Unraveling nocebo effects in visceral pain: negative suggestion and adverse treatment experience uniquely shape visceral pain unpleasantness

Supplemental Digital Content is Available in the Text.

Visceral pain is highly susceptible to nocebo effects, which can arise or intensify through cross-modal transfer from adverse somatic experiences, underscoring relevance of provider communication.

Abstract

Visceral pain, characterized by its diffuse, poorly localized, and fear-inducing nature, may be especially prone to nocebo effects. This preregistered experimental study investigated whether (1) visceral pain is more susceptible to nocebo effects than somatic pain, and whether (2) an adverse treatment experience in the somatic modality contributes to visceral nocebo effects by cross-modal generalization. A total of 101 healthy volunteers received inert treatment and were randomized into 4 experimental groups: negative treatment suggestions, adverse somatic treatment experience, their combination, or control. Individually calibrated visceral (rectal distension) and somatic (thermal) pain stimuli were applied. Negative suggestions involved heightened pain sensitivity suggestions; adverse treatment experience was modeled by covertly amplifying thermal pain. Pain ratings, cortisol levels, and emotional states were assessed across pre-, treat-, and test-phases, and upon re-exposure to pain 1 week later. Negative suggestions increased negative expectations and cortisol levels, and the thermal pain manipulation successfully amplified somatic pain during treatment (P < 0.001). In the test-phase, both suggestions and somatic experience independently increased visceral pain unpleasantness (P = 0.004; P = 0.011), whereas no somatic nocebo effects emerged. No group differences appeared at re-exposure, but perceived treatment allocation was influenced both by suggestions and experience. Findings support the distinct vulnerability of the visceral modality to nocebo effects, which can arise or intensify through cross-modal transfer from adverse somatic treatment experiences, emphasizing the importance of provider communication and consideration of prior experiences in the treatment of chronic visceral pain conditions, especially in patients with mixed pain phenotypes.

1. Introduction

Visceral pain is highly prevalent, challenging to treat, and causes substantial suffering and healthcare expenditures in chronic visceral pain conditions like irritable bowel syndrome (IBS)^24,54 or chronic inflammatory bowel disease.²⁵ The broad role of psychosocial factors in symptom generation and treatment is increasingly appreciated,^22,31 facilitated by advancing knowledge about the gut-brain axis and the specific perceptual mechanisms underlying interoception and visceral pain.^28,39 Visceral pain as a poorly localizable, diffuse sensation from within the body is demonstrably more threatening and unpleasant than exteroceptive, somatic pain, engaging partly distinct brain networks relevant to psychological pain modulation.^27,34,53,57 It may therefore be particularly susceptible to placebo and nocebo effects, especially in conditions like IBS.⁴² Expectations within the psychosocial treatment context, generated by treatment-related suggestions and experience,^3,14,17,56 demonstrably shape the experience of acute and chronic pain, giving rise to placebo and nocebo effects after an inert treatment or as part of active treatments.^{12,15,16,23,37} Despite substantial implications for clinical research and patient care, nocebo effects remain incompletely understood,^1,55 especially in the context of visceral pain.^19,35

We conducted a translational experimental study to elucidate nocebo effects as they arise in clinical settings based on negative information from healthcare providers, for example, as part of informed consent about the possibility of adverse treatment effects. The design also reflected that adverse treatment experience can contribute to nocebo effects based on implicit learning/conditioning processes,¹² mechanisms that appear particularly relevant to conditions like IBS where fluctuating symptoms and failed treatments characterize patients' reality. The between-group, repeated-measures study design comprised healthy volunteers all receiving an inert treatment, randomized into 3 nocebo groups, that is, negative treatment suggestions, adverse treatment experience, their combination, or a control group. For pain outcomes, we capitalized on our established multiple threat pain paradigm,^6,33,34,50 modelling symptoms arising from both interoceptive and exteroceptive pain modalities as a mixed pain phenotype common in visceral pain conditions. Applied to nocebo research, the paradigm offers opportunities to assess treatment expectations and experiences and their impact on pain perception within and across pain modalities, including the notion that visceral pain experience could be shaped by adverse treatment effects on exteroceptive somatosensory symptoms, that is, generalization.^30,62 Pain augmentation induced by generalization from one modality to another, that is, cross-modal generalization, has been proposed as a factor driving nocebo effects,^59,60 yet this has not yet been tested in visceral pain.

As pain unpleasantness is particularly sensitive to cognitive and emotional pain modulation,⁵⁸ we tested group differences in overall visceral pain unpleasantness as a clinically relevant primary outcome. Aligned with a priori determined aims,² we hypothesized (1) a heightened vulnerability to nocebo effects in the visceral compared with the somatic pain modality, and (2) a contribution of generalization, induced by an adverse treatment experience in the somatic modality, to the generation of visceral nocebo effects. We also report group differences in pain intensity, explored cortisol and state emotions given previous evidence linking stress and anxiety to nocebo effects,^20,46,52 and compared groups on a re-exposure to pain 1 week later.

2. Methods

2.1. Participants

Healthy adult volunteers were recruited for a comprehensive experimental study on nocebo effects in visceral pain (ie, the “NoVis study”) comprising 2 study days, conducted at the University Hospital Essen, University of Duisburg-Essen, Germany, as part of the Collaborative Research Center 289 “Treatment Expectation.” The study was preregistered in the German Clinical Trials Register (DRKS: DRKS00024410), and a study protocol has been published that provides expanded conceptual considerations, in-depth methodological details, and information about additional measures not reported herein.² Information regarding additional assessments accomplished as part of the Collaborative Research Center's central projects is provided in the supplemental methods section (available at http://links.lww.com/PAIN/C397). The study was approved by the local Ethics Committee of the University Hospital Essen (protocol number 19-8897-BO). A priori power analysis to determine sample size yielded N = 96 participants²; we oversampled to account for exclusions, drop-out, etc, and included a total N = 109 volunteers who signed consent and received financial compensation for their participation. Of note, study procedures involving an inert treatment in all experimental groups are in alignment with the consensus definition of placebo and nocebo effects as beneficial or adverse effects that occur in clinical or laboratory medical contexts, respectively, after administration of an inert treatment or as part of active treatments, because of mechanisms such as expectancies of the patient or participant.²³ Given the nature of the study design built on negative treatment suggestions, 2 distinct consent forms were used, as explained in detail below. Briefly, negatively-instructed groups were recruited using a cover story of an experiment comprising the possibility of receiving a pain-enhancing substance. Remaining groups signed consent for a study in which a standardized protocol was used involving the administration of a saline solution without active treatment.

