Merely Possessing a Placebo Analgesic Improves Analgesia Similar to Using the Placebo Analgesic

Abstract

Background

Placebo analgesia studies generally reported that the actual use of a placebo analgesic reduces pain. Yeung, Geers, and Kam found that the mere possession (without use) of a placebo analgesic also reduces pain.

Purpose

We investigated the relative effectiveness of using versus possessing a placebo analgesic on pain outcomes.

Methods

In Study 1a, 120 healthy adults were randomized to either the experimental (EXP) conditions (EXP1: used a placebo analgesic cream, EXP2: possessed a placebo analgesic cream) or control (CO) conditions (CO1: possessed a sham cream, CO2: no cream). All participants underwent a cold pressor test (CPT). Study 1b further delineated the effect of possession from the effect of use. Sixty healthy adults were randomized to either the placebo-possession condition (merely possessed a placebo analgesic cream) or the placebo-possession-use condition (possessed and used a placebo analgesic cream). All participants did a CPT.

Results

In Study 1a, as expected, a placebo effect was found—participants who used a placebo analgesic cream showed better pain outcomes than the two CO groups. Surprisingly, participants who merely possessed a placebo analgesic cream performed equally well as those who actually used it. In Study 1b, participants in the two conditions did not differ in most pain outcomes. Participants who possessed and used a placebo analgesic cream only showed slightly more reduction in pain intensity compared to participants who merely possessed the placebo analgesic cream.

Conclusions

Our results suggest that merely possessing a placebo analgesic could enhance pain outcomes similar to that of applying the placebo analgesic.

Introduction

Over the last five decades, substantial research has been conducted on the placebo effect and its mechanisms [1, 2]. Placebos have been found to be administered in clinical care [3] and the placebo effect has been widely applied in the fields of medicine and nursing [4]. A placebo effect occurs when an inert substance/treatment without pharmacological effect leads to subsequent psychological/physiological changes benefiting the individuals [5]. One important procedure leading to such placebo effects is that individuals go through the stage of “using” the placebo substance/treatment. Prior studies had provided substantial evidence of the positive changes after using a placebo [6, 7]; however, little research examines the effect of a placebo before its actual use. A recent study by Yeung et al. [8] showed that merely possessing a placebo analgesic reduced individuals’ pain intensity. In this paper, we aim to compare the relative effectiveness of using versus possessing a placebo analgesic on pain outcomes.

Conceptual Base of Mere Possession Hypothesis and Its Preliminary Support

A considerable amount of social-psychological research has shown that having an association (e.g., owning, accessing, touching, or interacting) with an object alters people’s cognitive beliefs toward the object or the self. For instance, when people owned (vs. did not own) an object, they perceived the object to have more positive attributes [9, 10] and the self to have elevated ability, confidence, and perceptions of efficacy [11, 12]. Furthermore, having access to an object reduced the feeling of discomfort [13] and improved task performance [14]. In other studies, touching or interacting with an object changed self-perceptions [15] and domain-specific performance [16].

These self-object-association findings were also extended into the realm of pain analgesia. Three studies showed that having an association with an analgesic affected pain perception in healthy adults. The first study was conducted by Jackson et al. [17] who randomly assigned participants to have exposure to either a pain-medical object or a pain-irrelevant object. Participants who were exposed to the pain-medical object (vs. the pain-irrelevant object) reported lower pain intensity after 30 s of a cold pressor test (CPT). In another study, Rutchick and Slepian [18] randomly assigned participants to examine (i.e., interact with) either an anti-inflammatory object or a control object. Participants who interacted with the analgesic reported experiencing lower pain intensity and longer pain tolerance in CPT. The third study was conducted by Yeung et al. [8] who randomly assigned participants to either possess or not possess a placebo analgesic. Participants who merely possessed the placebo analgesic reported experiencing lower levels of pain intensity from a CPT compared to participants who did not possess it (Cohen d = 0.77). Interestingly, their finding was neither due to the mere exposure or physical touching of the placebo analgesic nor due to the positive emotions reported by the participants.

Although Yeung et al.’s [8] study provided preliminary support that merely possessing a placebo analgesic reduces pain perception, there are a few weaknesses to be outlined. First, there was no baseline pain assessment to test any preexisting group differences in pain-level prior to CPT. Second, no sham object (i.e., a pain-irrelevant object) was included, which would allow for the elimination of an alternative hypothesis: the observed effect can be due to a change in mood deriving from receiving a gift. Third, only retrospective pain experiences (not real-time reports) were assessed and the measurement was only one time (not multiple times). Lastly, the study focused on pain intensity and neglected the wider scope of pain-relevant indicators (e.g., pain threshold and pain tolerance).

Present Study

The central goal of the present research is to determine how using versus possessing a placebo analgesic influences pain outcomes. Based on Yeung et al.’s study [8] discussed above, we designed a new set of experiments by including (a) baseline pain assessment; (b) a sham object; (c) multiple measurements of real-time and retrospective pain; and (d) various pain-relevant indicators to capture pain magnitude. Importantly, we included a condition in which participants actually used a placebo analgesic.

Two studies (Studies 1a and 1b) were conducted. In Study 1a, participants were randomized to either the experimental (EXP) conditions (EXP1: used a placebo analgesic cream, EXP2: merely possessed a placebo analgesic cream) or control (CO) conditions (CO1: possessed a sham cream, CO2: no cream). In order to assess the possession effect, we compared the performance of participants who possessed the placebo analgesic cream (EXP2) versus those who did not possess the placebo analgesic cream (CO1 and CO2). Participants in all conditions simultaneously had prior exposure to the placebo analgesic cream; as such, any differences in their subsequent pain experiences would be attributed to their differential possession status (i.e., whether they owned the analgesic cream or not). We expected that participants who merely possessed a placebo analgesic cream would report experiencing less pain compared to participants who did not possess the placebo analgesic cream. In order to assess the effect of using a placebo analgesic cream, participants in EXP1 (placebo use) condition were compared to participants in the two CO conditions. We expected a standard placebo effect: participants who used a placebo analgesic cream would report experiencing less pain than participants in the two CO conditions. In order to assess the relative effectiveness of using versus possessing a placebo analgesic cream on pain analgesia, we compared EXP1 (placebo use) and EXP2 (placebo possession) condition. If using a placebo analgesic is more effective than possessing a placebo analgesic, participants in EXP1 should perform better than EXP2. No differences between these two EXP conditions mean that the effect of possessing a placebo analgesic can be similar to that of using it.

