Journal Logo

Research Paper

Placebo effect in children: the role of expectation and learning

Gniß, Silke; Kappesser, Judith; Hermann, Christiane*

Author Information
doi: 10.1097/j.pain.0000000000001811

1. Introduction

Classical conditioning and expectations are well-known underlying mechanisms of placebo hypoalgesia in adults.10,11,13 Both mechanisms are not distinct but interlinked assuming that learning experiences evoke expectations which in turn activate the placebo network.3,13,26,46 Most studies implicitly or explicitly combine conditioning and expectations when investigating placebo hypoalgesia.1,5,14,15,27,37 To examine the unique effect of both, it is necessary to disentangle them. However, there are only a few studies which have investigated conditioned placebo effect without positive expectations and/or positive expectations without conditioning.1,27,53 Even fewer have tested for differences between both mechanisms.53 Voudouris et al.52 found “conditioning” to induce a stronger placebo effect than “expectancy” and no difference in the extent of the placebo effect between “conditioning” and “conditioning and expectancy”.

With regard to children, no study has investigated the differential impact of conditioning and expectations. This is not surprising since there are only 3 studies attempting to experimentally induce a placebo hypoalgesia in children.21,30,55 The first study struggled with floor effects.21 The second study found a much stronger expectation-induced placebo hypoalgesia in 6- to 9-year-old children when compared with adults.29,30 The third study found no differences between 10- to 15-year-old children and adults, when the placebo effect was induced by conditioning and expectations.55 The inconsistent results of the 2 latter studies may be due to methodological differences (participants' age range, method to induce placebo hypoalgesia). Taken together, the limited evidence suggests that there is a placebo effect in children. However, it remains unclear how strongly it is affected by conditioning and/or expectations and whether there are age-related effects. The strength of the placebo effect may decrease with increasing age because of developmental processes such as the reductions of suggestibility2,7,30,38,45 or prefrontal maturation.6,20,29

Particularly in children, the person raising the expectation may impact on the strength of the placebo effect. Such persons, “informants,” are usually physicians or parents. Both can influence pain-relevant outcomes by their appearance, communication, or general behavior.9,19,23,25,43,47,49 With regard to parents, their education about pain management and presence during painful procedures is strongly recommended; yet, confidence in the empirical evidence because of its quality is regarded as low.40,48

Little is known about the role of learning in the development of conditioned placebo hypoalgesia. Although the experience of a meaningful pain reduction is a prerequisite for the development of a conditioned placebo hypoalgesia, the extent of this experience is possibly associated with the extent of the placebo hypoalgesia.

Placebo hypoalgesia involves not only changes in explicit pain behavior but also in peripheral physiological reactions.35 Heart rate and electrodermal activity were generally shown to increase during pain experiences.28,32 Accordingly, this increase was reduced in most of the few studies investigating physiological changes in the placebo context.41 To the best of our knowledge, no study has investigated the association between placebo hypoalgesia and peripheral physiological changes in children yet.

In this study, we aim to (1) disentangle expectation and conditioning as explanatory mechanisms of placebo hypoalgesia, (2) examine age-related differences in self-report and psychophysiological correlates associated with the placebo effect, (3) investigate the impact of the informant on expectations, and (4) examine the impact of the experienced temperature reduction during conditioning to better understand the development of a conditioned placebo hypoalgesia.

2. Methods

2.1. Participants

Participants were recruited by flyers at local primary and secondary schools (children and adolescents) and by mass mails at the local university (adults). With regard to the aims of the study, a cover story was used for all participants (ie, adults, children, adolescents, and their parents) in which the purpose of the study was described as a scientific comparison of 2 analgesic creams (see 2.2).

Before the experiment, potential participants (in case of children and adolescents their parents) were screened telephonically. Potential participants were excluded when they met one or more of the following criteria: (1) chronic diseases (incl. chronic pain), (2) developmental disorders, (3) current psychological problems (assessed by parts of structured clinical interviews; adult participants: affective disorders, substance-related, and addictive disorders by Mini-DIPS34; child and adolescent participants: affective disorders, anxiety disorders, conduct disorder, and attention-deficit hyperactivity disorder by parent version of Children-DIPS44), (4) insufficient knowledge of the German language, (5) skin disease on the forearms, and (6) acute pain or analgesic medication on the day of the experiment. The final sample size consisted of 172 children (52.33% female) between 6 and 17 years (M = 11.62 ± 3.21 years) and 32 adults (50% female) between the age of 19 and 29 years (M = 21 ± 2.06 years).

Financial reimbursement was 10 Euros per hour for children and 8 Euros per hour for adults. Instead of financial reimbursement, adult student participants could also receive course credit for their participation. All adult participants were students. Since the experiment took part in a laboratory on university campus, they were paid the usual reimbursement amount per hour. Children were paid an additional 2 Euros per hour to cover the additional time needed to come to the university campus in the company of one of their parents.

The study was approved by the local ethics committee of the Department for Psychology and Sports Science of the Justus-Liebig-University Giessen. Consistent with the cover story, all participants and their parents gave consent for participation in a research study on the comparison of 2 well-established analgesic creams. Before obtaining consent, all participants and their parents were told that participants could withdraw at any time during the experiment without any disadvantages. For the children, information about the experiment was provided using age-appropriate wording. After the experiment, all participants and their parents were debriefed in detail about the deception and its necessity. They were informed about the true aims of the study and explained the study's relevance. After having been debriefed, all participants and children's parents gave their informed consent for use of their (child's) data.