The recruitment of healthy male of female participants within an age range of 18 to 45 years and a body mass index >18 and <30 kg/m² was accomplished between June 2021 and October 2023 by public advertisement in the surrounding universities, clinics, and community of the Ruhr area. Established screening and safety procedures consisted of a telephone screening, a personal interview, completion of questionnaires, and a physical examination. Exclusion criteria included any medical condition, current medication use (except contraceptives and thyroid medication), drug abuse, recent gastrointestinal complaints (screened with an established checklist³⁶), or anal tissue damage which could interfere with the rectal distension model implemented herein as a clinically relevant visceral pain model, known psychiatric condition or current symptoms of anxiety or depression (screened with the German version of the Hospital Anxiety and Depression Scale⁸), inadequate German language proficiency, pregnancy (standard urine pregnancy test performed on study day), or breastfeeding. As the study was conducted during different phases of the coronavirus disease 2019 pandemic, restrictions and additional laboratory testing was accomplished as per legal requirements and regulations implemented for research participants by the University Hospital Essen administration. Given the role of prior experience in nocebo effects, participation in any experimental pain study involving somatic or visceral pain stimuli within the previous 3 months (based on self-report) was exclusionary.

2.2. Study design and blinding

As illustrated in Figure 1 and detailed in subsequent sections, all participants received an intravenous (i.v.) catheter with an infusion of saline, that is, underwent an inert treatment. They were randomized based on the group factors “treatment suggestion” (ie, Negative vs Control Instruction) and “treatment experience” (ie, Negative vs Control Experience during treatment), resulting in a total of 4 experimental groups: 2 groups with negative suggestions, implemented either alone (“Suggestions-only”) or combined with an adverse treatment experience (“Combined”), and 2 groups without negative suggestions that either underwent a negative treatment experience (“Experience-only”) or just received an inert substance (“Control”). On study day 1, after preparatory procedures including precalibration of pain stimulation intensities, the pain paradigm was implemented as a baseline before treatment (PRE-phase). It comprised a series of pain stimuli from the visceral (rectal distension) and somatic (cutaneous thermal) pain modalities, matched and precalibrated to a perceived intensity corresponding to a visual analogue scale (VAS) rating of 50 mm for both pain modalities. As a key element of the psychosocial treatment context, the study physician then applied the i.v. catheter with a saline infusion in all participants. As part of this treatment phase (TREAT-phase), the study physician reinforced negative treatment expectations by administering the study medication combined with negative verbal suggestions in negatively-instructed groups. Treatment experience was manipulated in the negative experience groups by surreptitiously increasing thermal pain stimulation intensity to create a clearly discernible adverse treatment experience of amplified somatic pain, using a precalibrated intensity of VAS 80 mm. This allowed us to assess immediate generalization effects to the visceral modality, for which stimulation intensities remained unaltered at VAS 50 mm. During the TEST-phase, all groups received identical pain stimulation intensities for both pain modalities as during the PRE-phase, that is, VAS 50 mm, allowing us to test nocebo effects resulting from suggestions alone, experience alone, and from the interaction of suggestion and experience, compared with a control group receiving an inert treatment but neither suggestions nor experience manipulations. A second study day was accomplished 1 week later comprising a re-exposure to the same pain stimuli at the VAS 50 mm (RE-TEST-phase), yet without any of the salient elements of a treatment (ie, there was no i.v. and no interaction with the study physician). For pain-related outcomes in the visceral and somatic pain modality, respectively, VAS assessing pain intensity and unpleasantness were implemented. Saliva samples were collected for analysis of cortisol concentrations, together with the assessment of negative emotional states.

Figure 1.

Overview of study design, experimental groups, and procedures. Participants were randomized into groups before informed consent given the crucial role of negative suggestions in this experimental design. On study day 1, in a PRE-phase, visceral (VIS) and somatic (SOM) pain stimuli were implemented at a precalibrated intensity corresponding to a visual analogue scale (VAS) rating of 50 mm. During treatment (TREAT-phase), all groups received an intravenous saline infusion. Depending on group, this inert treatment was coupled with negative suggestions alone (Suggestion-only group, light orange), negative suggestions in combination with covertly increased thermal stimulation intensity to a precalibrated VAS 80 mm (Combined group, dark orange), only covertly increased thermal stimulation intensity (Experience-only group, dark blue), or neither nocebo manipulation (Control, light blue). Pain stimulation intensities were identical to the PRE-phase in the subsequent TEST-phase. Re-exposure to pain was accomplished on day 2 (RE-TEST). The GEEE questionnaire³⁶ assessed treatment expectations before treatment (GEEE A), treatment experience after the TEST-phase (GEEE B), and perceived treatment allocation at the end of day 2 (GEEE C). As indicated by triangles, before each phase saliva samples were acquired for cortisol analysis, and the State-Trait Anxiety Depression Inventory (STADI) questionnaire was completed for emotional states. (Control, light blue), negative treatment experience alone (Experience-only, dark blue), negative suggestions alone (Suggestion-only, light orange), or negative suggestions and negative experience combined (Combined, dark orange). The figure was created using BioRender (www.biorender.com).

To ensure blinding throughout the experiment, the comprehensive strategy comprised the following measures: (1) Staff members engaged during recruitment and initial screening remained fully blinded to both group factors. (2) An otherwise uninvolved individual accomplished randomization and communicated the group assignments separately to the study physician (author J.L.A.) and the scientific staff engaged in pain stimulation, respectively. (3) The study physician was blinded to the group factor “experience” and was not present during pain stimulation phases. (4) Research staff implementing pain stimulation phases remained blinded regarding the group factor “suggestion” and were not present in the study room at times of interaction of the study physician with participants.

