Change in pain expectations but no open‐label placebo analgesia: An experimental study using the heat pain paradigm

Open‐label placebos (OLP) prescribed without deception and with a convincing rationale have been shown to evoke powerful treatment effects. Patients’ treatment expectations seem to influence the magnitude of the effect.


| INTRODUCTION
Pain-related disability is one of the most common medical conditions worldwide (James et al., 2018), impairing the quality of life (Lamé et al., 2005).Low back pain and headache are two of the three most common causes of years lived with disability (James et al., 2018).It is followed by huge socioeconomic burdens through high healthcare costs, job loss, absenteeism and presenteeism (Breivik et al., 2013).
Often, commonly prescribed pharmaceutical treatments for pain conditions do not show superior efficacy compared to placebo controls (Machado et al., 2015).Others achieve only small effects at the cost of more side effects (Roelofs et al., 2008).Efficacy of placebo treatments has been demonstrated in various medical conditions and mental disorders including fibromyalgia (Chen et al., 2017), irritable bowel syndrome (Kaptchuk et al., 2008), migraine (Macedo et al., 2008), osteoarthritis (Zhang et al., 2008) or depression (Kirsch et al., 2002).
Prescription of placebos in daily medical care is a common phenomenon, often without the patient's informed consent (Howick et al., 2013;Hróbjartsson & Norup, 2003).This practice is ethically intolerable as it violates the patient's autonomy (Blease et al., 2016).In recent years, open-label placebos (OLP) have been discussed as an alternative.OLPs, which are provided with full transparency and disclosure, have shown clinically meaningful effects on various medical conditions and mental disorders (von Wernsdorff et al., 2021).In chronic pain, an RCT with 97 patients suffering from chronic low back pain found moderate-to-large effect sizes for OLP compared to treatment as usual (Carvalho et al., 2016), and these effects remained stable after 5 years (Carvalho et al., 2021).Similar results for OLP were found by other research groups for chronic back pain (Kleine-Borgmann et al., 2019) and irritable bowel syndrome (Kaptchuk et al., 2010).Moreover, effect sizes for placebo analgesia are moderate to large in studies on placebo mechanisms (Vase et al., 2009).
Previous research suggested that effects of OLP may depend on treatment expectations and the provision of a convincing rationale for why and how placebos might work (Buergler et al., 2023;Charlesworth et al., 2017;Locher et al., 2017;Spille et al., 2023), possibly because a compelling rationale promotes positive treatment expectations.Yet, little is known about how precisely a convincing rationale should be designed and provided to foster positive expectations and enhance the effects of OLP.
Drawing on a recent study examining the effects of OLP in depression (Friehs, Rief, et al., 2022), this study investigated whether two conceptually different OLP rationales could increase pain tolerance and reduce pain intensity and unpleasantness in a standardized heat pain experiment.In a sub-clinical sample of people who selfreported reoccurring or persistent pain for at least the last 3 months, we compared a personal-emotional (PE) and a scientific-matter-of-fact (SM) rationale with a control group that received the same placebo without any rationale (CG).In this pre-post design, we hypothesized an increase in pain tolerance and a decrease in expected and experienced pain intensity and unpleasantness for PE and SM, but not for the CG.

| Study design
From November 2021 to April 2022, we conducted a laboratory experiment at the Department of Adult Clinical Psychology and Psychotherapy at the University of Kaiserslautern-Landau (RPTU), Landau, Germany.It was preregistered at aspre dicted.org (https:// aspre dicted.org/ 7d4sw.pdf).The study was approved by the local Ethics committee of the Department of Psychology at the University of Kaiserslautern-Landau (RPTU), Landau, Germany (2020-246).Prior to participation, written informed consent was obtained from each participant.All experimental procedures were in accordance to the ethical guidelines of the German Psychological Society (DGPs, 2016).