2.2. Experimental design

2.2.1. Age

Since the aim of this experiment was to investigate age-related differences in the placebo effect, we aimed for a broad age range in children and adolescents (6-17 years). The lower age criterion of 6 years was set because the pain self-report scale used in this study is recommended for ages of 6 years onwards.51 To allow for a direct comparison between children, adolescents, and adults, we also included a group of adults (≥18 years) in the present sample. To determine age groups, Piagetian conceptualizations were taken into account (children of 7-12 years: concrete operational stage; adolescents from 13 years onwards: formal operational stage) as were age groups used in previous studies (younger children: 6-9 years and older children 10-15 years).30,55 To reduce within-group variance in the child and adolescent age groups, and to discriminate better between middle childhood, early, and late adolescence, 3 age groups with an age range of 4 years were defined: younger children (6-9 years), older children (10-13 years), adolescents (14-17 years), and, additionally, adults (>18 years).

2.2.2. Placebo induction

To obscure the fact that we were investigating placebo effects, a cover story was used for all participants including their parents. This cover story implied that participants' task was to compare each of 2 “analgesic” creams (NODOLOR and ELANICIN) to a nonanalgesic control cream. The comparison was not a direct comparison (NODOLOR vs ELANICIN), but each cream was compared successively to a control cream: NODOLOR in the expectancy-based paradigm and ELANICIN in the classical conditioning paradigm. The control cream was introduced as a “nonanalgesic moisturizer.” In fact, all 3 creams were SUPER-VISC High Viscosity Electrolyte-Gel (EASYCAP GmbH, Herrsching, Germany) refilled in 3 different tubes. The order of the expectancy-based and the classical conditioning paradigm was randomized across participants as were the sides of the arm (left/right) on which each of the creams was applied to.

2.2.3. Expectancy-based paradigm and informants

In the expectancy-based paradigm, informants introduced and applied the analgesic cream (NODOLOR) and the nonanalgesic moisturizer. To avoid gender effects of the informants, the creams were introduced and applied either by the children's mothers or by 1 of 4 female confederates being introduced as a physician and wearing a white coat. Participating children were randomly assigned to 1 of these 2 conditions (“mother” or “physician”). If the mother could not attend the experiment, the informant was always chosen to be “physician” (N = 26). For adult participants, the application was always conducted by the “physician”.

“Physicians” were trained to use standardized instructions when introducing and applying the cream. All mothers received a leaflet before the beginning of the expectancy-based paradigm in which the relevant information about the cream was summarised. Mothers were asked to explain the efficacy of the cream to their children in their own words but to try to cover all the information given in the leaflet. Investigators were present during the explanation and application of the cream and provided assistance whenever necessary. After the expectancy-based paradigm, investigators also rated mothers' performances on an 11-point numerical rating scale (NRS) ranging from 0 “not at all accurate and reliable” to 10 “very accurate and reliable.” On average, mothers' performance was rated as 5 (mean: 4.97, SE: 0.33; median: 5.25; mode: 7).

Instructions of “physicians” were identical to information given to mothers. Both types of informants told participants that NODOLOR was a highly effective analgesic cream and that the moisturizer would not have an analgesic effect. Informants also suggested how participants might feel the effectiveness of NODOLOR (“Perhaps you feel already that your arm is getting numb or cold. Sometimes it starts to prickle. Some participants feel the effect not until the heat pain is applied. In any case the pain will hurt less on this arm with the NODOLOR cream!”). The cream NODOLOR was applied on one forearm and the moisturizer on the other. It was randomly balanced on which arm (left or right) the cream NODOLOR was applied and on which arm the application of heat pain started.

2.2.4. Classical conditioning paradigm

In the classical conditioning paradigm, the experimenters told participants that one of the creams administered would be an “analgesic cream” (ELANICIN) and the other one a moisturizer without any analgesic effect. To avoid expectancy effects, participants were not told which cream was applied on which forearm and both creams were presented in identical white tubes. Participants were asked to concentrate on what they perceived on their forearms. Furthermore, they had to tell the investigator at the end of the paradigm on which forearm they thought the “analgesic” cream had been administered. To simulate a pain alleviating effect, temperature was reduced (−4.0°C) on one randomly chosen forearm during the conditioning phase (see “trial structure”).

2.2.5. Pain stimuli

Painful phasic heat stimuli were applied with a 3*3-cm Peltier element-based advanced thermal stimulator thermode (PATHWAY Pain & Sensory Evaluation System V 3.5 advanced thermal stimulator/CHEPS Combi-System, Medoc Ltd, Ramat Yishai, Israel) on both forearms. To determine the experimental temperature, (1) the individual heat pain threshold was established by applying the method of limits (baseline temperature: 32°C; acceleration rate: 1°C/s; deceleration rate: 8°C/s). Eight heat pain thresholds were measured on each arm, and the individual heat pain threshold was calculated as the mean of the last 10 thresholds (5 of each arm). (2) Approximately 1.5°C were added to the individual heat pain threshold. (3) Painful stimuli of this temperature had to be rated as at least “medium painful,” defined as a minimum of 30/100 on a visual analogue scale (VAS), by participants. If this criterion was not met, the temperature was increased by 1.5°C as long as the stimulus was rated as at least 30/100 on a VAS.

This experimental temperature lasted for 0.7 seconds (baseline temperature: 32°C; acceleration/deceleration rate: 8°C/second). Mean experimental temperature was 44.0 ± 3.5°C for the children and 47.5 ± 2.8°C for the adults. For safety and ethical reasons, a maximal temperature of 49°C (children) and 51°C (adults) was defined which must not be exceeded.

2.2.6. Trial structure

The experiment consisted of 5 phases: classical conditioning paradigm with (1) conditioning baseline phase, (2) conditioning phase, and (3) conditioning test phase and expectancy-based paradigm with (4) expectancy baseline phase and (5) expectancy test phase (Fig. 1). With the exception of the conditioning phase, temperature was identical on both forearms during all phases.