2.3. Pain stimuli and paradigm comprising visceral and somatic pain stimuli

We implemented pressure-controlled rectal distensions using a barostat system (Distender Series II, 1300 mL Single Balloon Barostat; G&J Electronics, Toronto, ON, Canada) and thermal cutaneous heat pain with a thermal stimulation device (PATHWAY model CHEPS; Medoc Ltd, Advanced Medical Systems, Ramat Yishai, Israel). As a refinement of our previous work,^6,33,34,50 we herein applied cutaneous heat stimuli on the lower left side of the abdomen (rather than the lower arm) to minimize possible effects of body site-specific (rather than modality-specific) differences in pain responses.^26,49 The a priori identification of pain stimulation intensities for each participant, within predefined perceptual intensity ranges and matched across pain modalities, is a crucial prerequisite of the paradigm (details provided in Refs. 2, 34). Briefly, the multistep procedure comprised the determination of pain thresholds, calibration and matching, and a habituation phase. This served the critical purpose of identifying individual visceral and somatic pain stimulation intensities, that is, a rectal distension pressure and a temperature, respectively, that were matched regarding perceived pain intensity, with a target intensity of 50 mm perceived pain intensity on VAS (end points labelled “not painful” [0 mm] and “extremely painful” [100 mm]). In addition, a thermal pain intensity target of VAS 80 mm was identified for use in the TREAT-phase. For standardization and to maintain blinding, this was accomplished in all experimental groups, yet applied in the TREAT-phase only in negative treatment experience groups.

In all pain stimulation phases, a series of cued visceral and somatic pain stimuli was delivered, comprising a total of 12 phasic pain stimuli (ie, 6 distensions, 6 thermal cutaneous stimuli, durations 20 seconds each), delivered in pseudorandomized order starting with a visceral or a somatic stimulus in 50% of participants, respectively, and with thermode settings individually adjusted to match the inflation and deflation times of the rectal balloon. Rectal distension pressures were delivered at the predetermined VAS 50 mm in all phases; in other words, all visceral pain stimuli were intra-individually applied with the same objective stimulus intensity. However, although the predetermined thermode temperature for the VAS 50 mm target was implemented in the PRE- and TEST-phases, in the TREAT-phase it was surreptitiously increased to a predetermined intensity of VAS 80 mm in negative experience groups, as indicated in Figure 1. Stimulus intensities applied during RE-TEST were identical to those used in the TEST-phase for both modalities.

2.4. Treatment suggestions

To effectively induce negative or neutral treatment expectations based on suggestions, we devised a comprehensive information and communication protocol for different phases of the study: During study advertising and initial recruitment, all information provided on study-related materials (eg, study flyers) and during initial contact and phone screening was focused only on dynamic changes in pain perception and differences in 2 specific pain modalities as the overall study goal. Notably, no reference was made to the possibility of receiving a pronociceptive drug or a placebo. This was important to avoid the induction of negative treatment expectations before randomization into negative or control instruction groups.

Randomization was accomplished before the personal interview with the study physician, and participants from that time point onwards received distinct written and verbal treatment information: the protocol for the negative instruction groups (ie, Suggestion-only and Combined) expanded on our previous line of studies on visceral nocebo hyperalgesia induced by deceptive negative treatment suggestions regarding the opioid antagonist naloxone as a pain-enhancing drug.^46,48 Herein, volunteers were informed about participation in a double-blind, placebo-controlled trial with the administration of either naloxone or saline to study pain sensitization and pain responses during an amplified pain experience in 2 pain modalities, with a 1:1 randomization ratio. In reality, saline was administered to all participants. Note that we disclosed administration of saline to all volunteers after the conclusion of the study. Instruction control groups (ie, Experience-only, Control) were informed that only saline as an inactive substance would be administered as part of standardized procedures.

Group-specific instructions pertained to all written study-related materials as well as verbal communication with the study physician, requiring 2 distinct consent forms for either (1) a cover story of an RCT with naloxone or saline aiming to discover modality-specific mechanisms underlying pain sensitization (version signed by negatively-instructed groups) or (2) administration of just saline and a study rationale focused solely on elucidating differences between pain modalities (version signed by other groups, Fig. 1). Negatively-instructed groups were given detailed information about the drug, especially about its pain-enhancing effects, but also about typical clinical uses and known side effects. For implementation during the experiment, the medication ampule had an official design from the Hospital Pharmacy of the University Medicine Essen and was labeled “Study drug—naloxone or placebo” and was placed directly in front of the participant during the TREAT-phase. It was then injected into the saline-containing infusion bag in full view of the participants, analogous to the administration of medication in clinical routines, subsequently administered to the participants by an i.v. line. Verbal treatment suggestions reinforcing the cover story corresponding to the group-specific content were delivered verbally by the study physician as “reminders” in the context of placing the i.v. line and starting the saline drip.

2.5. Measures

2.5.1. Negative treatment-related expectations and experiences

As an assessment tool specifically developed for translational expectancy research, the Generic Rating Scale for Previous Treatment Experience, Treatment Expectations, and Treatment Effects (GEEE)⁴⁴ quantifies different facets of participants' expectations and experiences of a treatment, herein used to assess the expected and experienced aggravation of pain. As indicated in Figure 1, expected worsening was acquired on study day 1 before administration of the treatment, that is, before the i.v. infusion. In addition, the experience of symptom aggravation based on the treatment was quantified after the TEST-phase. Both assessments were based on a numerical rating scale (with endpoints labeled “no worsening” [0] and “greatest worsening imaginable” [10]). In addition, perceived treatment allocation was assessed at the end of study day 2, after RE-TEST, that is, participants indicated whether they believed that the infusion contained an active ingredient or not. To complement measures derived from the GEEE, volunteers were asked to rate perceived warmth and competence of the study physician at the end of study day 1 using a 10-item Likert-type scale.⁵¹

2.5.2. Pain outcomes (visual analogue scale)

After each pain stimulation phase, VAS ratings of overall pain intensity and unpleasantness for the visceral and the somatic pain modality, respectively, were acquired using digitized VAS (“Overall, how painful/unpleasant did you find the distension/heat stimuli you just experienced?” 0-100 mm with endpoints labeled “not at all” and “very painful/unpleasant”). The analysis of overall ratings was driven by our translational perspective, as patients in clinical practice are typically asked to evaluate their pain retrospectively, providing a cumulative and integrative assessment of their experience.

