Placebo Analgesia

Introduction

For a patient experiencing pain, the perception that an effective treatment has been administered is often sufficient to produce significant analgesia. To the extent that the analgesia is due to the psychobiological effects of the treatment process, as opposed to an active property of the treatment, the person can be said to have experienced a placebo analgesic response. The actual treatment manipulation can take a variety of forms: a dummy tablet, nasal spray, surgical procedure, magnetic treatment, or topical cream.

Whatever the form, the most critical determinants of the analgesic efficacy of a placebo are (1) the presence of sensory cues that have been associated with effective treatment or pain relief in the past and (2) the expectation of pain relief.

The concept of placebo implies that there is a mismatch between what the patient expects and the treatment’s actual intrinsic efficacy. If the patient believes that the placebo treatment may be effective, positive expectations of analgesia are created, and these expectations are linked to pain relief. This, in turn, implies that the patient is deceived. However, it is also possible to create placebo effects through conditioning, the process of learning that sensory cues associated with the treatment context are linked with pain relief. In such cases, placebo effects may be independent of the patients’ conscious belief ( ), and it is thus possible in some cases to obtain placebo effects without explicit deception ( ).

It may also be possible to obtain expectancy (or treatment context–related benefits) in the context of active drug treatment, as is the case with patient-controlled analgesia ( ) and as demonstrated by studies of overt versus hidden drug treatment ( ). In these cases, even though the active treatment cannot be called a placebo in the strict sense, the analgesic response that it elicits in the patient may be said to have a placebo component. Finally, the effects of expectations might, in some cases, interact with the active pharmacological mechanisms of drug treatments to produce synergistic effects ( ). One striking example is a study of the drug proglumide, a cholecystokinin antagonist shown to relieve pain better than placebo alone when given overtly, with full patient awareness, but performs more poorly than placebo when given without patients’ awareness ( ).

In our view, placebo effects can thus arise in several situations—concomitant with conscious expectations about treatment ( ), following conditioning of pain relief with explicit sensory cues (with or without awareness of expectation; ), and associated with the psychosocial context and ritual that surrounds treatment ( ). It is true that according to this definition the conceptual lines dividing placebo effects and effects of psychological therapies are blurred ( ), but this definition respects the common origin of these effects in the brain of the patient. The brain mechanisms of the various psychological influences on pain, as well as whether they arise from common or distinct sources, are an empirical matter.

Although sham treatments can produce a powerful analgesic effect, in a typical clinical situation it is not usually obvious whether the improvement observed in a patient is due to a placebo response. As we describe later, this is often true even when the patient is known to have received a placebo treatment. Failure to appreciate this point has created confusion about what effects placebos do and do not have. Because of this, the first part of this chapter deals at length with definitions and with the phenomena that are most commonly confused with placebo analgesic responses. The second part of the chapter focuses on recent advances in understanding the neurobiology of the placebo response.

Terminology

The term placebo is most likely derived from the Latin stem placebit (“it will please”). Since the beginning of medicine, health professionals have knowingly deceived patients by giving them sham treatments—sometimes with well-meaning intentions and other times for self-serving purposes. However, the longevity of many treatments with probably no active effects ( ) and the success of physicians who prescribed them suggest that patients must have attributed some benefit to these ministrations. Furthermore, perhaps it is possible that the psychobiological effects can confer concrete physiological health benefits and even be an integral part of treatment. For example, cite the classic ancient Chinese text on acupuncture, the 1st-century bce Yellow Emperor’s Inner Classic, as saying, “if a patient does not consent to therapy [acupuncture] with positive engagement, the physician should not proceed as the therapy will not succeed.” The study of placebo treatments, which have been selected to have no direct therapeutic benefit, is the study of the psychobiological effects of the treatment itself.

In this chapter we differentiate between the placebo, the placebo effect, and the placebo response. The placebo itself is a dummy treatment such as sham surgery or a sugar pill. The placebo effect is an observable difference between groups that is attributable to the efficacy of the placebo—for example, the difference in mean treatment effect between a group that has received a placebo treatment and one that received no treatment. The placebo analgesic response refers to the pain relief in an individual that results from the expectation of effectiveness of the therapeutic intervention.

The terminology surrounding placebo research can be confusing inasmuch as some authors use the term “placebo response” to mean any type of improvement in a placebo group in a clinical trial, even if that improvement is related to statistical artifacts such as sampling bias and regression to the mean or to the natural history of a clinical condition. Here, we reserve the term “response” for an active neurobiological process that occurs as a result of placebo treatment. Thus, from the standpoint of understanding mechanisms, it is the placebo response of the individual that is the most interesting and informative object of study.

