Placebo analgesia: cognitive influences on therapeutic outcome

The therapeutic response to a drug treatment is a mixture of direct pharmacological action and placebo effect. Therefore, harnessing the positive aspects of the placebo effect and reducing the negative ones could potentially benefit the patient. This article is aimed at providing an overview for clinicians of the importance of contextual psychosocial variables in determining treatment response, and the specific focus is on determinants of the placebo response. A better understanding of the physiological, psychological, and social mechanisms of placebo may aid in predicting which contexts have the greatest potential for inducing positive treatment responses. We examine the evidence for the role of psychological traits, including optimism, pessimism, and the effect of patient expectations on therapeutic outcome. We discuss the importance of the patient-practitioner relationship and how this can be used to enhance the placebo effect, and we consider the ethical challenges of using placebos in clinical practice.

better predicted by the patient's individual beliefs and negative expectations regarding a drug's eff ect (nocebo eff ects) than by the specifi c pharmacological properties of the drug [4] itself.
Placebo and nocebo eff ects are therefore centrally impor tant in medical practice, whether these are intentionally used by health-care practitioners or not. A more widespread recognition of this fact may enable and encourage patient-physician interactions that are conducive to positive treatment outcomes. Such interactions are important in determining a number of factors that may aff ect outcome, including patients' mood, their concepts of disease, treatment expectations, and their willingness to endure therapeutic side eff ects. However, this knowledge should be taken in the context of the observation that, in the US, 50% of patients leave after a visit with their doctor without an adequate understanding of what the doctor has told them [5]. Th erefore, there is clearly an opportunity for improv ing therapeutic responses to current treatments and a need for more research into and greater under standing of what factors aff ect those responses.
Until recently, placebo analgesia has been regarded as a nuisance phenomenon in clinical trials. Th is has shifted to the idea that studying placebo eff ects allows us to gain insight into the mechanisms of the endogenous control of pain. Research over the last few decades has improved our understanding of the neuropsychology and neurobiology that underlie placebo and nocebo eff ects. Th ese insights have potentially far-reaching implications for future research and clinical practice. Further understanding of these mechanisms may aid in predicting which contexts (physiological, psychological, and social) have the greatest potential for inducing positive treatment responses. Th is may help to improve the design of clinical trials. It may also help to understand treatment mechanisms by allowing the separation of drug-specifi c responses from non-specifi c (contextually mediated) responses.
Th is article is aimed at providing an overview for clinicians of the importance of contextual variables in determining treatment response, and the specifi c focus is on placebo response. We will examine psychological

Abstract
The therapeutic response to a drug treatment is a mixture of direct pharmacological action and placebo eff ect. Therefore, harnessing the positive aspects of the placebo eff ect and reducing the negative ones could potentially benefi t the patient. This article is aimed at providing an overview for clinicians of the importance of contextual psychosocial variables in determining treatment response, and the specifi c focus is on determinants of the placebo response. A better understanding of the physiological, psychological, and social mechanisms of placebo may aid in predicting which contexts have the greatest potential for inducing positive treatment responses. We examine the evidence for the role of psychological traits, including optimism, pessimism, and the eff ect of patient expectations on therapeutic outcome. We discuss the importance of the patient-practitioner relationship and how this can be used to enhance the placebo eff ect, and we consider the ethical challenges of using placebos in clinical practice.
traits that can lead to variability in the placebo response of the patient, the practitioner's role in amplifying the placebo response, the physiological mechanisms of placebo in relation to patient and practitioner variables, and fi nally how the placebo eff ect can potentially be used for patient benefi t.

Psychological variables that infl uence placebo responses
Placebo responses are highly variable between individuals, having a range of responses from 0% to 100% depending on the context [6]. Responses to a placebo treatment are independent of age and social and physical demographics, but recent evidence suggests that gender may play a part in placebo response rates [7]. However, psychological variables appear to be much better predictors of placebo responses [8].