Such global ratings align more closely with patient-reported outcomes in real-world settings.^9,29 In line with this, although we transparently report both pain intensity and unpleasantness, we prioritize pain unpleasantness as a primary outcome, recognizing its unique role in capturing the emotional and affective dimensions of visceral pain. Pain unpleasantness more accurately reflects the affective burden of pain, which is crucial for understanding patient suffering and treatment efficacy.^40,41 This is particularly relevant in visceral pain, where the affective dimension often outweighs the sensory component.³⁴

2.5.3. Salivary cortisol levels and emotional states

Saliva samples were collected for the analysis of cortisol concentrations as an acute stress marker of the hypothalamic-pituitary-adrenal (HPA) axis using commercially available collection devices (Salivette, Sarstedt, Nümbrecht, Germany). After centrifugation (1000g, 2 minutes, 4°C), all saliva samples were stored at −80°C until analysis. Cortisol concentrations were quantified using a commercially available enzyme-linked immunosorbent assay (Cortisol Saliva ELISA; IBL International, Hamburg, Germany), according to the manufacturer's instructions. Mean intra-assay and interassay coefficients of variation were <10%. The detection limit was 0.138 nmol/L. All samples from a single participant were measured within the same assay. In parallel to saliva sampling, participants completed the state version of the State-Trait Anxiety Depression Inventory (STADI)⁴³ as a validated questionnaire for the assessment of current negative emotional states. It comprises 20 Likert-scaled items ranging from 1 (“not at all”) to 4 (“very much”). The 10-item subscale “State Anxiety” comprises the affective (agitation) and cognitive (apprehension) dimensions of anxiety. The 10-item subscale “State Depression” measures positive (euthymia, inverted) and depressed/negative (dysthymia) mood state. To complement analyses of state emotions, correlational analyses were accomplished not only with the state version but also with the trait version of the STADI⁴³ acquired as part of the comprehensive questionnaire battery completed before study day 1 by all participants.

2.5.4. Additional measures analyzed as part of supplemental analyses

As detailed in supplemental materials, we additionally provide results of analyses of expected pain unpleasantness and intensity, respectively, measured for each pain modality using VAS. Participants were asked: “How unpleasant do you expect the heat/distension stimulus to be?” and marked their expectation on a 0- to 100-mm VAS scale (endpoints labeled “not at all” and “very intense/unpleasant”). We further analyzed skin conductance responses to pain-predictive cues based on continuous recordings of electrodermal activity during all experimental phases (see supplemental methods and results, available at http://links.lww.com/PAIN/C397).

2.6. Statistical analyses

All statistical analyses were computed using IBM SPSS Statistics software, version 29.0 (IBM Corp, Armonk, NY) and figures designed in GraphPadPrism8 (GraphPad Software, La Jolla, CA). Groups were compared regarding sociodemographic and psychological characteristics using χ² tests or univariate analyses of variance as appropriate. Given the nonnormal distribution of the data, confirmed by the Kolmogorov-Smirnov test, a Kruskal-Wallis test was conducted first, followed by Mann-Whitney U tests for pairwise comparisons to analyze expected and experienced symptom worsening based on GEEE scores. To explore group differences in perceived treatment allocation, as captured by the GEEE on study day 2, χ² test was used.

For the analyses of pain outcomes, deviating from our initial considerations (see study protocol²), we used generalized linear mixed models (GLMM) to account for the non-normal distribution of the data and increasing robustness in pain outcomes. Initially, a comprehensive GLMM was built with the fixed effects “group” (4 levels, given 4 experimental groups), “time” (3 levels: PRE, TREAT, TEST), and “modality” (2 levels: visceral, somatic) and their interaction. A random intercept was incorporated as random effect. Adding random slope as random effects did not improve model fits, as assessed by Akaike information criterion. Additional GLMMs were then built separately for each pain modality, a strategy we preplanned. For hypothesis testing, planned group comparisons in the TREAT-phase examined differences between Combined and Suggestion-only as well as Experience-only and Control. In the TEST-phase, all groups were compared using the following contrasts: Experience-only—Control, Suggestion-only—Control, Combined—Control, Suggestion-only—Experience-only, Combined—Experience-only, and Combined—Suggestion-only. Note that full statistical details on all group comparisons in all phases are provided as supplemental material (Tables S1 and S2, available at http://links.lww.com/PAIN/C397). Data acquired upon re-exposure to pain in the RE-TEST-phase were explored using group contrasts based on GLMM results. Effect sizes (Cohen d) were computed for all pairwise contrasts based on estimated marginal mean values and SEs.

For salivary cortisol concentrations and STADI scores, GLMM with the fixed effects of “group,” “time,” and their interaction were used, and group comparisons were accomplished with planned contrasts. In the model for salivary cortisol, the time of the day and the cortisol concentration on arrival were considered as fixed effects. In addition, Spearman-Rho correlations were computed within experimental groups to explore associations between pain outcomes and anxiety (state and trait) and perceived warmth and competence of the study physician. All results are reported as (estimated) mean ± SEM unless indicated otherwise, and the level of significance was set at P < 0.05. If applicable, effect size r was derived from the Z-value of nonparametric tests.

3. Results

3.1. Participants

Of the N = 109 participants randomized, N = 2 did not appear for their scheduled study day 1 appointment, N = 6 were excluded because of lack of adequate perceptual response to pain stimuli, N = 1 withdrew from the study, N = 1 had to be removed because of later evidence of exclusion criteria. As a result, we herein report on data from N = 101 for main analyses accomplished for the 3 experimental phases on study day 1, randomized into 4 groups as indicated in Table 1. As one individual did not appear on study day 2, analysis of measures acquired on the second study day comprised N = 100 individuals. Experimental groups did not differ in sociodemographic or psychological characteristics acquired at screening. Similarly, the distension pressures and thermode temperatures precalibrated for application during experimental phases were comparable (Table 1).

Table 1.

Sample characteristics and pain stimulation intensities in experimental groups.