| Sample
Participants were recruited via the psychotherapeutic outpatient clinic (WiPP) and student email lists, posters and flyers at the University of Kaiserslautern-Landau (RPTU), Landau, Germany.The study was advertised as a 'study on psychological influences on pain perception'.We aimed at a heterogeneous sub-clinical sample of participants with any form of self-reported reoccurring or persistent pain with or without any form of current treatment.Participants were recruited from both the town population and the student population.The inclusion criteria were as follows: self-reported forms of reoccurring or persistent pain for at least the last 3 months (for headache and back pain at least once per week), age ≥ 18 years, sufficient fluency in German language and sufficient eyesight.We defined subclinical as follows: participants who, in self-report, fulfil the criteria for chronic pain (i.e.reoccurring or persistent pain for at least the last 3 months).Exclusion criteria were as follows: skin pathologies, neuropathies or any sensory abnormality affecting the thermal or tactile modality, diabetes, allergic reactions to heat, pregnancy, breastfeeding, acute pain medication or alcohol intake in the last 12 h, intake of illegal drugs in the last 2 weeks, completed studies of medicine or psychology and psychology students after bachelor degree (to avoid participation with previous knowledge about placebo effects).If participants took any form of pain-influencing medicine, they were asked to hold that treatment for 12 h prior to the laboratory session.Participants received a financial compensation of 8€ or course credit.

| Procedure
The experimental procedure was similar to other studies using the heat pain paradigm in placebo research (Kube et al., 2020;Locher et al., 2017).People interested in participation were first screened in person or via telephone.During screening, basic information about the study aims and the procedure were given, inclusion and exclusion criteria were checked.If people were eligible, an appointment for the laboratory experimental session was scheduled.During the screening, the provider behaved in a neutral but friendly way, trying to encourage participation without persuading participants.Questions were answered without revealing any further information prior to study participation.Before arrival, participants were randomly assigned to one of the three conditions (personal-emotional (PE) vs. scientific-matter-of-fact (SM) vs. control group (CG)) and the provider followed the respective behavioural interaction script throughout the laboratory session.On arrival, participants were checked for the additional exclusion criteria referring to the last few days/hours (i.e.medication intake, consumption of alcohol or illegal drugs), read and signed the study information and informed consent and a blood pressure test was conducted to establish a medical laboratory setting, while adhering to the restrictions resulting from the COVID-19 pandemic.Afterwards, a heat pain trial was performed to familiarize participants with the basic experimental procedure (see below).In the following baseline measurement, participants rated different questionnaires on demographics, their general pain symptoms and related coping and impairment (see below).The baseline assessment of heat pain tolerance, expected and experienced pain intensity and pain unpleasantness followed.Participants in the PE and SM conditions received a placebo ointment and the respective expectation rationale (as detailed below).Participants in the CG received the same ointment, but it was stated that the ointment would only be used to improve the quality of the pain measurement, without any expectation rationale.The ointment was applied by the provider.After the application of the ointment, the post assessment of heat pain tolerance, expected and experienced pain intensity and pain unpleasantness was conducted.This was followed by questionnaires on the perceived warmth and competence of the provider, trait anxiety, openness for experiences and a manipulation check.The experimental session was closed by the debriefing of participants about the true aims and deceptions of the study, and payment of the allowance.For an overview of the experimental procedure, see Figure 1.