Active Placebo Responses Versus Statistical Artifacts

The placebo response is widely misunderstood, in part because of modern clinical trial methodology and in part because of lack of understanding of the proximate mediating causes of clinical improvement. In clinical trials, the use of placebo treatment comparison groups is commonplace. The idea is to control for non-specific factors related to administration of the treatment and to the patient’s perception of the treatment. Frequently, in clinical trials of pain and a variety of other disorders, patients in the placebo group improve ( ; ; ). The confusion begins with the assumption that the reason that such patients improve is because they received a placebo. This assumption is often unwarranted.

Patients in the placebo group might improve for several reasons. First, they might show improvement that would have happened with no treatment at all because of the natural history of the disease. Second, patients tend to enroll in trials or treatment when pain is at its worst, thereby resulting in apparent improvement with time as a consequence of regression to the mean. Third, patients may benefit from the positive psychosocial context of being enrolled in a study, which usually means increased medical attention, care, and assessment, as well as, increasingly, additional social support from other patients connected through Internet-based social media (the classic Hawthorne effect is a related phenomenon; ).

To illustrate the problem, consider the common condition of idiopathic headache. In most people the headaches that they experience will arise and subside completely without treatment. Thus any treatment given at the peak of headache severity (or no treatment) will tend to be followed by improvement. This is true whether the treatment is a starch pill or an active analgesic. To assess whether the placebo treatment had any actual psychobiological effects, it is necessary to compare improvement in a placebo treatment group with that in a no-treatment group ( Fig. 27-1 ). This comparison can estimate the magnitude of the placebo effect (i.e., the benefit specifically attributable to taking the placebo), whereas improvement in the placebo group reflects a composite of many factors.

The various types of artifacts that can be mistaken for active placebo responses are described in more detail elsewhere ( ), but it is worthwhile to elaborate briefly on regression to the mean, a pervasive problem illustrated in Figure 27-2 . Imagine patients in a clinical trial for treatment of irritable bowel syndrome (IBS), a condition in which symptoms fluctuate over time but may be stable for a period of years ( ). In this thought experiment, imagine that there is no change in the average symptom severity over time, only fluctuation around a stable value. Patients tend to enroll when symptoms are relatively severe, as marked by the arrows in Figure 27-2 A. Because the symptoms fluctuate around a stable mean value, symptom severity on subsequent measurements will tend to be closer to the mean, and therefore the symptoms will appear to improve over time ( Fig. 27-2 B). Thus, even if there is no true improvement in the population over time, the time of study enrollment is not randomly sampled with respect to symptoms, and there is apparent improvement ( Fig. 27-2 C).

The clinical significance of regression to the mean in chronic pain is illustrated by the work of . They conducted a population- and clinic-based study of people with temporomandibular disorders in which 147 patients who had been referred for treatment of “facial ache or pain in the jaw muscles, the joint in front of the ear or inside the ear (excluding infection)” were compared with 95 community cases identified in a random sample telephone survey of individuals who reported the same complaints but did not seek treatment. All subjects rated their pain severity at study entry and 1 year later. Pain severity at 1 year was much less than at entry for both the treated and untreated groups. The greatest improvement occurred in those with the highest level of pain at study entry, and when the subjects were matched for initial pain severity, no difference in pain levels at the 1-year follow-up was noted between the treated and untreated groups. However, at 4–6 weeks, many patients with temporomandibular disorders in the clinic group reported improvement and attributed their improvement to the treatment received.

Thus, improvement in placebo-treated groups in clinical trials is confounded by both natural healing processes and statistical artifacts such as regression to the mean. In a typical randomized placebo-controlled clinical trial of headache treatment, large numbers of patients in the placebo control group report improvement (e.g., ). Based on such improvement, it is frequently stated that a certain percentage of subjects or patients in a treatment trial are placebo responders. Indeed, Beecher’s oft quoted survey of clinical analgesic trials, from which he concluded that an average of 30% of patients respond to placebo treatments ( ), is based on just such an estimate. In fact, assessing the benefit from taking a placebo requires comparison with a no-placebo group, which controls for natural history, regression to the mean, and other effects of enrolling in the study or manipulation of the type of placebo intervention (e.g., ).

Evidence for Placebo Analgesia