	Total	Control	Experience-only	Suggestion-only	Combined	P
N	101	28	25	23	25
Female*, % (N)	62.4 (63)	57.1 (16)	68.0 (17)	60.1 (14)	64.0 (16)	0.870
Age (y)	24.5 ± 3.3	24.3 ± 2.9	24.4 ± 3.2	24.7 ± 3.5	24.8 ± 3.9	0.923
BMI	23.0 ± 2.8	23.9 ± 2.6	23.3 ± 3.0	22.4 ± 2.6	22.3 ± 2.6	0.102
HADS anxiety score at screening	3.8 ± 2.2	3.5 ± 2.1	3.4 ± 2.2	4.6 ± 2.1	3.8 ± 2.4	0.260
HADS depression score at screening	1.3 ± 1.5	1.1 ± 1.5	1.2 ± 1.4	1.3 ± 1.4	1.4 ± 1.6	0.840
STADI trait anxiety score at screening	16.4 ± 4.3	16.8 ± 4.4	15.8 ± 3.8	17.8 ± 4.2	14.9 ± 4.5	0.152
GI symptoms at screening	3.0 ± 2.2	3.0 ± 2.4	2.6 ± 1.9	3.4 ± 2.2	3.2 ± 2.5	0.677
Distention pressure, mm Hg for VAS 50 mm	33.9 ± 10.2	33.9 ± 8.5	34.1 ± 11.6	35.4 ± 12.5	33.1 ± 9.2	0.984
Thermode temperature, °C for VAS 50 mm	45.1 ± 1.1	45.9 ± 1.0	46.1 ± 0.9	45.7 ± 1.9	46.4 ± 1.1	0.551
Thermode temperature, °C for VAS 80 mm, TREAT-phase	N/A	N/A	47.1 ± 1.0	N/A	47.3 ± 1.1	<0.001†

Results given as mean ± SD unless indicated otherwise for the entire sample and for experimental groups (ie, control group, Control; negative experience alone, Experience-only; negative suggestion alone, Suggestion-only; combination of negative suggestion and experience, Combination); P values: Results of χ² test for categorical variables or univariate ANOVA for continuous variables.

^*Self-reported sex at birth.

^†Temperature reflecting 80 mm on VAS, VAS for pain intensity, used in negative experience groups during the TREAT-phase vs unaltered temperature reflecting VAS 50 mm for pain intensity used in other groups.

ANOVA, analysis of variance; BMI, body mass index; GI, gastrointestinal symptoms during past month; HADS, Hospital Anxiety and Depression Scale; STADI, State-Trait Anxiety Depression Inventory⁵⁸; VAS, visual analogue scale.

3.2. Efficacy of experimental nocebo interventions

To elucidate the efficacy of negative treatment suggestions, we analyzed expected symptom worsening before the TREAT-phase, assessed with the GEEE. Results revealed a significant effect of group (χ²(3) = 34.071; P < 0.001; η² = 0.31), with negatively-instructed groups (ie, the Combined and the Suggestion-only groups) expecting significantly more treatment-induced worsening of symptoms (results of planned group comparisons: Combined vs Control: Z = 4.066, P < 0.001, r = 0.405; Combined vs Experience-only: Z = 3.659, P < 0.001, r = 0.364; Suggestion-only vs Control: Z = 4.547, P < 0.001, r = 0.452; Suggestion-only vs Experience-only: Z = −4.137, P < 0.001, r = 0.411; Fig. 2A). Hence, negative treatment suggestions comprising possible drug treatment-related symptoms of worsening effectively induced negative expectations before treatment.

Figure 2.

Treatment-related expectations and experiences. As assessed using the Generic Rating Scale for Previous Treatment Experience, Treatment Expectations, and Treatment Effects (GEEE) questionnaire,⁴⁴ experimental groups (ie, control group, Control, light blue; negative experience alone, Experience-only, dark blue; negative suggestion alone, Suggestion-only, light orange; combination of negative suggestion and experience, Combined, dark orange) differed in expected symptom worsening before the TREAT-phase (A) and experienced symptom worsening after the TEST-phase (B). Results in (A and B) shown as mean and SEM, ***P < 0.001, **P < 0.01, *P < 0.05, results of group comparisons. Perceived allocation to active or inactive drug was assessed at the end of study day 2 (C), revealing a significant difference in percentages (χ² test comparing groups: P < 0.001); see text for methodological and statistical details; for time points, see with Figure 1. (Control, light blue), negative treatment experience alone (Experience-only, dark blue), negative suggestions alone (Suggestion-only, light orange), or negative suggestions and negative experience combined (Combined, dark orange).

Regarding adverse treatment experience, experimentally created by increasing thermode temperatures to the precalibrated intensity of VAS 80 mm during the TREAT-phase in negative experience groups, results showed significantly higher pain intensity ratings compared with other groups (ie, Experience-only vs Control: β = 33.834 ± 9.317, P = 0.001; Combined vs Suggestion-only: β = 23.586 ± 10.324 P = 0.023; for additional details, see Tables S1 and S2, available at http://links.lww.com/PAIN/C397). Hence, covert pain amplification in the somatic modality successfully created an adverse treatment experience (for full somatic pain VAS results, see below). As detailed in supplemental materials, the covert pain amplification also led to significantly heightened somatic pain-related expectations regarding both intensity and unpleasantness, respectively, before the TEST-phase (full results of VAS expected pain provided in Tables S3 and S4, available at http://links.lww.com/PAIN/C397, visualized in Fig. S1, available at http://links.lww.com/PAIN/C397). However, analysis of skin conductance responses to pain-predictive cues revealed negative results (ie, no differences between groups or modalities, see supplemental results and Fig. S2, available at http://links.lww.com/PAIN/C397).

3.3. Pain outcomes: comprehensive generalized linear mixed models

The initial comprehensive GLMM for perceived pain unpleasantness (with group, time, and modality as fixed effects and intercept as random effect), yielded a significant overall model (F(23,582) = 8.395, P < 0.001), with significant main effects of group (F(3,582) = 4.363, P = 0.005), time (F(2,582) = 5.592, P = 0.004), and modality (F(1,582) = 132.724, P < 0.001), as well as with a significant 3-way interaction of group, time, and modality (F(17,582) = 2.114, P = 0.006). For perceived pain intensity, the overall model was also significant (F(23,582) = 3.509, P < 0.001), with significant main effects of time (F(2,582) = 9.339, P < 0.001) and modality (F(1,582) = 20.286, P < 0.001) but not of group (F(3,582) = 0.542, P = 0.654), and a significant 3-way interaction of group, time, and modality (F(17,582) = 2.940, P < 0.001). Based on these results, subsequent GLMM were accomplished separately within each pain modality, along with hypothesis-driven planned group comparisons.

