Abstract Background Social modeling has the capacity to shape treatment outcomes, including side effects. Purpose This study investigated the influence of social modeling of treatment side effects, gender, and participant empathy, on side effects of a placebo treatment. Methods Ninety-six participants (48 females) completed a study purportedly investigating the influence of modafinil (actually placebo) on alertness and fatigue. The participants were randomly seated with a male or female confederate and saw this confederate report experiencing side effects or no side effects. Participant empathy was assessed at baseline. Changes in modeled and general symptoms, and misattribution of symptoms, were assessed during the session and at 24-hr follow-up. Results During the experimental session, seeing side effect modeling significantly increased modeled symptoms (p = .023, d = 0.56) but not general or misattributed symptoms. Regardless of modeling condition, female participants seated with a female model reported significantly more general symptoms during the session. However, response to social modeling did not differ significantly by model or participant gender. At follow-up, the effect of social modeling of side effects had generalized to other symptoms, resulting in significantly higher rates of modeled symptoms (p = .023, d = 0.48), general symptoms (p = .013, d = 0.49), and misattributed symptoms (p = .022, d = 0.50). The experience of modeled symptoms in response to social modeling was predicted by participants’ levels of baseline empathy. Conclusions Social modeling of symptoms can increase the side effects following treatment, and this effect appears to generalize to a broader range of symptoms and symptom misattribution over time. Higher baseline empathy seems to increase response to social modeling. Symptoms, Side effects, Placebo, Nocebo, Social modeling Introduction The experience of unpleasant symptoms or side effects following medical treatment is exceptionally common. Recent evidence indicates that between 40% and 100% of all side effects are caused by “non-specific” factors associated with the treatment context, rather than by the active ingredients in the treatment itself . A large proportion of reported side effects are thought to be due to the nocebo effect: the experience of unpleasant physical symptoms or outcomes in response to an inert agent (e.g., a placebo treatment), caused by psychological mechanisms including negative expectations [2, 3]. While rates vary by treatment and disease, as many as 75% of placebo-treated patients in randomized controlled trials report unpleasant side effects, with many such patients withdrawing from studies because of these symptoms . Some of these reported symptoms may also be commonly experienced physical symptoms that are misattributed to the treatment, for example headaches, back pain, fatigue, and sleep difficulty [5, 6]. The experience of treatment side effects (whether they are causally related to the treatment or not) results in patient distress, increased use of medical services, and treatment nonadherence or discontinuation [3, 7]. Social modeling of treatment outcomes has emerged as an important contributor to the formation of both placebo and nocebo effects [8, 9]. In contrast to nocebo effects, placebo effects are the beneficial effects of a treatment, caused not by the drug itself but by psychological mechanisms (including positive expectations) that are elicited through the treatment context [10, 11]. Seeing others’ response to treatment can influence the viewer’s expectations about how effective the treatment is, and conversely, the modeling of adverse events can influence expectations about treatment side effects and cause an increase in nocebo responding . Research has explored this question both in pain paradigms [13–15] and other experimental setups involving either placebo medications, ointments, or exposure to other inert agents described as being potentially harmful [16–19]. Viewer expectations appear to mediate the influence of social information on the nocebo effect , although social modeling can contribute to nocebo effects when the modeling occurs outside of conscious awareness by subliminally viewing a model’s facial expressions in response to painful stimuli . Social modeling also occurs on a wider scale when the media’s coverage of health scares increases reported symptoms over a large section of the population. This has been shown in a study of adverse event reporting following a change in the formulation of thyroid medication, where following television news stories of patients’ adverse symptoms to the new medication formulation, the rates of reporting of adverse reactions increased dramatically . Media coverage of specific side effects was matched by those symptoms being reported in the following month’s adverse reaction reports, suggesting a close alignment between the type of symptom modeled and the response in the viewer . The social modeling of symptoms is also implicated as an important method of transmission in cases of mass psychogenic illness (MPI; ). One such incident occurred in Melbourne Airport in 2005, where 57 people became ill after a “gas leak” . The episode started with a dramatic index case of a woman fainting in a visible central location, which was later attributed to stress and lack of food. Subsequent victims reported feeling generally unwell, with some feeling faint and nauseous. Most casualties were female staff who worked in the airport terminal, and symptoms spread via line of sight, sound, and information, rather than by direct contact. Extensive testing of both the airport and patients could not find evidence of any toxic agent, and most symptoms resolved rapidly once patients left the building. MPI symptoms are thought to occur through nocebo-like processes driven by social modeling, negative expectations, and to heightened anxiety . As seen in the Melbourne Airport incident, MPI episodes are more common in closed social settings and are often set off by a dramatic initial case of someone falling sick. The symptoms are transmitted mostly by line of sight or communication from the initial case and usually resolve relatively quickly. Reviews of MPI have identified that women are more likely to be affected than men , suggesting that women may be more influenced by the modeling of symptoms in their social environment. To date, research exploring the impact of model and viewer gender on nocebo effects has been equivocal. In some studies, female participants paired with a female model demonstrated enhanced nocebo effects [16, 17]; however, these studies included only a female model. In another, nocebo effects were enhanced following modeling by a male confederate compared with a female . Across all three of these studies, face-to-face modeling was used. A follow-up to the study by Mazzoni et al.  also found that social modeling increased symptom reporting, and that both male and particularly female participants reported more symptoms in the presence of a same-gender confederate, independent of social modeling. Understanding whether the model or viewer gender influences nocebo effects following social modeling will add to current understanding of MPI episodes, in particular why such