Figure 1.
Figure 1.:
Experimental design.

Figure 2 displays a detailed trial sequence with rating and stimulation times. Experimental control was realized with Presentation 16.3 (Neurobehavioral Systems, Inc, Berkeley, ON, Canada).

Figure 2.
Figure 2.:
Trial sequence between 2 pain stimuli (PS).

2.3. Outcome measures

2.3.1. Subjective pain intensity ratings

Visual analogue scales (VASs) were chosen as they have been validated for use in children from the age of 6 years as well as in adults.51 Participants rated their pain experience after each painful stimulus on a 101-point VAS ranging from 0 “no pain” to 100 “worst pain imaginable.” Participants had maximally 30 seconds for each pain rating. Rating pain on the VAS was practiced with all participants before the experiment started.

2.3.2. Subjective physician plausibility rating of adult participants

Plausibility of physicians was rated by adult participants on an 11-point NRS ranging from 0 “not at all plausible” to 10 “very plausible.”

2.3.3. Electrocardiogram (ECG)

The ECG was continuously recorded with 2 pregelled Ag/AgCl disc surface electrodes (Ø 10 mm; Megro, Wesel, Germany). After cleaning the recording sites with alcohol, the electrodes were placed on the right clavicle and the lowest rib on the left. The grounding electrode was placed on the left clavicle. Owing to technical problems, the recording was done either with the BrainVision Analyzer 2 Software (sample rate: 1000 Hz; Brain Products, Munich, Germany) or with the Portable Biosignal Recorder VARIOPORT-B and VARIOGRAPH Version Software 4.79 (sample rate: 1024 Hz; BECKER MEDITEC, Karlsruhe, Germany).

R-Peaks were offline detected with a customized MATLAB program (MathWorks, Inc, Natick, MA). R-Peak detection was manually checked, and data were excluded, when R-Peaks could not be identified due to artefacts. Mean interbeat intervals (IBIs) were calculated for time intervals of 500 ms starting 2 seconds before stimulus onset and ending 8 seconds after stimulus onset. Three parameters were calculated: (1) baseline (defined as mean IBI during second 2 before stimulus onset), (2) maximum heart rate deceleration (defined as difference between baseline and the maximum IBI between 1.5 and 4.5 seconds after stimulus onset), and (3) maximum heart rate acceleration (defined as difference between baseline and the minimum IBI between 4 and 7 seconds after stimulus onset). Changes in heart rate responses were determined by taking into account findings on heart rate deceleration and acceleration during aversive conditioning.31 The time windows for deceleration/acceleration were determined based on inspection of the grand average of the overall IBI changes.39

2.3.4. Electrodermal activity (EDA)

The electrodermal activity was measured either with BrainVision Analyzer 2 Software (sample rate: 1000 Hz; Brain Products) or with the Portable Biosignal Recorder VARIOPORT-B and VARIOGRAPH Version Software 4.79. For recording, two 10-mm Ag/AgCl disc surface electrodes (European Headquarters, GE Medical Systems Information Technologies GmbH, Freiburg, Germany) filled with TD-246 Isotonic Electrolyte Paste (EASYCAP GmbH) were fixed at the thenar and hypothenar of the left hand which had been cleansed with water. Data were 1 Hz low pass filtered and screened for artefacts. Two parameters were determined: (1) baseline (defined as mean skin conductance level during second 2 before stimulus onset) and (2) skin conductance response (defined as difference between baseline and the maximal amplitude during 1 and 6 seconds after stimulus onset) which was range corrected following the procedure of Lykken.33

2.4. Procedure

The experiment took place in a psychophysiological laboratory at the local university. After welcoming participants, they were guided into the experimental room and familiarized with the heat pain device. Children, their accompanying parents, and adults gave informed consent for participation. Participants were seated comfortably in an armchair, and the electrodes for the peripheral physiological recordings were applied. After preparation had been finished, participants' parents, if present, were accompanied to the waiting room.

The experiment was run from an adjacent room. Experimenters only joined participants in the experimental room when giving standardized instructions or when having to change the thermode from one forearm to the other. Otherwise, communication with and monitoring of participants occurred using video camera and intercom.

The experiment started with determining the warmth threshold (baseline temperature: 32°C, acceleration rate: 1°C/second, deceleration rate: 8°C/second) 8 times on each arm to familiarize the participants with the method of limits. Next, the experimental temperature was adjusted (see 2.2) followed by baseline, learning, and test trials.

After the experiment, the electrodes were removed, and participants were fully debriefed and explained why they had been deceived at the beginning with regard to study aims. All participants (and their parents) signed a form confirming that they had been debriefed and received their reimbursement. The experiment lasted about 1.5 hours.

2.5. Data analysis

All data were analyzed with IBM SPSS Statistics 22 (IBM, Armonk, NY). A significance level of P < 0.05 was chosen for the Pearson product–moment correlation coefficients and the analyses of variance (ANOVAs). One-factorial comparisons (t test or one-factorial ANOVAs) keeping the second (or second and third) factor constant were used as post hoc tests.

2.5.1. Missing data

Twenty-two participants could not be included in the final sample either because their pain thresholds exceeded the maximum temperature inducible by the heat pain device (max. temperature 49°C for children and adolescents and 51°C for adults) that was used (10 children and 6 adults), or 6 participants could not be analyzed because of technical problems with the experimental procedure (2 children and 4 adults).