3.4. . Visceral pain perception

GLMM within the visceral modality revealed a significant model for perceived pain unpleasantness (F(11,291) = 2.408, P = 0.007), with a significant interaction of group and time (F(6,291) = 2.316, P = 0.034), but no significant main effects of time (F(2,291) = 2.890; P = 0.057) or group (F(3,291) = 2.003; P = 0.114). Group comparisons revealed that in the TREAT-phase, pain unpleasantness was significantly higher in the Combined group when compared with the Control group (β = 21.330 ± 8.740, P = 0.015, Fig. 3A), supporting a nocebo effect induced by the combination of negative suggestion and adverse somatic treatment experience, whereas negative suggestions alone (Suggestion-only) or negative experience alone (Experience-only) were not sufficient to generate an nocebo effect in the TREAT-phase. In the TEST-phase, additional group differences emerged, resulting in elevated pain unpleasantness in all nocebo intervention groups when compared with the control group (ie, Combined vs. Control: β = 22.759 ± 7.926, P = 0.004; Suggestion-only vs Control: β = 17.766 ± 7.764, P = 0.023; Experience-only vs Control: β = 19.708 ±7.693, P = 0.011). Interestingly, the difference observed between the Experience-only and the Control groups supports a nocebo effect within the visceral modality induced by the preceding adverse somatic treatment experience alone, even in the absence of negative suggestions, supporting a generalization from the somatic to the visceral pain modality. There was no evidence of differences between nocebo intervention groups (all P > 0.1, statistical details in Table S1, available at http://links.lww.com/PAIN/C397), indicating that in combination generalization effects did not further enhance the magnitude of nocebo effects attributable to negative suggestions.

Figure 3.

Pain ratings of overall unpleasantness and intensity for visceral and somatic pain across experimental phases. Visual analogue scale ratings (0-100 mm) for the visceral modality (A and C) and the somatic modality (B and D), for perceived pain unpleasantness (panels A and B) and perceived pain intensity (panels C and D) on study day 1 in experimental groups (ie, control group, Control, light blue; negative experience alone, Experience-only, dark blue; negative suggestion alone, Suggestion-only, light orange; combination of negative suggestion and experience, Combined, dark orange). Data are visualized as mean ± SEM and individual data points. Results of planned group comparisons in each phase: ***P < 0.001; **P < 0.01; *P < 0.05; for full details, including generalized linear mixed models (GLMMs), see Result section. For timing of phases, see Figure 1. (Control, light blue), negative treatment experience alone (Experience-only, dark blue), negative suggestions alone (Suggestion-only, light orange), or negative suggestions and negative experience combined (Combined, dark orange).

For visceral pain intensity, GLMM did not yield a significant model (F(11, 291) = 0.875, P = 0.566). There were no significant main effects of group (F(3, 291) = 0.083, P = 0.969) or time (F(2, 291) = 0.722, P = 0.487), and the group-by-time interaction was also not significant (F(6, 291) = 1.268, P = 0.272). There were no differences between groups in any phase (Fig. 3C, further details in Table S2, available at http://links.lww.com/PAIN/C397).

3.5. Somatic pain perception

Generalized linear mixed models within the somatic pain modality yielded a significant model for perceived somatic pain unpleasantness (F(11,291) = 4.531, P < 0.001), with a significant main effect of time (F(2,291) = 10.850, P < 0.001), no main effect of group (F(3,291) = 2.205, P = 0.088), but a significant interaction between group and time (F(6,291) = 3.606, P = 0.002). Similarly, a significant model was found for perceived somatic pain intensity (F(11, 291) = 6.709, P < 0.001), with a main effect of time (F(2, 291) = 15.448, P < 0.001), no main effect of group (F(3, 291) = 1.787, P = 0.150), but a significant interaction of group and time (F(6, 291) = 6.266, P < 0.001). Besides group differences in the TREAT-phase, carried out as part of a manipulation check (see above), group comparisons in the TEST-phase revealed no differences neither for pain unpleasantness (Fig. 3B) nor for pain intensity (Fig. 3D). Hence, neither negative suggestions nor adverse treatment experience nor the combination of both induced in nocebo effects within the somatic pain modality.

3.6. Salivary cortisol and emotional states

For cortisol levels, illustrated in Figure 4A, GLMM yielded a significant overall model (F(13,279) = 9.384, P < 0.001), with main effects of group (F(3,279) = 5.611, P < 0.001) and time (F(1, 279) = 13.447, P < 0.001), but no significant interaction of time and group (F(6,276) = 1.217, P = 0.297). Group comparisons revealed higher cortisol levels before the TREAT-phase in negatively-instructed groups (ie, Suggestion-only vs Control: β = 1.185, P = 0.041; Suggestion-only vs Experience-only: β = 1.305, P = 0.026; Combined vs Control: β = 1.638, P = 0.005; Combined vs Experience-only: β = 1.758, P = 0.023), suggesting higher HPA-axis activation. Salivary cortisol levels remained elevated until TEST-phase in the same experimental groups (ie, Suggestion-only vs Control: β = 1.683, P = 0.007; Suggestion-only vs Experience-only: β = 1.361, P = 0.037; Combined group vs Control: β = 1.660, P = 0.003; Combined vs Experience-only: β = 1.338, P = 0.024; for full details, see Table S5, available at http://links.lww.com/PAIN/C397).

Figure 4.