episodes seem to disproportionately affect women. Participant empathy has also been indicated as a contributing factor to placebo and nocebo effects following face-to-face social modeling [14, 27] but not when modeling is via video . To date, the influence of empathy on the experience of nocebo effects in response to social modeling has been examined only in pain paradigms, and it is unclear whether these effects extend to other symptom reports. Empathy is commonly assessed using the Interpersonal Reactivity Index (IRI; ), which comprises four facets of empathy: empathic concern, personal distress, perspective taking, and fantasy. Empathic concern is primarily linked with heightened placebo effects, whereas nocebo hyperalgesia has been linked with personal distress [14, 27]. Further investigation of these different aspects of empathy has the potential to aid understanding of underlying mechanisms of social modeling, for example, whether nocebo effects are primarily facilitated by feelings of sympathy for the model (as in empathic concern) or the experience of heightened personal distress in response to viewing someone else experience unpleasant symptoms . Understanding how heightened empathy influences responses to social modeling may also inform strategies for reducing nocebo effects, for example, relaxation or stress reduction in the case of heightened personal distress. The current research aims to investigate the impact of social modeling of side effects on symptoms experienced following placebo treatment, the role of participant and model gender in this process, as well as trait viewer empathy. To assess whether the effect of social modeling is specific to the symptoms reported by the confederate, the study also assessed the influence of social modeling on specifically modeled symptoms. To date, research has primarily investigated the influence of social modeling either on specific pain outcomes or socially modeled symptoms alone. This does not allow for investigation of the possibility that social modeling may influence a broader range of symptoms than just those that are specifically modeled. Therefore, in this study, we specifically investigated whether modeling also affected general symptoms and the misattribution of symptoms. Methods Design Potential participants were invited to take part in research investigating the effectiveness of the intranasal delivery of two different doses of modafinil (a cognitive enhancing or “smart” drug) in improving alertness and cognitive performance. All nasal sprays delivered as part of the study were placebos (isotonic saline solution). The research was approved by the University of Auckland Human Participants Ethics Committee (reference number 016206). Written informed consent to participate in the study as described to participants was obtained before participants took part in the experimental protocol. Because deceptive information was provided about the study, participants were fully debriefed after the completion of data collection, including all 24-hr follow-up assessments. This was done to maintain the integrity of the study and to minimize the risk that other potential participants might be made aware of the true study aims and procedures by a participant who had already been debriefed. During the debriefing, participants received information about the true nature and aims of the study and the need for deception. Participants were given the opportunity to ask questions and to withdraw their data if they wished. No participants chose to withdraw their data after debriefing. Participants Participants were 96 adults recruited from the wider University of Auckland community and included undergraduate and postgraduate students, as well as staff. Because the impact of both participant and model gender was under investigation in the current research, an equal number of male and female participants were recruited (n = 48 each). Based on previous research, it was estimated that a total sample size of 94 participants was needed to detect a 40% increase in symptom reporting following social modeling, based on achieving power of 0.8 with an alpha level of .05 using a negative binomial regression analysis and accounting an R2 of .65 from the other predictors [14, 25, 29]. The participants had a mean age of 21 years (SD = 2.92), ranging from 18 to 36. Most of the participants identified as being of Asian (59%) or European (27%) ethnicity, with a smaller proportion identifying as Maori or Pacific Island (6%) and other (7%) which primarily comprised Middle Eastern participants. Inclusion criteria required participants to be aged 18 years or over and to be able to read and write in English. Potential participants were excluded in line with the cover story if they had any conditions (including blood pressure over 140/90 mmHg or heart rate over 100 bpm at baseline) or were taking any medications that contraindicated the use of modafinil. Participants were reimbursed NZ$50 for their time. Participants were also asked to complete a 24-hr online follow-up to assess symptoms and attribution of these symptoms as side effects. In total, 93 participants (97%) completed the 24-hr follow-up, although the three participants who did not were all male and in the no modeling control condition. Procedure Participants attended a 1-hr experimental session at a Clinical Research Centre associated with the University of Auckland School of Medicine and a 10-min online follow-up questionnaire 24-hr after their experimental session. Participants were provided with a Participant Information Sheet containing information about the study, as well as verbal information about the research. Participants provided written informed consent to take part in a study of modafinil. In line with this information, participants’ blood pressure and heart rate were measured twice at baseline (before and after baseline questionnaire completion). Participants were block randomized by gender to be seated with either a male or female confederate and to view the confederate report experiencing side effects (modeling condition) or no difference (control condition) after ostensibly taking the same treatment as the participant. Random assignment to conditions was carried out by a researcher not otherwise involved in the study, using the random number generator function in Microsoft Excel. The experimenter was blind to the participants’ assigned condition during recruitment, booking, and baseline procedures but because of the study design was aware of assigned condition after the modeling procedure took place. Until the modeling procedure, only the confederate was aware of the participant’s group randomization, and allocation concealment was achieved using opaque envelopes that were opened by the confederate only once the participant had started the study session with the primary experimenter. The experimenter followed a study script for all participants to keep each session consistent, blood pressure and heart rate measures were taken using an automatic ambulatory blood pressure monitor, and participants completed all questionnaires in pen and paper format without the input of the experimenter. Participants completed a baseline questionnaire assessing their physical symptoms, trait empathy, alertness