The peripheral psychophysiological recordings of n = 8 children and n = 1 adult in the final sample could not be analyzed because of technical problems in both paradigms. Moreover, in the conditioning paradigm, peripheral physiological recordings are missing for n = 1 child during the test and for n = 1 adolescent during the baseline phase. Consequently, these 2 participants were excluded for analyses of the relevant phases.

To identify participants who did (not) detect the reduced temperature during conditioning trials, change scores were defined for the conditioning phase to investigate changes in subjective pain intensity ratings (mean VAS score during baseline phase–mean VAS score during conditioning phase on placebo and control arm). Learning was defined as the difference in these 2 change scores larger than zero. Altogether, 39 participants (n = 25 children [age 6-13], n = 9 adolescents [age 14-17] and n = 5 adults) did not detect the reduced temperature during conditioning trials and, consequently, were excluded from further analyses for the conditioning paradigm.

2.5.2. Statistical analyses Manipulation check: order effect of expectancy-based and conditioning paradigm

To investigate whether the order of expectancy-based and conditioning paradigm had an effect on results, 2 (order: expectancy-based first vs conditioning first) × 2 (cream: placebo vs control) mixed-factorial design ANOVAs were conducted for subjective ratings and psychophysiological correlates. Because of multiple comparisons, P-values of these ANOVAs were Bonferroni corrected (0.05/12). Assessment of the placebo effect in the expectancy-based and the conditioning paradigm

Mean scores were calculated for subjective pain reports, heart rate, and electrodermal activity for the arm with the placebo cream (Pla) and the arm with the control cream (Con). These were used to calculate change scores for each arm (ΔPla and ΔCon: mean test phase–mean baseline phase for both paradigms and, additionally, mean conditioning phase–mean baseline phase for the conditioning paradigm) as measure of the placebo effect, thus controlling for possible baseline differences. Higher values in subjective pain ratings and lower values in psychophysiological correlates indicate a placebo effect. The difference of these change scores (ΔPla − ΔCon) was calculated and used as a measurement when correlations were used and to ease interpretation of results when investigating the role of the informant. Age effect

Two (cream: placebo vs control) × 4 (age: younger children, older children, adolescents, and adults) mixed-design ANOVAs were conducted for each dependent variable and both paradigms. In the expectancy-based paradigm, only participants whose informant had been the physician were included to keep the impact of the informant constant. In the conditioning paradigm, the conditioning phase was included in addition to the baseline and test phase. Role of informant

A 2 (cream: placebo vs control) × 2 (informant: mother vs physician) × 3 (age: younger children, older children, and adolescents) mixed-design ANOVA was conducted. Role of learning for conditioned placebo hypoalgesia

To investigate the relation between perceived pain reduction in the conditioning phase and the extent of the placebo hypoalgesia in the test phase, differences in change scores were calculated for each dependent variable (ΔPla − ΔCon) and correlated with each other.

3. Results

3.1. Manipulation check: order of the expectancy-based and the conditioning paradigm

The order of the expectancy-based and the conditioning paradigm was randomized across participants. The order of paradigms did not have an effect on all but one outcome measure. For the skin conductance response corrected according to Lykken,33 means were higher when the expectancy-based paradigm was conducted first (F(1; 188) = 35.011, P < 0.001). Inspection of the means suggests that this is only due to higher skin conductance responses to pain stimuli on both arms in the expectancy-based paradigm in participants starting with this paradigm. Possibly, this is accounted for by greater arousal related to the uncertainty regarding the extent of analgesia to be expected in light of no previous experience of attenuated pain (ie, conditioning).

3.2. Placebo effect subjective pain report for the expectancy-based and the conditioning paradigm

Placebo effects were found in each of both paradigms (Fig. 3A for the expectancy-based paradigm and Fig. 3B for the conditioning paradigm). The change score is significantly higher for the arm with the placebo cream than the arm with the control cream in both paradigms (P-values ≤ 0.001). The effect size, however, is slightly higher in the expectancy-based paradigm (nearly medium size) than in the conditioning paradigm (small effect).

Figure 3.
Figure 3.:
Mean change scores on VAS 0 to 100 for both arms in the (A) expectancy-based and the (B) conditioning paradigm.* p <= 0.001.

3.3. Age effects

3.3.1. Age-related placebo effects in subjective pain reports

A significant main effect for “cream” was found in both paradigms with pain intensity ratings decreasing more on the arm with the placebo cream than on the arm with the control cream. Age-related placebo effects in subjective pain reports were found in both paradigms although not as straightforward as expected (Table 1 and Figs. 4 and 5). There was neither a main effect for “age” nor an interaction between “cream” and “age.” Accordingly, Bonferroni-corrected post hoc tests did not reveal differences between age groups. Effect sizes, however, show small effects for adolescents and adults and nearly medium effects for younger and older children (Figs. 4 and 5) indicating a decrease in the placebo effect with increasing age.

Table 1
Table 1:
Results of the 2-factorial ANOVA with repeated measurement investigating the effects of cream and age on the subjective pain report in both paradigms.
Figure 4.
Figure 4.:
Mean change scores on VAS 0 to 100 for placebo and control cream and the 4 age groups in the expectancy-based paradigm. Error bars indicate SE.
Figure 5.
Figure 5.:
Mean change scores on VAS 0 to 100 for placebo and control cream and the 4 age groups in the conditioning paradigm. *P < 0.01; error bars indicate SE.

3.3.2. Age-related effects in psychophysiological correlates

Descriptive data of heart rate and skin conductance response and related age effects are summarized in Tables 2 and 3.

Table 2
Table 2:
Mean (and SEs) for psychophysiological correlates in both paradigms.
Table 3
Table 3:
Results of the 2-factorial ANOVA with repeated measurement investigating the effects of cream and age on heart rate deceleration, heart rate acceleration, and skin conductance response in the expectancy-based and the conditioning paradigm.