Research into the psychological context of the placebo response has focused largely on the role of treatment expectations [9]. Treatment expectations, which may be based partly on past experiences with individual physicians and treatments, have a major eff ect on the therapeutic response. If treatment experiences have been negative and frustrating, these may compound over time and shape the expectation for future treatments. Furthermore, accompanying negative mood states, especially in patients with chronic conditions [10], lend themselves to generating negative treatment expecta tions. In these situations, drug effi cacy competes with the negative treatment expectations of the patient. Negative expectations can modulate or, in the worst case, completely abolish positive therapeutic eff ects of drugs. For example, Bingel and colleagues [11] demonstrated the eff ect of positive and negative information about a drug. Th ey found that positive treatment expectations substantially enhanced the analgesic benefi t of remifentanil, a potent μ-agonist. Negative treatment expectation interfered with the analgesic potential of remifentanil to the extent that the eff ect of the potent analgesic was completely abolished.
Th ere is also growing evidence that personality may aff ect the placebo response. Th e main personality traits for which there is evidence of an eff ect are optimism, pessimism, trait anxiety, and neuroticism [12][13][14][15][16][17]. Dispositional optimism and pessimism are habitual styles of expecting good or bad outcomes in life and therefore can be regarded as a dispositional bias in expectation.
Optimists demonstrate an attentional bias for positive information [18] and, even when faced with negative information, will tend to reframe the information in positive ways [19]. Optimism correlates negatively with trait anxiety and neuroticism and positively with reported use of positive coping strategies in general. Scheier and Carver [20] suggest that the general positive expectations associated with optimists lead to persistence and striving toward goals in the face of adversity. Optimism may therefore infl uence the extent to which a patient, given a placebo treatment, persists in the treatment and interprets it positively.
A recent study [17] showed that dispositional optimism predicted the reproducibility of experimental placebo analgesia in an experiment in which experimental placebo analgesia (in response to an inert cream) was tested on two separate occasions. (Importantly, dispositional optimism correlated highly with the change in pain over the two experimental sessions.) It is possible that optimists, having received the 'treatment' once, were more likely to expect a subsequent positive treatment response.
In contrast, pessimists are more likely than optimists to be infl uenced by negative or unpleasant expectations. Th is is because the pessimist dispositional tendency is to believe that negative events are likely to occur, and pessimists are even more negatively reactive when given such information [21]. Th ere is evi dence for a relationship between dispositional pessimism and nocebo responce in a study by Geers and colleagues [14]. A pill was given to healthy volunteers who had been divided into optimists and pessimists to see whether providing negative expectations about the treatment would make them feel worse. Th e authors found that pessimists were more likely to produce a nocebo response.
In addition to expectations and the positive or negative biases in expectation, a psychological (but not necessarily orthogonal) factor that is known to infl uence the response to pain is anxiety. Th e bi-directional relationship between anxiety reduction and analgesia is crucial to the pain response, and anxiety reduction commonly leads to reduced pain experience [22,23]. Some theories of placebo analgesia have included anxiety reduction as a central mediator [24]. However, it is necessary to diff erentiate between state and trait anxiety as both have a distinct impact on the placebo eff ect. State anxiety is an immediate and often transient response to an external stressor. Th is form of anxiety refl ects an acute psychological state of expectation in response to specifi c contextual cues. A study on placebo reproducibility showed that low state anxiety was a moderator of placebo response and a predictor of how an individual would respond on repeat exposure to the same placebo [17]. In this study, the reduction in anxiety that preceded the reduced anticipation of pain in the second placebo session suggests that the reduction in anxiety may be mediating the positive eff ects on expectations. A similar result was obtained by Vase and colleagues [25] in patients with irritable bowel syndrome. Th ey showed a positive correlation between state anxiety reduction and pain tolerance following placebo administration.
In contrast, trait anxiety relies on the susceptibility of an individual's personality to experience anxiety in a way that is less dependent on environmental context and more driven by internal factors [26]. Unlike state anxiety, trait anxiety usually has no correlation with pain tolerance following placebo treatment, possibly because trait anxiety is not as externally manipulable as state anxiety [27,28].

The patient-practitioner relationship
At least a part of the benefi t of some therapies may depend on the non-specifi c verbal and physical interaction -termed a 'healing ritual' -that takes place between a physician and patient. Th is benefi t, which is part of a placebo eff ect, has been considered important in explaining the use of alternative medicines and treatments used in chronic pain for which the physical or physiological mechanisms are not fully understood.
It has been thought that the quality of the physicianpatient relationship is integral to positive outcomes, but until now, data to confi rm such beliefs have been hard to fi nd. Th rough a landmark study, a research team from Th omas Jeff erson University (Philadelphia, PA, USA) has been able to quantify a relationship between physicians' empathy and their patients' positive clinical outcomes, suggesting that a physician's empathy is an important factor associated with clinical success [29].