and fatigue, and demographic information. They were then given standardized information about the study nasal spray (described as modafinil), including that the effect of modafinil would be to increase wakefulness and cognitive function and reduce tiredness. Participants were informed that modafinil use can result in adverse effects, and that common side effects include nausea, headache, nervousness, anxiety, rhinitis, diarrhea, back pain, insomnia, dizziness, and dry mouth, as well as increasing heart rate and blood pressure (information from actual modafinil drug information sheets). After this, participants self-administered one spray into each nostril and were taken to a waiting area for 10 min (the duration of time they were informed it would take the modafinil to have an effect). Participants were seated with either a male or female model (confederate) in the waiting area after taking the nasal spray, and this model was already present when participants arrived. The model was described as being another participant who was also taking part in the study that day. Participants were additionally randomly assigned to either the control condition or the side effect modeling condition. Approximately 7 min into the waiting period, the experimenter asked first the confederate and then the participant, how they were feeling after taking the modafinil spray. In the control condition, the confederate reported feeling “fine” and “not really any different.” In the side effect modeling condition, the confederate reported “not feeling so great” and that they were experiencing a headache and dizziness. Following the modeling procedure, the participant was returned to the study room to continue the experimental session. Once back in the study room, the experimenter explained and conducted the cognitive tests. Blood pressure and heart rate were measured again immediately before and after the cognitive test administration. Finally, participants completed the post-medication questionnaire, including an assessment of any symptoms experienced and whether the participant attributed these symptoms as being side effects of the modafinil medication, as well as questions about alertness and fatigue. Participants completed a 24-hr follow-up questionnaire over the internet the following day, which again asked about any symptoms they had experienced since completing the experimental session and whether they attributed these as being modafinil side effects. Materials and Measures Models The models in the study were one male and one female who were the same age (mid 20s), with similar hair color, build, and were both of European ethnicity. The models underwent training before the study and memorized brief scripts for each scenario to ensure consistency in their responses in both the side effect modeling and the neutral control conditions. In addition to this, the models met at least once a week during the study to recalibrate their responses. The study experimenter, who was present for the modeling procedure, also provided them with regular feedback to enhance intermodel and intramodel consistency. Physical Symptoms Participants’ self-reported experience of 38 physical symptoms was assessed at baseline, post-medication, and 24-hr follow-up using a modified version of the Generic Assessment of Side Effects scale  with additional common physical symptoms . Symptoms from the original scale that were unlikely to be experienced over a short time duration (e.g., diarrhea, insomnia) were removed. Participants were asked to rate their experience of each of the listed symptoms over the previous 20 min (baseline), since taking the modafinil spray (post-medication) and since leaving the study session (24-hr follow-up). Symptoms were rated on a scale from 0 (not present) to 3 (severe). Participants were also asked (at post-medication and follow-up) whether they attributed any symptoms they had experienced as being medication side effects. To assess possible nocebo symptoms that participants experienced after taking the modafinil (placebo) spray, scores were created reflecting the total number of modeled (headache and dizziness) and general (other nonmodeled) symptoms that increased (either because they were new or had increased in intensity) between baseline and post-medication and baseline and 24-hour follow-up. Scores reflecting the number of misattributed symptoms at both post-medication and follow-up were also calculated by summing the number of unchanged symptoms (i.e., those symptoms that were rated the same at baseline as they were at either post-medication or follow-up) that participants subsequently attributed to the medication. Because the misattribution of modeled symptoms happened infrequently (in 5.2% and 3.2% of participants at post-medication and at follow-up, respectively), modeled and general misattributed symptoms were combined as one outcome variable, primarily comprised of misattributed general symptoms. This strategy to assess misattribution processes was chosen for two reasons. First, while the placebo spray contained no active ingredient, new and increased intensity symptoms were likely caused by a nocebo process resulting from treatment administration, thus attributing these symptoms to the treatment was in line with participants’ subjective experiences. In contrast, attributing symptoms that were already present at baseline demonstrates a clearer case of misattribution. Second, to attribute a symptom to a particular cause, the symptom must first be experienced. Thus, the total number of symptoms that participants attribute to the treatment is necessarily closely linked to the total number of symptoms reported (in the current study, rs = .64 to .94, ps < .001), making it difficult to tease apart attribution and nocebo symptom processes. Additional analyses assessing the impact of modeling and gender on the number of attributed symptoms (modeled and general) show an identical pattern of results to new and increased intensity symptoms reported below. The attribution of unchanged symptoms to the placebo treatment offers a clearer insight into attribution processes alone. Empathy In line with previous social modeling literature [13, 14], participant empathy was assessed using the IRI . This scale comprises four subscales of seven items each, with a total of 28 items. Participants rate each item on a five-point Likert-type scale ranging from “does not describe me well” to “describes me very well.” All four subscales were used in the current study: empathic concern, perspective taking, personal distress, and fantasy. Previous research has found associations between placebo effects following social modeling and the empathic concern subscale of the IRI. After reverse coding of relevant items, scores for each subscale were calculated by summing participant responses, with higher scores reflecting higher empathy in each domain. Heart Rate and Blood Pressure Participant heart rate and blood pressure were assessed using an iHealth wireless blood pressure monitor. Assessments were carried out at baseline (before and after baseline questionnaire completion) and post-test (immediately after the final test and again 5 min later). A mean of the two baseline and two post-test measurements for heart rate, systolic