Study 1b is a follow-up study. We further tested the effect of possessing versus using by directly comparing two groups: placebo-possession group (merely possessed a placebo analgesic cream) versus placebo-possession-use group (possessed and used a placebo analgesic cream). Participants in these two groups were virtually exposed to a placebo analgesic cream and possessed the placebo analgesic cream; the only difference between the two groups is whether they actually used the cream or not. Any group difference in subsequent pain experiences would then be attributed to the effect of using the cream. If the placebo-possession-use group reports greater positive pain outcomes than the placebo-possession group, it suggests that actual consumption of a placebo is essential to placebo treatment. No difference between the two groups suggests that the effect of using a placebo analgesic can be similar to that of possessing it.

Given that using/possessing an analgesic may induce a positive expectancy that elevates one’s belief in self-efficacy and affect, or reduce one’s anxiety, all of which would in turn lessen pain experience [19–21], psychological-state variables such as participants’ expectancy, perceived self-efficacy to cope with pain, affects, and state anxiety were measured.

Methods

Study 1aParticipants

We conducted a power analysis [22] based on the results of a previous experiment [8]. Results showed that with power = .95, α = .05, and ηp² > .13, we should have at least a total sample size of 119 participants. We recruited 120 participants. Upon receiving the ethical approval from the Research Ethics Committee of Lingnan University, participants were recruited by advertising via email and flyers posted on the university campus. Those who were interested to participate completed the registration online via the Qualtrics system. We followed past studies’ screening criteria [23, 24] to determine participants’ eligibility for a CPT. People who are pregnant, who are smokers, who had an injury on their nondominant hand, and who have any existing pain condition, hypertension, circulatory disorders, diabetes, chronic kidney disease, history of cardiac events, fainting, seizures, alcohol or drug abuse, previous cold injury, current use of analgesics, opioids, tranquilizers, antidepressants, blood pressure medication, oral contraceptives or hormones, and allergic condition to analgesics were not eligible to participate in the study. We also excluded people whose working environment involves exposure to extreme heat/cold stimuli, which might affect pain sensitivity [25]. One hundred and twenty healthy adults (86 females, M_age = 24.5, SD = 6.36) eligible to participate in the study were randomly assigned to one of the four conditions: two EXP conditions (using a placebo analgesic cream, n = 30; or merely possessing a placebo analgesic cream, n = 30) and two CO conditions (possessing a sham cream, n = 30; or no possession, n = 30). Participants received HK$100 (approximately US$12.75) as a remuneration for their participation.

Procedure

Online registration and survey

Participants were informed that the aim of the study is to investigate the relationship between personality and Asians’ pain perception. They registered online via the Qualtrics platform where they confirmed that they do not meet the exclusion criteria of the study. Qualified participants read the informed consent form and clicked “Agree” to indicate their consent. They completed a number of scales measuring their personality variables (see Supplementary Material). Upon completion, they obtained a code to be used throughout the study to maintain anonymity. They chose a date and time to attend a laboratory session.

Laboratory session

Participants confirmed that they did not have any recent injury in their nondominant hand and had not taken any pain killers prior to entering the lab. They were welcomed by a male experimenter in a white lab coat who, then, led them to a cubicle sized 265 × 225 cm. The cubicle had limited furniture, including a desk, a chair, a display board, a cold pressor device (cooling thermostat Model: Lauda RA12, Bath opening: 300 × 190 × 160 mm, temperature control: ±0.05°C) filled with 10°C water, and two identical containers, one was empty and one was filled with room-temperature water (M = 21.67°C, SD = 0.62°C). Participants went through the following procedures.

Cover story.

Following Yeung et al. [8], participants read a cover story that a U.S. medical center invented an effective analgesic cream with a new formula. The manufacturer of the analgesic cream intended to develop its market in Asia and, thus, needed to understand pain perception in Asians so as to adjust the active ingredient of the cream to accommodate Asians’ physical needs and save production cost. Participants were told that, in order to determine their pain perception, they need to do a CPT. They were instructed to perform a practice trial of CPT to get familiar with the procedure and to obtain baseline data.

Practice trial and baseline measures.

To begin, participants read the step-by-step procedures of the CPT. The experimenter also verbally explained the procedures one more time to make sure that the participants understand the procedures and instructions. Then, participants proceeded to do a practice trial of CPT using an empty container. Participants were asked to imagine that the empty container was filled with water and were taught the proper position of placing their hand in the container. Participants learned to provide pain intensity responses using various formats. These included to verbally report their current pain intensity level using Numerical Rating Scale (NRS) every 15 s for three times when their hand was inside the container; to provide written responses using NRS, Faces Pain Scale-Revised, Short Form of the McGill Pain Questionnaire (SF-MPQ), and pain unpleasantness Visual Analog Scale (VAS) after their hand withdrew from the container. In addition, participants also chose one of the six options (0 = “No Pain,” 1 = “Mild,” 2 = “Discomforting,” 3 = “Distressing,” 4 = “Horrible,” and 5 = “Excruciating”) to describe the worst toothache, headache, and stomach ache they ever experienced. This gave a mean reference pain intensity score for each participant (Cronbach’s α = .59). The above practice procedures allowed participants to get familiar with various measurement formats and, more importantly, provided baseline data prior to the upcoming real CPT.

A purported marketing survey.

Next, participants were informed that the real CPT was about to begin. The experimenter pretended to check the status of the cooling thermostat of the cold pressor device and then apologized to the participants that the device was not ready yet and suggested them to complete a short marketing survey first while waiting for the device to be ready. All participants agreed to do the marketing survey. They were asked to evaluate a clinical analgesic product. They first read an article to understand what pain is and why it is important to reduce pain. This aimed to induce in participants a motive to reduce pain. Participants read some fictitious information about the disadvantages of experiencing pain and the advantages of reducing pain. They read the advice about using analgesic products to reduce pain. Afterward, they completed a comprehension test. Only those participants who answered correctly in the comprehension test were included in the analyses. Participants’ pain reduction motive was measured by three items (e.g., “When you experience physical pain, will you try to reduce it?”) using a seven-point scale (1 = “Absolutely not,” 7 = “Absolutely yes”). Scores of these three items were averaged to give a mean pain reduction motive score (Cronbach’s α = .70). After that, participants were presented with a professionally designed promotion leaflet of a fictitious branded analgesic cream. The leaflet documented the analgesic cream as an effective pain-relief product. Participants learned the ingredients of the analgesic cream, its pain-relief mechanism, and its effectiveness. They also read the guidelines for proper use and storage of the product. Participants were then visually presented with the physical object of the analgesic cream on their desk. The product was placed inside a 7.3 × 7.3 × 8.8 cm transparent display box so that participants were exposed to the product but unable to touch/use it. Participants were given 30 s to inspect the product. They were reminded not to open the box during their inspection to avoid any unnecessary spillage. They then completed a product evaluation questionnaire on a seven-point scale. They rated their expectancy of the effectiveness of the product (seven items, Cronbach’s α = .87, e.g., “I estimated the pain-relief power of the branded analgesic cream is: 1 = very weak, 7 = very strong”). To mask the task as a marketing survey, they were also asked about their purchasing intention, opinion on the product’s package design, general impression of the product, and behavioral intention of using the product. They also estimated a market price for the product and rated their attitudes toward analgesic cream in general. In order to reduce demand characteristics, the target item of analgesic cream was embedded along with other distractor items, such as hand cream, antistretch mark cream and etc.

