Abstract The discovery of the endogenous systems of analgesia has produced a large amount of research aimed at investigating their biochemical and neurophysiological mechanisms and their neuroanatomical localization. Nevertheless, the neurobiological acquisitions on these mechanisms have not been paralleled by behavioural correlates in humans—in other words, by the understanding of when and how these endogenous mechanisms of analgesia are activated. Until recent times one of the most studied behavioural correlates of endoge- nous analgesia was stress-induced analgesia, in which the activation of endogenous opioid systems is known to be involved. By contrast, today the placebo analgesic effect represents one of the best-described situations in which this endogenous opioid network is naturally activated in humans. Therefore, not only is placebo research helpful towards improving clin- ical trial design and medical practice, but it also provides us with a better understanding of the endogenous mechanisms of analgesia.
Keywords Placebo · Endogenous opioids · Expectation · Conditioning · Clinical trials
Endogenous Mechanisms of Analgesia
Placebo administration has been found to activate the endogenous mecha- nisms of analgesia, and today the placebo effect represents one of the most interesting models to help us understand when and how these mechanisms are activated (Petrovic et al. 2002; Wager et al. 2004; Colloca and Benedetti
2005). As most of our knowledge about placebo analgesia is based on opioid mechanisms, particular attention has been paid to the endogenous opioid sys-
tems. However, non-opioid mechanisms are also involved in placebo analgesia (Colloca and Benedetti 2005), although nothing is known about them. A re- cent comprehensive review of the endogenous mechanisms of analgesia by
Millan (2002) clearly shows the complexity of this endogenous network, which includes both opioid and non-opioid systems. Thus, the opioid systems repre- sent a ﬁrst step towards understanding the intricate mechanisms underlying
the placebo analgesic effect.
Opioid receptors can be found throughout the brain, brainstem and spinal cord (Pfeiffer et al. 1982; Wamsley et al. 1982; Atweh and Kuhar 1983; Sadzot et al. 1991; Fields and Basbaum 1999; Fields 2004). These receptors may exert analgesic effects through different mechanisms (Jensen 1997), such as mod- ulation at the spinal level and/or control of cortical and brainstem regions. The modulation of the spinal cord is one of the best described (Fields and Basbaum 1999; Fields 2004). The opioid system in the brainstem comprises different regions like the periaqueductal grey (PAG), the parabrachial nuclei (PBN) and the rostral ventromedial medulla (RVM) (Fields and Basbaum 1999; Fields 2004). Although the opioid receptors are less characterized in the cor- tex, autoradiographic studies indicate high concentrations of opioid receptors in the cingulate cortex and prefrontal cortex (Wamsley et al. 1982; Pfeiffer et al. 1982; Sadzot et al. 1991), and one of the highest levels of opioid receptor binding has been found in the anterior cingulate cortex (ACC) (Vogt et al.
1993). Studies performed by means of positron emission tomography (PET) and the radioactive opioid 11C-diprenorphine conﬁrm previous animal and human autoradiography ﬁndings (Jones et al. 1991; Willoch et al. 1999, 2004). In addition, opioid receptor agonists such as remifentanil and fentanyl have
been shown to act on several regions known to be involved in pain processing and containing high concentrations of opioid receptors (Firestone et al. 1996; Adler et al. 1997; Casey et al. 2000; Wagner et al. 2001; Petrovic et al. 2002).
One of the main problems of this endogenous opioid network is to under- stand when and how it is involved in analgesic effects. For example, it has long
been known that fear and stress play an important role in the activation of the endogenous opioid system (Willer and Albe-Fessard 1980; Fanselow 1994). It has also been hypothesized that several other contexts may be important for
the activation of the endogenous opioid systems (Fields and Basbaum 1999; Price 1999). In this regard, today the placebo response is probably the best-
described situation in which this endogenous opioid network is naturally ac- tivated in humans (Colloca and Benedetti 2005). Therefore, placebo research is also important for the understanding of the endogenous mechanisms of analgesia.
