A summary of key information from all papers is presented in Electronic Supplementary Material 1, whereas key categories and corresponding identified themes and subthemes are presented in Table 1.
Domains of Nocebo Effect
Across all studies, the nocebo effect was successfully invoked in the domains of pain sensation [6, 24,25,26,27,28,29,30,31,32], nausea [33, 34], non-specific symptoms [21, 35, 36], perception of cognitive abilities [1, 37, 38], skin dryness , touch [6, 39], itching , headache , and cardiorespiratory function . The nocebo effect was not observed in the satiety  and dizziness  domains. In two studies (one in the area of pain  and another in the area of nausea ), a nocebo response was not confirmed at the level of the whole sample, but rather was observed among participants exhibiting certain characteristics only. It should also be noted that, across studies, not only was the nocebo effect invoked in different domains but also used in different procedures. As such, these findings point to the universality of nocebo as well as to the importance of participant characteristics and experimental circumstances in invoking the nocebo effect.
Conditioning. Research examining the role of conditioning in the onset of nocebo has identified the conditions under which the nocebo effect can be invoked as well as those that contribute to its severity. Several studies [25, 28, 29, 31] demonstrated that nocebo can be invoked using psychological stimuli alone (e.g., abstract visual stimuli) as opposed to pharmacological stimuli (e.g., a placebo ointment). In cases of a pharmacologically induced nocebo effect, such as inactive creams or sprays, it is not possible to completely rule out the influence of previous learning (i.e., expectations regarding these substances). In contrast, the nocebo effect was also invoked in these studies using stimuli for which participants did not have any associations with unpleasant symptoms, such as horizontal or vertical stripes , light of a certain color , or a human face  displayed on a screen.
The findings of the studies included in this review confirm that a number of general conditioning laws can be applied to the nocebo effect. For example, a longer conditioning process produces a more robust nocebo effect and one that is more resistant to extinction . This finding can be explained by the increased predictability of events under this condition and has also been observed for the placebo effect . Interestingly, the role of conditioning duration was more important for non-painful tactile stimuli than for painful stimuli, a finding that perhaps points to the importance of rapid learning about stimuli associated with pain or physiological discomfort. In addition, two studies within a same paper demonstrated that the nocebo effect can be generalized to similar stimuli , where the expression of nocebo for the original stimuli for which conditioning was performed (e.g., pictures of animals) and new stimuli that fall into the same conceptual category (a second group of pictures of animals) practically does not differ . Such generalization was more pronounced if the magnitude of the original nocebo response was greater. This is consistent with the assumption that generalization is an active cognitive process that requires explicit learning . Conversely, one study demonstrated that nocebo can be invoked even without the participants’ consciousness using implicit conditioning . In this study, painful and non-painful stimuli were administered after the presentation of a series of different visual stimuli (abstract black characters on a white background with red lines in the foreground that served as distractors). All visual stimuli that preceded painful stimulation were different from each other, but shared one difficult-to-notice characteristic (symmetry of black characters) that distinguished them from the visual stimuli that preceded painless stimulation (in which all were characterized by asymmetry of black characters — counterbalanced). Although the participants could not report the difference between these two groups of visual stimuli upon request, the nocebo effect was nevertheless evident. Two additional studies further demonstrated that the nocebo effect can also be activated subliminally. After conditioning with a supraliminal stimulus (presentation of a male face), nocebo was observed during later subliminal presentation of the same conditioned stimulus, but not for a control stimulus (another male face) .
Together, the findings of these studies have a number of implications for people experiencing chronic pain. Firstly, we might assume that repeated conditioning of the contextual stimuli that accompany chronic pain, results in nocebo pain which is stronger and more resistant to extinction. In addition, as a result of generalization, implicit and subliminal conditioning processes, pain can also be activated or intensified by stimuli that were not present during experienced pain or that were not consciously registered. As such, the onset of pain may seem very unpredictable and uncontrollable, thus contributing to feelings of helplessness and stress .