blood pressure, and diastolic blood pressure was calculated and used in subsequent analyses. Alertness and Fatigue In line with the cover story, participants’ self-reported levels of alertness and fatigue were assessed at baseline and again post-test using the vigor and fatigue subscales from the Profile of Mood States questionnaire . There were no significant effects of social modeling, or participant or model gender, or any interactions between these factors, on alertness (ps > .23) or fatigue (ps > .24) after participants took the placebo spray. Statistical Analysis All statistical analyses were carried out using SPSS version 22. The influence of social modeling condition, participant gender, model gender on symptom outcomes (modeled, general, and misattributed symptoms), and their interactions were assessed using negative binomial regression with maximum likelihood estimation. This approach was chosen because these outcomes comprised count data (i.e., the number of modeled and general symptoms that were new or increased in intensity, and the number of symptoms that remained unchanged but were misattributed to the placebo treatment). Significant interaction effects were followed up with Bonferroni-corrected pairwise tests. Negative binomial regression was also used to assess the influence of trait empathy on symptom outcomes. The interaction term between social modeling condition and each empathy subscale was entered into the model, with significant interaction effects indicating that trait empathy regression slopes were significantly different between the two modeling conditions. Where this interaction was significant, individual regression slopes for each modeling condition were examined. Heart rate and blood pressure outcomes were assessed using 2 × 2 × 2 (condition by participant gender by model gender) Analysis of Covariance (ANCOVA) analyses controlling for respective baseline scores. An alpha level of 0.05 was used. Results At baseline, 88% of participants reported experiencing one or more symptoms, with a mean of 3.82 symptoms (SE = 0.31). The most commonly reported symptoms at baseline were drowsiness (47%), dry mouth (35%), and fatigue (32%). Only 12% of participants reported headache, and 7% reported dizziness at baseline. After taking the placebo nasal spray, 95% of participants reported at least one symptom (M = 4.25, SE = 0.32), with headache (25%) and dizziness (26%) increasing markedly, while fatigue and drowsiness (25% and 28%, respectively) decreased in line with the cover story. At 24-hr follow-up, 81% of participants reported one or more symptoms (M = 5.15 symptoms, SE = 0.66), with headache (27%) and dizziness (23%) still elevated from baseline, and fatigue (36%) and drowsiness (38%) the most commonly reported symptoms. Other symptoms that showed marked increases following placebo administration included nausea, chest pain, heart palpitations, abdominal pain, tremor or shaking, blurred vision, and numbness or tingling sensations. Post-Medication Modeled Symptoms There was a significant effect of side effect modeling on increases in modeled symptoms of headache and dizziness, χ2(1) = 6.77, p = .009 (Fig. 1). Seeing the model report these side effects resulted in larger increases in these specific symptoms in the social modeling condition (M = 0.57, SE = 0.12) compared with the neutral control condition (M = 0.20, SE = 0.07; d = 0.56). Neither participant gender (p = .12), model gender (p = .67), nor interactions between these factors (ps > .05), significantly influenced the modeled symptom outcomes. Fig. 1. View largeDownload slide Bar graph showing the mean (SE) number of new or increased intensity modeled symptoms of headache and dizziness in the neutral control and side effect modeling groups. **p < .01. Fig. 1. View largeDownload slide Bar graph showing the mean (SE) number of new or increased intensity modeled symptoms of headache and dizziness in the neutral control and side effect modeling groups. **p < .01. Post-Medication General Symptoms There was no significant effect of social modeling on general symptoms, χ2(1) = 1.57, p = .21. However, the effect of participant gender on new or increased intensity general symptoms was significant, χ2(1) = 8.52, p = .004. Female participants experienced significantly more symptoms (M = 2.79, SE = 0.30) than males (M = 1.71, SE = 0.22; d = 0.59). There was also a significant interaction between participant and model gender, χ2(1) = 6.03, p = .014. Independent of the social modeling condition they were assigned to, female participants seated with a female model experienced significantly more general symptoms (M = 3.46, SE = 0.49) than either male participants seated with the same female model (M = 1.41, SE = 0.27; d = 0.74, p < .001) or female participants seated with a male model (M = 2.25, SE = 0.37; d = 0.40, p = .048). Post-Medication Misattributed Symptoms Neither modeling, participant gender, nor model gender had a significant influence on misattribution of symptoms as side effects during the study session, ps > .10. There was, however, a significant interaction between participant and model gender, χ2(1) = 7.76, p = .005. Female participants seated with a female model misattributed significantly more symptoms as medication side effects (M = 0.83, SE = 0.20) than either male participants seated with the same female model (M = 0.17, SE = 0.09; d = 0.86, p = .003) or female participants seated with the male model (M = 0.29, SE = 0.12; d = 0.74, p = .021). Male participants seated with the male model (M = 0.54, SE = 0.16) misattributed significantly more symptoms as side effects than male participants seated with the female model (d = 0.58, p = .044). There were no other interaction effects (ps > .54). Follow-Up Modeled Symptoms At follow-up, participants in the side effect modeling group were, again, significantly more likely to have experienced new or increased-intensity modeled symptoms (M = 0.53, SE = 0.12) compared with those in the neutral control condition (M = 0.20, SE = 0.07), χ2(1) = 5.20, p = .023, d = 0.48 (Fig. 2, left panel). Neither participant gender alone, model gender alone, nor the interaction of any of the factors, was a significant predictor of modeled symptoms at follow-up (ps > .05). Fig. 2. View largeDownload slide Bar graphs showing the mean (SE) number of new or increased intensity modeled (left), general (center), and misattributed (right) symptoms in the neutral control and side effect modeling groups at 24-hr follow-up. *p < .05. Fig. 2. View largeDownload slide Bar graphs showing the mean (SE) number of new or increased intensity modeled (left), general (center), and misattributed (right) symptoms in the neutral control and side effect modeling groups at 24-hr follow-up. *p < .05. Follow-Up General Symptoms At 24-hr follow-up, side effect modeling also had a significant influence on the number of new or increased intensity general symptoms, χ2(1) = 6.22, p = .013 (Fig. 2, center panel). Participants who experienced side effect modeling during the study session reported significantly more of these symptoms at follow-up (M = 3.51, SE = 0.63) than participants in the neutral control group (M = 1.78, SE = 0.37; d = 0.49). There was also a significant effect of participant gender, χ2(1) = 