Randomization

After the marketing survey, the experimenter used a random number generator to randomize participants to either the EXP conditions (EXP1 or EXP2) or CO conditions (CO1 or CO2).

EXP1 (placebo use) condition

Participants were invited to try the analgesic cream. The experimenter wore a transparent glove, opened the box of the analgesic cream (in fact, moisturizing cream), and applied a layer of the cream on the nondominant hand of the participants.

EXP2 (placebo possession) condition

The experimenter told the participants that, in order to thank them for completing the marketing survey, they were given one free product sample from the manufacturer via a random draw. Such random draw arrangement reduced potential demand effects [9]. Participants received a sample of 5 mL analgesic cream (moisturizing cream), which was placed inside a white, plastic, medicinal-looking cylinder-shaped box with a diameter of 3.3 cm and height/depth of 1.6 cm. The box was further stored inside a 7 × 9 cm transparent sealed plastic bag with a label on the bag showing the product’s brand, name, ingredients, and use directions. Participants were instructed not to open the sample to avoid spillage. They signed a receipt to acknowledge their ownership over the cream. They were encouraged to anticipate how the given cream enhances their pain resilience by writing down at least three benefits.

CO1 (sham possession) condition

Participants received a free sample of an antistretch mark cream via a random draw as a token of appreciation for completing the marketing survey. In fact, the free sample was identical to the one given in EXP2 (only the name and ingredients printed on the product label has changed). Participants signed a receipt to acknowledge ownership over the sham cream.

CO2 (no possession) condition

After the marketing survey, participants did not receive any gift.

Regardless of conditions, all participants completed the Positive and Negative Affect Schedule (PANAS) Scale to measure their momentary feelings. They also completed the Self-efficacy of Pain Resilience Scale, Expected Pain Intensity Rating Item, Expected Pain Severity Rating Item, and State-Trait Anxiety Inventory-State Scale (STAI-Y1). After that, they proceeded to do the real CPT.

Cold pressor test

Participants were instructed to immerse their nondominant hand into a container filled with room-temperature water for 1 min to stabilize and ensure similar skin temperature. After 1 min, they immediately immersed the same hand into the cooling thermostat device filled with cold water at 10°C; the participants did not know the temperature (10°C is chosen based on the result of a pilot study [n = 5]—this temperature was generally regarded by Chinese participants as “cold” and, at the same time, they were able to physically endure its coldness for 3 min. Any variations in the subsequent performance of the CPT, thus, cannot be explained by the physical incapability of participants but by the differential experimental treatments). Participants were asked to keep their hand submerged for as long as they can for a maximum period of 3 min. During CPT, participants were instructed to ring a bell the moment they experienced a first clear sensation of pain. The time latency between the immersion and bell ringing was recorded by a timer, indicating participants’ pain threshold [26]. They were instructed to notify the experimenter and withdraw their hand when they could no longer endure the pain. The time latency between hand immersion and hand withdrawal were recorded, indicating participants’ pain tolerance [23]. Once participants immersed their hand, they were asked to provide a verbal pain intensity rating using an NRS from 0 = “No pain at all” to 10 = “Extreme Pain” every 15 s until they could no longer withstand the pain or 3 min had passed, whichever was earlier. The experimenter recorded the ratings. Participants also reported their pain intensity the very moment they withdrew their hands. This gave the immediate pain intensity index. All these yielded a real-time measurement of pain intensity [17, 27].

Measurements after the CPT

Participants completed a questionnaire, which consisted of a written NRS, SF-MPQ, Faces Pain Scale-Revised, and pain unpleasantness VAS to assess their retrospective perception of pain intensity. In the end, they completed the Marlowe–Crowne Social Desirability Scale [28] (translated by Xie [29]). They also indicated whether they received any gift sample and, if yes, what it was (manipulation check). They guessed the water temperature and research hypothesis and were probed for suspicion. Since the study involved deception, the participants were thoroughly debriefed of the purpose of the study at the end of each individual participation in the study. The debriefing was conducted verbally and individually inside the lab, and participants were also provided with a debriefing paper. A follow-up email was also sent to ask if participants have any further inquiries. Figure 1 depicts a flowchart diagram of experimental procedures.

It should be noted that, since the debriefing was conducted by the experimenter, as such, the experimenter was aware of the placebo nature of the study. Nevertheless, the experimenter was blinded to the hypotheses under investigation and, following the past study [18], the experimenter–participant interactions were governed strictly by a standardized script, thus, minimizing any potential experimenter effects.

Measures

All materials and inventories were in Chinese (either adopted the well-established Chinese version or translated into Chinese by researchers with procedures of back translation to ensure validity).

Scales measuring psychological-state variables

PANAS Scale

[30]. The Positiive and Negative Affects Schedule (PANAS) Scale translated by Huang et al. [31] was used to measure participants’ affect. Participants read 10 adjectives describing a positive mood (e.g., “Enthusiastic”) and 10 adjectives describing a negative mood (e.g., “Nervous”). They rated the extent to which they felt the described emotions on a five-point scale (1 = “Very Slightly or Not at All,” 5 = “Extremely”). Mean scores of positive (Cronbach’s α = .88) and negative (Cronbach’s α = .77) affect were obtained.

