If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Corresponding author: Dr Emily Hird, Department of Psychosis Studies, Institute of Psychiatry, Psychology and Neuroscience main building, 16 De Crespigny Park, London, SE5 8AF,
Psychosis is characterised by unusual percepts and beliefs in the form of hallucinations and delusions. Antipsychotic medication, the primary treatment for psychosis, is often ineffective and accompanied by severe side-effects, but research has not identified an effective alternative in several decades. One reason that clinical trials fail is that psychosis patients tend to show a significant therapeutic response to inert control treatments, known as the placebo effect, which makes it difficult to distinguish drug effects from placebo effects. Conversely, in clinical practice, a strong placebo effect may be useful as it could enhance the overall treatment response. Identifying factors that predict large placebo effects could improve the future outlook of psychosis treatment. Biomarkers of the placebo effect have already been suggested in pain and depression, but not in psychosis. Quantifying markers of the placebo effect would have the potential to predict placebo effects in psychosis clinical trials. What is more, the placebo effect and psychosis may represent a shared neurocognitive mechanism in which prior beliefs are weighted against new sensory information to make inferences about reality. Examining this overlap could reveal new insights into the mechanisms underlying psychosis and indicate novel treatment targets. We provide a narrative review of the importance of the placebo effect in psychosis and propose a novel method to assess this.
). Few pharmacological alternatives to current antipsychotics have been identified in several decades, which is concerning considering the variable efficacy and severe side-effects of antipsychotics (
). Rather than a lack of treatment response, a key reason for the lack of progress may be that patients show a significant placebo response in clinical trials (
The placebo response refers to symptom improvements that result after administration of an inactive treatment, including the placebo effect, natural history, and regression to the mean (
). The placebo effect describes symptom improvements attributable to neurobiological and psychological processes associated with beliefs about treatment. Placebo effects can alter pain (
Cognitive theories provide a framework to explain the placebo effect, proposing that the brain interprets a noisy world through beliefs built on prior experience (
). Experienced reality thus reflects the brain’s ‘best guess’ of the world, which can be misleading, as seen in the placebo effect. For example, the longstanding belief that a medication is an effective painkiller is unlikely to change during an occasional placebo treatment. Instead, the brain’s pain modulatory system may inhibit sensory inputs to reconcile the mismatch between “what is expected” and “what is felt”, at least until beliefs are updated. Intriguingly, the odd perceptions and beliefs which characterise psychosis are also associated with a conflict between beliefs and sensory input; for example, delusions represent strong beliefs that are maintained despite contradictory evidence (
). While the placebo effect reflects an adaptive effect of learned beliefs, psychosis reflects a divergence in these. This shared mechanism could explain the marked placebo effect in psychosis (
) and could represent a treatment target for psychosis.
In randomised controlled trials (RCTs) placebo responses are seen as a source of noise to be accounted for. However, placebo effects represent an intriguing phenomenon with a well-established scientific following (
) which indicates the importance of beliefs in our experience of reality. Placebo effects and treatment effects modulate the same endogenous processes; for example, like analgesic treatment, placebo treatment is associated with endogenous release of opioids (
Identifying factors that predict the individual placebo effect could help to enhance treatment success, either by improving the methods of clinical trials through the detection of individuals likely to show a strong placebo effect, or by allowing the placebo effect to be harnessed in-the-clinic to enhance the effect of drug treatments. Methods to identify predictors of placebo responses in psychosis have been explored using meta-regression (
). but more precise individual predictors of placebo effects have not yet been identified in psychosis.
We discuss shared neurocognitive processes behind psychosis and the placebo effect and propose that a more precise approach should identify neurocognitive measures of belief-updating.
The placebo response is present in antipsychotic trials
The placebo response in psychosis is significant and clinically relevant, and its magnitude varies between individuals in patients with psychosis (table 1). Placebo responses reduce the likelihood of finding treatment effects in antipsychotic trials, and their magnitude have increased over the last six decades(
). One meta-analysis showed that on average, in RCTs of antipsychotics in the 1960s, scores on the Brief Psychiatric Rating Scale (BPRS) in patients receiving placebo worsened by 3.5 points pre to post treatment. In contrast, in the 2000s, the average placebo-treated patient improved by 3.2 BPRS points (
). See table 1 for a summary of meta-analyses evaluating the extent of placebo responses in antipsychotic RCTs, with 27-59% of schizophrenia patients showing a placebo response of more than 25% PANSS score reduction (
Early Placebo Improvement Is a Marker for Subsequent Placebo Response in Long-Acting Injectable Antipsychotic Trials for Schizophrenia: Combined Analysis of 4 RCTs.