A further issue is whether the information a patient is given makes a diff erence to the outcome. Pollo and colleagues [30] conducted a study in a clinical setting to investigate the diff erences between the double-blind and a deceptive paradigm. Post-operative patients were treated for pain on request for three consecutive days with a basal infusion of saline solution. Th e symbolic meaning of the saline solution varied in three diff erent groups of patients. Th e fi rst group was given no specifi c information, the second was told that the infusion could be either a potent analgesic or a placebo (double-blind), and the third group was told that it contained a potent pain killer (deception). Th e placebo eff ect of the saline infusion was measured by recording the doses of buprenor phine requested over the three days of treatment. Th e authors found a decrease in buprenorphine intake with double-blind administration and an even greater decrease with the deceptive administration of saline. Th e time course of the post-operative pain was the same in the three groups over the three-day period of treatment. Th is demon strates that the same analgesic eff ect was obtained with diff erent doses of buprenorphine [30], and dose require ment depended on the information provided to patients. Th is study is an exemplar, in a real-world clinical setting, of the eff ect of positive information on the therapeutic eff ect of a drug.

Physiological mechanisms of endogenous pain control which are relevant to placebo
To many clinicians and researchers, the evidence for psychological factors infl uencing the treatment response may seem unconvincing without knowledge of the physio logical mechanisms by which these eff ects can manifest. Here, we review the neural circuitry in the brain for which there is currently the strongest evidence for a mediating role in placebo analgesia.
Evidence has been mounting over the past 30 years for a central role of descending pain modulatory circuits, notably the endogenous opioid system, in mediating placebo analgesic responses. Th e endogenous descending pain modulatory circuit consists of the midbrain periaqueductal grey (PAG), the rostral ventral medulla (RVM), and the spinal cord [31]. Th e PAG integrates input from the limbic forebrain (including the amygdala and pregenual cingulate cortex) and the diencephalon with ascending input from the dorsal horn [30]. Early studies described the PAG-RVM system as a descending inhibitory control that plays a role in endogenous analgesia or in creating suffi cient spinal gain for pain sensory signal detection. It is now clear that the descending control is bi-directional and includes facilitatory mechanisms [32]. Th e fi nal output of this system is determined by the dynamic balance between inhibition and facilitation which can be altered in diff erent behavioral, emotional, and patho logical states. Two of the candidate neuromodulatory systems are the endogenous opioid and serotonin (5-hydroxy tryptamine, or 5-HT) systems.
Until now, much of the evidence for the role of endogenous opioids and 5-HT in the modulation of pain processing has arisen mainly from animal experiments. Th ese studies suggest that the serotonergic system works together with the opioid system to mediate a nociceptive gateway within the central nervous system via a descending network of serotonergic spinal-raphe projections [33,34]. Th e actions of pain transmission are further mediated via serotonergic projections to the spinal dorsal horn from brain regions such as the pregenual cingulate cortex, thalamus, hypothalamus, PAG, RVM, and raphe magnus [33,35,36].
In humans, neuroimaging has provided evidence that the endogenous opioid system is central to mediating placebo eff ects on pain [37][38][39]. Th ese studies provide evidence that placebo analgesia is associated with activation of the endogenous opioid system and with μ-opioid receptors within a number of brain regions, including prefrontal, limbic, and brainstem regions. Furthermore, the changes in activity in these brain regions are related to reductions in the physical and emotional aspects of the pain experience, indicating that variation in endogenous opioid transmission relates to variances in placebo eff ects across individuals. A recent functional magnetic reso nance imaging (fMRI) study has shown activity within all key regions of the descending pain modulatory system − rostral anterior cingulate cortex (rACC), hypothalamus, PAG, and RVM − during placebo analgesia, and this activity was signifi cantly decreased when the μopioid receptor antagonist naloxone was present during the placebo intervention [40]. Coupling between rACC and PAG was signifi cantly increased during placebo analgesia but in the presence of naloxone was not diff erent versus control. Positron emission tomography imaging has also been used to determine the regional activation of endogenous opioid neurotransmission during placebo analgesia [41]. Expectation-induced placebo analgesia was associated with marked activation of μopioid receptor-mediated neurotransmission in an extensive set of brain regions [41], including anterior cingulate cortex. Further more, opioid-related activities in several brain regions within this network correlated with changes in specifi c self-report measures of placebo analgesia, such as pain intensity and unpleasantness, as well as subjects' emotional states.