The Placebo Effect
Methodology to Reveal Real Placebo Effects
The placebo effect is the effect that follows the administration of an inert medical treatment (the placebo), be it pharmacological or not. In order for a placebo effect to be demonstrated, several other phenomena have to be ruled out (Table 1), as the placebo itself is not always the cause of the effect that is observed (Benedetti and Colloca 2004; Colloca and Benedetti 2005; Pollo and Benedetti 2004). For example, most painful conditions show a spontaneous variation in pain intensity that is known as natural history (Fields and Levine
1984). If a subject takes a placebo just before his pain starts decreasing, he may believe that the placebo is effective although that decrease would have occurred anyway. This is not a placebo effect but a misinterpretation of the
cause–effect relationship. Regression to the mean represents another example. This is a statistical phenomenon in which individuals tend to receive their initial pain assessment when pain is near its greatest intensity, but their pain level is
likely to be lower when they return for a second assessment (Davis 2002). In this case also, the improvement cannot be attributed to any intervention they might have undergone. A further source of confusion is represented by false-positive
errors made by the patient according to signal detection theory (Allan and Siegel 2002). In other words, a patient may erroneously detect a symptomatic relief in response to an inert treatment. The ambiguity of the symptom intensity may also lead to biases following verbal suggestion of beneﬁt. Sometimes
it is a co-intervention, such as the mechanical stimulation produced by the insertion of a needle to inject an inert solution, which is responsible for the reduction of pain (Pollo and Benedetti 2004). All these examples show that,
although an improvement may occur after the administration of a placebo, the placebo is not necessarily the cause of the effect that is observed.
In a clinical trial all these factors are present. Therefore, clinical trials do
not representa good model to study and to understand the mechanisms of the placebo effect. For example, it has been shown that many results obtained in the clinical trial setting (Hrobjartsson and Gotzsche 2001) differ from those obtained in the laboratory setting where strictly controlled conditions are adopted (Vase et al. 2002). In the laboratory setting, for instance, it is possible to comparea no-treatment group witha placebo group so that the spontaneous
remission of the symptom can be identiﬁed. The difference between the no- treatment group and the placebo group represents the real placebo effect. In the laboratory setting other confounding variables can be controlled as well (Benedetti and Colloca 2004; Colloca and Benedetti 2005), like regression to the mean, which can be ruled out by using experimental pain. Likewise, symptom detection ambiguity and subjects’ biases can be eliminated through the measurement of objective physiological parameters. Finally, the effects of a co-intervention can be identiﬁed by using the appropriate experimental model.
The term placebo effect is generally considered to be different from placebo response. The former has been used to refer to any improvement in the con-
dition of a group of subjects that has received a placebo. By contrast, the latter refers to the change in an individual caused by a placebo manipulation. However, these two terms can be used interchangeably.
The Role of Expectation, Motivation and Conditioning
When all the factors described above have been ruled out, the real placebo re- sponse is a psychobiological phenomenon whereby several brain mechanisms can be identiﬁed (Table 1). It is important to realize that the administration of a placebo involves the complex psychosocial context around the patient,
such as the doctor’s words, the hospital environment, the sight of a therapeutic apparatus and the like. Therefore, the study of the placebo effect is basically the study of the psychosocial context that surrounds the patient.
It has long been known that the placebo effect involves both cognitive factors and conditioning mechanisms. On the one hand the deceptive administration of a placebo treatment can lead the subjects to believe that the treatment is effective, such that the anticipation and expectation of analgesia lead to a sig- niﬁcant placebo analgesic effect (Amanzio and Benedetti 1999; Benedetti et al.
1999b; Price et al. 1999; Pollo et al. 2001). Some studies show that different ver- bal instructions lead to different expectations and thus to different responses, and this plays a fundamental role in the placebo effect (Kirsch 1985; Kirsch and Weixel 1988; Kirsch 1990, 1999; Price and Fields 1997; Pollo et al. 2001). This occurs not only with placebos but even with active drugs such as epinephrine (Schachter and Singer 1962), amphetamine and chloral hydrate (Lyerly et al.
1964), and carisoprodol (Flaten et al. 1999). In other words, a typical drug effect can sometimes be reduced if the subjects are misinformed on the action
of the drug and, similarly, new side-effects can be elicited with the appropriate verbal instructions. Motivation and desire for pain relief has also been found to play a role in placebo analgesia (Price and Fields 1997; Vase et al. 2003).