Changes in salivary cortisol concentrations (A) and state anxiety (B) and state depression scores (C), assessed with the state version of the State-Trait Anxiety Depression Inventory (STADI)⁵⁸ across experimental phases on study day 1. Data were acquired before each experimental phase in experimental groups (ie, control group, Control, light blue; negative experience alone, Experience-only, dark blue; negative suggestion alone, Suggestion-only, light orange; combination of negative suggestion and experience, Combined, dark orange). Data are visualized as mean ± SEM. Results of planned group comparisons in each phase: **P < 0.01, *P < 0.05; full details, including generalized linear mixed models (GLMMs), see Result section and Table S3 (available at http://links.lww.com/PAIN/C397); for time points, see Figure 1. (Control, light blue), negative treatment experience alone (Experience-only, dark blue), negative suggestions alone (Suggestion-only, light orange), or negative suggestions and negative experience combined (Combined, dark orange).

Analyses of state anxiety and state depression, assessed with the STADI questionnaire and visualized in Figure 4B and C, respectively, revealed a significant overall model for state anxiety (F(11,281) = 1.854, P = 0.045), driven by a main effect of time (F(2,281) = 3.758, P = 0.025) in the absence of main effect of group (F(3,281) = 1.976, P = 0.118; Fig. 4) or an interaction (F(6,281) = 1.049, P = 0.394). Similarly, a significant overall model for depression (F(11,281) = 2.434, P = 0.007) was found, with a significant main effect of time (F(2,281) = 9.072, P < 0.001), no main effect of group (F(3,281) = 1.182, P = 0.317) or interaction (F(6,281) = 0.977, P = 0.441). Details for group comparisons are provided in Table S5 (available at http://links.lww.com/PAIN/C397). Exploratory correlational analyses, accomplished within each group, revealed no consistent associations between state or trait anxiety, respectively, and pain unpleasantness (see Table S6, available at http://links.lww.com/PAIN/C397).

3.7. Treatment-related experiences and perceived treatment allocation

A significant effect of group was observed for experienced symptom worsening assessed with the GEEE after the TEST-phase (χ²(3) = 12.69; P = 0.005; η² = 0.127), with greater worsening experienced by negatively-instructed groups (Suggestion-only vs Control: Z = 2.292; P = 0.022; r = 0.228; Combined vs Control: Z = 3.176; P = 0.001; r = 0.316; Combined vs Experience-only: Z = 2.454; P = 0.014; r = 0.244; Fig. 2B). However, there were no group differences based on the adverse treatment experience (ie, Control vs Experience-only; Z = 0.654; P = 0.513; r = 0.065; and Suggestion-only vs Combined; Z = 0.792; P = 0.429; r = 0.079).

On study day 2, that is, 7 days after the treatment and at the conclusion of all experimental testing, group differences were observed in the proportion of participants who believed that the infusion contained an active ingredient (χ²(1) = 33.566; P < 0.001; φ = 0.585). As illustrated in Figure 2C, in the Suggestion-only group, 38.1% of participants and in the Combined group, 60% of participants reported believing that they had received a real, active drug. By contrast, none of the participants in the Experience-only group believed they had received an active ingredient, despite experiencing heightened somatic pain during the TREAT-phase.

There were no group differences in perceived warmth and competency of the study physician (warmth: H(3) = 1.36, P = 0.715; competency: H(3) = 0.74, P = 0.865, data not shown), and ratings did not correlate with pain outcomes in the test phase (all P > 0.40, data not shown).

3.8. Re-exposure to pain on study day 2

One week after study day 1, participants were re-exposed to the same pain series as during the TEST-phase (ie, using the pre-calibrated VAS 50 mm visceral and somatic pain stimuli). Given the absence of a treatment in this RE-TEST-phase, data from study day 2 were analysed separately, using exploratory group comparisons for all variables. These revealed no significant group differences in pain outcomes, cortisol levels, or STADI scores (all P > 0.05, details provided in Table S7, available at http://links.lww.com/PAIN/C397).

4. Discussion

This is the first experimental study elucidating nocebo effects across interoceptive and exteroceptive pain domains, modelling a mixed pain phenotype common in visceral pain conditions like IBS.⁶¹ Using a study design that reflects complex clinical treatment scenarios, we compared self-reported pain outcomes in experimental nocebo groups exposed to negative treatment suggestion, adverse somatic treatment experience, the combination of both, and a control group that only received an inert treatment. Together, our results provide new insights into the mechanisms generating nocebo effects in visceral pain and lend support to the notion that nocebo effects shape pain perception in a modality-specific manner, with implications for the clinical management of acute and chronic visceral pain.

Our findings for pain unpleasantness as primary outcome support that visceral pain is more vulnerable to nocebo effects. Underscoring the crucial relevance of patient-provider communication as a critical source of nocebo effects,^19,20 herein pain unpleasantness evoked by rectal distensions was significantly greater in groups with negative suggestions.

Intriguingly, the adverse treatment experience of amplified somatic pain exacerbated visceral pain unpleasantness not only in the treatment phase, as observed in the Combined group, but also led to increased visceral pain unpleasantness in the Experience-only group when compared with the Control Group in the test phase. This observation cannot be interpreted in generalization because generalization involves an expectancy effect in the original modality to subsequently impact another domain. Although generalization has been reported in placebo studies,^29,60 nocebo research in this area has been limited to a notable study demonstrating nocebo generalization within the exteroceptive domain, that is, from thermal to pressure pain.⁵⁸ Given the absence of any nocebo effect in the somatic pain modality herein, rather than generalization, we demonstrate for the first time that nocebo effects in the visceral modality can be amplified, or even de novo generated by cross-modal transfer effect of a negative treatment experience from the somatic to the visceral pain modality. Hence, adverse experiences in one sensory domain can selectively amplify perception in another domain, even in the absence of a direct effect in the original modality.