Despite the above advances in our understanding of the physiological mechanisms of placebo analgesia, there are signifi cant gaps in our knowledge. While there is evidence that a network of brain regions is involved in placebo analgesia (including prefrontal, cingulate, orbitofrontal, limbic, and brainstem regions) and that responses in these brain regions are substantially driven by changes in expectation [42], it is not known precisely where in the brain expectations are generated and maintained or by what mechanism these expectations come to act on the descending pain modulatory circuit, including the endogenous opioid system. Although the prefrontal cortices play an important role in mediating placebo analgesia, there is little evidence outside of the placebo literature for their role in expectation eff ects on pain. One possible explanation is that the prefrontal cortices are not involved in the expression of expectation but rather in their generation through processes of conditioning and learning. fMRI studies have identifi ed a common pre frontal cortical network that is involved in conditioned placebo analgesia [43,44] and that consists of areas that are also important in memory and recall.
Further gaps in our knowledge of the mechanisms of placebo are in relation to the role of neurotransmitter systems other than opioids. Because of its central role in the central modulation of pain processing, the 5-HT neurotransmitter system is one candidate, and there is emerging evidence for a role of this system in placebo response. Carriers of the S allele of the 5-HT transporter (5-HTT) gene have been linked with a range of anxietyrelated personality traits, such as self-reported neuroticism and agreeableness [45][46][47]. Furmark and colleagues [48] found an association between the human genes responsible for driving serotonergic activation of the amygdala during social anxiety and placebo-induced reduc tions in stress. It is currently not known whether this anxiety-related amygdala sensitivity would have an impact on placebo analgesic responses.
Th ere is also interest in further understanding the possible role of the principal stress response pathway, the hypothalamic-pituitary-adrenal (HPA) axis, in placebo response. Th e HPA axis has a close relationship to negative psychological states such as anticipatory anxiety [49]. Although little evidence links the HPA axis with the placebo response, its activation has been linked to the nocebo response. Benedetti and colleagues [50] showed that administration of an inert substance coupled with a negative verbal suggestion upregulates the HPA axis. It is unclear whether placebo and nocebo treatments would have opposing eff ects on the HPA axis. In rodents, endogenous opioids have been shown to exert both inhibitory and facilitatory eff ects on HPA activity. One pathway for this inhibition is an eff ect on cortisol, which is both a product and an inhibitor of the HPA axis. Morphine, for example, was shown to activate adrenocortical release indirectly via the hypothalamus and anterior pituitary and directly via the adrenal gland. In humans, however, endogenous opioids predominantly inhibit HPA axis activity, whereas high doses of the opioid receptor antago nist, naloxone, activate the HPA axis. Th ese fi ndings suggest a possible interaction between the opioid system and HPA axis as part of the placebo response, and opioid release inhibits stress response pathways. Although currently there is lack of direct evidence to support this hypothesis, the association of placebo response and reductions in anxiety makes this an important area for future study.

Enhancing the placebo eff ect in clinical practice
Th ere is scope for enhancing the eff ect of positive informa tion regarding the content of a treatment in everyday clinical practice for the treatment of pain. Th e psychological and physiological benefi ts of this were outlined in the previous sections. Physicians may deliberately give patients treatments they believe to be placebos more often than one might expect, as suggested by several studies worldwide. Forty-fi ve percent of 231 Chicago metropolitan area physicians who were on medical school faculties and who were given a web-based question naire affi rmed that they gave placebos to their patients. Reasons provided by these physicians for giving placebos included complying with a patient's wishes and avoiding confl ict (70%), applying a perceived 'placebo eff ect' (48%), avoiding changing another doctor's prescription (40%), avoiding telling the patient that all treatment options were exhausted (40%), testing whether a condition was 'functional' or 'organic' (25%), and 'other' (9%) [51].
Th ere is, however, controversy about whether giving a placebo as a treatment is ethical. Any ethical assessment to promote placebo eff ects in clinical practice requires knowledge of the clinical advantages of inducing placebo eff ects [2] and how placebo eff ects can be promoted without deception.