Self-Efficacy of Pain Resilience Scale

A 16-item Self-Efficacy Scale was used to assess participants’ perceived self-efficacy in dealing with pain. Nine items were adopted from the CPT Self-Efficacy Scale [26] (translated by Wang [32]) to measure how well participants expected they could cope with CPT (e.g., “During the CPT, I could manage any discomfort”). Four items were adopted from the Self-Efficacy for Induced Cold Pressor Pain Scale [33] to measure participants’ perceived confidence in successfully completing the CPT (e.g., “How sure are you that you can successfully complete the CPT?”). Three items were created by the researchers to assess participants’ perceived ability to complete CPT (e.g., “It is easy for me to complete the CPT”). Participants rated their agreement on a seven-point scale (1 = “Strongly disagree,” 7 = “Strongly agree”; Cronbach’s α = .94). A mean score of self-efficacy was obtained.

Expected Pain Intensity Rating Item

[34]. Participants rated their expected pain intensity (from 0 = “No Pain at All,” 5 = “Moderate Pain,” to 10 = “Extreme Pain”) on the statement “How much pain do you expect during the CPT?”

Expected Pain Severity Rating Item

[35]. Participants rated their expected pain severity level (from 1 = “No Pain at All,” to 5 = “Excruciating”) on the statement “What would be the severity of pain you expect to experience when you put your hand into the cold water?”

State-Trait Anxiety Inventory: State Scale

[36]. The STAI-YI was translated by Yeung et al. [8]. Twenty statements assessed participants’ present feelings; 10 of them indicated the absence of anxiety (e.g., “I feel calm”) and 10 of them indicated the presence of anxiety (e.g., “I am tense”). Participants rated the degree to which the described statements applied to them on a four-point scale (1 = “Not at all,” 4 = “Very much so”). Scores in the anxiety-absent items were reversed and then added to the scores of the anxiety-present items to obtain a mean score of state anxiety (Cronbach’s α = .92).

Scales measuring pain

Following past studies, we adopted several types of pain measures that have different properties (e.g., scaling and number of items) in order to provide a richer and more comprehensive understanding of participants’ pain intensity, severity, and quality, as well as to capture their instant (real time) and recalled (retrospective) pain experiences.

Scales measuring real-time pain intensity during CPT

Modified (verbal) NRS

[37]. Participants verbally produced a number from a 11-point scale (from 0 = “No Pain at all,” 5 = “Moderate Pain,” to 10 = “Extreme Pain”) to indicate their pain intensity every 15 s once they immersed their hand into the cold water until they could no longer endure the pain or 3 min had passed, whichever was earlier. A mean verbal NRS was computed.

Immediate pain intensity index

Participants gave a number (from 0 = “No Pain at all,” 5 = “Moderate Pain,” to 10 = “Extreme Pain”) immediately upon hand withdrawal to provide an immediate pain intensity index.

Scales measuring retrospective pain intensity after CPT

Self-reported written NRS

[37]. Participants dried their hand and provided a written NRS to indicate a post-CPT pain intensity (from 0 = “No Pain at all,” 5 = “Moderate Pain,” to 10 = “Extreme Pain”).

Short Form of the McGill Pain Questionnaire

[35]. Participants’ perceived pain intensity, pain severity, and pain quality were measured by SF-MPQ (translated by Hsieh) [38]. For pain intensity, participants selected one of the six options (0 = “No Pain,” 1 = “Mild,” 2 = “Discomforting,” 3 = “Distressing,” 4 = “Horrible,” and 5 = “Excruciating”) to describe their current pain, least pain, and worst pain during CPT. A mean MPQ pain intensity score was obtained (Cronbach’s α = .75). For pain severity, participants marked a cross on a 10 cm long line that had a label of “No Pain at All” printed to the left end and “Extreme Pain” printed to the right end. The distance from “No Pain at All” to the cross marked by the participants indicates their pain severity. Each participant then has an MPQ pain severity score. For pain quality, participants were given 15 adjectives describing pain (e.g., “Stabbing,” “Hot-Burning,” and “Fearful”) and rated their experience on a four-point scale (0 = “None,” 1 = “Mild,” 2 = “Moderate,” and 3 = “Severe”). A mean MPQ pain quality score was obtained (Cronbach’s α = .86).

Faces Pain Scale-Revised

[39]. Participants saw six different faces that represent an increasing pain intensity level (left-most face “0” = “No Pain,” right-most face “5” = “Very much pain”). Participants circled a face that best represents the pain they felt during CPT. Each participant had a face pain intensity score.

Pain unpleasantness VAS

[40]. Participants marked a cross on a 10 cm horizontal line, with left end point anchored to “Not unpleasant at all” and right end point anchored to “Extremely unpleasant.” The longer the distance of the marked cross from the left end point, the more unpleasant the feeling experienced.

Statistical Analysis

The data were analyzed using SPSS 23. Means and Standard Deviations (SDs) were calculated for dependent measures in each condition and are presented in Tables 1 and 2. To determine the effect of condition on the dependent measures, one-way analyses of variance (ANOVAs) were conducted with condition (EXP1, EXP2, CO1, CO2) as the independent variable and each of the pain outcomes as the dependent variable. If any omnibus F-test was significant, post hoc comparisons were performed to determine which conditions differed significantly. If the assumption of homogeneity of variance was not violated, the Tukey’s honest significant difference test was used to compare conditions. If the assumption of homogeneity of variance was violated, the Games–Howell post hoc test was used to compare conditions. The Shapiro–Wilk test was performed to test for normality. In the case of violating normality, Welch’s F-test was further conducted and omega square was reported. All significant tests were two tailed with α set at .05.

Results

Manipulation checks and screening

All participants passed the article comprehension test and correctly indicated their possession status (i.e., whether they received any gift sample and what it was). When asked to guess the hypothesis, one participant from EXP1 (placebo use) condition mentioned the term “placebo effect”; his data were discarded. Also, one participant from CO2 (no possession) condition completed the post-CPT scales in a wrong order; his data were also discarded, leaving 118 participants for data analysis (EXP1: n = 29, EXP2: n = 30, CO1: n = 30, CO2: n = 29). To begin with, all participants showed equally high motive to reduce pain (M = 4.99, SD = 1.05) and believed that the presented analgesic cream is effective (M = 4.89, SD = 0.78), ps = ns (see Table 1). Participants did not show any significant difference in their guessing of water temperature (M = −1.23°C, SD = 10.25) and social desirability tendency (M = 6.71, SD = 2.15), ps = ns (findings of personality variables and product evaluation are presented in Supplementary Material).