Manipulation of Expectations. Research in which nocebo was invoked by manipulating expectations emphasizes the importance of emotions in the onset of the nocebo effect. Specifically, it has demonstrated that people exposed to nocebo treatment report greater levels of anxiety , worry , arousal and a lower sense of dominance , and a general worsening of mood . Similarly, a larger nocebo effect was found among individuals with a more pronounced fear of pain, but this relationship was mediated by increased stress in these individuals . Negative emotions are assumed to provoke increased attention to bodily symptoms , which makes it more likely that such symptoms will be noticed and, in turn, attributed to a given treatment. In contrast, induction of a positive mood appears to preclude the later onset of the nocebo effect , perhaps by diverting attention from symptoms or changing expectations about the likelihood that negative symptoms will occur .
In an attempt to shed further insight into the role of expectations, a number of studies directly measured participants’ expectations regarding the nocebo effect. In one study, nocebo instruction influenced not only participants’ expectations but also their performance and subjective experience . In a second study, nausea expectations predicted the experience of nausea . In two studies, nocebo instruction had no effect on expectations or on symptom experience or performance [38, 43]. In one of these studies , however, a moderate correlation was found between expectations about task performance and subsequent perception of changes in performance following treatment (where no difference in objective performance measures was observed). Thus, expectations play an important role in the emergence of the nocebo effect, regardless of whether they stem from experimental manipulation or not. In one interesting study , an objective electrophysiological measure whose early phase is associated with anticipation — contingent negative variation (CNV) — was used as a measure of pain expectation. This study demonstrated that the nocebo cue produced a high amplitude CNV potential while the placebo sign produced a low amplitude potential. A similar and more recent study investigated the impact of the specificity and valence of memory and symptom-related expectations on placebo and nocebo . Although individuals with higher memory specificity and more positive expectations regarding the future onset of itching exhibited a stronger placebo response, no such patterns were found for nocebo, a finding that was attributed to the robustness of the nocebo effect.
On the whole, there is evidence of the importance of expectations in the occurrence of nocebo, both when expectations are measured by participant report and using electrophysiological measures. However, further research is needed to better understand the specific characteristics of expectations that predict a stronger nocebo effect. For example, the nocebo response may be the result of the interaction between general expectations about the action of pharmacological or psychological treatments gained from previous experiences and specific expectations about symptoms and symptom intensity that are acquired during the experimental procedure. This question certainly deserves further research attention.
Comparison of Conditioning and Expectation Manipulation. In studies comparing conditioning and expectation manipulation procedures, the nocebo response was more pronounced when both procedures were used when compared to only one [30, 31], suggesting the presence of an additive effect.
Levels of Nocebo Response. Analysis of results across a number of studies offer the interesting possibility to compare the occurrence of nocebo at various levels, including sensory, affective, physiological and behavioral levels. For example, perception of more severe pain (the sensory component of nocebo) was found in studies where conditioning was performed explicitly [30, 31], but not when it was conducted implicitly . In contrast, greater perceived discomfort with pain (affective component) was evident with both explicit [25, 31] and implicit conditioning . Studies measuring the physiological component of the nocebo response found higher skin conductivity following presentation of the nocebo cue compared to that measured following the placebo sign [25, 31]. Interestingly, extinction occurred very quickly in the case of explicit conditioning , while the physiological component of nocebo was resistant to extinction in the case of implicit conditioning . Another study found that cortisol was increased in the nocebo group when compared with a control group (physiological level), but there were no differences in subjective pain ratings between the two groups (sensory level) . Research involving behavioral measures produced inconsistent results. Namely, while the nocebo effect was not evident in a study that measured reaction time required to terminate a painful stimulus , another study demonstrated that men had shorter reaction times when terminating a nausea-inducing rotation stimulus after receiving a nocebo instruction . It should be noted, however, that participants in the first study were instructed to terminate the stimulus as soon as possible (reaction time), while in the second study, they were told to terminate the rotation when they could no longer endure it (tolerance time). Finally, in a somewhat atypical study, the nocebo effect was evident in physiological and behavioral levels . In this study, half of the participants were told that they had a genetic risk for reduced exercise capacity , while the other half were told they had a protective gene. This nocebo information had a negative effect on both cardiovascular measures (physiology) and running endurance (behavior). It should be noted, however that, in this study, each participant was given “personalized nocebo information” (i.e., that tests indicated that he or she was at increased genetic risk for reduced exercise capacity) in contrast to receiving information that most people experience side effects following specific treatment. Secondly, no concrete treatment was applied in this study and, as such, does not speak to the nocebo effect in the narrower sense. This onset of symptom exacerbation following the expectation of such exacerbation without any special treatment has been called a nocebo-related effect . In another study by the same authors  in which participants were told that they had either a risk or protective gene for satiety, a placebo effect was evident, but the nocebo effect was not observed.