). Placebo responses have been getting more powerful over recent decades, suggesting that it is a key concern for studies developing new treatments for psychosis (
Table 1a summary of meta-analyses reviewing the magnitude of placebo responses and average predictors of placebo responses in RCTs for antipsychotic medication
Early Placebo Improvement Is a Marker for Subsequent Placebo Response in Long-Acting Injectable Antipsychotic Trials for Schizophrenia: Combined Analysis of 4 RCTs.
Placebo response (>25% PANSS score reduction) was identified in 47% of patients (per-protocol analysis) or 27% of patients (last-observation-carried-forward analysis)
Percent reduction in the PANSS total score at week 1 (+) Lower PANSS G12 item score (+)
Mean placebo response was 6.25 PANSS units Placebo responses increased by an average of 2.74 PANSS units per decade since 1970.
Lower age (+) Shorter duration of illness (+)
More recent publication year (+) Larger study sample size (+) More study sites (+) Use of the PANSS rather than the BPRS scale to measure response (+) Shorter wash-out phases (+) Shorter trial duration (+)
Identifying factors that predict individual placebo responses could improve treatments in psychosis, by improving treatment detection on clinical trials, or by enhancing treatment effects in-clinic. Various meta-analyses have attempted to identify predictors of individual placebo responses across antipsychotic trials (table 1). As table 1 shows, patient-centred predictors are themed around disease duration, symptom severity, age, and prior treatment, while study-centred predictors mainly involve the number of arms or sites, trial duration, sample size, industry sponsorship, type of randomisation (thus perceived likelihood of getting active treatment), type of measure, and how recent the study is. A qualitative review of 31 meta-analyses and systematic reviews of over 500 RCTs looked across different psychiatric disorders to identify variables associated with the placebo response. Of 20 variables examined, three were consistently associated with placebo effects across disorders: lower symptom severity, more recent trials, and unbalanced randomisation (more patients randomly assigned to drug than to placebo) (
Are placebo and drug-specific effects additive? Questioning basic assumptions of double-blinded randomized clinical trials and presenting novel study designs.
). This suggests that in contrast to an additive effect of drug and placebo components of a treatment, there can be an interaction where the treatment response is greater than the sum of its parts (
)) (figure 1). For example, in pain, positive beliefs can enhance, and negative beliefs can abolish, the analgesic effect of an opioid agonist, and both are associated with changes in the endogenous pain modulatory network (
). Consequently, results in the active treatment arm could be influenced by individual placebo effects. Data indicate a decrease in antipsychotic drug-placebo differences from 1960 to now (
) and a recent meta-regression analysis showed that the increasing placebo response was the only predictor, other than industry sponsorship, of antipsychotic drug effect sizes across 167 RCTs (
Sixty years of placebo-controlled antipsychotic drug trials in acute schizophrenia: Systematic review, Bayesian meta-analysis, and meta-regression of efficacy predictors.
). A heterogenous placebo response could increase the likelihood of trial failure, creating the impression that newer therapies are less efficacious than older therapies, increasing costs for drug development and disincentivising industry from investing in new treatments (
The studies in table 1 suggest a picture of a typical placebo responder ‘profile’; however, they represent post-hoc averages of placebo responses across studies, which are contaminated by artefacts, including regression to the mean (where extreme measurements tend to be followed by measurements that are closer to the mean) (
). Identifying the mechanisms and individual predictors of placebo effects could allow more precise prediction of the magnitude of the placebo effect, which would be necessary to reliably predict placebo effects in clinical trials or in the clinic.
The most convincing evidence for the presence of a placebo effect in RCTs for psychosis is that the probability of receiving treatment (which will influence participant’s beliefs about the probability of receiving treatment) predicts outcomes. This has been shown across 4 meta-analyses, reflected in the association of placebo responses with more treatment arms, unbalanced randomisation, and a higher probability of receiving treatment (
). Another hint is that a longer trial duration appears to be associated with a decreased response in the placebo arm of trials, as found across 4 meta-analyses (
). This could reflect the fact that the influence of a participant’s beliefs might decrease over time due to extinction (the loss of a conditioned response over time due to the absence of reinforcement), causing a decreased placebo effect. However, given the potential for statistical contamination in clinical trials, these findings should only be taken as suggestive.