In the expectancy-based paradigm, significant age effects were found for maximum heart rate acceleration and skin conductance response. Post hoc tests showed that maximum heart rate acceleration was higher for younger children than for all other groups (P-values < 0.01) and skin conductance response was higher for younger children than for adolescents and adults (P-values ≤ 0.05).

In the conditioning paradigm, significant age effects were found for maximum heart rate acceleration and skin conductance response. Post hoc tests showed higher acceleration for younger than for older children, for younger children than for adults, and for adolescents than for adults (P-values < 0.05) and higher skin conductance responses for younger children than for adolescents or adults and for older children than for adolescents (P-values ≤ 0.05). In addition to age effects, effects for cream were found for maximum heart rate deceleration (stronger deceleration on the arm with the placebo cream) and acceleration (weaker acceleration on the arm with placebo than with control cream).

3.4. Role of the informant

When investigating the impact of the informant (mother or “physician”) on children and adolescents, results, displayed in Table 4, reveal a significant interaction between “cream” and “age” (F(2, 166) = 5.32, P = 0.01, = 0.06) as well as a significant main effect for the factor “cream” (F(1, 166) = 27.59, P < 0.001, = 0.14). Post hoc tests reveal that change scores of the placebo and the control cream differed only for the younger and older children (t(54) = 4.22, P < 0.05, d = 0.57; t(60) = 3.17, P < 0.05, d = 0.41) but not for adolescents (t(55) = 1.07, P = 0.87, d = 0.14).

Table 4
Table 4:
Results of the 3-factorial ANOVA with repeated measurement investigating the role of the informant in addition to the variables cream and age on the subjective pain report in the expectancy-based paradigm.

Owing to the relevance of the role of the informant, we decided to explore possible differences between age groups separately for each informant despite a nonsignificant omnibus test. Two 1-factorial ANOVAs with the difference of the change scores (ΔPla − ΔCon) as dependent variable and age as independent variable were conducted separately for each type of informant. Figure 6 shows that the decline was significant for those children whose mothers were the informant (F(2, 80) = 5.36, P < 0.01, = 0.12). Bonferroni-corrected post hoc tests revealed a difference between younger children and adolescents (P-value <0.01). No significant differences were found for children who were informed and treated by a physician (F(2, 86) = 0.69, P = 0.51, = 0.02), and no differences between mother and physician as informant were found for any age group in Bonferroni-corrected post hoc t-tests.

Figure 6.
Figure 6.:
Results of explanatory ANOVAs and post hoc t tests investigating the role of the informant on subjective pain ratings separately for mother and “physician” and each age group. ANOVA, analysis of variance.

3.5. Role of learning for conditioned placebo response

Associations between the amount of perceived pain reduction during the conditioning trials and the amount of the placebo hypoalgesia in the test trials were examined and are displayed in Table 5. For subjective pain report, perceived pain reduction during conditioning trials correlated positively with placebo hypoalgesia in test trials (r = 0.54, P < 0.001). This relationship declined with age. Positive correlations were also shown for maximum heart rate acceleration (r = 0.19, P = 0.01, ie, greater heart rate acceleration in response to pain stimuli on the control arm as compared to the placebo arm), although no systematic age effects were found.

Table 5
Table 5:
Correlations between perceived pain reduction in the conditioning trials and amount of placebo hypoalgesia in the test trials for the whole sample and separately for age groups for subjective pain report and psychophysiological correlates.

4. Discussion

The first aim of this study was to disentangle expectation and conditioning as explanatory mechanisms of placebo hypoalgesia. Given that research on the placebo effect in children is still scarce,45,54 it is reassuring that a placebo effect was found in each paradigm adding support to the assumption that children and adolescents like adults demonstrate robust placebo responsiveness, and that expectancies as well as conditioning play an important role.30,45,54,55 This has important clinical implications as conditioning as well as expectancies can and should be harnessed for supportive therapeutic application.

In adults, research on disentangling conditioning and expectation is also limited.1,27,53 One study having explicitly examined the effects of both mechanisms on placebo hypoalgesia found conditioning alone to be as effective as conditioning in combination with expectation.53 Looking at the effect sizes in this study, a slightly different picture emerges. Expectancies resulted in a somewhat larger effect size than conditioning (0.454 vs 0.304), indicating that expectancies induce a stronger placebo effect than conditioning. However, this result requires replication and cautious interpretation because methodological differences in the conditioning paradigm such as the type of stimuli, the number of conditioning trials, or the experienced temperature reduction may change results. Furthermore, unlike in typical placebo conditioning paradigms, our participants were not informed on which arm the “analgesic” cream was administered. This possibly made it more difficult to associate the cream with the temperature reduction.

The second aim of the study was to investigate placebo hypoalgesia with regard to age. For subjective pain reports, no significant interactions between “age” and “cream” were found in either paradigm (expectancy-based: P = 0.51, conditioning: P = 0.26). However, when looking at the results of exploratory post hoc tests, effect sizes for younger and older children (expectancy-based: d = 0.42, d = 0.45; conditioning: d = 0.41, d = 0.42) were almost twice as high as they were for adolescents and adults (expectancy-based: d = 0.23, d = 0.23; conditioning: d = 0.26, d = 0.23). Higher placebo effects in younger children in comparison with adults were suggested by Krummenacher et al.29,30 They found higher difference scores between control and placebo hypoalgesia condition for younger children and compared those to healthy adults tested in a different paradigm29 (3.6 and 5.6 times higher). Furthermore, results of the study by Wrobel et al.,55 using conditioning and expectancies to induce a placebo effect, support the assumption that placebo hypoalgesia decreases with age during childhood since they found a comparable hypoalgesic effect in 10- to 15-year-old children and adults. Taken together, these results add up to a surprisingly consistent picture suggesting the existence of age-specific differences in placebo hypoalgesia with stronger effects in children and smaller effects in adolescents and adults. This age-specific pattern seems to be independent of the mechanism with which placebo hypoalgesia is induced.

