Underpredicting Pain: An Experimental Investigation Into the Benefits and Risks

Kaya J. Peerdeman; Andrew L. Geers; Delia Della Porta; Dieuwke S. Veldhuijzen; Irving Kirsch

Disclosures

Pain. 2021;162(7):2024-2035. 

In This Article

Study 2

To better understand the effects of strongly underpredicting pain, we replicated study 1 while adding a third experimental condition in which a medium underprediction of pain was given (Medium underprediction condition). This comparison allowed us, among others, to assess possibly reduced assimilation of experienced pain intensity upon strong underprediction (rather than contrast effects) and to compare the effects on trust. Furthermore, to increase the chances of contrast effects and to address new regulations, several changes were made in the experimental procedure and manipulations (eg, increased intensity of heat pain stimuli).

Methods Study 2

Participants. As in study 1, we included 41 healthy participants per condition, resulting in a total sample size of 123 participants. Eligibility criteria were the same as in study 1.

Procedure. The experiment was conducted at the same location as study 1, from March to October 2018. The study protocol was approved by Leiden University's Psychology Research Ethics Committee (CEP18-0306_133) and preregistered at the Netherlands Trial Register (NL7033).

The procedure was the same as for study 1, with several exceptions. Participants now filled in the screening and other online questionnaires in the laboratory, in response to updates of European privacy regulations. Furthermore, the target temperature was set to now correspond to a high pain intensity rating of 7 to max 8 on the pain intensity NRS (M = 48.4°C, SD = 1.5°C) to increase the discrepancy between the predicted and actual stimulus intensity and thereby the chance of contrast effects. Also, the repeated presentations of the low and high stimulus were replaced by a single presentation of a high reference stimulus to simplify the comparison participants had to make and to minimize premanipulation exposure to correct instructions about stimulus intensity as this might also influence the likelihood of contrast effects. The thermode was now moved to a novel spot on the arm before the reference stimulus, instead of after the reference stimulus as was the case in study 1. Moreover, we now obtained ratings of expected pain intensity, certainty of pain expectation, fear of pain, and pain unpleasantness scores also for the reference stimulus to serve as comparators in our analyses. Medoc software issues were still present for many participants, and stimuli were at peak for 4.0 to 10.4 seconds (M = 5.6, SD = 2.1 seconds). See Figure 3 for an overview of the procedure.

Figure 3.

Overview of the procedure with all measures taken in study 2. NRS, numerical rating scale; VS, verbal suggestion.

Verbal Suggestions of Pain Intensity. The condition-specific verbal suggestions were slightly rephrased. Participants in the Strong underprediction condition were told "You'll now get 4 low stimuli. These are a lot lower than the reference stimulus. Based on your responses before, it is very likely that you will experience them as nonpainful."

Participants in the Medium underprediction condition were told: "You'll now get 4 medium stimuli. These are noticeably lower than the reference stimulus. Based on your responses before, it is very likely that you will experience them as only mildly painful.

"Participants in the Correct prediction condition were told: "You'll now get 4 high stimuli. These are equally high as the reference stimulus. Based on your responses before, it is very likely that you will experience them as highly painful."

The condition-specific suggestion was repeated before giving 4 additional trials: "You will now again get 4 low/medium/high stimuli. It is very likely that you'll experience them as nonpainful/only mildly painful/highly painful."

Measures. The same pain rating scales, autonomic measures, and questionnaires were used as in study 1. In addition, participants rated how competent, sympathetic, and honest they perceived the experimenters to be on the 0 to 10 visual analogue scale. The questionnaires had acceptable to good internal consistency (LOT-R α = 0.77; Generalized Trust Scale α = 0.78; PCS α = 0.89; SMS-R ability to modify self-presentation α = 0.81; SMS-R sensitivity to others α = 0.78; STAI-T α = 0.91; baseline STAI-Ss α = 0.80; postintervention STAI-Ss α = 0.80).