Experimentally inducing placebo effects removes the issue of statistical artefacts associated with placebo responses. There is substantial evidence for an altered impact of beliefs in psychosis (discussed further below), but Schmack et al. carried out what is to our knowledge the only experimental investigation of a placebo-like paradigm in psychosis. In a training session, participants viewed a set of visual stimuli which were biased in terms of their direction of motion. Participants were informed that the motion was ambiguous and that any perceived bias was caused by the glasses through which they viewed the stimuli. In a subsequent testing session, participants continued to wear the glasses and were asked to rate the direction of motion of truly ambiguous stimuli, which allowed assessment of the effect of beliefs associated with the placebo-like glasses on perception of motion. In healthy people, delusionality was associated with a stronger influence of placebo-like beliefs and increased fronto-sensory functional connectivity (
). Interestingly, schizophrenia patients exhibited a reduced effect of placebo-like beliefs compared to healthy controls, but a positive correlation between the effect of beliefs and delusionality, and enhanced belief-related fronto-sensory connectivity compared to controls (
). This suggests that while patients with schizophrenia have a decreased tendency to acquire new placebo-like beliefs, delusionality is associated with an increased influence of these beliefs. This provides a hint that delusional individuals might exhibit a stronger placebo effect, but also this is not consistent across all individuals with psychosis. We discuss this further below.
The cognitive basis of psychosis appears to overlap with that of the placebo effect
Most research into the neurocognitive basis of the placebo effect focuses on pain and depression. See supplementary table 1 for a summary of papers assessing individual predictors of the placebo effect in these modalities. Given the paucity of research into the neurocognitive basis of the placebo effect in psychosis, it is difficult currently to assess an empirical link between the two phenomena. However, in the following section we summarise potential cognitive similarities between the placebo effect and key characteristics of psychosis, which are worthy of further investigation.
There is substantial evidence that beliefs shape the placebo effect, as has been established in placebo analgesia (
). Placebo analgesia is stronger when people have more reliable beliefs, and this is associated with activity in the periaqueductal gray, key areas for signalling prediction error when errors in beliefs are detected (
). Placebo effects can be driven by different levels of computation: both unseen (subliminally presented) cues and perceived (supraliminally presented) cues evoke a comparable influence on pain perception (
). Classical conditioning, - where the presentation of a cue alongside an unconditioned stimulus such as reduced pain causes the cue to become conditioned to the stimulus(
) – is one way to induce the placebo effect. More pairings between treatment cue and a real drug increases the magnitude and duration of placebo analgesia, in line with a conditioning effect (
). The type of conditioning is important: extinction is eliminated after a partial reinforcement schedule compared to a full reinforcement schedule in placebo analgesia (
Alterations in belief-updating at different stages of psychosis may influence the nature of the placebo effect in these patients. One study assessed how cognitive beliefs (induced by written presentation of a phoneme) or perceptual beliefs (induced by an accompanying visual input of lip movements) influenced perception of ambiguous auditory input in participants at clinical high-risk of psychosis (CHR) and first-episode psychosis (FEP) patients. FEP patients relied more on cognitive beliefs than CHR participants or healthy controls (
) which we predict would cause these individuals to show a stronger placebo effect to cognitive cues, such as an overt instruction that a treatment will cause improvement. CHR participants in this study exhibited a weaker effect of perceptual beliefs (
) which we predict would cause a weaker placebo effect in response to perceptual cues, such as a change in sensation correlated with a treatment. The severity of psychosis symptoms and associated cognitive changes could also influence the placebo effect. For example, paranoid individuals are thought to expect more environmental volatility, be more rigid in these beliefs, switch choices excessively in anticipation of reversals on a reversal-learning task. These participants rely more on their beliefs about volatility than on actual feedback, and confuse errors as a signal that the task has changed (
). Because these individuals rely more on their prior beliefs and less on environmental information, we predict that they may be slower to acquire new placebo-related beliefs in the first place, but that once these beliefs are established, they would express a greater magnitude of placebo effects. Alterations in learning in psychosis could also influence the latency of the placebo effect. For example, psychosis patients exhibit decreased reversal-learning (
) which could mean placebo treatment initially influences outcomes less because patients do not learn associations quickly, but we predict would also cause a more persistent effect of beliefs once they have been acquired.
