, pp 153169 4. Ekman PE, Friesen WV: Facial Action Coding System. Palo Alto, CA, Consulting Psychologists Press, 1978 5. Fischer AA: Reliability of the pressure algometer as a measure of myofascial trigger point sensitivity. Pain 28:411414, 1987 6. France CR, France JL, al’Absi M, Ring C, McIntyre D: Catastrophizing is related to pain ratings, but not nociceptive flexion reflex threshold. Pain 99:459-463, 2002 7. Hadjistavropoulos HD, Craig KD, Hadjistavropoulos T, Poole GD: Subjective judgments of deception in pain expression: Accuracy and errors. Pain 65:251-258, 1996
9. Hadjistavropoulos T, LaChapelle DL, Hadjistavropoulos HD, Green S, Asmundson GJG: Using facial expressions to assess musculoskeletal pain in older persons. Eur J Pain 6:179-187, 2002 10. Hadjistavropoulos T, LaChapelle DL, MacLeod FK, Hale C, O’Rourke N, Craig KD: Cognitive functioning and pain reactions in hospitalized elders. Pain Res Manage 3:145-151, 1998 11. Hadjistavropoulos T, LaChapelle DL, MacLeod FK, Snider B, Craig KD: Measuring movement-exacerbated pain in cognitively impaired frail elders. Clin J Pain 16:5463, 2000 12. Hale CJ, Hadjistavropoulos T: Emotional components of pain. Pain Res Manage 2:217-225, 1997 13. Hill ML, Craig KD: Detecting deception in pain expressions: The structure of genuine and deceptive facial displays. Pain 98:135-144, 2002 14. Kappesser J, Williams AC: Pain and negative emotions in
376 the face: Judgements by health care professionals. Pain 99:197-206, 2002 15. Labus JS, Keefe FJ, Jensen MP: Self-reports of pain intensity and direct observations of pain behavior: When are they correlated? Pain 102:109-124, 2003 16. LaChapelle DL, Hadjistavropoulos T, Craig KD: Pain measurement in persons with intellectual disabilities. Clin J Pain 15:13-23, 1999 17. LeResche L, Dworkin SF: Facial expressions of pain and emotions in chronic TMD patients. Pain 35:71-78, 1988 18. Patrick CJ, Craig KD, Prkachin KM: Observer judgements of acute pain: Facial action determinants. J Pers Soc Psychol 50:1291-1298, 1986 19. Price DD, McGrath PA, Rafii A, Buckingham B: The validation of visual analogue scales as ratio scale measures for chronic and experimental pain. Pain 17:45-56, 1983
Self-Report and Facial Expression of Pain 20. Prkachin KM: The consistency of facial expression of pain: A comparison across modalities. Pain 51:297-306, 1992 21. Prkachin KM, Berzins S, Mercer SR: Encoding and decoding of pain expressions: A judgement study. Pain 58:253-259, 1994 22. Prkachin KM, Craig KD: Expressing pain: The communication and interpretation of facial pain signals. J Nonv Behav19:191-205, 1994 23. Prkachin KM, Currie NA, Craig KD: Judging nonverbal expressions of pain. Can J Behav Sci 15:43-73, 1983 24. Prkachin KM, Mercer SR: Pain expression in patients with shoulder pathology: Validity, properties and relationship to sickness impact. Pain 39:257-265, 1989 25. Stevens SS: The psychophysics of sensory functions. Am Scientist 48:226-253, 1960