The effect sizes of adolescents and adults are lower than those of children, however, relate nicely to results of other placebo studies27 in which similar effect sizes were found for pain ratings of healthy adults using different instructions. One reason for the lack of statistical significance is probably the high within-group variances in our study attenuating statistical power. However, high variances are common in the placebo literature27,30 because of various individual and contextual factors impacting on the placebo effect. Age is associated with several developmental changes such as the reduction in suggestibility3,7,30,38,45 or the increasing maturation of the prefrontal cortex6,20,29 which add to variance.

Age effects were not only investigated using subjective pain reports but also psychophysiological correlates. To control for age-related differences in the absolute levels of heart rate and skin conductance due to maturation,17 only changes from baseline scores were analyzed. In both paradigms, younger children showed greater heart rate acceleration and higher skin conductance response. This pattern of autonomic response is typically observed in response to acute experimental pain or aversive stimuli used as unconditioned stimuli38 and suggests a greater defensive response in young children.

In addition to age-related effects, in the conditioning paradigm, a placebo effect was found for heart rate acceleration (ie, lower for pain stimuli on the placebo than on the control arm). Hence, pain stimuli on the placebo arm were associated with an attenuated defensive response to the pain stimulus consistent with anticipated lowered pain. Initial heart rate deceleration was less pronounced for stimuli applied on the control than the placebo arm. In response to conditioned stimuli, such an initial heart rate deceleration is typically considered as an orienting response and tends to be greater for aversive conditioned stimuli.31 In this study, pain stimuli on the placebo arm might elicit a greater orienting response due to the expectation of pain relief. Yet, it cannot be ruled out that the attenuated heart deceleration observed for the control arm might result from initial heart rate deceleration being overlaid by the onset of the subsequent pronounced heart rate acceleration.39 That the effects for cream were primarily found in the conditioning paradigm may reflect stronger implicit learning because of the conditioning procedure.22 Taken together, it seems worthwhile to include psychophysiological correlates when investigating placebo hypoalgesia, as a means to more objectively demonstrate the placebo effect.

The third aim of this study was to investigate the role of the informant regarding the rise of expectancies, given that parent–patient–physician interactions have an impact on the placebo effect.45 No significant effects were found in the omnibus test. Owing to the informants' relevance, we explored potential differences separately for each type of informant. Owing to their explorative nature, these results need to be interpreted cautiously, yet they reveal that the strongest placebo effect was induced by the mother in the group of younger children. Given that suggestion interventions, thus arguably entailing expectancy effects, were found to be not efficacious with regard to needle pain,50 this finding seems surprising at first. However, and in addition to further methodological limitations (eg, relating to the content of the provided information24), none of the studies included in the review16,18,21 involved parents as informants. Possibly, mothers know their children better than a stranger, find better means to explain the efficacy of the treatment, and are, therefore, capable of inducing a stronger placebo effect. This is further in line with results and recommendations of a clinical practice guideline regarding pain during injections40,48 emphasizing the presence and education of parents.

Our fourth aim was to examine the association between learning and test trials to better understand the development of a conditioned placebo hypoalgesia. We found positive associations between learning and test trials, particularly for the children and subjective pain ratings as did Wrobel et al.55 Similarly, a lower defensive response to the pain stimuli during conditioning (ie, attenuated heart rate acceleration) was significantly correlated with the later placebo effect but also for maximum heart rate acceleration. Although these associations may seem trivial at first sight, they have important clinical implications. They emphasize the fact that a strong initial pain reduction is a necessary prerequisite for a strong placebo hypoalgesia.

There are some limitations associated with this study. One is the consideration of age as ordinal rather than a ratio variable. Possibly, the age effects found are due to the groups rather than to the age in years. Considering age as continuous variable, however, would assume a continuous and linear relationship (maturation) of children's cognitive capacities and stimulus-dependent body responses such as heart rate or skin conductance with age which seems unlikely since neither cognitive capacities nor body responses mature linearly. A second limitation is the number of nonlearning participants in the conditioning paradigm which is high (almost 20%), particularly in children. Successful conditioning crucially depends on less subjective pain. Reducing temperature may not necessarily result in less subjective pain, especially in children. To what extent temperature reduction is associated with less perceived pain may depend on the baseline temperature used to induce pain. Possibly, for some participants, the temperature reduction (4°C) may have been noticeable but not meaningful in terms of pain reduction. Therefore, participants with higher baseline pain ratings may have perceived a reduction of 4°C more strongly than participants with lower baseline ratings.8 Finally, the difference between the placebo effect in children and adolescents or adults could be associated with the influence of the context on the placebo effect.4,10,12,36,42 Our experiment took place in the department of psychology, and more than half of adult participants were psychology students. Hence, the experiment could have been more impressive for the children because they were not familiar with the building. Nevertheless, the plausibility of the physician was even by adults rated as 6.47 on a 0 to 10 NRS.

Taken together, results of this study showed that it is possible to induce placebo hypoalgesia separately by conditioning and by expectancies and that these effects are especially pronounced in children younger than 14 years. From a clinical viewpoint, it seems worthwhile to further explore the generation of a placebo hypoalgesia by parental verbal instructions and suggestions and by providing a strong initial pain reduction. Future studies should focus on ecological validity of studies, for example, by adding a group in which conditioning and expectancies are combined.