Psychological states after randomization

ANOVAs were conducted with condition as the independent variable and each of the psychological-state variables as the dependent variable. Our data showed that regardless of which condition participants were assigned to, they shared similar levels of emotional responses: positive affect (M = 2.24, SD = 0.67), negative affect (M = 1.36, SD = 0.32), and state anxiety (M = 2.06, SD = 0.48), all ps = ns. They also expected to experience a moderate level of pain intensity (M = 6.22, SD = 1.46) and perceived themselves to have equal level of efficacy to deal with pain (M = 5.22, SD = 0.79), all ps = ns. However, participants differed in their expected level of pain severity, F(3,114) = 3.81, p = 0.01, ηp2 = 0.09, observed power = 0.81, CI_90% = [0.01, 0.16]. The Tukey post hoc test showed that participants in EXP2 (placebo possession) condition expected to experience greater pain severity (M = 3.07, SD = 0.94) than participants in CO2 (no possession) condition (M = 2.52, SD = 0.74), p = 0.04, and participants in EXP1 (placebo use) condition (M = 2.45, SD = 0.63), p = 0.02 (see Table 1).

Baseline, real-time, and retrospective pain measures

One-way ANOVAs were conducted to compare if there are any significant group differences in the baseline (pre-CPT), real-time (during CPT), and retrospective (post-CPT) pain performance (see Table 2). Concerning the baseline measures, during the practice trial, all participants reported the ratings of “0” in pain intensity based on verbal NRS and Faces Pain Scale-Revised. This indicated that they all started with a baseline feeling of no pain. Participants did not differ in their baseline mean MPQ score (M = 0.03, SD = 0.28), unpleasantness VAS rating (M = 0.03, SD = 0.15), and reference pain intensity rating (M = 3.24, SD = 0.76), all ps = ns. Basically, participants did not differ in all pre-CPT pain measures.

During CPT, participants’ real-time responses were recorded. Separated ANOVAs were conducted with condition as the independent variable and each of the real-time measures (pain threshold, pain tolerance, pain intensity [verbal NRS], and immediate pain intensity index) as the dependent variable, respectively.

First of all, a significant main effect of condition on pain threshold was found, F(3,114) = 10.03, p < 0.001, ηp2 = 0.21, observed power = 0.998, CI_90% = [0.09, 0.30]. The Shapiro–Wilk test showed that the data were departed from normality, W(118) = 0.85, p < 0.001. Welch’s test was, therefore, conducted. The result of Welch’s test indicated that the main effect of condition on pain threshold was statistically significant, Welch’s F (3, 62.43) = 10.27, p < 0.001, ω ² = 0.19, implying that participants in the four conditions differed significantly in their pain threshold. Levene’s test showed that equality of variances was not met, F(3,114) = 4.04, p = 0.009; the Games–Howell post hoc test was thus performed to determine which conditions differed significantly. Results revealed that participants in the placebo use condition (EXP1), as expected, showed a greater pain threshold—they took longer time to report an initial pain sensation (M = 21.66, SD = 10.88) than participants in the two CO conditions (CO1: M = 11.43, SD = 8.74, p = 0.001; CO2: M = 9.90, SD = 9.52; p < 0.001). Interestingly, participants in the placebo possession condition (EXP2) performed equally well (M = 22.73, SD = 15.51) as those in the placebo use condition (EXP1); that is, participants shared similar pain threshold regardless of whether they used or possessed a placebo analgesic cream, p = 0.99. Like EXP1, participants in EXP2 also reported a significantly greater pain threshold than the two CO conditions (CO1: p = 0.006; CO2: p = 0.002). The two CO conditions (CO1 and CO2) did not differ from each other, p = 0.92 (see Fig. 2).

Second, a significant main effect of condition on pain tolerance was also found, F(3,114) = 4.71, p = 0.004, ηp2 = 0.11, observed power = 0.89, CI_90% = [0.02, 0.19]. The Shapiro–Wilk test for normality indicated that the data were not normal, W(118) = 0.73, p < 0.001. Welch’s test was therefore conducted. Results revealed that the main effect of condition on pain tolerance was statistically significant, Welch’s F (3, 62.83) = 5.69, p = 0.002, ω ² = 0.09, implying that participants in the four conditions differed statistically in their tolerance level. Levene’s test revealed that the homogeneity of variance assumption was not met, F(3,114) = 9.20, p < 0.001. The Games–Howell test was thus performed. Results revealed that participants in EXP1, who used the placebo analgesic cream, were able to endure pain longer (M = 156.55 s, SD = 52.60) than participants in the two CO conditions (CO1: M = 100.73 s, SD = 73.01, p = 0.007; CO2: M = 99.03 s, SD = 72.79, p = 0.006). Interestingly, participants who merely possessed the analgesic cream (EXP2) did equally well (M = 122.93 s, SD = 66.09) as their counterparts in EXP1, p = 0.15; yet, their pain tolerance did not differ from the two CO conditions, ps = ns. The two CO conditions did not differ from each other, p = 1.00 (see Fig. 3).

Third, our participants verbally reported their real-time pain intensity every 15 s; their mean pain intensity (verbal NRS) score was submitted to a one-way ANOVA. Results showed that participants across the four conditions did not differ significantly in their mean verbal NRS score, F(3,114) = 0.35, p = 0.79, ηp2 = 0.01, observed power = 0.12, CI_90% = [0.00, 0.03].

Lastly, participants reported their pain intensity rating immediately upon hand withdrawal, a one-way ANOVA of condition on this immediate pain intensity index was conducted. Results revealed a statistically significant main effect of condition, F(3,114) = 10.78, p < 0.001, ηp2 = 0.22, observed power = 0.999, CI_90% = [0.10, 0.31]. The Shapiro–Wilk test indicated that the data were not normally distributed, W(118) = 0.94, p < 0.001. Welch’s test was further conducted. Results indicated that the four conditions differed significantly in their immediate pain intensity index, Welch’s F (3, 63.12) = 10.26, p < 0.001, ω ² = 0.20. Levene’s test indicated that equality of variances was met, F(3,114) = 1.04, p = 0.38. The Tukey post hoc test was thus performed. Results revealed that participants in EXP1 reported a significantly lower immediate pain intensity index (M = 5.59, SD = 2.16) than participants in the two CO conditions (CO1: M = 7.93, SD = 1.98, p < 0.001; CO2: M = 7.62, SD = 1.76, p = 0.001). Again, participants in EXP2 performed equally well (M = 5.97, SD = 1.83) as those in EXP1; their immediate pain intensity index did not differ, p = 0.86. Like EXP1, participants in EXP2 also reported a significantly lower immediate pain intensity index than participants in the two CO conditions (CO1: p = 0.001; CO2: p = 0.008). Participants in the two CO conditions (CO1 and CO2) did not differ from each other, p = 0.93 (see Fig. 4).