6.91, p = .009. Independent of social modeling condition, female participants reported significantly more general symptoms (M = 3.58, SE = 0.63) than males (M = 1.75, SE = 0.36; d = 0.51). There were no other significant main or interaction effects (ps > .06). Follow-Up Misattributed Symptoms There was a significant effect of social modeling on symptom misattribution at 24-hr follow-up, χ2(1) = 5.277, p = .022 (Fig. 2, right panel). Participants in the side effect modeling condition (M = 0.68, SE = 0.12) misattributed significantly more symptoms to the placebo medication than those in the neutral control group (M = 0.31, SE = 0.09; d = 0.50). There were no other significant main or interaction effects on symptom misattribution at follow-up (ps > .09). Empathy Male and female participants’ IRI subscale scores were compared using independent samples t-tests. Female participants had significantly higher scores for empathic concern (Mf = 10.66, SEf = 0.26; Mm = 9.56, SEm = 0.29), t(142) = −2.83, p = .005 and personal distress (Mf = 7.17, SEf = 0.42; Mm = 5.86, SEm = 0.35), t(137.23) = −2.39, p = .018. On the fantasy subscale, females also had higher scores, but this difference did not reach significance (Mf = 11.75, SEf = 0.49; Mm = 10.47, SEm = 0.46), t(142) = −1.90, p = .059. Male and female participants did not differ significantly on the perspective taking subscale, (Mf = 14.38, SEf = 0.37; Mm = 13.93, SEm = 0.39), t(142) = −0.83, p = .407. Negative binomial regression analyses were carried out to assess the influence of participant empathy on response to social modeling. Separate analyses were conducted for modeled symptoms, general symptoms, and misattributed symptoms reported during the session and at the 24-hr follow-up. The interaction between social modeling condition and each of the four IRI subscales were entered into each model, with significant interactions indicating that the regression slopes in the control and social modeling conditions differed (Table 1). Table 1 Table Showing the Wald Chi-Square Statistics (p-values) for Interactions Between Social Modeling Condition and Empathy Modeled Symptoms General Symptoms Misattributed Symptoms Post-medication Empathic concern 0.53 (0.77) 5.93 (0.052) 3.38 (0.19) Perspective taking 8.21 (0.017) 4.08 (0.13) 3.69 (0.16) Personal distress 0.14 (0.93) 6.86 (0.032) 8.02 (0.018) Fantasy 1.07 (0.59) 0.84 (0.66) 4.56 (0.10) Follow-up Empathic concern 0.35 (0.84) 2.00 (0.37) 5.22 (0.07) Perspective taking 6.90 (0.032) 0.84 (0.66) 0.04 (0.98) Personal distress 1.97 (0.37) 1.07 (0.59) 0.68 (0.71) Fantasy 3.52 (0.17) 1.56 (0.44) 2.78 (0.25) Modeled Symptoms General Symptoms Misattributed Symptoms Post-medication Empathic concern 0.53 (0.77) 5.93 (0.052) 3.38 (0.19) Perspective taking 8.21 (0.017) 4.08 (0.13) 3.69 (0.16) Personal distress 0.14 (0.93) 6.86 (0.032) 8.02 (0.018) Fantasy 1.07 (0.59) 0.84 (0.66) 4.56 (0.10) Follow-up Empathic concern 0.35 (0.84) 2.00 (0.37) 5.22 (0.07) Perspective taking 6.90 (0.032) 0.84 (0.66) 0.04 (0.98) Personal distress 1.97 (0.37) 1.07 (0.59) 0.68 (0.71) Fantasy 3.52 (0.17) 1.56 (0.44) 2.78 (0.25) Bold values indicate significance at p < .05. View Large Table 1 Table Showing the Wald Chi-Square Statistics (p-values) for Interactions Between Social Modeling Condition and Empathy Modeled Symptoms General Symptoms Misattributed Symptoms Post-medication Empathic concern 0.53 (0.77) 5.93 (0.052) 3.38 (0.19) Perspective taking 8.21 (0.017) 4.08 (0.13) 3.69 (0.16) Personal distress 0.14 (0.93) 6.86 (0.032) 8.02 (0.018) Fantasy 1.07 (0.59) 0.84 (0.66) 4.56 (0.10) Follow-up Empathic concern 0.35 (0.84) 2.00 (0.37) 5.22 (0.07) Perspective taking 6.90 (0.032) 0.84 (0.66) 0.04 (0.98) Personal distress 1.97 (0.37) 1.07 (0.59) 0.68 (0.71) Fantasy 3.52 (0.17) 1.56 (0.44) 2.78 (0.25) Modeled Symptoms General Symptoms Misattributed Symptoms Post-medication Empathic concern 0.53 (0.77) 5.93 (0.052) 3.38 (0.19) Perspective taking 8.21 (0.017) 4.08 (0.13) 3.69 (0.16) Personal distress 0.14 (0.93) 6.86 (0.032) 8.02 (0.018) Fantasy 1.07 (0.59) 0.84 (0.66) 4.56 (0.10) Follow-up Empathic concern 0.35 (0.84) 2.00 (0.37) 5.22 (0.07) Perspective taking 6.90 (0.032) 0.84 (0.66) 0.04 (0.98) Personal distress 1.97 (0.37) 1.07 (0.59) 0.68 (0.71) Fantasy 3.52 (0.17) 1.56 (0.44) 2.78 (0.25) Bold values indicate significance at p < .05. View Large During the session, the influence of perspective taking on modeled symptoms differed significantly between the social modeling and control conditions. In the control condition, perspective taking did not influence modeled symptoms, B = 0.13, Wald 95% confidence interval (CI) (−0.06, 0.33), p = .17, whereas in the social modeling condition, participants’ trait perspective taking significantly predicted their experience of modeled symptoms, B = 0.21, Wald 95% CI (0.06, 0.36), p = .006. The influence of personal distress on other symptoms and misattributed symptoms also differed significantly between the social modeling conditions. In the control condition, personal distress did not influence either general symptoms, B = 0.04, Wald 95% CI (−0.03, 0.11), p = .25, or misattributed symptoms, B = 0.05, Wald 95% CI (−0.09, 0.18), p = .51. Personal distress significantly predicted both the experience of general symptoms, B = 0.09, Wald 95% CI (0.02, 0.16), p = .018, as well as misattributed symptoms, B = 0.16, Wald 95% CI (0.05, 0.28), p = .006, in response to social modeling of side effects. At 24-hr follow-up, only the influence of perspective taking on modeled symptoms differed significantly between the social modeling conditions. In the control condition, personal distress did not predict the experience of modeled symptoms, B = 0.03, Wald 95% CI (−0.16, 0.23), p = .73. In contrast, in the social modeling condition, participant perspective taking was a significant predictor of modeled symptoms, B = 0.18, Wald 95% CI (0.05, 0.32), p = .009. Heart Rate and Blood Pressure Participants were told that the modafinil spray they were taking could have the additional side effects of raising blood pressure and heart rate. There was not a significant effect of social modeling on systolic blood pressure, F(1, 87) = 1.28, p = .26, d = 0.20. However, there was a significant interaction between social modeling and model gender, F(1, 87) = 6.45, p = .013, d = 0.55 (Fig. 3). Side effect modeling by the female model resulted in significantly higher systolic blood pressure (M = 114.79 mmHg, SE = 1.25) than in the neutral control group with the same female model (M = 110.20 mmHg, SE = 1.25; p = .012, d = 0.55). In comparison to the female model, side effect modeling by the male model resulted in lower systolic blood pressure (M = 111.19 mmHg, SE = 1.25; p = .044, d = 0.46). There were no other significant main or interaction effects for systolic blood pressure (ps > .21). There were no significant main or interaction effects of social modeling, participant gender, or model gender on diastolic blood pressure (ps > .13). Fig. 3. View largeDownload slide Bar graph showing the mean (SE) systolic blood pressure (left) and heart rate (right) in the neutral control and side effect modeling groups with male and female models. *p < .05. Fig. 3. View largeDownload slide Bar graph showing the mean (SE) systolic blood pressure (left) and