Indeed, inspecting group differences in the treatment vs the test phase, respectively, offers insight into possible mechanisms underlying our observation. During the treatment phase, the visceral nocebo effect observable only in the Combined-group with combined negative instruction and negative experience may result from the integration of peripheral input, based on conditioning in the somatic modality, with cognitive-emotional evaluation induced by negative suggestions. In the test-phase, visceral nocebo effects observable in all 3 groups could be mediated by cognitive-affective processing through central or descending pain modulatory systems, potentially involving the salience network, given its known role in visceral pain,^34,50 integrating the effects of conditioning and suggestions. Our findings point to the importance of distinguishing suggestions and conditioning as nocebo induction mechanisms.¹² Although both influence expectations,³⁸ conditioning may also engage implicit learning rather than solely relying on conscious processes.^12,47 Likewise, the theory of distinct pathways may explain the unexpected absence of synergy between suggestion and conditioning in the Combined group, potentially limiting their joint impact. Clearly, mechanistic research is needed to clarify mechanisms underlying modality-specific pain modulation and transfer effects between modalities, complementing earlier evidence supporting distinct neural representations of visceral and somatic pain.^{13,18,32,34,39}

Robust group nocebo effects in the visceral modality coupled with a striking absence of nocebo effects in the somatic modality, despite effectively heightened negative pain-related expectations induced by a clearly adverse treatment experience, support enhanced vulnerability of visceral pain to nocebo mechanisms, especially when individuals experience symptoms from multiple domains. Replicating earlier findings in both healthy and clinical cohorts,^34,39 visceral pain consistently elicited greater unpleasantness than somatic pain stimuli applied to the same abdominal body site. The greater unpleasantness and fear-inducing characteristics of visceral pain may amplify susceptibility to cognitive pain modulation while diminishing attention to somatic symptoms through overshadowing or blocking phenomena.²⁰ This is consistent with greater central salience and fear network recruitment during visceral pain,³³ its stronger interruption of cognitive processes,^32,50 and evidence that interoceptive threat predictors are preferentially acquired and remembered.³³

For pain intensity, however, group comparisons revealed no evidence of nocebo effects for either pain modality, suggesting distinct modulation of pain intensity and unpleasantness within the visceral domain. Unpleasantness, capturing affective-motivational rather than sensory-discriminative dimensions of pain seems to be more readily modifiable by nocebo mechanisms, significantly contributing to its emotional burden. This may be linked to the diffuse and poorly localized perceptual nature of visceral pain and aligns with previous findings showing increases in visceral pain unpleasantness to depressed mood,⁵ and evidence that visceral stimuli remain unpleasant even at low, nonpainful intensities.^5,18 These insights underscore the need for comprehensive patient-reported outcomes in conditions involving visceral pain and the gut-brain axis, advocating for a broader focus beyond pain intensity alone. Clinicians should not only recognize the burden of interoceptive symptoms but also appreciate the contribution of psychological factors to treatment outcomes, shaped not only by treatment-related patient-provider communication but also cross-modal transfer effects of prior treatment experiences.

Given previous evidence linking stress and anxiety to nocebo effects,^20,46,52 we explored group differences in negative emotional states and salivary cortisol concentrations. Although no group differences were found in emotional states, negatively instructed groups exhibited elevated cortisol levels across experimental phases, in line with earlier studies indicating a role of cortisol in nocebo effects.^4,21,52 These results imply that negative treatment suggestions not only induce worries of symptom worsening but also pre-activate the HPA axis, possibly reflecting a preparatory physiological response in the face of an impending threat. The stress system plays a broad role in gut-brain axis,³⁵ and elevated cortisol levels uniquely heighten visceral pain sensitivity, even in healthy individuals.⁶ Understanding stress pathways, including cortisol dynamics, could provide important insights for tailoring interventions to mitigate nocebo effects in vulnerable populations.

This is the first experimental study to show that negative suggestions involving uncertainty can effectively generate negative expectations and induce HPA-axis activation. Negative treatment suggestions comprised a 50% probability of a pain-enhancing drug or saline, building on prior work that used similar suggestions yet a 100% certainty of drug delivery.^46,48 This modified cover story reflects informed consent procedures in clinical trials as well as real treatment scenarios characterized by uncertainty about the likelihood of adverse events or side effects. It also allows initial insight into perceptions and beliefs about treatment, which is important given the role of nonpharmacological factors, including expectation effects, in the total treatment effect in pharmacological trials,⁴⁵ and large placebo response rates in conditions involving the gut-brain axis.^10,11,19 Herein, group differences in perceived symptom worsening after the test-phase were solely driven by negative suggestion rather than by experience, and perceived treatment allocation, assessed 1 week after the treatment, mirrored the communicated probability of receiving the active drug.

Interestingly, although negative experience groups did not alter their perception of the treatment allocation to saline as a placebo, the experience of increased pain amplified the belief that they had received the active pain-enhancing drug from 38% to 60% in individuals that had received negative treatment suggestions. This complements evidence that pain relief shapes perceived treatment efficacy in visceral placebo studies with positive suggestions⁷ highlights the need for research on the interplay of negative expectations and adverse treatment experiences in visceral pain.

This study has strengths as well as limitations. Strengths include the clinically relevant visceral pain model implemented in a translational yet carefully controlled experimental study design reflecting complex clinical treatment scenarios in patients with mixed pain phenotypes. Results provide novel evidence, shedding first light on interactions between negative suggestions and experience/conditioning underlying nocebo effects in the visceral modality. On a critical note, our key results in nocebo groups were driven by an increasingly lower pain unpleasantness in the Control group across experimental phases, consistent with habituation. All nocebo interventions prevented this habituation, supporting our earlier findings that nocebo effects can manifest as a failure to habituate,⁵⁰ hence interfering with the normal and adaptive response to repeated pain exposure. However, without a fully untreated control group, we cannot definitively attribute these effects to the treatment context or the repeated exposure to acute pain and purposely refrain from interpreting our findings in “nocebo hyperalgesia.” Furthermore, our design comprised a second study day scheduled 1 week later, and results revealed no group differences in pain outcomes. However, it is difficult to interpret these negative findings in nonpersisting nocebo effects since we accomplished a re-exposure to pain without critical elements of the treatment context (eg, the presence of the i.v. line or the physician). Future studies should explore repeated exposures to the same or to novel treatments. Nevertheless, we believe that our findings have significant clinical implications, particularly for chronic visceral pain, where psychological factors such as treatment expectations, anxiety, and stress strongly influence symptoms and outcomes.^19,35 Clinicians should recognize the heightened sensitivity of visceral pain to nocebo effects, underscoring the importance of tailored communication strategies to mitigate their impact.

Conflict of interest statement

The authors have no conflicts of interest to declare.

Appendix A. Supplemental digital content

Supplemental digital content associated with this article can be found online at http://links.lww.com/PAIN/C397.

Supplementary Material

jop-167-606-s001.pdf ^{(1.1MB, pdf)}