The context around a therapy may act not only through expectation and con-
scious anticipatory processes. There are some lines of evidence indicating that, at least in some circumstances, the placebo response is a conditioned response due to repeated associations between a conditioned stimulus (e.g. shape and colour of pills and tablets) and an unconditioned stimulus (the active substance inside the pills and tablets) (Gleidman et al. 1957; Herrnstein 1962; Siegel 1985,
2002; Ader 1997; Voudouris et al. 1989, 1990; Wickramasekera 2001). In this case, it is the context itself that is the conditioned stimulus. However, even by considering a typical conditioning procedure, it has been shown that a condi- tioned placebo response can result from conditioning but is actually mediated by expectancy (Montgomery and Kirsch 1997). In fact, conditioning would lead to the expectation that a given event will follow another event, and this occurs on the basis of the information that the conditioned stimulus provides about the unconditioned stimulus (Reiss 1980; Rescorla 1988).
A recent study has investigated some physiological functions that are af- fected by placebos through anticipatory conscious processes and some other functions that undergo an unconscious mechanism of conditioning (Benedetti et al. 2003b). For example, verbally induced expectations of either analgesia or hyperalgesia antagonize completely the effects of a conditioning procedure in experimentally induced pain. By contrast, verbally induced expectations of ei- ther an increase or decrease of growth hormone (GH) and cortisol do not have any effect on the secretion of these hormones. However, if a pre-conditioning is performed with sumatriptan, a serotonin agonist that stimulates GH and inhibits cortisol secretion, a signiﬁcant GH increase and cortisol decrease can be found in plasma after placebo administration. These ﬁndings indicate that
verbally induced expectations have no effect on hormonal secretion whereas they affect pain. This suggests that placebo responses are mediated by condi- tioning when unconscious physiological functions, like hormonal secretion, are involved, whereas they are mediated by expectation when conscious phys- iological processes, like pain, come into play. Thus the placebo effect is a phe- nomenon that can be learned either consciously or unconsciously, depending on the system that is involved.
The Neurobiology of the Placebo Effect
The recent explosion of research on the neurobiological mechanisms of the placebo effect is paying dividends, as we are beginning to understand the cascade of events that occur following the administration of a placebo along with verbal suggestions of clinical beneﬁt (Fig. 1). Most of our neurobiologi- cal knowledge comes from the ﬁeld of pain and analgesia, and several lines of evidence suggest that placebo administration activates the endogenous mecha- nisms of analgesia described in Sect. 1. Today we know that placebo analgesia is mediated by both opioid and non-opioid mechanisms, depending on different factors.
An important step in understanding the mechanisms of placebo-induced anal- gesia was made in the clinical setting when Levine et al. (1978) provided evi- dence that placebo analgesia is mediated by endogenous opioids. Other studies subsequently further conﬁrmed this hypothesis (Grevert et al. 1983; Levine and Gordon 1984; Benedetti 1996). In addition, cholecystokinin was found to in- hibit placebo-induced analgesia, as cholecystokinin antagonists are capable of potentiating the placebo analgesic effect (Benedetti et al. 1995, 1996). In fact, cholecystokinin is an anti-opioid peptide which antagonizes endogenous opioid neuropeptides, so that its blockade results in the potentiation of opioid effects (Benedetti 1997).
Fields and Levine (1984) were the ﬁrst to hypothesize that the placebo re- sponse may be subdivided into opioid and non-opioid components. In fact, they suggested that different physical, psychological and environmental situa- tions could affect the endogenous opioid systems differently. This problem was recently addressed by Amanzio and Benedetti (1999), who showed that both expectation and a conditioning procedure could result in placebo analgesia. The former is capable of activating opioid systems whereas the latter activates speciﬁc sub-systems. In fact, if the placebo response is induced by means of strong expectation cues, it can be blocked by the opioid antagonist naloxone.
Fig. 1 Events that might occur following the administration of a placebo along with verbal suggestions of analgesia. A descending pain inhibitory network that involves the rostral anterior cingulate cortex, the orbitofrontal cortex, the periaqueductal grey and the pons/medulla is activated. Endogenous opioids might inhibit pain through this descending network and/or other mechanisms. The endogenous opioids also act
at the level of the respiratory centres. The β-adrenergic sympathetic system is also
inhibited during placebo analgesia, but the mechanism is not known (it might in-
volve either reduction of the pain itself or direct action of endogenous opioids). Non- opioid mechanisms also play a role. Cholecystokinin (CCK) counteracts the effects of the endogenous opioids thus antagonizing placebo analgesia. Placebo can also af- fect serotonin-dependent hormone secretion, mimicking the effect of the analgesic drug sumatriptan. GH, growth hormone; ACTH, adrenocorticotropic hormone; 5HT, serotonin
Conversely, if the placebo response is induced by means of a prior conditioning with a non-opioid drug, it is naloxone-insensitive.