One of the main advantages that placebo treatment could off er is a reduction in the need for chemically active medications. Th is would not only benefi t the patient by reducing side eff ects but also be a cost benefi t to health-care systems. A placebo might off er the theoretical advantage of an inexpensive treatment that would not cause adverse drug reactions or interactions with other medications, potentially avoiding complications of polypharmacy. Th e problems with polypharmacy in older patients were recently highlighted by Fox and colleagues [52]. Th ey showed that patients taking a cocktail of drugs with anticholinergic properties had increases in cognitive impairment and mortality. A signifi cant number of the patients were taking selective serotonin reuptake inhibitors for depression, a condition proven to be responsive to placebo treatment [53,54]. Th is is one example of a condition in which placebos might be used to reduce the need for drugs. Although we cannot currently advocate the use of placebos as treatments, there needs to be an exploration of the therapeutic options of using placebos more systematically. For placebos to be used in place of antidepressants, there would need to be fi rm evidence from clinical trials that they are at least as eff ective, carry no risk to the patient, and conform to ethical standards and guidelines in clinical care.
Despite these potential advantages in the use of placebos clinically, it is unclear whether a recommen dation of a treatment intended to promote the placebo eff ect can be made without deception in a way that does not undermine its therapeutic potential. Finniss and colleagues [2] give the example of acupuncture treatment for a patient who has chronic back pain and who has not been helped by standard medicine. Th e authors suggest that the prescribing physician could give the following disclosure: 'Recently, acu puncture has been shown to be no more eff ective than sham acupuncture, but both produce substantially greater symptom improvement in patients with chronic low back pain compared with those patients who receive no treatment or conventional therapy. It is possible that acupuncture works by a psycho logical mechanism that promotes self-healing, known as the placebo eff ect' [2]. A patient who received this disclosure and subsequently got better after undergoing acupunc ture might nonetheless develop a false belief about why it worked. Th is does not mean that the patient has been deceived, nor is there any intention to mislead the patient. Th ere is simply the intention to encourage the patient to initiate their own internal mechanisms of self-healing. Another example of the nondeceptive use of placebo was provided by Kaptchuk and colleagues [3]. Th ey reported an open-labeled placebo study in patients with irritable bowel syndrome. Patients were informed that they would be given a placebo treatment and were told that 'placebo pills have been shown in rigorous clinical testing to produce signifi cant mind-body self-healing' [3]. Patients reported significantly higher global improvement in their irritable bowel syndrome symptoms in comparison with the notreatment control group.
Patient-dependent infl uences may be as important as direct practitioner infl uences on the placebo eff ect in terms of how patients receive information provided to them by a health-care practitioner. As discussed already, psychosocial factors are important determinants of placebo response. A logical step would therefore be to enhance the response to placebos by infl uencing these factors. Th is may be possible by using psychosocial interventions, such as cognitive behavioral therapy, which could help alleviate psychological barriers such as neuroticism and negative expectations. Patients who fi t a particular psychological profi le that predicts poor treatment response would particularly benefi t from this approach. Although psychological treatments have been shown to be eff ective in reducing pain and disability in themselves, it is not known whether the benefi ts of psychological programs are partly a result of enhancing placebo mechanisms. For example, if a psychological treatment can reduce the tendency to generate negative cognitions, it may also improve the ability of the patient to generate positive expectations regarding a physical or pharmacological treatment. It is not known to what extent this determines the success of a psychological treatment and this is an area that requires further study.

Conclusions
Placebo analgesia is a robust psychological and neurophysiological phenomenon that appears to be dependent largely on expectation. However, further research is needed to better understand how the contextual factors that cause placebo response might actually have an impact on specifi c symptoms such as pain. What is clear is that placebo mechanisms can and should be enhanced to maximize the eff ect of currently available therapeutic agents. We assert that it is unethical not to recognize and promote the self-healing opportunities that placebo mecha nisms provide and that it is also unethical to deceive or intentionally mislead patients. We need to resolve this paradox. By providing positive and comprehensible information to the patient, fostering empathic patient-practitioner relationships, or using cognitivebehavioral intervention where necessary, placebo mechanisms could be potentially enhanced without the need for deception. In the future, it may also be possible to act directly on the physiological mechanisms mediating placebo responses by using pharmacological therapies.