Statistical Analyses

The same analysis strategy was used as in study 1, with 2 deviations. First, because we now compared 3 instead of 2 conditions, omnibus AN(C)OVAs in which condition with 3 levels was an IV were followed up by AN(C)OVAs comparing the Strong underprediction condition with (1) the Correct prediction condition and (2) the Medium underprediction condition. For readability, only results of pairwise comparisons are reported but only in cases of significant omnibus effects. Second, we now included the corresponding ratings for the reference stimulus as a covariate for analyses of all pain outcomes.

Results Study 2

Data Availability. Table Supplemental digital content 2 (available at http://links.lww.com/PAIN/B269) presents the descriptives (M ± SD) per condition of all reported measures. The data, analysis scripts, and results files are available through https://osf.io/24mu5/ (doi: 10.17605/OSF.IO/24MU5).

Participants. Four participants were excluded during the screening in the laboratory. All 123 eligible participants completed participation (50.4% female; M age = 22.1, SD = 2.9; 38.2% Dutch nationality; median baseline pain = 0.0, interquartile range = 0.0).

Effects of Verbal Suggestions on Pain Intensity. Pairwise ANCOVAs indicated that the strong underprediction of pain led to significantly lower pain intensity of the first postsuggestion trial than the correct prediction, F(1, 79) = 48.21, P < 0.001, = 0.38, although it did not significantly differ from the medium underprediction, F(1, 79) = 3.56, P = 0.063, = 0.04 (Figure 4A). Sensitivity analyses with stratification variable sex or stimulus peak duration as an additional covariate and without covariates yielded the same conclusions for the comparison of the Strong underprediction and Correct prediction conditions but indicated significantly lower pain intensity after strong vs medium underprediction when stimulus peak duration was included as a covariate, F(1, 72) = 4.54, P = 0.037, = 0.06, or when no covariates were included, F(1, 80) = 4.02, P = 0.048, = 0.05.

Figure 4.

(A) Experienced pain intensity ratings (M ± SE) upon the reference stimulus and each of the 8 postsuggestion trials, and (B) Expected pain intensity ratings (M ± SE) before the reference stimulus and each of the 8 postsuggestion trials, per condition, in study 2. ref, reference stimulus; VS, verbal suggestion.

An exploration of pain intensity during all postsuggestion trials indicated main effects of condition, F(1, 79) = 25.22, P < 0.001, = 0.20, and trial, F(3.34, 264.15) = 5.33, P < 0.001, = 0.01, for the comparison of the Strong underprediction and Correct prediction conditions, reflecting continuously lower pain in the Strong underprediction condition as compared to the Correct prediction condition and slight sensitization in these conditions. For the comparison of the Strong and Medium underprediction conditions, there was an interaction effect, F(2.96, 231.14) = 3.13, P = 0.027, = 0.01, but no clearly distinct patterns were observed.

Expected Pain Intensity. The verbal suggestions successfully manipulated expected pain intensity. Pairwise ANCOVAs indicated that the strong underprediction led participants to expect significantly less pain for the first postsuggestion trial than the correct prediction, F(1, 79) = 243.55, P < 0.001, = 0.76, and the medium underprediction, F(1, 79) = 52.62, P < 0.001, = 0.40 (Figure 4B).

Regarding trial 2, pairwise ANCOVAs indicated that participants in the Strong underprediction condition, who had experienced a violation of their initial expectation of low pain, also expected the following trial to be less painful than participants who received a correct prediction, F(1, 79) = 57.00, P < 0.001, = 0.42, but not when compared with a medium underprediction.

An exploration of expected pain intensity during all trials indicated a main effect of condition, F(1, 79) = 72.83, P < 0.001, = 0.37, and an interaction effect, F(3.56, 281.04) = 33.66, P < 0.001, = 0.14, for the comparison of the Strong underprediction and Correct prediction conditions. Participants in the Strong underprediction condition consistently expected less pain than participants in the Correct prediction condition, particularly directly upon the suggestions. For the comparison of the Strong and Medium underprediction conditions, there was an interaction effect, F(4.12, 321.31) = 11.06, P < 0.001, = 0.06, reflecting differences between the conditions directly upon the suggestions alone.

Other Pain Outcomes

Certainty of Pain Expectation. A pairwise ANCOVA comparing certainty of pain expectations for all postsuggestion trials in the Strong underprediction and Correct prediction conditions indicated main effects of condition, F(1, 79) = 26.71, P < 0.001, = 0.17, and trial, F(4.73, 373.61) = 3.88, P = 0.002, = 0.02, and an interaction effect, F(4.73, 373.61) = 2.42, P = 0.038, = 0.01 (Figure Supplemental digital content 9, available at http://links.lww.com/PAIN/B269). Participants in the Strong underprediction condition were less certain than participants in the Correct prediction condition of their expectations across trials, particularly upon repetition of the verbal suggestion. For the comparison of the Strong and Medium underprediction conditions, a main effect of stimulus, F(5.18, 404.27) = 6.00, P < 0.001, = 0.04, and an interaction effect were observed, F(5.18, 404.27) = 2.58, P = 0.024, = 0.01, reflecting lower certainty only upon repetition of the strong underprediction suggestion.

Fear of Upcoming Pain. A pairwise ANCOVA comparing fear of upcoming pain for all postsuggestion trials in the Strong underprediction and Correct prediction conditions indicated a main effect of condition, F(1, 79) = 14.14, P < 0.001, = 0.10, and an interaction effect, F(4.05, 319.56) = 13.07, P < 0.001, = 0.06 (Figure Supplemental digital content 10, available at http://links.lww.com/PAIN/B269). Participants in the Strong underprediction condition had less fear of pain for those trials closely following the suggestions than participants in the Correct prediction condition. For the comparison of the Strong and Medium underprediction conditions, an interaction effect was observed, F(3.63, 283.19) = 2.93, P = 0.025, = 0.01, reflecting lower fear upon strong underprediction for the first stimulus alone.

Pain Unpleasantness. A pairwise ANCOVA comparing pain unpleasantness for all postsuggestion trials in the Strong underprediction and Correct prediction conditions indicated main effects of condition, F(1, 79) = 7.42, P = 0.008, = 0.06, and trial, F(4.14, 327.22) = 2.75, P = 0.027, = 0.01, reflecting continuously lower pain unpleasantness in the Strong underprediction condition as compared to the Correct prediction condition, and a slight increase in pain unpleasantness over time in both groups (Figure Supplemental digital content 11, available at http://links.lww.com/PAIN/B269). For the comparison of the Strong and Medium underprediction conditions, an interaction effect was observed, F(3.78, 295.02) = 3.04, P = 0.020, = 0.01, reflecting slight increases in pain unpleasantness over trials in the Strong underprediction condition alone.

Autonomic Responses

Heart Rate. A pairwise ANCOVA comparing heart rate for all postsuggestion trials in the Strong underprediction and Correct prediction conditions indicated an interaction effect, F(7, 483) = 5.75, P < 0.001, = 0.02 (Figure Supplemental digital content 12, available at http://links.lww.com/PAIN/B269), reflecting that participants in the Strong underprediction condition had a lower heart rate than participants in the Correct prediction condition during the first trials. For the comparison of the Strong and Medium underprediction conditions, a main effect of condition was observed, F(1, 74) = 6.29, P = 0.014, = 0.06, reflecting continuously lower heart rate upon the strong underprediction as compared to the medium underprediction.

Skin Conductance Level. A pairwise ANCOVA comparing skin conductance level for all postsuggestion trials in the Strong underprediction and Correct prediction conditions indicated an interaction effect, F(3.11, 217.97) = 10.31, P < 0.001, = 0.02 (Figure Supplemental digital content 13, available at http://links.lww.com/PAIN/B269), reflecting that participants in the Strong underprediction condition had a lower skin conductance level than participants in the Correct prediction condition during the first trial alone. For the comparison of the Strong and Medium underprediction conditions, no significant effects were observed.

Skin Conductance Response. Pairwise ANCOVAs comparing the amplitude of the first skin conductance response for all postsuggestion trials indicated main effects of condition for the comparisons of the Strong underprediction with both the Correct prediction condition and the Medium underprediction condition, (F(1, 71) = 5.46, P = 0.022, = 0.03 and F(1, 73) = 10.57, P = 0.002, = 0.05, respectively), reflecting continuously higher responses in the Strong underprediction condition than in the other conditions (Figure Supplemental digital content 14, available at http://links.lww.com/PAIN/B269). Sensitivity analyses using the extended latency windows confirmed the main effect of condition for the comparison of the Strong and Medium underprediction conditions alone, F(1, 74) = 11.42, P = 0.001, = 0.09, while also showing a main effect of trial, F(5.7, 421.49) = 2.38, P = 0.031, = 0.01, reflecting a slight decrease over time in these conditions (Figure Supplemental digital content 15, available at http://links.lww.com/PAIN/B269).

Psychological Outcomes

Descriptives (Table Supplemental digital content 2, available at http://links.lww.com/PAIN/B269) and pairwise ANOVAs indicated that participants were aware of the condition-dependent instruction–experience (mis)match (strong underprediction vs correct prediction: F(1, 80) = 34.64, P < 0.001, = 0.30; strong vs medium underprediction: F(1, 80) = 13.52, P < 0.001, = 0.14). This was also largely supported by the perceived (mis)matches of expected and experienced pain, although there was no significant difference between the strong underprediction and correct prediction (strong underprediction vs correct prediction: F(1, 80) = 1.33, P = 0.252, = 0.02; strong vs medium underprediction: F(1, 80) = 6.78, P = 0.011, = 0.08).

Notably, pairwise ANOVAs indicated that trustworthiness of the experimenter who provided the verbal suggestion was rated significantly lower upon a strong underprediction than upon a correct prediction, F(1, 80) = 14.96, P < 0.001, = 0.16, but not as compared to a medium underprediction. In addition, pairwise ANOVAs indicated that honesty of this experimenter was rated lower upon a strong underprediction than upon either a correct prediction, F(1, 80) = 32.24, P < 0.001, = 0.29, or a medium underprediction, F(1, 80) = 4.86, P = 0.030, = 0.06.

Finally, pairwise ANOVAs indicated that participants rated a higher sense of owing it to the experimenter to report less pain in the Strong underprediction condition as compared to the Correct prediction condition, F(1, 80) = 11.44, P = 0.001, = 0.13. No significantly differential effect of the suggestions on state anxiety was observed, nor were there significant between-group differences in the other measures (eg, observation of discrepancy between reference and postsuggestion trials, trust in the other experimenter, perception of experimenters as competent and sympathetic, and other response bias measures).

Moderation by Psychological Characteristics

Comparisons of the Strong underprediction and Correct prediction conditions indicated that the effect of the suggestions on experienced pain intensity during the first postsuggestion trial was significantly moderated by dispositional optimism, b = −0.17, SE = 0.06, t = −2.71, P = 0.008, reflecting that participants who reported lower optimism showed larger pain reductions upon the strong underprediction than those who reported higher optimism (pain intensity Strong underprediction condition at lower vs upper quartile LOT-R = 4.5 vs 5.3). In line, a significant moderation by trait pain catastrophizing, b = 0.07, SE = 0.03, t = 2.28, P = 0.025, reflected that participants who reported higher catastrophizing showed larger pain reductions upon the strong underprediction than those who reported low catastrophizing (pain intensity Strong underprediction condition at lower vs upper quartile PCS = 5.3 vs 4.6). Furthermore, but in contrast to the finding of study 1, a significant moderation by state anxiety, b = 0.05, SE = 0.03, t = 2.05, P = 0.043, reflected that participants who reported higher state anxiety showed larger pain reductions upon the strong underprediction than those who reported low state anxiety (pain intensity Strong underprediction condition at lower vs upper quartile STAI-Ss = 5.2 vs 4.7). No other moderations were observed.

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....