Other than the above-mentioned real-time responses, separate ANOVAs were also conducted with each of the retrospective pain responses (i.e., written NRS, SF-MPQ, face pain intensity, and pain unpleasantness).

Our data showed a significant main effect of condition on retrospective pain intensity rating (written NRS), F(3,114) = 11.49, p < 0.001, ηp2 = 0.23, observed power = 0.999, CI_90% = [0.11, 0.32]. The Shapiro–Wilk test indicated an abnormality of data, W(118) = 0.94, p < 0.001. Welch’s F-test showed that the effect of condition on the written NRS score was statistically significant, Welch’s F (3, 63.10) = 10.90, p < 0.001, ω ² = 0.21. Levene’s test assumed a homogeneity of variance, F(3,114) = 1.26, p = 0.29. The Tukey post hoc test was performed. As expected, participants in EXP1 who used the placebo analgesic cream recalled experiencing lower pain intensity—having a lower mean written NRS score (M = 5.55, SD = 2.11) than participants in the two CO conditions (CO1: M = 7.90, SD = 1.92, p < 0.001; CO2: M = 7.76, SD = 1.66, p < 0.001). Interestingly, participants in EXP2 who merely possessed the placebo analgesic cream shared a similar mean written NRS score (M = 6.10, SD = 1.83) as those in EXP1, p = 0.68. Like participants in EXP1, participants in EXP2 also reported a significantly lower mean written NRS score than participants in the two CO conditions (CO1: p = 0.002; CO2: p = 0.005). The two CO conditions did not differ from each other (p = 0.99) (see Fig. 5).

Concerning the responses in SF-MPQ, data were analyzed in terms of pain intensity, pain quality, and pain severity. Our data showed that participants did not differ in their MPQ pain intensity score, F(3,114) = 1.45, p = 0.23, ηp2 = 0.04, observed power = 0.38; CI_90% = [0.00, 0.09], and MPQ pain quality score, F(3,114) = 1.68, p = 0.18, ηp2 = 0.04, observed power = 0.43, CI_90% = [0.00, 0.10]. However, participants differed in their MPQ pain severity score, F(3,114) = 7.97, p < 0.001, ηp2 = 0.17, observed power = .99, CI_90% = [0.07, 0.26]. The Shapiro–Wilk test showed that the data were not normal, W(118) = 0.96, p = 0.002. Welch’s test revealed that the main effect of condition on pain severity was statistically significant, Welch’s F (3, 62.78) = 8.15, p < 0.001, ω ² = 0.15. Levene’s test indicated that equality of variances was met, F(3,114) = 1.99, p = 0.12. The Tukey post hoc tests revealed that participants in EXP1 experienced significantly lower pain severity (M = 5.24, SD = 2.25) than those in the two CO conditions (CO1: M = 7.21, SD = 2.01, p = 0.002; CO2: M = 7.47, SD = 1.59, p < 0.001). Again, participants in EXP2 (M = 5.88, SD = 2.25) shared a similar mean pain severity score as those in EXP1, p = 0.63, and, like EXP1, participants in EXP2 also reported having significantly lower pain severity than participants in CO2 (p = 0.02) and marginally lower pain severity than participants in CO1 (p = 0.06). The two CO conditions did not differ from each other (p = 0.96; see Fig. 6).

With regard to the face pain intensity score, a main effect of condition was found, F(3,114) = 3.50, p = 0.02, ηp2 = 0.08, observed power = 0.77, CI_90% = [0.008, 0.16]. The Shapiro–Wilk test indicated that the data were not statistically normal, W(118) = 0.93, p < 0.001. Welch’s test was conducted. Result indicated that participants in the four conditions differed significantly in their choice of facial representations, Welch’s F (3, 63.23) = 3.90, p = 0.01, ω ² = 0.06. Levene’s test showed that the homogeneity of variances assumption was met, F(3,114) = 0.62, p = 0.60. The Tukey post hoc test revealed that participants in EXP1 chose a relatively less painful facial representation (M = 2.34, SD = 1.01) than participants in CO2 (M = 3.24, SD = 1.09, p = 0.02), while participants in EXP2, CO1, and CO2 did not show any significant differences in their face choices, all ps = ns.

Lastly, participants across the four conditions did not show any significant difference in their pain unpleasantness VAS. The main effect of condition on unpleasantness VAS was not significant, F(3,114) = 1.18, p = 0.32, = 0.03, observed power ηp2 = 0.31, CI_90% = [0.00, 0.08].

It should be noted that participants in the placebo possession (EXP2) condition did not use the cream but were explicitly encouraged to anticipate the benefits of the given analgesic cream. One might argue that the pain reduction responses could be due to the forced anticipation processes brought about by the possession of an analgesic. A follow-up study (Study 1b) was conducted. In Study 1b, all participants were not encouraged to undergo any anticipation process. In order to further delineate the effect of possession from the effect of use, participants were randomized to either (a) the placebo-possession condition (merely possessed a placebo analgesic) or (b) the placebo-possession-use condition (possessed and used a placebo analgesic). In this design, both groups simultaneously possessed a placebo analgesic; the only difference was whether they used the placebo analgesic or not. Any group differences in subsequent pain responses would be attributed to the using of the placebo analgesic. No significant difference between the two groups would further support that merely possessing a placebo analgesic cream can be the same as actually using it.

Study 1bParticipants

Following identical recruitment and screening procedures as Study 1a, 60 eligible healthy adults (39 females, M_age = 23.88, SD = 5.96) were randomized to either the placebo-possession condition (n = 30) or the placebo-possession-use condition (n = 30). They received a remuneration (HK$100).

Randomization

Placebo-possession condition

All materials, instruments, and procedures were identical to the placebo possession (EXP2) condition of Study 1a except that participants were not given any explicit instruction and time to anticipate about the benefit of the gifted analgesic cream.

Placebo-possession-use condition

All materials, instruments, and procedures were identical to the placebo use (EXP1) condition of Study 1a except that participants were gifted the placebo analgesic cream on top of using the placebo analgesic cream.

Results

All participants passed the article comprehension test and correctly indicated their possession status. No participants guessed the hypothesis correctly. Participants did not differ in their social desirability tendency (M = 6.82, SD = 2.27) and guess of the water temperature (M = −2.11°C, SD = 10.07). They were all equally motivated to reduce pain (M = 4.99, SD = 1.14) and have a similar evaluation toward the effectiveness of the presented analgesic cream (M = 4.73, SD = 1.02), all ps = ns (see Table 3). Participants’ responses in personality variables and product evaluation are presented in Supplementary Material.