heart rate (right) in the neutral control and side effect modeling groups with male and female models. *p < .05. A similar pattern of results was found for heart rate. There was not a significant main effect of social modeling on heart rate, F(1, 87) = 0.74, p = .39, d = 0.20, nor were there effects of participant or model gender (ps > .60). There was a significant interaction between social modeling and model gender, F(1, 87) = 5.84, p = .018, d = 0.51 (Fig. 3). Side effect modeling by the female model resulted in significantly higher heart rate readings (M = 72.79 bpm, SE = 1.17) than with the same female model in the neutral control condition (M = 68.96 bpm, SE = 1.17; p = .023; d = 0.51) or side effect modeling by the male model (M = 69.36 bpm, SE = 1.17; p = .041, d = 0.46). There were no other main or interaction effects (ps > .18). Discussion Social modeling clearly can be an important influence on the experience of treatment side effects. Social modeling of side effects increased reported adverse symptoms in viewers. This effect was specific to modeled symptoms over the short term, but generalized by follow-up to also impact general symptoms and increased the likelihood that symptoms would be misattributed to the treatment. At both time points, female participants reported significantly more new or increased intensity general symptoms. In addition, female participants seated with a female model were the highest reporters of these symptoms, and misattributed significantly more unchanged symptoms as being medication side effects, but only during the study session. Empathy appears to have played a role in the modeling process, with perspective taking and personal distress scores predicting modeled and other symptomatic outcomes, respectively, during the study session, and perspective taking alone predicting modeled symptoms at 24-hr follow-up. Although social modeling increased reported side effects, and female participants seated with a female model showed heightened reporting of general symptoms, there were no interactions between social modeling condition and either model or participant gender. These results indicate that the female participants were not significantly more susceptible to the effects of social modeling nor was the female confederate more effective in transmitting symptoms to the viewers. An interaction between social modeling and model gender was seen in participants’ objectively assessed physiological treatment side effects, namely blood pressure and heart rate. Participants were informed that side effects of the medication included both increased heart rate and blood pressure. Those participants who viewed the female (but not the male) model report that treatment side effects were significantly more likely to show this response. Women are typically more emotionally expressive than their male counterparts . While not assessed, the female model in the study may have expressed higher levels of distress when reporting symptoms, resulting in heightened viewer physiological responses . The differences in the objective physiological outcomes and the subjective symptom reports could point to two distinct types of information being modeled. First, the emotional content of the message containing threat information, which was more effectively displayed by the female model. Second, the informational content of the message, including the specific symptoms reported as well as likelihood of side effects. This information may increase nocebo symptoms by increasing attention toward and expectations of possible side effects. The current study extends previous work by investigating the role of social modeling of side effects in a medication-use context and outside of a pain paradigm. It is also novel in assessing side effect responding in three domains—specifically modeled symptoms, other general symptoms, and unchanged yet misattributed symptoms. Previous research has generally not made such distinctions—either grouping all symptoms together or examining only modeled symptoms and often not considering the role of attribution. The results with regard to specific symptoms are in line with previous research which indicates that social modeling results in increased nocebo effects soon after the social modeling procedure [13, 14, 16, 17]. Similar to results reported by Mazzoni et al. , female participants seated with a female model reported symptoms at significantly higher rates, regardless of social modeling condition. Being seated with another female may have similar effects to being seated in front of a mirror, which has been shown to increase internal self-focus and, thus, symptom reporting in women only . The results of the current study and those of Mazzoni et al.  indicate that the episodes of MPI may be more likely to affect women in part because groups comprised primarily of women experience more physical symptoms. Women seated with another woman in the current study also misattributed more unchanged symptoms to the placebo treatment, indicating that similar misattribution processes may also contribute to MPI episodes. More generalized effects of social modeling on symptoms appeared only at the 24-hr follow-up assessment. This may indicate that specific expectations about modeled symptoms were formed soon after social modeling, whereas more generalized negative expectations about treatment developed over a longer period. Exposure to social modeling may have contributed to more general recall that the medication carried a risk of side effects, such that participants remembered this broad threat information at follow-up, in addition to or instead of information about specifically modeled symptoms . General negative expectations contribute to the experience of adverse effects of medical treatments—patients who hold negative beliefs about medicines (e.g., that medicines do more harm than good) report significantly more treatment side effects than those who hold more positive beliefs . Similarly, patients who believe that they are particularly sensitive to the effects of medicines also experience more treatment side effects . Such general negative beliefs, as well as heightened anxiety, may have also contributed to the misattribution of unrelated symptoms to the placebo treatment [39, 40]. However, expectations about treatment outcomes were not assessed in the current study, and future research would benefit from investigating both expectations about specific side effects, as well as more general negative beliefs about medicines. In line with previous research using a pain paradigm, viewer empathy was associated with increased symptom reporting . While the current study points to a driving role of perspective taking in increased reporting of modeled symptoms both during the study session and at the 24-hr follow-up, previous research found that personal distress predicted increased pain in response to social modeling. In the current study, the personal distress component of empathy also significantly predicted increased general and misattributed symptoms during the study session. However, as these symptoms were not elevated above the control condition at this time point, the role of personal distress in driving symptom experience following social modeling remains unclear. Perspective