Negative expectations about treatment can worsen pain and itch sensations, a phenomenon known as the nocebo effect. These effects are experimentally induced to study how negative experiences aggravate symptoms. Nocebo responses are defined as a significant increase in sensation after a nocebo treatment compared to a control. While nocebo effects can worsen existing sensations, they don't necessarily create sensations from nothing. Research on conditioned allodynia (persistent pain/itch without stimuli) has yielded mixed results, but some studies show that conditioned effects can transfer between sensations. Negative expectations are typically induced through classical conditioning, verbal suggestions, or both. Classical conditioning creates implicit associations between an inert treatment and sensation aggravation, while verbal suggestions explicitly provide negative information. This review aimed to systematically investigate the magnitude of nocebo effects and their contributing factors, given the diversity in methods employed in previous research.

Existing research consistently demonstrates the role of learning in nocebo effects; verbal suggestions seem particularly potent when combined with conditioning. A previous meta-analysis (up to 2013) showed moderate to large nocebo effects, larger with combined learning methods. However, more recent studies and a comprehensive updated review were needed to examine how various learning methods affect nocebo effect magnitude. Reviews also show nocebo effects across various sensations (pain, itch) due to instructional or associative learning, highlighting the potency of negative associations, especially when patient-clinician communication is poor. This study addresses the need for a comprehensive examination of the impact of various factors on nocebo response magnitudes, including the type of sensation, stimulus modality, stimulation intensity, and the length of conditioning paradigms used in different experimental setups.

This systematic review and meta-analysis followed the PRISMA statement (2009) and Cochrane recommendations (2020), preregistered on ClinicalTrials.gov (ID: NCT04387851). PubMed, PsycINFO, EMBASE, and Cochrane CENTRAL were searched (initial search: March 18, 2019; repeated searches in June 2020 and July 2021). The review included original, peer-reviewed, controlled experimental studies on healthy human participants, focusing on cutaneous sensations (pain and/or itch) experimentally induced on the skin. Studies were included if they used verbal suggestions indicating worsened pain/itch from a (sham) treatment; studies merely conditioning pain responses without explicit treatment associations were excluded. Only studies with control groups or within-subject controls allowing nocebo effect calculations were included. Observational learning studies were excluded due to insufficient numbers. Study selection involved independent screening of titles/abstracts and full texts by two authors, resolving disagreements with a third. Data extraction (one author, checked by another) included intervention details, study population, sensation type, pain/itch rating data, stimulation type, outcome expectations, quality assessment information, and meta-analysis data. Risk of bias (RoB) was assessed using Marcuzzi et al.'s (33) method for quantitative sensory testing studies. Statistical analyses used Comprehensive Meta-Analysis software and R, employing random-effects models for meta-analyses due to design heterogeneity. Effect sizes (Hedges' g) were calculated from means and standard deviations, with standardized pain/itch scores (0-10 scale). Subgroup analyses compared Hedges' g based on learning methods, sensory stimulation type, and nocebo measurement timing. Meta-regression examined the impact of conditioning length, measurement timing, stimulus intensity, and RoB on nocebo magnitudes.

The meta-analysis included 37 distinct nocebo studies (40 study arms: 30 pain, 10 itch). Most studies (18 pain, 4 itch) used classical conditioning with verbal suggestions; others used verbal suggestions alone. RoB was generally low. Nocebo effects showed moderate to large magnitudes (Hedges' g between 0.26 and 0.71 across four pooled effects: verbal suggestions in pain, conditioning plus verbal suggestions in pain, verbal suggestions in itch, and conditioning plus verbal suggestions in itch). For pain, combining conditioning and verbal suggestions yielded larger effects (g = 0.71) than verbal suggestions alone (g = 0.63). In itch, verbal suggestions alone had a larger effect (g = 0.53) than combined methods (g = 0.26), although the number of itch studies was small. Different pain administration methods (thermal, electrical) yielded comparable effects. Meta-regression found no significant association between nocebo magnitudes and conditioning length, stimulus intensity, or RoB. Measuring nocebo effects in the first evocation trials showed larger effects than measuring the mean across all trials. This is attributed to the start of extinction after the discontinuation of learning paradigms.

The meta-analysis confirmed the moderate to large magnitudes of experimentally induced nocebo effects, ranging from 0 to 4 points on a 0-10 scale. Heterogeneity was moderate, primarily influenced by the timing of measurement (first trials showing larger effects). Nocebo magnitudes are comparable to placebo effects in other meta-analyses, but the negativity of nocebo paradigms may lead to stronger learned effects than positive expectations in placebo studies. Aversive learning is prioritized over neutral or positive information, potentially due to evolutionary reasons. In pain, combining conditioning and verbal suggestions yielded larger effects, suggesting experiential learning is more potent than suggestions alone. However, in itch, the small number of studies limits conclusions. Pain studies showed larger nocebo effects than itch, potentially because pain is perceived as a greater threat. Variability in nocebo responses was not explained by sensory stimulation type/intensity or conditioning length; this might be due to ceiling effects in learning or the dominant influence of verbal suggestions. This variability might reflect differences in experimental designs, highlighting the need for consistent and standardized experimental paradigms.

This meta-analysis quantified nocebo effect magnitudes in pain and itch for distinct learning paradigms. It replicated the stronger effect of combined conditioning and verbal suggestions for pain and established small to moderate effects for itch. Factors like conditioning length and sensory stimulation did not explain effect heterogeneity. This underscores the need for standardized experimental designs and a deeper investigation into verbal suggestions’ effects on learned nocebo effects.

The limited number of itch studies hampers firm conclusions regarding itch-specific factors. Inconsistent reporting of demographic factors and other variables (experimenter demeanor, study setting) prevented their analysis. While RoB was low, the utilized assessment tool might have missed some bias aspects.