Regional placebo analgesic responses can be obtained in different parts of the body (Montgomery and Kirsch 1996; Price et al. 1999), and these responses are naloxone-reversible (Benedetti et al. 1999b). If four noxious stimuli are applied to the hands and feet and a placebo cream is applied to one hand only, pain is reduced only on the hand where the placebo cream had been applied. This effect is blocked by naloxone, suggesting that the placebo-activated en- dogenous opioids have a somatotopic organization (Benedetti et al. 1999b). An additional study supporting the involvement of endogenous opioids in placebo analgesia was performed by Lipman et al. (1990) in chronic pain patients. These authors found that those patients who responded to placebo showed higher concentrations of endorphins in the cerebrospinal ﬂuid compared with those patients who did not respond.
A likely candidate for the mediation of placebo-induced analgesia is the opioid neuronal network described in Sect. 1 (Fields and Price 1997; Fields
and Basbaum 1999; Price 1999). A recent brain imaging study that used PET supports this hypothesis (Petrovic et al. 2002). These authors found that both a placebo and the rapidly acting opioid agonist remifentanil affect similar regions in the cerebral cortex and in the brainstem, thus indicating a re-
lated mechanism in placebo-induced and opioid-induced analgesia. In par- ticular, the administration of a placebo induced the activation of the rostral anterior cingulate cortex (rACC) and the orbitofrontal cortex (OrbC). More-
over, there was a signiﬁcant co-variation in activity between the rACC and the lower pons/medulla, and a sub-signiﬁcant co-variation between the rACC and the PAG, thus suggesting that the descending rACC/PAG/brainstem pain-
modulating circuit is involved in placebo analgesia. In another study using functional magnetic resonance imaging (fMRI), it was also found that brain activation patterns in the prefrontal cortex changed in anticipation to analgesia following placebo administration (Wager et al. 2004).
Placebo-activated endogenous opioids have also been shown to produce a typical side-effect of opioids, that is, respiratory depression (Benedetti et al.
1998; Benedetti et al. 1999a). After repeated administrations of analgesic doses
of buprenorphine in the post-operative phase, which induces a mild decrease of ventilation, a placebo is capable of mimicking the same respiratory depres- sant response. This respiratory placebo response can be blocked by naloxone, indicating that it is mediated by endogenous opioids. Thus, placebo-activated opioid systems act not only on pain mechanisms but also on the respiratory centres.
The involvement of other systems during placebo analgesia is further sup- ported by recent work in which the sympathetic and parasympathetic control of the heart were analysed during placebo analgesia (Pollo et al. 2003). In the clinical setting, it was found that the placebo analgesic response to a noxious stimulus was accompanied by a reduced heart rate response. In order to in-
vestigate this effect from a pharmacological viewpoint, the same effect was reproduced in the laboratory setting. It was found that the opioid antagonist naloxone completely antagonized both placebo analgesia and the concomitant
reduced heart rate response, whereas the β-blocker propranolol antagonized the placebo heart rate reduction but not placebo analgesia. By contrast, both placebo responses were present during muscarinic blockade with atropine,
indicating no involvement of the parasympathetic system. A spectral analy- sis of the heart rate variability for the identiﬁcation of the sympathetic and parasympathetic components showed that the β-adrenergic low-frequency spectral component was reduced during placebo analgesia, an effect that was
reversed by naloxone. These ﬁndings indicate that opioid-mediated placebo analgesia also affects the cardiovascular system, although we do not know whether the placebo-activated endogenous opioids inhibit the sympathetic
system directly or—through the reduction of pain—indirectly.
Recently, some non-opioid mechanisms have begun to be explored, although nothing is known about non-opioid mediators of placebo analgesia. In fact, the release of neurotransmitters, like dopamine, has been identiﬁed in other illnesses such as Parkinson’s disease. Nonetheless, it is important to emphasize that the integration between the ﬁndings in the ﬁeld of pain and those obtained in other pathological conditions is necessary and essential to better understand the intricate mechanisms of the placebo effect.