Manipulation of participants, expectations about task performance allows for a comparison of nocebo response using both subjective and objective measures. The studies included in this review suggest that such manipulation mainly affects subjective measures, while the situation is less clear for objective performance measures. In a study in which participants were told that a nasal spray increases or decreases arousal , group differences were evident in participants’ assessments of their performance on attention and working memory tasks, while no differences on objective task performance measures were observed. Similarly, in another study , participants receiving the instruction that transcranial magnetic stimulation causes a left-sided attention deficit reported concentration difficulties during a visual search task following false stimulation. In contrast, the opposite effect was observed on objective measures of left-sided attention, where the nocebo group exhibited improved performance. The authors attribute this latter result to the greater effort exerted by the nocebo group, while the placebo group may have expected “help” from the treatment. Conversely, in a previously mentioned “genetic risk” study , nocebo information affected participants’ performance (running endurance) but not subjective impression (running duration before reporting fatigue and feeling hot). Obviously, not all types of performance can be equally affected by effort and participants are not equally familiar with all types of performance deterioration (e.g., left-sided attention deficit ). Also, for performances more affected by emotions, it may be more easy to produce the nocebo effect. Additional research measuring the occurrence of nocebo using various measures and across differing levels and domains would contribute to a better understanding of the mechanisms underlying the nocebo effect and the conditions necessary for its occurrence.
Nocebo Risk Factors
The question of whether some people are more prone to placebo and nocebo effects than others is one that has long interested researchers, but one for which there is not yet a clear answer. The papers included in this review, albeit representing only a small number of studies, demonstrated that a number of individual characteristics are associated with a more pronounced nocebo response. In one study, participants who believed they were sensitive to radiation reported more symptoms and symptoms of higher intensity when they believed radiation had been activated . People with more pronounced somatosensory amplification reported a higher intensity of tactile stimulation when a false Wi-Fi signal was applied, but only if they had viewed a video about the harmfulness of this signal . Anxiety and depression also appear to be associated with a more pronounced nocebo response for painful stimuli , but this was the case only in shorter conditioning, whereas the effect of personality diminished in a longer conditioning. Finally, individuals with greater fear of medical pain reported more severe pain after applying a nocebo cream . Because this result was not observed for any of the five-factor-model personality traits nor for fear of other types of pain, the authors of this study concluded that arousal by trait relevant situational cue is necessary for expression of that trait.
Surprisingly, none of the studies included in this review examined the nocebo effect among health-anxious individuals. Health anxiety is a consequence of the belief that normal bodily sensations or changes are a sign of serious illness . Highly anxious people are prone to heightened attention to their body and bodily sensations and could therefore be more prone to the nocebo effect. There are sporadic findings that people with high health anxiety are more likely to report drug side effects , but more studies are necessary. Another characteristic that might be relevant to nocebo is anxiety sensitivity — a fear of anxiety itself because of the belief that anxiety can have detrimental physical, mental, and social consequences . Among anxiety-sensitive individuals, the expectation of unpleasant symptoms and a consequent increase in anxiety might further heighten fear and related bodily sensations, thus leading to a stronger nocebo response. While some preliminary findings have suggested that anxiety sensitivity is associated with the nocebo effect , further research examining this topic is certainly needed.