taking in the context of empathy relates to the tendency of an individual to more readily experience a situation from the viewpoint of others . This appears to have contributed to participants’ experience of modeled symptoms in response to seeing them reported by the confederate in the current study. It may be that the activation of the mirror neuron system in response to social modeling may facilitate the experience of these same symptoms in the viewer . However, this has yet to be formally investigated. The influence of social modeling on participants’ heart rate and blood pressure diverge from the results relating to symptom reporting. At the start of the experimental session, participants were warned that the physiological side effects of the medication included elevated heart rate and blood pressure. Social modeling of side effects by the female confederate—regardless of the gender of the participant—resulted in increased participant heart rate and blood pressure. These findings provide evidence for a physiological nocebo-like effect in response to social modeling. In previous studies where participants were told that the treatment they received would act to decrease heart rate and blood pressure, this decrease was subsequently realized [16, 42]. This provides preliminary evidence that the treatment context can either increase or decrease these physiological parameters, depending on the information provided. However, why this response occurred only in response to the female model in the current study is unclear, and the effects of the individual models cannot be ruled out. The current study is limited by the lack of a no treatment control group. Such a group would have enabled the assessment of the presence and extent of a nocebo effect. In addition, the design of the study meant that the experimenter was present during the social modeling procedure and, thus, was not blind to group assignment after this point. However, the experimenter was blind to group assignment for all baseline procedures and assessments, assessment of symptoms was carried out using a pen-and-paper questionnaire completed by the participant without the input of the experimenter, and heart rate and blood pressure were assessed using automatic blood pressure monitors requiring no interpretation on the part of the experimenter. Only one model of each gender was employed in this study. While this provided consistency and reduced variance in appearance and other behaviors, it raises the question of whether the results might be somewhat model-specific or whether they would generalize to other models. Previous similar studies have included the social modeling of at least four different symptoms [16–18], and the current study results may have been influenced by the use of a less intense modeling procedure involving the confederate reporting only two symptoms. The use of a placebo nasal spray rather than a tablet may have influenced participants’ beliefs about the strength of the treatment and thus the study outcomes. A systematic investigation of the influence of the route of treatment administration and the number of modeled symptoms, would provide important contributions to the nocebo and social modeling literature. The results of the current study demonstrate that social modeling has a specific effect on modeled symptoms in the short term but that these effects generalize to other general symptoms, as well as increase the misattribution of unchanged symptoms, over a longer duration. Social modeling can occur in face-to-face circumstances in interactions with friends and family, through social media and other online sources (including patient support websites) and through the news media [12, 23, 43]. These results highlight the importance of considering the potential for observation of others and communication between participants in clinical trial designs, and the need to minimize these factors to achieve accurate results about treatment effects. In the context of medical care, social modeling processes have the potential to influence the overall symptom burden experienced by patients, and may contribute to additional care-seeking, unnecessary medications to manage symptoms, reduced quality of life, and treatment nonadherence or discontinuation . The current results highlight the importance of developing strategies to minimize the impact of learning about other peoples’ unpleasant side effects on symptom experience, and of doing so before the influence of social modeling generalizes to a broader range of symptoms and to symptom misattribution. Acknowledgments This research was supported by an Auckland Medical Research Foundation project grant awarded to Dr Faasse (grant number: 3710108), an unrestricted research grant from Pharmac (the New Zealand Government’s Pharmaceutical Management Agency) to Professor Petrie and Dr Faasse, and a University of Auckland Summer Research Scholarship awarded to Brian Yeom under the supervision of Dr Faasse. Compliance with Ethical Standards Authors’ Statement of Conflict of Interest and Adherence to Ethical Standards The authors declare that they have no conflict of interest Authors' Contributions Kate Faasse and Keith J. Petrie were responsible for conception and design; Brian Yeom, Bryony Parkes, James Kearney, and Kate Faasse were responsible to acquisition of data; Kate Faasse was responsible for analysis and interpretation of data, and drafting the manuscript; all authors were responsible for critically revising the article, and for final approval of the version to be published. Ethical Approval All procedures performed in this research involving human participants were in accordance with the ethical standards of the University of Auckland Human Participants Ethics Committee (approval number 016206) and with the 1964 Helsinki declaration and its later amendments. Informed consent Informed consent was obtained from all individual participants included in the study. References 1. Mahr A, Golmard C, Pham E, Iordache L, Deville L, Faure P. Types, frequencies, and burden of nonspecific adverse events of drugs: Analysis of randomized placebo-controlled clinical trials. Pharmacoepidemiol Drug Saf . 2017;26:731–741. 2. Faasse K, Petrie K. The nocebo effect: Patient expectations and medication side effects. Postgrad Med J . 2013; 89: 540– 546. Google Scholar CrossRef Search ADS PubMed 3. Barsky AJ, Saintfort R, Rogers MP, Borus JF. Nonspecific medication side effects and the nocebo phenomenon. JAMA . 2002; 287: 622– 627. Google Scholar CrossRef Search ADS PubMed 4. Papadopoulos D, Mitsikostas DD. Nocebo effects in multiple sclerosis trials: A meta-analysis. Mult Scler . 2010; 16: 816– 828. Google Scholar CrossRef Search ADS PubMed 5. Rief W, Barsky AJ, Glombiewski JA, Nestoriuc Y, Glaesmer H, Braehler E. Assessing general side effects in clinical trials: Reference data from the general population. Pharmacoepidemiol Drug Saf . 2011; 20: 405– 415. Google Scholar CrossRef Search ADS PubMed 6. Petrie K, Faasse K, Crichton F, Grey A. How common are symptoms? Evidence from a New Zealand national telephone survey. BMJ Open . 2014; 4: e005374. Google Scholar CrossRef Search ADS PubMed 7. Shedden Mora M, Nestoriuc Y, Rief W. Lessons learned from placebo groups in antidepressant trials. Philos Trans R Soc B . 