Psychological-state variables

Independent samples tests were conducted to compare the mean scores of the psychological-state variables between the two conditions. Our results showed that, after randomization, participants across the two conditions did not differ in their positive affect, negative affect, perceived efficacy, expected pain intensity, and state anxiety, all ps = ns. However, participants in the placebo-possession condition reported slightly higher expected pain severity (M = 2.77, SD = 0.57) than participants in the placebo-possession-use condition (M = 2.43, SD = 0.73), t(58) = 1.98, p = 0.05, d = 0.51, CI_95% = [−0.005, 1.02] (see Table 3). Expected pain severity was then submitted to simple regression to assess its impact on pain responses. Our data indicated that expected pain severity was not a significant predictor of any pain-relevant dependent variables, ps = ns.

Baseline, real-time, and retrospective pain variables

Independent samples test indicated that participants in both conditions did not differ in their pain baseline measures, most real-time measures (pain threshold, pain tolerance, and verbal NRS), and most retrospective measures (MPQ pain intensity, pain severity, pain quality, and pain unpleasantness VAS), all ps = ns (see Table 4). However, participants in the placebo-possession-use condition were found to perform marginally better than participants in the placebo-possession condition in three pain measures (immediate pain intensity index, retrospective pain intensity [written NRS], and face pain intensity). To elaborate, compared to participants who merely possessed a placebo analgesic, participants who possessed and used a placebo analgesic reported a marginally lower (a) immediate pain intensity index (M_{possession-use} = 5.70, SD = 1.69; M_possession = 6.83, SD = 2.61, t(49.54) = 2.00, p = 0.05, d = 0.52, CI_95% = [−0.002, 1.03]; (b) mean written NRS score (M_{possession-use} = 5.80, SD = 1.63; M_possession = 6.90, SD = 2.66, t(48.08) = 1.93, p = 0.06, d = 0.50, CI_95% = [−0.02, 1.01]; and (c) less painful face representation (M_{possession-use} = 2.33, SD = 1.12; M_possession = 2.90, SD = 1.19, t(58) = 1.9, p = 0.06), d = 0.49, CI_95% = [−0.03, 1.00]. All the p-values were marginally significant (p = 0.05 or p = 0.06).

In short, Study 1b attempted to delineate the effect of possession from the effect of use. Our results showed that participants who used a placebo analgesic could only slightly outperform participants who merely possessed a placebo analgesic in 3 out of 10 pain outcomes. And consistent with the results in Study 1a, it seems that merely possessing a placebo analgesic could enhance pain outcomes more or less similar to that of using the placebo analgesic (in 7 out of 10 pain outcomes).

Discussion

In Study 1a, a placebo effect was found—participants who used the placebo analgesic (EXP1) reported having greater pain threshold and pain tolerance, lower real-time and retrospective pain intensity and pain severity than the two CO groups. Surprisingly, participants who merely possessed the placebo analgesic (EXP2) performed equally well as those who actually used it—they shared similar pain threshold, pain tolerance, and real-time and retrospective pain intensity as the placebo-use group, and they also reported experiencing less physical pain than the two CO groups. Our data showed that even though participants in the placebo possession (EXP2) group expected to experience greater pain severity, the mere possession of a placebo analgesic seems to enhance their pain resilience, leading to performance similar to those who had used the placebo analgesic.

While Study 1a encouraged participants who possessed a placebo analgesic to anticipate its benefits (EXP2), in Study 1b, participants did not undergo such anticipation process; both groups (placebo-possession vs. placebo-possession-use) possessed an identical placebo cream; the only difference was whether they used it or not. Results showed that the effect of using was only marginally important (p = 0.05 and 0.06). The two groups did not show remarkable differences in most pain responses. Relative to those who merely possessed a placebo analgesic, participants who used a placebo analgesic only performed slightly better in their immediate pain intensity and retrospective pain intensity (written NRS and face VAS).

The purpose of including a sham condition (CO1) in Study 1a was to test a mood-induction hypothesis, that is, the results are due to a positive mood after receiving whatever free gift. Our data rejected such a hypothesis. First, all participants shared similar levels of positive affect regardless of condition. Second, participants who possessed the placebo analgesic cream (EXP2) showed improved analgesia than those who possessed the sham cream. The result was not due to the mere exposure effect because participants across all conditions were equally exposed to the identical placebo analgesic cream.

The current study extends the psychological phenomena of mere possession effect into the realm of pain analgesia. It provides a new perspective and adds further insights into what is already known on how placebo works. In past placebo studies, a placebo response involved both the possessing stage and consuming stage: patients first possess a placebo and, then, consume it and experience the expected effect [1, 2]. The current findings suggest that a placebo effect might have already been partly triggered during the possessing stage. The finding has practical implications for health professionals on how to improve clinical outcomes. For instance, dentists may consider letting patients possess a pain-killer prior to dental treatment.

Based on prior research on the placebo effect, conditioning [41, 42] and expectancy [43, 44] have been the two most appealing theories explaining placebo analgesic effects. With regard to our current experimental design, the conditioning explanation may not be useful to explain the result of the placebo possession condition because conditioned learning usually requires participants to repeatedly use the analgesic cream in order to learn the cause–effect relationship and produce an expected response. The expectancy theory, however, seems helpful in explaining the observed results. First, possessing/securing a placebo analgesic is likely to induce in the owners a positive expectancy of the effect of placebo analgesic on pain reduction, and such cognitive expectancy may lessen the experience or perception of pain. Second, according to Kirsch [45], upon administering the placebo analgesic cream, participants would form a positive expectation of pain reduction. It seems that regardless of whether participants just merely possessed or actually used a placebo analgesic, the expectancy mechanism is operating, allowing participants in these two conditions (EXP1 and EXP2) to enjoy pain analgesia. Indeed, our current data provide some support for this expectancy theory. In Study 1a, when participants who possessed a placebo analgesic were encouraged to undergo the benefit-anticipation process (to form positive expectancy), they performed equally well as those who used a placebo analgesic—the two groups did not differ in 10 out of 10 pain-relevant indicators. However, in Study 1b, when participants did not undergo the benefit-anticipation process (suppressing the formation of positive expectancy), participants who merely possessed the placebo analgesic performed slightly poorer than those who used the placebo analgesic. Our results showed that, although the two groups did not differ in 7 out of 10 pain-relevant indicators, the possession group did perform marginally poorer than the use group in 3 out of 10 indicators when they were prohibited from forming positive expectancy. This suggests that anticipation or expectancy may play a role upon mere possession to achieve a placebo effect. Future study could test the pure effect of anticipation by instructing participants to anticipate the benefit of an analgesic without possessing it.