In some studies, gender differences in nocebo response were observed. In general, research demonstrated that women report more pain after the nocebo treatment [24, 26, 27]. For example, when participants were taught to associate three different scenarios presented through virtual reality (e.g., an office) with pain, vibration or lack of stimulation, a nocebo response was observed among women only . In this study, women reported a higher pain experience for the stimuli applied in the earlier pain-related scenario than for the control scenario, while there was no difference between the pain-related and vibration-related scenarios. In another study reporting the nocebo effect among women only , women reported more severe pain when they had been told that a short sound stimulus preceding painful stimuli (prepulse) would intensify the pain experience. Interestingly, this increase occurred for all stimuli following nocebo instruction, regardless of whether a prepulse was present or not. Again, this finding is indicative of the role of anxiety, which likely increased following instruction, in the onset of a nocebo response . In another study examining gender differences, conditioning proved to be more effective in inducing nocebo for nausea among women, while manipulation of expectations was more effective among men . A similar finding has been demonstrated elsewhere and is similarly true for the placebo effect . The authors argue that this result reflects gender differences in learning capacities . Furthermore, although the men participating in this study had previously terminated nausea-induced rotation following nocebo instructions, they did not report a greater number of symptoms, perhaps suggesting that some of the observed gender differences stem from a socially conditioned willingness to acknowledge symptoms . Another proposed explanation for the more frequent occurrence of nocebo among women is that women have more stable negative expectations, worry more, and are more problem-oriented than men .
The only study examining nocebo in a clinical sample points to some unique aspects of this population. In this study, women with cancer reported more cognitive impairment when they had been told that chemotherapy could result in such impairment, but only if they exhibited heightened awareness of the stigma associated with their disease . This finding indicates the importance of social cognition in this population. However, this finding pertains to the reporting of symptoms that appeared before the nocebo instruction was given and, as such, is perhaps a result of a change in the interpretation of previous events or even event recall in accordance with the instructions provided. Arguably, if participants had been asked to assess their own performance on a cognitive task within an experiment, the effect might have been even more pronounced. In any case, there is certainly a lack of research examining the nocebo effect in clinical samples, including those with both somatic and mental illnesses.
How Is the Nocebo Effect Reduced or Prevented?
A number of studies demonstrated that various procedures performed prior to invoking nocebo were effective in its prevention. In one study, invoking a nocebo response through conditioning was more difficult following latent inhibition . This involved the presentation of a conditioned stimulus prior to the conditioning process without an appropriate unconditioned stimulus. In this study, participants were first exposed to false stimulation that allegedly causes nausea, followed by real stimulation that caused nausea and then again to the false stimulation . At the latter false stimulation, participants reported nausea of the same intensity as the control group, who received false stimulation only (no nocebo induction). In contrast, a nocebo response was evident in a third group that underwent the classic nocebo conditioning procedure (actual simulation, followed by false stimulation). The authors argue that the efficacy of latent inhibition in preventing the nocebo effect is a result of reduced attention to the conditioned stimulus because of its decreased novelty . Interestingly, the outcome following latent inhibition was the same even when participants were told that this procedure would be performed and for what purpose . In another study, the aforementioned induction of positive mood was demonstrated to prevent a nocebo response . Finally, findings of one study indicated that invoking a placebo response may also diminish the subsequent nocebo effect . However, the reverse is not true, where evoking a nocebo response did not reduce the subsequent occurrence of a placebo response .
Conversely, some strategies did not prove to be effective in reducing or preventing a nocebo response. Specifically, studies that aimed to reduce the nocebo response by differently formulating or re-framing information given to participants regarding side effects did not produce impressive results [23, 49]. For example, in one study , one group of participants were given information regarding drug side effects in the standard manner, while a second group were told that side effects were a sign that the drug was working. All participants were administered the real medication. Although no differences were found between groups, there were indications that such a procedure would be useful in the prevention of nocebo among people with negative beliefs regarding certain drugs . Another strategy that has not been demonstrated to be effective in reducing the nocebo effect is reducing the price of a given medication. In fact, marketing research has shown that a lower price for a medicinal cream resulted in a stronger nocebo response among participants who had previously been exposed to information about the side effects of such creams, as compared to a higher price . In light of the readily available information about side effects that is included in medications’ instructions for use, this finding suggests that the frequency and number of side effects could increase when a given medication goes on sale. Indeed, price itself appears to act as a placebo or nocebo by creating certain expectations about a given product .