| Experimental conditions
Each experimental condition received a different rationale, shaping the expectations for the placebo application, with an accompanying interaction style throughout the laboratory session.The PE versus SM rationales were adapted from the ones used in a previous study (Friehs, Rief, et al., 2022).In addition to different verbal suggestions, the two styles also differed in the way the provider interacted with the participants, since previous research has shown that the perceived warmth and competence of a provider modulates the placebo effect (Howe et al., 2017).All conditions received the same ointment (a standard basic ointment with a small amount of oil of thyme to achieve the typical smell of ointment, produced by a local pharmacy).This ointment evoked powerful placebo effects in previous studies (Kube et al., 2020).One advantage of using an ointment as a placebo (as opposed to tablets, for instance) is that it allows researchers to plausibly explain that it may work quite quickly.
In the PE condition, the rationale was designed to create an emotion-based expectation rather than an expectation based on scientific facts.Each participant was informed that they would receive a placebo, meaning that the ointment would not contain any active ingredient.It was explained that people who receive a placebo often experience similar effects as people who receive an actual medication.The rationale suggested that the placebo ointment may make participants less sensitive to pain and could thus enable the participant to better cope with thermal stimulations.This rationale was underlined by the presentation of an anecdotal personal report: Participants were informed that a young woman, who participated in a similar previous study, reported less pain and was able to tolerate pain better.Based on this positive anecdotal 'evidence', the participant was told that they too could experience positive effects of the placebo ointment.Similar to previous OLP studies, it was stated that it may be helpful to be open minded towards the use of the placebo ointment, but that it is also normal for participants to have doubts about placebos (Kaptchuk et al., 2010).In the PE rationale, each participant was addressed personally in colloquial speech, including references to how the placebo might affect them.In terms of the provider's behaviour, the PE condition was designed to match its rationale and provide a highly warm and friendly interaction, including small-talk, smiling, active listening, directly addressing the participant, room for questions, eye contact and the provider directly facing the participant.
In the SM condition, the rationale was designed to highlight the general efficacy of placebos and evoke expectations based on scientific facts.Participants were informed that they would receive a placebo, meaning that the ointment would not contain any active pharmaceutical agent.It was explained that scientific studies showed similar effects for placebo treatments and actual medication.The rationale suggested that the placebo may have a desensitizing effect against pain and could thus enable people to cope better with thermal stimulations.Furthermore, participants were informed about the potential mechanisms of action of placebos.Specifically, it was explained that in terms of classical conditioning, the application of a placebo that looks and feels like a certain medication can activate a positive autonomic bodily reaction, similar to the reaction to an actual medication.Similar to the PE condition, participants were informed that it may be helpful but not necessary to be open minded towards placebos.In the SM rationale, participants were addressed in precise, yet impersonal language, using scientific terminology.In terms of provider behaviour, the SM condition was designed to match its rationale and provide a neutrally warm interaction, reflecting a professional distance without being rude or mean.Thus, the providers in the SM condition were instructed not to do small talk, not to smile, have little eye contact and to keep room for questions to a minimum.The provider did not address the participant personally (e.g. by saying 'one' instead of 'you') and mostly faced the computer instead of the participant.
In the CG condition, it was neutrally stated that the baseline measurement was flawed and in order to improve the measurement quality, an ointment would be needed.Participants were told that this can happen for some people and that the measurement would have to be repeated.In terms of the provider behaviour, the CG was similar to the PE condition, aiming for a highly warm and friendly interaction.This interaction style combined with a rationale evoked the strongest effects in a previous study (Howe et al., 2017) and thus allows to separately test the effects of rationale versus no rationale.Details of the OLP rationales can be obtained from Supplement C.
We aimed to keep the encounter similar across all conditions with regard to the perceived competence.Thus, we did not manipulate the provider's competence; instead, in all conditions, the setting was designed to be professional and competent, including a blood pressure test, medical posters on the walls, professional clothing and name tags, and the encounter included competent explanations and execution of procedures (Howe et al., 2017).Furthermore, the PE versus SM styles were designed to be similar in terms of length.We predicted that the PE versus SM style would differ in their perceived warmth, but not in their perceived competence.The perception of the two styles was pre-tested in a sample that was independent of the sub-clinical sample of the main study.Furthermore, we tested whether the two styles differ in their effects on participants' pain expectancies.The results of this pre-test are reported below.

| Pre-test: Perception of experimental conditions
In an online pre-test, 125 healthy participants (M Age = 26.56,SD Age = 9.164; 92 females, 33 males; 88% had the A-level degree or a higher education) were recruited at the University of Kaiserslautern-Landau (RPTU), Landau, Germany.In this pre-test, participants watched videotaped sequences from the three conditions (PE vs. SM vs. CG).In these videos, participants watched a person undergoing the experimental procedure described above.Participants were asked to imagine themselves participating in the laboratory experiment and to rate their expected pain intensity and pain unpleasantness at baseline (T1) and after the placebo application (T2), as well as the perceived warmth and competence of the provider.

| Pain stimuli
We used a standardized heat pain paradigm: the suprathreshold method of the Thermo Sensory Analyser (TSA-II, Medoc Israel) that provides thermal stimulations.The TSA-II is a safe and commonly used device for the study of pain sensation and analgesic effects.The procedure followed in this study is the same as in previous placebo studies using a heat pain paradigm (Kube et al., 2020;Locher et al., 2017).The thermode of the TSA-II was placed on two different locations of the proximal volar forearm on the non-dominant side.The locations were changed between the baseline and post-measurement to prevent sensitization and habituation effects (Emerson et al., 2014).Before the baseline measurement, participants were familiarized with the heat stimuli and the process of terminating heat stimulations in a single trial run.