2011; 366: 1879– 1888. Google Scholar CrossRef Search ADS 8. Colloca L, Miller F. The nocebo effect and its relevance for clinical practice. Psychosom Med . 2011; 73: 598– 603. Google Scholar CrossRef Search ADS PubMed 9. Webster RK, Weinman J, Rubin GJ. A systematic review of factors that contribute to nocebo effects. Heal Psychol . 2016;35:1334–1355. 10. Schedlowski M, Enck P, Rief W, Bingel U. Neuro-bio-behavioral mechanisms of placebo and nocebo responses: Implications for clinical trials and clinical practice. Pharmacol Rev . 2015; 67: 697– 730. Google Scholar CrossRef Search ADS PubMed 11. Klosterhalfen S, Enck P. Neurophysiology and psychobiology of the placebo response. Curr Opin Psychiatry . 2008; 21: 189– 195. Google Scholar CrossRef Search ADS PubMed 12. Faasse K, Petrie K. From me to you: The effect of social modeling on treatment outcomes. Curr Dir Psychol Sci . 2016; 25: 438– 443. Google Scholar CrossRef Search ADS 13. Vogtle E, Barke A, Kroner-Herwig B. Nocebo hyperalgesia induced by social observational learning. Pain . 2013; 154: 1427– 1433. Google Scholar CrossRef Search ADS PubMed 14. Swider K, Babel P. The effect of the sex of a model on nocebo hyperalgesia induced by social observational learning. Pain . 2013; 154: 1312– 1317. Google Scholar CrossRef Search ADS PubMed 15. Vogtle E, Kroner-Herwig B, Barke A. Nocebo hyperalgesia: Contributions of social observation and body-related cognitive styles. J Pain Res . 2016; 9: 241– 249. Google Scholar CrossRef Search ADS PubMed 16. Faasse K, Grey A, Jordan R, Garland S, Petrie K. Seeing is believing: Impact of social modeling on placebo and nocebo responding. Heal Psychol . 2015; 34: 880– 885. Google Scholar CrossRef Search ADS 17. Lorber W, Mazzoni G, Kirsch I. Illness by suggestion: Expectancy, modeling, and gender in the production of psychosomatic symptoms. Ann Behav Med . 2007; 33: 112– 116. Google Scholar CrossRef Search ADS PubMed 18. Mazzoni G, Foan L, Hyland ME, Kirsch I. The effects of observation and gender on psychogenic symptoms. Heal Psychol . 2010; 29: 181– 185. Google Scholar CrossRef Search ADS 19. Broderick JE, Kaplan-Liss E, Bass E. Experimental induction of psychogenic illness in the context of a medical event and media exposure. Am J Disaster Med . 2011; 6: 163– 172. Google Scholar PubMed 20. Koban L, Wager T. Beyond conformity: Social influences on pain reports and physiology. Emotion . 2015; 16: 24– 32. Google Scholar CrossRef Search ADS PubMed 21. Egorova N, Park J, Orr SP, Kirsch I, Gollub RL, Kong J. Not seeing or feeling is still believing: Conscious and non-conscious pain modulation after direct and observational learning. Sci Rep . 2015; 5: 16809. Google Scholar CrossRef Search ADS PubMed 22. Faasse K, Cundy T, Petrie K. Throxine: anatomy of a health scare. BMJ . 2010; 340: 20– 21. 23. Faasse K, Gamble G, Cundy T, Petrie K. Impact of television coverage on the number and type of symptoms reported during a health scare: A retrospective pre-post observational study. BMJ Open . 2012; 2: e001607– e001607. Google Scholar CrossRef Search ADS PubMed 24. Bartholomew RE, Wessely S. Protean nature of mass sociogenic illness: From possessed nuns to chemical and biological terrorism fears. Br J Psychiatry . 2002; 180: 300– 306. Google Scholar CrossRef Search ADS PubMed 25. Bartholomew RE. “Mystery illness” at Melbourne Airport: Toxic poisoning or mass hysteria? Med J Aust . 2005; 183: 564– 566. Google Scholar PubMed 26. Boss LP. Epidemic hysteria: A review of the published literature. Epidemiol Rev . 1997; 19: 233– 243. Google Scholar CrossRef Search ADS PubMed 27. Hunter T, Siess F, Colloca L. Socially induced placebo analgesia: A comparison of a pre-recorded versus live face-to-face observation. Eur J Pain . 2014; 18: 914– 922. Google Scholar CrossRef Search ADS PubMed 28. Davis MH. Measuring individual differences in empathy: Evidence for a multidimensional approach. J Pers Soc Psychol . 1983; 44: 113– 126. Google Scholar CrossRef Search ADS 29. Faul F, Erdfelder E, Buchner A, Lang A-G. Statistical power analyses using G*Power 3.1: Tests for correlation and regression analyses. Behav Res Methods . 2009; 41: 1149– 1160. Google Scholar CrossRef Search ADS PubMed 30. Rief W, Glombiewski J, Barsky AJ. Generic assessment of side effects: GASE. 2009. Available at http://www.gase-scale.com/. Accessibility verified July 13, 2016. 31. Davis MH. A mulitdimensional approach to individual differences in empathy. JSAS Cat Sel Doc Psychol . 1980; 10: 85. 32. McNair P, Lorr M, Droppleman L. POMS Manual . 2nd ed. San Diego: Educational and Industrial Testing Service; 1981. 33. Kring AM, Gordon AH. Sex differences in emotion: Expression, experience, and physiology. J Pers Soc Psychol . 1998; 74: 686– 703. Google Scholar CrossRef Search ADS PubMed 34. Wild B, Erb M, Bartels M. Are emotions contagious? Evoked emotions while viewing emotionally expressive faces: Quality, quantity, time course and gender differences. Psychiatry Res . 2001; 102: 109– 124. Google Scholar CrossRef Search ADS PubMed 35. Ingram RE, Cruet D, Johnson BR, Wisnicki KS. Self-focused attention, gender, gender role, and vulnerability to negative affect. J Pers Soc Psychol . 1988; 55: 967– 978. Google Scholar CrossRef Search ADS PubMed 36. Reyna VF. A theory of medical decision making and health: Fuzzy trace theory. Med Decis Mak . 2008; 28: 850– 865. Google Scholar CrossRef Search ADS 37. Horne R, Weinman J, Hankins M. The beliefs about medicines questionnaire: The development and evaluation of a new method for assessing the cognitive representation of medication. Psychol Heal . 1999; 14: 1– 24. Google Scholar CrossRef Search ADS 38. Horne R, Faasse K, Cooper Vet al. The perceived sensitivity to medicines (PSM) scale: An evaluation of validity and reliability. Br J Health Psychol . 2013; 18: 18– 30. Google Scholar CrossRef Search ADS PubMed 39. Petrie K, Moss-Morris R, Grey C, Shaw M. The relationship of negative affect and perceived sensitivity to symptom reporting following vaccination. Br J Health Psychol . 2004; 9: 101– 111. Google Scholar CrossRef Search ADS PubMed 40. Cameron LD, Leventhal H, Love RR. Trait anxiety, symptom perceptions, and illness-related responses among women with breast cancer in remission during a tamoxifen clinical trial. Heal Psychol . 1998; 17: 459– 469. Google Scholar CrossRef Search ADS 41. Lee Y-T, Tsai S-J. The mirror neuron system may play a role in the pathogenesis of mass hysteria. Med Hypotheses . 2010; 74: 244– 245. Google Scholar CrossRef Search ADS PubMed 42. Faasse K, Cundy T, Gamble G, Petrie K. The effect of an apparent change to a branded or generic medication on drug effectiveness and side effects. Psychosom Med . 2013; 75: 90– 96. Google Scholar CrossRef Search ADS PubMed 43. Bartholomew RE, Wessely S, Rubin GJ. Mass psychogenic illness and the social network: is it changing the pattern of outbreaks? J R Soc Med . 2012; 105: 509– 512. Google Scholar CrossRef Search ADS PubMed © Society of Behavioral Medicine 2018. All rights reserved. For permissions, please e-mail: firstname.lastname@example.org. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)
Annals of Behavioral Medicine – Oxford University Press
Published: Jan 25, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera