Examining the persistence of previously induced placebo and nocebo effects over the time-period of one week enabled the assessment of their temporal stability. Both placebo and nocebo effect were revealed to be relatively persistent after one week, even though there was a significant decrease in the strength of both placebo and nocebo effects in pain ratings. However, both effects showed a rebound of effect strength from the last trials of day 1 to the first trials of day 8 and underwent no extinction over the entire time course of the experiment. Expectations did not only affect behavioral ratings but were represented in anticipatory theta-to-alpha activity on day 8. The persistence of effects was largely dependent on the individual strength of placebo and nocebo effects on day 1, as participants that showed greater effects on day 1 also showed greater effects on day 8. We further investigated the neuronal correlates of the persistence of effects and found that a stronger persistence of placebo effects was predicted by larger placebo-induced modulation of fMRI activity in the amygdala during the anticipation of pain and in the right DLPFC and the bilateral anterior insula during pain processing on day 1. Conversely, the persistence of nocebo effects was correlated with larger nocebo-induced changes in fMRI activity in the thalamus during pain perception.
Our finding of relatively stable placebo and nocebo effects, both on the group and individual level, contribute to the existing body of knowledge on the highly persistent nature of placebo and nocebo effects in clinical settings [10] and expands the understanding of the longer-term stability of placebo effects [5, 17]. Importantly, we used a larger number of test trials per session than most previous studies [17–20], which led to participants constantly receiving sensory information that did not fit their beliefs on the objective level. Nevertheless, placebo and nocebo effects did not undergo extinction. Thus, our results further challenge classical learning models in the context of placebo and nocebo effects, which would predict that participants learn from their experiences and adjust their expectations over time, leading to a reduction of placebo and nocebo effects [8, 9], but highlight the need for more complex interpretations. One possible mechanism is self-reinforcing feedback loops, whereby participants tend to learn more from experiences that align with their expectations, leading to a stabilization of these expectations so that can withstand potentially invalidating information [8, 9]. At least within one session, this may explain the stability of effects.
It is important to note that the placebo and nocebo effects stayed stable between sessions and even increased between the last trials of day 1 and the first trials of day 8. The effects were further found to persist after the change in treatment context from day 1 (MRI lab) to day 8 (EEG lab), in line with previous findings [5]. This suggests that the induced beliefs regarding the efficacy of our treatment were not overwritten by the experience on day 1 but showed a rebound in strength. Other potential explanations for the lack of decrease in expectations include that our combination of verbal instructions and strong conditioning induced highly persistent initial beliefs [20, 31], or that there was a shift in expectations from being driven by beliefs to more unconscious associations over time [10]. However, the effects on expectation ratings were remarkably stable, which indicates that participants were aware of their beliefs, suggesting rather a conscious representation of expectations. As it has already been demonstrated that both conditioning [32, 33] and verbal instructions alone [5, 34] can lead to reliable placebo and nocebo effects, further studies could compare the long-term stability of more conscious (i.e. verbal instructions) versus unconscious methods of expectation induction (i.e. conditioning) to elucidate different factors that might affect the stability of expectation effects.
We observed no difference between placebo and nocebo effects both in terms of strength and stability. This is somewhat surprising given previous evidence suggesting that nocebo effects can be introduced more easily and are more robust against extinction [16, 19]. Nevertheless, it has to be considered that the majority of these findings were derived from studies that investigated the course of expectations in one session only [16, 18, 19]. To our knowledge, we are the first to examine the longer-term stability of negative pain-related expectations. The persistence of both placebo and nocebo effects may be related to the high number of conditioning stimuli that we used (10 per condition), as the number of conditioning stimuli is related to the strength of induced effects [20]. The differential stability of placebo and nocebo effects within shorter time frames has been attributed to higher arousal in nocebo compared to placebo blocks, which impedes learning from experiences and thereby stabilizes nocebo effects [16]. Here, conditions within blocks were pseudo-randomized, which could lead to a similarly high level of arousal for all conditions and therefore similarly low rates of extinction in both placebo and nocebo effects.
Interestingly, we found predictive neural activity for the stability of both placebo and nocebo effects. For placebo effects, this activity was located in areas associated with the DPMS during pain anticipation in the amygdala, and during pain processing in the DLPFC and the anterior insula. The DPMS is proposed to play a pivotal role in placebo and nocebo effects [2, 3]. The amygdala exhibits direct connections to classical placebo areas like the vmPFC and the periaquedactual grey (PAG) [2] and decreased activity in the amygdala has been associated with increased placebo effects [4]. It further has been reported to show predictive activity for participants perceiving stimuli as more painful despite receiving an explicitly neutral cream [35]. Additionally, a rightward asymmetry in the volume of subcortical limbic structures including the amygdala has been reported for placebo responders compared to non-responders [36]. These findings underline the importance of this structure for the placebo response and the individual differences in response quality. This could be mediated by the important role the amygdala has in threat assessment [37]. A reduced threat perception might lead to higher placebo effects and also to more stable placebo effects. As the amygdala is a key structure in associative learning [2, 38, 39], reduced activity could also be interpreted as a reduction in learning from experiences during the test phase, leading to more pervasive placebo effects.
The DLPFC has been linked to placebo effects in both anticipation of pain and during pain perception [22–24, 40, 41]. The DLPFC has a pivotal role in the top-down modulation of placebo effects [2]. This could possibly relate to the suppression of learning from prediction errors described by Schenk et al. [8]. Moreover, activity in the DLPFC has been reported to have predictive power for individual placebo responses [40]. In conjunction with our findings, this illustrates the importance of the DLPFC in not only predicting individual placebo responses, but also the individual persistence of these responses. The anterior insula also has often been reported to show increased activation in placebo conditions [2] and further has been reported to have predictive power for individual placebo responses [40]. It also has been marked it as an important node in the salience network [42], and it may also play an important part in evaluating the expected threat level [24, 43, 44]. Considering the involvement of the amygdala and the anterior insula in affective processing and associative learning, the persistence of placebo effects might depend on the affective evaluation of the stimuli and learning processes triggered by it, with activity in the amygdala mediating this evaluation during the anticipation of pain and the anterior insula mediating the evaluation during pain perception. This effect of the affective evaluation might be extended by a strong top-down modulation by the DLPFC [8].
The persistence of nocebo effects was predicted by increased activity in the thalamus. The thalamus has been reported to encode nociceptive information [45] and nocebo responders showed increased activation in the thalamus compared to non-responders [46]. At first glance, our finding could also be interpreted as an artifact of the thalamus simply encoding higher differences in pain perception between the nocebo and control condition on day 1, leading to higher effects on day 8, however, it has to be kept in mind that we already accounted for the effects on day 1. Thus, a higher nocebo effect on day 1 cannot explain the observed relationship. A more promising explanation is based on the thalamus being linked to conditioning processes regarding pain [47]. Thus, stronger activity in the thalamus could be interpreted as an indicator of a stronger conditioning effect that exerts influence on stimulus processing during the test phase or, in line with the concept of self-reinforcing feedback loops, a stronger learning process from expectation-consistent information [8, 9]. The thalamus therefore appears to play an additional role next to stimulus processing in the longer-term maintenance of nocebo effects.
Even though we found no significant behavioral differences between the persistence of nocebo and placebo effects, we see different neural patterns predicting the persistence of these effects. This fits the results of our previous study [24] with differential neural patterns for positive and negative expectations during pain perception. Additionally, this might indicate that positive and negative valences of expectations rely on differential mechanisms that support their up-keeping. Interestingly, while several areas are connected to placebo and nocebo effects during anticipation of pain and during pain processing [24], out of these only few regions seem to have predictive power for the long-term persistence of these effects.
On day 8, we observed an increased theta-to-alpha EEG power for negative compared to positive expectations during the anticipation of pain. This is in line with previous findings that indicated a high relevance of the pre-stimulus low frequency oscillatory activity that allow for the subsequent modulation of the pain perception [25–27, 43, 48]. Lower frequency activity, whether altered by spontaneous fluctuations or pain-related expectations, has been shown to affect the pain perception, although the direction of this modulation was inconsistent over studies [25–27, 48]. Nevertheless, this points towards an important role of alpha activity for the top-down signaling of expectations that ultimately lead to the modulation of bottom-up sensory information as observed in placebo and nocebo effects [25, 49]. As we observed these changes in neural activity one week after the original induction of expectations, this further supports the notion that expectations are not transient but instead result in true, lasting changes in perception.
Taken together, our findings show that the initial experience with a treatment can have long-lasting effects in the same and other contexts. This implies the need to reduce or at least reframe negative experiences in order for them not to impede future treatments. Conversely, this also means that positive experiences can have long-term consequences and should be harnessed. This raises important clinical questions. Notably, the persistence of nocebo effects could serve as one crucial gateway for the development of chronic pain. The emergence of chronic pain is hypothesized to follow a complex interaction of factors, gathered in an integrative psychobiological pain model proposed by Büchel [50]. This model synthesizes the most common approaches for understanding the emergence of chronic pain and underscores the pivotal role of expectations. Following this model, persisting alterations of pain perception by negative expectations (which could also be termed as a stable nocebo effect) could act as a significant driving force, accelerating a vicious cycle leading to chronic pain. However, our findings regarding the persistence of placebo effects suggest that the opposite could also occur, manifesting as a virtuous cycle. This underscores the importance of carefully handling the expectations of patients. The question further arises if the persistence of expectation effects on perception could be an important factor in other disorders. Persistent expectations could e.g. play a role in psychosis, in which the influence of expectations on perception has already been discussed [51].
Overall, our results show that both placebo and nocebo effects can remain stable over one week, and that the stability of the individual effects were determined by distinct neural correlates for positive and negative expectations. Especially for positive expectations areas related to learning, affective processing, and top-down control appear to predict the strength of responses one week after the initial induction of expectations.