Numerical ratings are used as a behavioural measure of placebo effect in analgesia (
). However, this approach is non-specific to the underlying cognitive mechanism. To more completely assess how beliefs, belief-updating, and precision influence outcomes in psychosis, the cognitive mechanism of the placebo effect can be formalised using computational frameworks which allow underlying mechanisms to be mathematically quantified. Rival models can be formally compared for fit with behaviour and the resulting parameters regressed against neural biomarkers to identify neurocognitive predictors of the placebo effect. These formal models can be applied to cognitive data to identify individuals likely to show a placebo effect and increase the chance of detecting treatments on clinical trials or enhance treatment effects in-clinic. See figure 2 for a diagram summary of the similarities between the placebo effect and psychosis, based on computational models of cognition.
Figure 2cognitive similarities between the placebo effect and psychosis.
). According to this framework, to navigate the world we learn contingencies between events to predict the value of future outcomes. These beliefs are updated based on the error between belief and outcome (prediction error). Reinforcement learning has been the subject of extensive research to formalise decision-making in humans (
). The placebo effect is inherently related to value processing, as symptom relief has a positive value, and the placebo effect has also been linked to reward processing (
). Reinforcement learning models have been used to model decision-making in psychosis groups, revealing disrupted prediction error signalling in the dopaminergic midbrain and striatum in CHR participants and FEP patients (
). These alterations could influence the individual placebo effect in psychosis populations.
Reinforcement learning does not explicitly account for how the uncertainty of beliefs or sensory information might change their influence on perception. Another computational framework which does account for this is predictive processing (
). Here, the brain makes inferences about the causes of noisy sensory inputs using an internal model of the world. Incoming sensory data are compared against beliefs, generating prediction errors, which refine higher-level beliefs, in an iterative hierarchical process related to the neural hierarchy of the brain (
). The influence of beliefs during inference is weighted by their statistical certainty (known as precision). An uncertain belief would have less impact than a certain belief. Recent advancements in neuroimaging such as laminar functional magnetic resonance imaging (fMRI) – which allows layer-specific neuroimaging(
). The placebo effect has been proposed to reflect this inference process, with the mean and certainty of beliefs shifting perception of a painful stimulus towards prior beliefs, and both cued and trait expectations (such as optimism about pain levels) affecting pain perception (
). There is evidence for an influence of precision on the placebo effect in pain, with most studies indicating that more precise beliefs have a stronger impact on perception (
Vase L, Robinson ME, Verne GN, Price DD (2003): The contributions of suggestion, desire, and expectation to placebo effects in irritable bowel syndrome patients. An empirical investigation. Pain 105: 17–25.
) which has revealed that patients with schizophrenia update their beliefs more to unexpected information and less to consistent information compared with healthy controls (
). One study used Pavlovian conditioning and computational modelling to show that people who hallucinate exhibit stronger beliefs, hallucinations are associated with increased variability in weighting between sensory evidence and perceptual beliefs, and patients with psychosis are less able to recognize an increasing volatility in contingencies (
). These results hint at a mechanism in which patients with psychosis might exhibit increased placebo effects because they update their beliefs either too much or too little, sub-optimally weight pre-existing beliefs against new information, and misjudge the level of volatility in the environment.
A neuroanatomical overlap between psychosis and placebo effects
The neural processes important for learning are implicated in the placebo effect and altered in psychosis. Dopamine is linked to prediction error signalling(
) and implicated both in the placebo effect and psychosis in areas important for reward processing: positron Emission Tomography (PET) shows that the placebo effect to pain is associated with dopamine release in the nucleus acccumbens(
The role of dopamine dysregulation and evidence for the transdiagnostic nature of elevated dopamine synthesis in psychosis: a positron emission tomography (PET) study comparing schizophrenia, delusional disorder, and other psychotic disorders [no. 11].
). In placebo analgesia, areas such as the dorsolateral PFC, rostral ACC and periaqueductal gray act as a descending pain control system to modulate pain perception top-down via the downregulation of nociceptive activity in sensory areas (
). These findings indicate a role for top-down processing in psychosis, and in the placebo effect, which are linked to an increased influence of beliefs on perception.
The observation of a cognitive and neural overlap between the placebo effect and psychosis is of particular clinical interest because characterising the mechanism driving the two phenomena may facilitate the identification of novel treatment targets for psychosis.