Conflict of interest statement

The authors have no conflicts of interest to declare.

Supplemental video content

A video abstract associated with this article can be found at


The authors thank all children and their parents as well as all adults for their participation, Laurens Berthold for his help with programming as well as Maike Schmidt, Denise Hillenbrand, Ruth Augustin, Johanna Weinberg, Maria Waligora, Inna Schillert, Antje Arlt, Sandra Lorek, Kerstin Hoffmeister, Madeleine Düker, Daniela Traut, and Sarah Kretschmer for their help with data collection.


[1]. Amanzio M, Benedetti F. Neuropharmacological dissection of placebo analgesia: expectation-activated opioid Systems versus conditioning-activated specific subsystems. J Neurosci 1999;19:484–94.
[2]. Benedetti F. The neurobiology of placebo analgesia: from endogenous opioids to cholecystokinin. Prog Neurobiol 1997;52:109–25.
[3]. Benedetti F. Placebo and the new physiology of the doctor-patient relationship. Physiol Rev 2013;93:1207–46.
[4]. Benedetti F, Carlino E, Pollo A. How placebos change the patient's brain. Neuropsychopharmacology 2010;36:339–54.
[5]. Bingel U, Lorenz J, Schoell E, Weiller C, Büchel C. Mechanisms of placebo analgesia: rACC recruitment of a subcortical antinociceptive network. PAIN 2006;120:8–15.
[6]. Casey BJ, Tottenham N, Liston C, Durston S. Imaging the developing brain: what have we learned about cognitive development? Trends Cogn Sci 2005;9:104–10.
[7]. Ceci SJ, Bruck M. Suggestibility of the child witness: a historical review and synthesis. Psychol Bull 1993;113:403–39.
[8]. Cepeda SM, Africano JM, Polo R, Alcala R, Carr DB. What decline in pain intensity is meaningful to patients with acute pain? PAIN 2003;105:151–7.
[9]. Chambers C, Craig K, Bennett S. The impact of maternal behavior on Children's pain experiences: an experimental analysis. J Pediatr Psychol 2002;27:293–301.
[10]. Colloca L, Benedetti F. Placebos and painkillers: is mind as real as matter? Nat Rev Neurosci 2005:545–52.
[11]. Colloca L, Benedetti F. How prior experience shapes placebo analgesia. PAIN 2006;124:126–33.
[12]. Colloca L, Klinger R, Flor H, Bingel U. Placebo analgesia: psychological and neurobiological mechanisms. PAIN 2013;154:511–4.
[13]. Colloca L, Miller FG. How placebo responses are formed: a learning perspective. Philos Trans R Soc B Biol Sci 2011;366:1859–69.
[14]. Colloca L, Petrovic P, Wager TD, Ingvar M, Benedetti F. How the number of learning trials affects placebo and nocebo responses. PAIN 2010;151:430–9.
[15]. Colloca L, Sigaudo M, Benedetti F. The role of learning in nocebo and placebo effects. PAIN 2008;136:211–8.
[16]. Eland JM. Minimizing pain associated with prekindergarten intramuscular injections. Issues Compr Pediatr Nurs 1981;5:361–72.
[17]. Fleming S, Thompson M, Stevens R, Heneghan C, Plüddemann A, Maconochie I, Tarassenko L, Mant D. Normal ranges of heart rate and respiratory rate in children from birth to 18 years of age: a systematic review of observational studies. The Lancet 2011;377:1011–8.
[18]. Fowler-Kerry S, Lander JR. Management of injection pain in children. PAIN 1987;30:169–75.
[19]. Frank NC, Blount RL, Smith AJ, Manimala MR, Martin JK. Parent and staff behavior, previous child medical experience, and maternal anxiety as they relate to child procedural distress and coping. J Pediatr Psychol 1995;20:277.
[20]. Gogtay N, Gied JN, Lusk L, Hayashi KM, Greenstein D, Vaituzis C, Nuget TF, Herman DH, Clasen LS, Toga AW, Rapoport JL, Thompson PM. Dynamic mapping of human cortical development during childhood through early adulthood. Proc Natl Acad Sci 2004;101:8174–9.
[21]. Goodenough B, Kampel L, Champion GD, Laubreaux L, Nicholas MK, Ziegler JB, McInerney M. An investigation of the placebo effect and age-related factors in the report of needle pain from venipuncture in children. PAIN 1997;72:383–91.
[22]. Hamm AO, Vaitl D. Affective learning: awareness and aversion. Psychophysiology 1996;33:698–710.
[23]. Hohmeister J, Demirakça S, Zohsel K, Flor H, Hermann C. Responses to pain in school-aged children with experience in a neonatal intensive care unit: cognitive aspects and maternal influences. Eur J Pain 2009;13:94–101.
[24]. Jaaniste T, Hayes B, von Baeyer CL. Providing children with information about forthcoming medical procedures: a review and synthesis. Clin Psychol Sci Pract 2007;14:124–43.
[25]. Kazory A. Physicians, their appearance, and the white coat. Am J Med 2008;121:825–8.
[26]. Kirsch I, Lynn SJ, Vigorito M, Miller RR. The role of cognition in classical and operant conditioning. J Clin Psychol 2004;60:369–92.
[27]. Klinger R, Soost S, Flor H, Worm M. Classical conditioning and expectancy in placebo hypoalgesia: a randomized controlled study in patients with atopic dermatitis and persons with healthy skin. PAIN 2007;128:31–9.