As described in Sect. 2.2, by using the analgesic drug sumatriptan, a sero- tonin agonist of the 5-HT1B/1D receptors that stimulates GH and inhibits cor- tisol secretion, it was shown that a conditioning procedure is capable of pro- ducing hormonal placebo responses. In fact, if a placebo is given after repeated administrations of sumatriptan, a placebo GH increase and a placebo cor- tisol decrease can be found (Benedetti et al. 2003b). Interestingly, verbally induced expectations of increase/decrease of GH and cortisol did not have any effect on the secretion of these hormones. Therefore, whereas hormone secretion is not affected by expectations, it is affected by a conditioning pro- cedure. Although we do not know whether these placebo responses are really mediated by serotonin, a pharmacological pre-conditioning appears to affect serotonin-dependent hormone secretion. These new ﬁndings may help inves- tigate non-opioid mechanisms in placebo analgesia.
The verbal instructions that induce expectations may have either a hopeful and trust-inducing meaning, eliciting a placebo effect, or a fearful and stressful meaning, inducing a nocebo effect (Hahn 1985, 1997; Benedetti and Amanzio
1997; Moerman 2002a, b). In a study performed in post-operative patients (Benedetti et al. 1997), negative expectations were induced by injecting an inert substance along with the suggestion that pain was going to increase.
In fact, pain increased and this increase was prevented by the CCK antago- nist proglumide. This indicates that expectation-induced hyperalgesia of these patients was mediated, at least in part, by CCK. The effects of proglumide were not antagonized by naloxone, which suggests that endogenous opioids were not involved. Since CCK plays a role in anxiety and negative expecta- tions themselves are anxiogenic, these results suggest that proglumide acted on a CCK-dependent increase of anxiety during the verbally induced negative expectations. Although this study analysed a nocebo procedure, the involve- ment of CCK in the nocebo effect might help better clarify the modulation of pain by both positive and negative expectations.
The release of endogenous substances following a placebo procedure is a phenomenon which is not conﬁned to the ﬁeld of pain, but it is also present in motor disorders such as Parkinson’s disease. As occurs with pain, in this case patients are given an inert substance (placebo) and are told that it is an anti-Parkinsonian drug that produces an improvement in their motor perfor- mance. Both clinical improvements (Shetty et al. 1999; Goetz et al. 2000; Pollo et a. 2002) and changes in the single-neuron ﬁring pattern (Benedetti et al. 2004) have been found after placebo administration in Parkinson’s patients. A study used PET in order to assess the competition between endogenous dopamine
and 11C-raclopride for D2/D3 receptors, a method that allows identiﬁcation of endogenous dopamine release (de la Fuente-Fernandez et al. 2001). This study
shows that placebo-induced expectation of motor improvement activates en- dogenous dopamine in the striatum of Parkinsonian patients, thus indicating
that endogenous substances other than opioids may be involved in the placebo response in other illnesses. De la Fuente-Fernandez and Stoessl (2002) argue that the expectation-induced release of dopamine in Parkinson’s disease is
related to reward mechanisms. According to these authors, dopamine release by expectation of reward, in this case the expectation of clinical beneﬁt, could represent a common biochemical substrate in many pathological situations, including pain. It is worth noting that there is an important interaction be-
tween dopamine and opioid systems, and that endogenous opioids are also involved in reward mechanisms (de la Fuente-Fernandez and Stoessl 2002).
Very recently, the neural mechanisms of placebo treatments have also been
studied in depression, although these studies need further research and conﬁr- mation because they did not include appropriate control groups. Depressed pa- tients who received a placebo treatment showed both electrical and metabolic changes in the brain. In the ﬁrst case, placebos induced electroencephalo- graphic changes in the prefrontal cortex of patients with major depression, particularly in the right hemisphere (Leuchter et al. 2002). In the second case, changes in brain glucose metabolism were measured by using PET in subjects with unipolar depression. Placebo treatments were associated with metabolic increases in the prefrontal, anterior cingulate, premotor, parietal, posterior insula, and posterior cingulate cortex, and metabolic decreases in the subgen- ual cingulate cortex, para-hippocampus and thalamus (Mayberg et al. 2002).
Interestingly, these regions also were affected by the selective serotonin reup- take inhibitor ﬂuoxetine, a result that suggests a possible role for serotonin in placebo-induced antidepressant effects.
Therefore, although very little is known on the role of non-opioid medi- ators in placebo analgesia, the investigation of the placebo effect in other illnesses demonstrates that other neurotransmitters—like CCK, serotonin and dopamine—may contribute to the modulation of the placebo responses in different circumstances and pathological conditions.