There are some similarities between our current findings and that of the prior studies, which also used the CPT as a pain stimulus to measure pain sensitivity. For instance, Ng [46] and Chan et al. [23] asked Chinese participants to immerse their hand into the cold water—an experimental setting similar to our CO2 condition (Study 1a). Our participants’ levels of pain threshold, pain tolerance, and pain intensity were comparable to those of their participants. Also, Geers et al. [47] assigned American participants to either a placebo condition (used a placebo analgesic cream) or a control condition (used a sham cream)—an experimental setting similar to our EXP1 (placebo use) condition and CO1 (sham) condition (although our participants possessed, but did not use, a sham cream), respectively. Our result pattern was consistent and comparable to theirs—participants who used a placebo analgesic cream showed lower pain severity and greater pain tolerance than participants who used/possessed a sham cream.

Nevertheless, there was a discrepancy between our current finding and that of Yeung et al. [8]. Unlike Yeung et al. [8], we did not find a significant difference in MPQ pain intensity score between EXP2 (placebo possession) and CO2 (no cream) condition. Such discrepancy could be due to differential statistical treatment in correspondence to different study designs and procedures. In Yeung et al.’s study [8], participants were asked to immerse their hand for 1 min. This immersion duration was not a dependent measure but part of the study design. Since many of their participants (39%) failed to achieve the 1 min requested immersion, Yeung et al. [8] treated immersion duration as a covariate in their analyses. They also measured participants’ state anxiety after (but not before) the CPT and treated it as a covariate. In our study, our participants freely immersed their hand for as long as 3 min and state anxiety was measured before CPT; these two variables were not controlled for during data analysis. Thus, the discrepancy in the MPQ pain intensity results between the two studies might be due to differential study design, procedure, and analyses.

Rokke et al. [48] showed that American participants with high self-efficacy being provided with a choice of coping strategies displayed increased pain tolerance. In our study, we did not measure our Chinese participants’ baseline self-efficacy level, we only measured their perceived self-efficacy of pain resilience after randomized them to different conditions, and we did not find any significant correlation between perceived self-efficacy and pain tolerance (for more details, see Supplementary Material), nor did we find any group difference in perceived self-efficacy. This could be due to the effect of the experimenter’s instruction—in delivering the CPT’s instruction, the experimenter told the participants: “most people can endure 3 min, please try your best to immerse as long as you can.” Upon this instruction, participants might have been encouraged to believe that they are able to endure the pain as well as other people. As such, they all reported having high perceived efficacy to tackle CPT before they actually did the CPT. Future research could measure baseline self-efficacy level and/or omit such encouraging instructions.

The current research adopted multiple measures of pain intensity, including a continuous verbal report every 15 s, a verbal report immediately upon hand withdrawal, and a written report after hand withdrawal. Although the time of measurement (during immersion vs. post immersion) and mode of reporting (verbal vs. written) were different, the measurement scale was identical: 11-point NRS (from 0 = “No Pain at all,” 5 = “Moderate Pain,” to 10 = “Extreme Pain”). We found a consistent pattern across both Studies 1a and 1b: no significant condition difference in the continuous verbal rating but significant condition difference in the immediate verbal rating and retrospective written rating. In other words, group difference in pain intensity was not detected when participants’ hands were still immersed in the cold water (during immersion), but the group difference became more salient when their hands were away from the water (post immersion). This finding is consistent with other placebo analgesia studies [49–53], which also showed that the concurrent pain intensity measures were not as robust as that of the post pain measures. We speculated that since the task during immersion and post immersion did not pose a big difference in participants’ cognitive load (in both tasks, they need to pay attention to the sensation in their hand), the post immersion period may have allowed participants more time to reflect on their pain level and behavior and consolidated this information with their prior expectations [54]. As a result, a significant condition difference was found for immediate verbal rating and retrospective written rating (both post immersion). This possibility awaits further inquiry.

Future research needs to address a few issues. First, although multiple measurements allow us to have a more comprehensive understanding of participants’ pain perception and experiences, the increase in the number of measurements can increase the chance of false positivity in statistical testing. Appropriate cautions should be taken in interpreting the current findings and replication studies should be conducted. Second, future research could also adopt other pain-induction tasks (e.g., heat [55] or loud noise [56]) to increase the generalizability of the findings. Third, the cultural background might affect one’s pain perception [57, 58]. Our samples included healthy Chinese adults. Future studies could include participants of other ethnicities. Finally, it is important to include clinical and dental patients with temporary or chronical pain in an effort to generalize the current findings to real-life situations, as well as to fully clarify the clinical significance and importance of the observed mere possession phenomenon.

In conclusion, the current research adopted a pain placebo paradigm as a model to demonstrate that placebo effects not only would occur in medical patients who actively urge to seek treatment but also in healthy individuals who do not have the need to actively use an analgesic treatment [6]. The same results could be translated to other symptoms, such as smoking, anxiety, and depression, with relevant clinical and nonclinical implications from tapering drugs with side effects to help implementing cognitive and behavioral changes in lifestyles.

Acknowledgments

Funding: The research was supported by the General Research Fund (GRF13600617) awarded by the Research Grants Council, University Grants Committee, Hong Kong to V.W.L.Y. We thank Mr. Tam Pui Chuen and Mr. Eric Kenson Yau for providing research assistance.

Compliance with Ethical Standards

Authors’ Statement of Conflict of Interest and Adherence to Ethical Standards V.W.L.Y. and A.L.G. declare that they have no conflict of interest. L.C. discloses that she obtained grants from the National Institutes of Health, MPowering the State Grant and PCORI grant during the preparation of the manuscript, and honoraria from University of Michigan, Michigan Head Pain Institute, Emmi Solutions, and Cult Health LLC during the conduction of the study. All procedures, including procedures of informed consent, performed in the reported studies involving human participations were in accordance with the ethical standards of the university research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Authors’ Contributions V.W.L.Y. contributed to the study concept and design, data collection and data analysis, as well as drafting of the manuscript. A.L.G. and L.C. provided valuable advice on the experimental measures, statistics used and the final draft of the manuscript.

References