| Measures
Primary outcomes were heat pain tolerance and expected and experienced pain.Other variables were used for sample description and secondary analyses.

| Heat pain tolerance
Heat pain tolerance was determined using the method of limits.Starting at 32°C, the heat stimulus increased at 0.5°C per second.Participants were asked to stop the stimulus at the moment when they could no longer tolerate the pain.The temperature at which they stopped the increasing heat stimulus is the individual pain tolerance.At 50.5°C, the heat stimulations stopped automatically to prevent damage to the skin.The cooldown rate was 8°C per second.By taking the mean average of three measurements of tolerance at baseline (T1) and post-assessment (T2), respectively, we ensured a reliable assessment of pain tolerance and control for fluctuations in single trials.
2.7.2 | Expected and experienced pain intensity and unpleasantness Before each pain tolerance assessment, participants were asked to indicate what pain intensity and unpleasantness they expected for the upcoming trial.Pain intensity and unpleasantness were both rated on a numeric analogue scale (NAS) ranging from 0 ('no pain sensation' / 'not unpleasant at all') to 100 ('most intense pain sensation imaginable' / 'the most unpleasant imaginable').After each pain tolerance assessment, participants were asked to rate their actually perceived pain intensity and unpleasantness using the same NAS.This allows for the measurement of the subjective pain experience of the participant.These measures are established in assessing pain in experimental studies (Price et al., 1994) and pain expectations in experimental placebo analgesia studies (Locher et al., 2017).Intensity measures the cognitive component of pain sensation, whereas unpleasantness measures its affective component (Price, 2000).As with pain tolerance, the intensity and unpleasantness ratings were averaged across the three trials.

| Reoccurring pain
Participants provided information on the location, intensity, unpleasantness and frequency of their reoccurring or persistent pain for at least the last 3 months.Pain intensity and unpleasantness for pain conditions were rated on the same NAS as described above.Frequency was rated on a numerical scale ranging from 1 ('daily') to 4 ('less than once per month').For every pain location, intensity, unpleasantness and frequency were rated separately.Furthermore, participants were asked to give information on intake of pain medication in the last 3 months (1 = 'daily' to 5 = 'none'), diagnosed pain conditions and current participation in treatments against pain.We did not differentiate between reoccurring and persistent pain.We used the Pain Disability Index (PDI; Pollard, 1984) in its German version (Dillmann et al., 1994).Using an 11-point scale from 0 ('no disability') to 10 ('total disability'), disability was assessed in seven life domains (family/ home responsibilities, recreation, social activities, occupation, sexual behaviour, self-care and life-support activity), with one item per life domain.The total score indicates the global level of disability, ranging from 0 to 70.For individuals with and without chronic pain, the PDI has shown good psychometric properties (Mewes et al., 2009;Soer et al., 2013).In the current sample, internal consistency of the PDI was McDonalds ω = 0.837.

| Pain-related catastrophizing
We used the German version (Meyer et al., 2008) of the Pain Catastrophizing Scale (PCS; Sullivan et al., 1995).In 13 items (e.g.'I feel I can't go on'), a 5-point Likert scale (0 = 'not at all', 4 = 'all the time') is used to assess pain catastrophizing.A higher total score (range: 0-52) indicates higher levels of pain catastrophizing.For individuals with and without chronic pain, the PCS has shown good psychometric properties (Van Damme et al., 2002).In the current sample, internal consistency of the PCS was McDonalds ω = 0.860.

| Pain coping
Pain-related cognitive and behavioural coping were assessed using the German Questionnaire for the assessment of pain processing (FESV; Fragebogen zur Erfassung der Schmerzverarbeitung; Geissner, 2001).It measures cognitive coping on the three subscales action planning competence, cognitive restructuring and experience of self-competence with four items each (e.g.'When I have pain, I tell myself that I can cope much better than before').Behavioural coping is measured on the three subscales mental distraction, counter steering activities and relaxation techniques (e.g.'When I am in pain, I distract myself by listening to good music').
Participants were asked to rate how the strategies applied to them personally, on a 6-point Likert-scale (1 = 'not at all' to 6 = 'completely').A higher total score (range: 4-24) indicates better coping.It showed good psychometric properties for individuals with pain conditions (Geissner, 2001).In the current sample, internal consistencies of the FESV subscales were McDonalds 0.647 ≤ ω ≤ 0.814.
2.7.7 | Warmth/Competence Our secondary outcomes were the participants ratings of the perceived warmth and competence of the providers.These were assessed using a German scale for the assessment of the Stereotype Content Model (Friehs, Aparicio Lukassowitz, & Wagner, 2022).It has shown good validity and reliability in rating perceived warmth and competence of occupational groups (e.g.physicians).For this study, participants were asked to rate how they perceived the provider.Each item was presented as the two extremes on a semantic differential (1-5): warmth was assessed using three items (unfriendly -friendly, ill-natured -good-natured and cold -warm) and competence using four (careless -thorough, incompetent -competent, lazy -hard working and inefficient -efficient).Higher mean values (range: 1-5) represent higher warmth / competence.In the current sample, internal consistency of the warmth subscale was McDonalds ω = 0.712 (Competence ω = 0.651).

| Further measures
We assessed other possible confounding variables, as both openness to experiences (Kube et al., 2020) and trait anxiety (Salzmann et al., 2022) have been shown to influence the placebo response.To measure the Big-Five personality trait openness to experiences, we used the respective subscale (6 items) of the NEO Five-Factor inventory in its German short version (Körner et al., 2008).Trait anxiety was measured, using the respective 10 items of the State-Trait Anxiety and Depression Inventory in its German version (Laux et al., 2013).In the current sample, internal consistency was McDonalds ω = 0.669 for openness to experiences and ω = 0.867 for trait anxiety.Furthermore, we assessed sociodemographic variables, including age, gender and education level.Similar to previous studies (Kube et al., 2020;Locher et al., 2017), we assessed the credibility of the treatment.Participants in the two OLP conditions (PE and SM) rated whether they believed that they received a placebo (1 = 'yes', 2 = 'likely', 3 = 'no'), and were asked to describe the term placebo in their own words.Participants were excluded if they did not believe that they received a placebo or could not describe the term in their own words.

| Provider training
A detailed script was used for both the telephone screening and the experimental procedure.The providers (three female student assistants) practiced the script several times with and without sham patients, and received video feedback.Special attention was paid to convincingly explain the rationales, and to practice switching between the rationales between subsequent sessions.Providers were trained in all three conditions and conducted experimental sessions for all three conditions.

| Randomization and blinding
After scheduling the appointment, participants were randomized into the three conditions (PE vs. SM vs. CG), prior to the experimental session.We used 'Research Randomizer' (Urbaniak & Plous, 2015) to randomly assign block-wise generated numbers from 1 to 3 (PE-CG) to the chronologically ordered list of participants.Because of the study design, the providers were aware of the allocation.Participants were kept blind to the existence of different conditions throughout the experiment.

| Statistical analyses
For each of our five primary outcome measures (pain tolerance, expected and experienced pain intensity and unpleasantness), we specified a 2 × 3 mixed factorial ANOVA with our predictors time (T1 = baseline; T2 = post-treatment) and condition (PE vs. SM vs. CG).Time and condition were allowed to interact to detect condition-dependent pre-post changes.
Warmth and competence were analysed using separate one-way-ANOVA with the factor condition.Using ANCOVA, we used change scores of expected pain intensity and unpleasantness, and pain tolerance to explore their influence on the other primary outcomes.Furthermore, we explored the influence of pain-related disability, pain catastrophizing, pain coping, openness to experiences and trait anxiety on our primary outcomes.According to Field (2017), we only conducted covariate analyses if the covariate did not differ between the conditions at baseline.The ANCOVA results are detailed in Supplement B.
Where applicable, we performed paired post hoc comparisons, using the Šidák correction for multiple testing.For information on our sample and possible baseline differences, we conducted separate one-way ANOVA with the dependent variables (DV) age, pain-related disability, pain catastrophizing, pain coping, openness to experiences, trait anxiety, reoccurring pain intensity / unpleasantness and the baseline primary outcomes.Furthermore, separate χ 2 tests for the DVs gender, education level and reoccurring pain locations and diagnosis were conducted.Kruskal-Wallis tests were conducted for frequencies of reoccurring pain and pain medication.In these analyses, the independent variable was condition.
Significance was determined at the p < 0.05 threshold.We report the effect sizes η p 2 in the ANOVA/ANCOVA, and g Hedges in the post-hoc comparisons.All statistical analyses were run using IBM SPSS version 28.

| RESULTS
3.1 | Sample characteristics and baseline differences A total of 76 individuals who self-reported reoccurring or persistent pain for at least the last 3 months participated in the experiment.Five participants were excluded from the analyses for the following reasons: four individuals from the OLP conditions (PE and SM) did not believe that they received a placebo, and one participant showed questionable answering patterns, like always choosing the leftmost answer.Thus, the final sample consisted of 71 individuals.Our sample was predominantly female (84.5%),M Age = 31.460(SD = 13.397;range: 19-70).No significant differences in demographic variables between the conditions could be observed (see Table 2).There were baseline differences between the conditions in the number of participants who suffered from reoccurring or persistent pain in the chest or limb, and intensity and unpleasantness of pain in the lower body (see Supplement A).All other variables showed no baseline differences.
With respect to reoccurring or persistent pain, 20 participants (28.2%) stated that they have a (chronic) pain diagnosis.Fourteen participants reported no intake of pain medication in the last 3 months (19.7%), 14 reported intake multiple times a week (19.7%),23 reported intake multiple times a month (32.4%) and 20 reported intake less than once per month (28.2%).Regarding ongoing pain treatment, 28 participants (39.4%) were undergoing treatment at the time of the study, of whom 24 participants received medical treatment (33.8%), 12 received psychotherapy (16.9%) and 11 received other forms of treatment (15.5%).A total of 25 participants reported reoccurring or persistent pain in one or two locations (35.2%), 33 in three or four (46.5%) and 13 in at least five locations (18.3%), with all participants naming at least one location with reoccurring or persistent pain for at least the last 3 months.The most common pain location was back pain (64.8%), followed by headache (57.7%), neck pain (53.5%) and pain during menorrhoea (43.7%; 51.7% of all females, respectively).Highest pain intensity was reported for lower body pain (M = 63.410,SD = 20.641) and lowest pain intensity for neck pain (M = 52.630,SD = 19.433).Highest pain unpleasantness was reported for limb pain (M = 60.330,SD = 28.073) and lowest pain unpleasantness for neck pain (M = 48.420,SD = 20.253).Pain-related disability (M = 27.859,SD = 13.600) was in the 90% percentile compared to a sub-clinical sample of people suffering from chronic pain (Mewes et al., 2009).Pain catastrophizing was high (M = 23.606,SD = 9.302) and above the clinical cut-off for moderate catastrophizing ( 20), but below the clinical cut-off for high catastrophizing (30; Sullivan, 2009).Participants' percentile ranks for pain coping were between 27 and 88 compared to a clinical sample of patients with pain diagnoses (Geissner, 2001).
Detailed information about the sample can be obtained from Table 2 and Supplement A.

| DISCUSSION
In a standardized heat pain experiment with a sub-clinical sample of participants who suffered from reoccurring or persistent pain for at least the last 3 months, we examined the effects of an open-label placebo (OLP) ointment.We assessed two different approaches to provide a convincing rationale (personal-emotional (PE) and a scientificmatter-of-fact (SM)), and compared it to a control group (CG) receiving no rationale.The manipulation check confirmed that rationales were perceived differently, with the PE being perceived as warmer, as hypothesized, and, although not intended, more competent than the SM.The manipulation successfully changed participants' pain expectations, lowering the expected pain intensity after placebo application in the PE and SM, but not in the CG.However, expected pain unpleasantness did not change in any condition.Despite a change in participants' pain expectations through different rationales, this study failed to find significant effects of OLP on pain tolerance and subjective pain experience compared to the CG.
Our findings suggest that evoking positive treatment expectations is not sufficient to elicit an OLP response.This is consistent with other research, showing that the effects of OLP on chronic back pain (Kleine-Borgmann et al., 2019), irritable bowel syndrome (Ballou et al., 2022) and allergic rhinitis (Kube et al., 2022;Schaefer et al., 2018) are not related to participants' treatment expectations.However, other studies did find that positive treatment expectations are related to the effects of OLP (El Brihi et al., 2019) and deceptive placebos (DP; Locher et al., 2017;Peerdeman et al., 2016;Rutherford et al., 2017;Schmitz et al., 2019;Watkinson et al., 2017).Hence, it might be that the magnitude of expectations change in the present study was not powerful enough to evoke an OLP effect.The lack of change in pain tolerance and experienced pain might be a small-sample power issue, as there were small descriptive changes from T1 to T2 with small-to-medium effect sizes between the conditions.Alternatively, it is possible that the effects of OLP in pain are in fact unrelated to positive treatment expectations.
The failure of OLP in this study is in contrast to a number of previous studies examining the effects of OLP in pain.A study using the same heat pain paradigm found that an OLP ointment provided with a convincing rationale reduced subjective pain experience more than an OLP ointment without a rationale (Locher et al., 2017).Another study using this heat pain paradigm found that an OLP ointment increased participants' pain tolerance compared to a no-treatment control group (Kube et al., 2020).However, our study differed from these two previous studies in terms of the sample characteristics: Whereas Locher et al. (2017) and Kube et al. (2020) both examined healthy participants, our study examined people with reoccurring and moderately distressing levels of pain.Conceivably, people who frequently suffer from pain are very much engaged with their normal pain, such that pain induced through the thermal stimulations was not relevant enough for them to evoke OLP effects.Besides, some participants received treatment for their pain which might have altered their pain processing, thereby preventing OLP effects.Smallsample subgroup analyses implicated that expectation change might depend on the severity of the underlying pain condition.Alternatively, people with reoccurring pain might have undergone failed treatments in the past, such that a negative anticipatory response was (unconsciously) elicited when taking the placebo ointment.However, studies examining clinical samples of patients suffering from chronic pain also found significant improvements through OLP (Carvalho et al., 2016;Kleine-Borgmann et al., 2019), yet not in an experimental setting and with the provision of placebo pills (as opposed to an ointment).Differences between placebo application routes were found (Blythe et al., 2022;Fernández-López et al., 2022;Meissner & Linde, 2018;Peerdeman et al., 2016), but its impacts have been discussed controversially (Fässler et al., 2015) or insubstantial as indicated by a recent meta-analysis (Buergler et al., 2023).Furthermore, it is possible that the information to receive 'only' a placebo provoked disappointment or other negatively associated feelings.This is a common phenomenon in participants assigned to control groups or placed into a group they did not prefer (Locher et al., 2017;McCambridge et al., 2014;Skingley et al., 2014).Moreover, it is possible that the CG in the present study was designed in a way that it was difficult for OLP to be superior to it.Unlike previous experimental studies (Kube et al., 2020;Locher et al., 2017), the CG received the same ointment as the two OLP groups and it is possible that this ointment influenced participants' pain perception (Bräscher et al., 2018), although it was not labelled as a potential treatment, but linked to technical necessities.It has been argued that placebos themselves may (unconsciously) provoke beneficial effects through associative learning mechanisms (Kaptchuk, 2018;Kaptchuk et al., 2020).It is possible that the failure of OLP relative to the CG merely reflects the (unintended) effectiveness of the CG.This is particularly pertinent since the CG received the same quality of interaction as the PE.
The failure of the two OLP versions from the present study is in line with a previous study examining the effects of the same OLP versions on experimentally induced sadness (Friehs, Rief, et al., 2022).In that previous study, only DP (labelled as 'fast-acting antidepressants'), but neither the PE-OLP nor the SM-OLP evoked significant effects over the CG.It might be that the current PE and SM rationales are not sufficiently convincing to facilitate deep processing of the presented information, such that participants processed them only superficially, thereby limiting the potential of placebo effects (Geers et al., 2019).Information that is consistent with prior knowledge can be processed superficially and still evoke effects (e.g.information on a medication as in DP), whereas new information that is less consistent with prior knowledge (e.g. the information that OLP can have a positive therapeutic effect) requires a deeper level of processing in order to have effects.This hypothesis would explain the occurrence of a DP effect in the previous study, but the absence of an OLP effect in both the previous and the present study.Future studies should compare new rationales to an established one (Kaptchuk et al., 2010).Another explanation might be that pain patients tend to overpredict new pain events, which might offset the potential effect of positive treatment expectations (Crombez et al., 1996).

| Strengths and limitations
This study was the first to test different rationale styles in a sub-clinical sample of people suffering from reoccurring or persistent pain in the context of OLP.In contrast to many other studies on placebo mechanisms conducted with healthy samples (Kube et al., 2020;Locher et al., 2017;Schneider et al., 2020), a sub-clinical sample provides insights into daily medical care effects.Furthermore, it allows investigating underlying mechanisms (e.g.expectations) in a controlled laboratory environment.Conversely, most OLP studies with clinical samples focus on repeated OLP application in a naturalistic environment mainly aiming at symptom improvement (Charlesworth et al., 2017;von Wernsdorff et al., 2021).We developed a highly standardized procedure (i.e.provider training with video feedback, scripted interactions and standardized setting) to ensure the validity of the study against confounding variables such as provider effects.We designed and optimized our rationale styles based on a previous study (Friehs, Rief, et al., 2022) and pre-tested it in a large sample.We advertised our study as a 'study on psychological influences on pain perception', and recruited in a university outpatient clinic and among students, which might have attracted a more heterogeneous sample than many other OLP studies that advertised 'novel mind-body treatments' which might attract mainly participants who are open minded towards new forms of interventions (Carvalho et al., 2016;Kaptchuk et al., 2010;Locher et al., 2017).We also compared our rationales to a CG that received the same procedure and placebo but without a rationale, allowing us to directly investigate the influence of expectations.Moreover, we successfully established a perception of warmth and competence in the providers; although, neutral warmth was statistically lower compared to high warmth, both rated high.
Notwithstanding these strengths, there are a number of limitations that need to be considered.Most importantly, the sample size implicates a reduced statistical power to detect meaningful effects and increases the risk for bias in parameter estimates.We recommend a replication with a larger, preferably clinical, sample to derive solid implications for clinical practice.By design, a sub-clinical sample incorporates confounding alternative explanations (e.g.heterogeneity in diagnoses, pain intensity and disability).Furthermore, we conducted an experimental procedure on acute pain perception, which questions the generalizability of our results and potential implications on (more often than not chronic) pain conditions.Our design might also have influenced our results in an unwanted way: Subjective pain intensity and unpleasantness ratings are dependent on the objective pain tolerance, as participants could stop the heat pain stimulus at any moment which might suppress potential changes in subjective pain intensity and unpleasantness.However, ANCOVAs on the pain intensity and unpleasantness ratings controlling for pain tolerance did not show any different results.In our rationales, we paired the content with different interaction styles, which might have impacted our results.Future studies could address this with a more appropriate 2 × 2 design.Other limiting factors might be potential selection biases and unblinding due to the small recruitment pool.Moreover, the ethics committee required us to include information about a possible application of an ointment without active ingredient in the informed consent.As this information was processed before the information disclosed by the rationales, participants might have questioned the reliability of the rationales from the beginning, preventing potentially larger changes in expected or experienced pain ratings.Possibly, differences between treatment providers influenced our results.Although these results are based on very small subgroup sample sizes, the findings correspond to results of other researchers (Daniali & Flaten, 2019).Finally, the actual suggestion of the rationales, designed to induce positive treatment expectations, was rather short (about 5 minutes).Although research shows that short interactions can evoke placebo effects (Schneider et al., 2020), most OLP studies had longer clinical encounters (Carvalho et al., 2016;Kaptchuk et al., 2010;Kleine-Borgmann et al., 2019).Expectations are composed of different conscious and sub-conscious dimensions (Laferton et al., 2017).Possibly, our rationales did not fully address all expectation dimensions or our expectation measurement assessed only a certain aspect of participants' expectations -resulting in biased measurements or incomplete expectation formation.

| Conclusion
Our rationale styles changed participants treatment expectations, but the open-label placebo did not show any effects over the control group receiving the same inert ointment without a rationale.This suggests that positive expectations not sufficient to elicit OLP effects on experimentally induced pain.

F
I G U R E 2 Results of perceived warmth and competence.CG, control group; PE, Personal-emotional rationale; SM, Scientific-matter-offact rationale.Error bars represent Standard Deviations.*Significant bootstrapped effects.