Future directions
To examine placebo effects in psychosis, it would be useful to examine in detail (
) belief-updating, across patient groups (specifically CHR participants, FEP patients, and schizophrenia patients), and their associations with delusions or hallucinations. Computational modelling should be used to delineate differences at different levels of belief (from high-level cognitive beliefs to low-level perceptual beliefs), as well as parameters that might influence placebo effects such as expected volatility and learning rate. A key line of enquiry would be to examine the association between these variables and placebo analgesia effects induced by a conditioning paradigm (
), in patients with psychosis and assess whether the influence of these factors differ in patients compared to of healthy controls. It would be particular of interest to assess whether cues can be paired with physiological responses in psychosis. After repeated administration of a dopamine agonist, a placebo intervention elicits both a clinical response and thalamic response in Parkinson’s patients (
Quantifying and predicting the likely placebo effect in individuals could allow us to account for placebo effects in clinical trials, or to enhance treatments in-clinic. Given the complex and heterogenous nature of the placebo effect (see Supplementary table 1), the prediction of placebo effects could be improved using multivariate data (
). Identifying individuals likely to show a strong placebo effect early on could be used randomise those less likely to respond to placebo to the treatment arm in RCTs and allow us to account for individuals likely to show a strong placebo effect in the analysis. Identifying individuals likely to experience a placebo effect would remove the ethical barrier of providing a placebo treatment with unknown efficacy when other medications are available, and a placebo treatment could be administered alongside other medications to enhance treatment effects. See supplemental section for further detail on future directions in predicting and harnessing the placebo effect.
Limitations and challenges
This paper does not constitute a systematic review. Hence, we can only claim to provide a subjective narrative of the potential importance of the placebo effect in psychosis and how to assess this. Furthermore, we draw on findings in pain research to make inferences about the placebo effect in psychosis, but there is no direct evidence linking the neurocognitive basis of pain and psychosis. Unlike research into the placebo effect in pain, and to some extent in depression (see supplementary table 1), currently there is a paucity of empirical data describing predictors of the placebo effect in psychosis. A key challenge for this research area is to develop a better understanding of the shared neurocognitive mechanisms between pain, depression and psychosis, and to assess mechanisms underlying the placebo effect in psychosis.
Conclusion
Contemporary accounts suggest that rather than being a passive receiver of information, the brain constructs reality through learning. Placebo effects demonstrate the influence of beliefs on mental and physical states, and psychosis demonstrates the impact of beliefs going awry. We argue that the two phenomena might involve the same neurocognitive mechanism, as suggested by overlapping neurocognitive correlates. The placebo effect in psychosis is significant, heterogeneous, and forms a significant component of the treatment response; this is likely to decrease the probability of finding new treatments. Therefore, a critical goal to improve outcomes in psychosis is to identify predictors of the placebo effect. In pain and depression, cognitive and neural predictors of placebo effects are themed around belief-updating, in line with experimental data showing that the placebo effect involves these neural mechanisms. Individual predictors of placebo effects have not yet been experimentally investigated in psychosis. The identification of these predictors is important because this information could be used to improve the likelihood of identifying much-needed new treatments. Further, the employment of these precision psychiatry techniques could be used to harness the placebo effect to improve outcomes in the clinic and could reveal insights into the mechanisms underlying psychosis and new targets for treatment.
Acknowledgements
Stefan Leucht honoraria as a consultant and/or advisor and/or for lectures from Alkermes, Angelini, Eisai, Gedeon Richter, Janssen, Lundbeck, Lundbeck Institute, Merck Sharpp and Dome, Otsuka, Recordati, Rovi, Sanofi Aventis, TEVA, Medichem, Mitsubishi. Karin Jensen received funding from Pro Futura at Riksbankens Jubileumsfond, Sweden. All other authors report no biomedical financial interests or potential conflicts of interest.
Early Placebo Improvement Is a Marker for Subsequent Placebo Response in Long-Acting Injectable Antipsychotic Trials for Schizophrenia: Combined Analysis of 4 RCTs.
Are placebo and drug-specific effects additive? Questioning basic assumptions of double-blinded randomized clinical trials and presenting novel study designs.
Sixty years of placebo-controlled antipsychotic drug trials in acute schizophrenia: Systematic review, Bayesian meta-analysis, and meta-regression of efficacy predictors.
Vase L, Robinson ME, Verne GN, Price DD (2003): The contributions of suggestion, desire, and expectation to placebo effects in irritable bowel syndrome patients. An empirical investigation. Pain 105: 17–25.
The role of dopamine dysregulation and evidence for the transdiagnostic nature of elevated dopamine synthesis in psychosis: a positron emission tomography (PET) study comparing schizophrenia, delusional disorder, and other psychotic disorders [no. 11].
00016-2&id=gr1.jpg){kind=link}
00016-2&id=gr2.jpg){kind=link}
00016-2&id=gr3.jpg){kind=link}