[28]. Koenig J, Jarczok MN, Ellis RJ, Hillecke TK, Thayer JF. Heart rate variability and experimentally induced pain in healthy adults: a systematic review. EJP 2014;18:301–14.
[29]. Krummenacher P, Candia V, Folkers G, Schedlowski M, Schönbächler G. Prefrontal cortex modulates placebo analgesia. PAIN 2010;148:368–74.
[30]. Krummenacher P, Kossowsky J, Schwarz C, Brugger P, Kelley JM, Meyer A, Gaab J. Expectancy-induced placebo analgesia in children and the role of magical thinking. J Pain 2014;15:1282–93.
[31]. Lipp OV. Human fear learning: contemporary procedures and measurement. In: Craske MG, Hermans D, Vansteenwegen D, editors. Fear and learning: from basic processes to clinical implications. Washington: American Psychological Association, 2006:37-mue52.
[32]. Loggia ML, Juneau M, Bushnell CM. Autonomic responses to heat pain: heart rate, skin conductance, and their relation to verbal ratings and stimulus intensity. PAIN 2011;152:592–8.
[33]. Lykken DT, Venables PH. Direct Measurement of skin conductance: a proposal for standardization. Psychophysiology 1971;8:657–72.
[34]. Margraf J. Mini-DIPS: Diagnostisches Kurz-Interview bei psychischen Störungen. Berlin Heidelberg: Springer, 2013.
[35]. Meissner K, Bingel U, Colloca L, Wager TD, Watson A, Flaten MA. The placebo effect: advances from different methodological approaches. J Neurosci 2011;31:16117–24.
[36]. Miller FG, kaptchuk tj. The power of context: reconceptualizing the placebo effect. JRSM 2008;101:222–5.
[37]. Montgomery GH, Kirsch I. Classical conditioning and the placebo effect. PAIN 1997:107–13.
[38]. Morgan AH, Hilgard ER. Age differences in susceptibility to hypnosis. Int J Clin Exp Hypnosis 2008;21:78–85.
[39]. Mueller EM, Sperl MFJ, Panitz C. Aversive imagery causes de novo fear conditioning. Psychol Sci 2019;30:1001–15.
[40]. Pillai Riddell R, Taddio A, McMurtry CM, Shah V, Noel M, Chambers CT. Process interventions for vaccine injections: systematic review of randomized controlled trials and quasi-randomized controlled trials. Clin J Pain 2015;31(10 suppl):S99–108.
[41]. Pollo A, Vighetti S, Rainero I, Benedetti F. Placebo analgesia and the heart. PAIN 2003;102:125–33.
[42]. Price DD, Finniss DG, Benedetti F. A comprehensive review of the placebo effect: recent advances and current thought. Annu Rev Psychol 2008;59:565–90.
[43]. Rehman SU, Nietert PJ, Cope DW, Kilpatrick AO. What to wear today? Effect of doctor's attire on the trust and confidence of patients. Am J Med 2005;118:1279–86.
[44]. Schneider S. Diagnostisches Interview bei Störungen im Kindes- und Jugendalter. Berlin Heidelberg: Springer, 2009.
[45]. Simmons K, Ortiz R, Kossowsky J, Krummenacher P, Grillon C, Pine D, Colloca L. Pain and placebo in pediatrics: a comprehensive review of laboratory and clinical findings. PAIN 2014;155:2229–35.
[46]. Stewart-Williams S, Podd J. The placebo effect: dissolving the expectancy versus conditioning debate. Psychol Bull 2004;130:324–40.
[47]. Street RL, Gordon H, Haidet P. Physicians' communication and perceptions of patients: is it how they look, how they talk, or is it just the doctor? Soc Sci Med 2007;65:586–98.
[48]. Taddio A, McMurtry CM, Vibhuti S, Pillai Riddell R, Chambers CT, Noel M, MacDonald NE, Rogers J, Bucci LM, Mousmanis P, Lang E, Halperin SA, Bowles S, Halpert C, Ipp M, Asmundson GJG, Rieder MJ, Robson K, Uleryk E, Antony MM, Dubey V, Hanrahan A, Lockett D, Scott J, Bleeker EV, HELPinKids&Adults. Reducing pain during vaccine injections: clinical practice guideline. Can Med Assoc J 2015;187:975–82.
[49]. Tates K, Meeuwesen L. Doctor–parent–child communication. A (re)view of the literature. Soc Sci Med 2001;52:839–51.
[50]. Uman LS, Chambers CT, McGrath PJ, Kisely S. A systematic review of randomized controlled trials examining psychological interventions for needle-related procedural pain and distress in children and adolescents: an abbreviated cochrane review. J Pediatr Psychol 2008;33:842–54.
[51]. von Baeyer CL. Children's self-reports of pain intensity: scale selection, limitations and interpretation. Pain Res Manage 2006;11:157–62.
[52]. Voudouris NJ, Peck CL, Coleman G. Conditioned response models of placebo phenomena further support. PAIN 1989;38:109–16.
[53]. Voudouris NJ, Peck CL, Coleman G. The role of conditioning and verbal expectancy in the placebo response. PAIN 1990;43:121–8.
[54]. Weimer K, Gulewitsch MD, Schlarb AA, Schwille-Kiuntke J, Klosterhalfen S, Enck P. Placebo effects in children: a review. Pediatr Res 2013;74:96–102.
[55]. Wrobel N, Fadai T, Sprenger C, Hebebrand J, Wiech K, Bingel U. Are children the better placebo analgesia responders? An experimental approach. J Pain 2015;16:1005–11.

Placebo effect; Classical conditioning; Expectation; Children; Age; Heart rate; Skin conductance

Supplemental Digital Content

© 2020 International Association for the Study of Pain