Despite the lack of guidance available for practitioners, extensive polypharmacy has become the primary method of treating patients with severe and chronic mood, anxiety, psychotic or behavioral disorders. This ground-breaking new book provides an overview of psychopharmacology knowledge and decision-making strategies, integrating findings from evidence-based trials with real-world clinical presentations. It adopts the approach and mind-set of a clinical investigator and reveals how prescribers can practice 'bespoke psychopharmacology', tailoring care to the individualized needs of patients.
Placebo and Nocebo Effects
The art of medicine consists in amusing the patient while nature cures the disease.
If a person is (a) poorly, (b) receives treatment intended to make him better, and (c) gets better, then no power of reasoning known to medical science can convince him that it may not have been the treatment that restored his health.
▢ Understand the clinical importance of the placebo effect, its magnitude and known clinical and neurobiological (e.g., pharmacogenetic) determinants across major psychiatric disorders, and how it differs from “no treatment”
▢ Differentiate relatively high- versus low- placebo-responsive forms of psychopathology
▢ Recognize the controversy within the literature about rising placebo response rates in clinical trials and how these may influence “failed” rather than “negative” study outcomes
▢ Understand the role of baseline severity as a factor influencing drug–placebo response rates across psychiatric disorders
▢ Describe the nocebo effect and its known clinical determinants
▢ Describe strategies to minimize placebo response rates in psychopharmacology clinical trials
If the placebo effect is not the bane of every psychopharmacologist’s existence, it probably should be. Placebo responses largely negate all rules of pharmacodynamics, undermine theories about drug mechanisms of action, ruin clinical trials by causing failed (rather than negative) findings that mask the true potential for otherwise promising compounds, inflate costs for drug research and development, and generally give a black eye to neuroscience-based explanations for psychopathology. They also lend humility to clinicians’ assumptions that psychopharmacology reliably holds the upper hand when dealing with any and all matters of mental illness. In this chapter we will review known clinical features and correlates (if not actual predictors) of placebo responsivity across major psychiatric conditions, and offer guidance about how clinicians can anticipate, recognize and manage placebo effects – rather than ignore, dismiss, or otherwise struggle against them.
“Placebo” literally means “I shall please” and historically has been a descriptor for any treatment intended mainly to provide psychological benefit rather than physiological efficacy. For the purposes of this discussion we use the term to define a pharmacodynamically inert or psychotropically inactive substance. That said, we would challenge the frequent assertion that a placebo is a substance having no intended therapeutic value; on the contrary, it is precisely because placebo effects account for so annoyingly large a proportion of the variability in psychiatric treatment outcomes that clinicians must understand their therapeutic role. This means recognizing characteristics of a placebo versus pharmacodynamically mediated response, anticipating factors that could predispose to placebo effects, and – for clinicians, but not clinical trialists – capitalizing on their potential contribution to overall outcome whenever and wherever opportunity permits. Strangely enough, investigators strive intentionally to minimize placebo effects in RCTs while clinicians strive unintentionally to maximize them in real-life practice. Particularly in highly treatment-resistant patients, the clinician’s ability to breathe life into an otherwise inert substance – in no small part via the therapeutic alliance – may be the very glue that keeps fragile patients engaged and able to persevere despite the pull toward demoralization and defeat.
“Regression to the mean” means that repeated measurements of a variable within a studied sample will eventually fall back (”regress”) to the mean value of the population from which the sample was derived.
Perhaps the most obvious and important criticism one can levy against the placebo effect is the claim that it does not exist – or, more specifically, that it is no more than an artifact of a phenomenon called regression to the mean. This is the statistical observation that when extreme values on a complex measure of interest (e.g., symptom severity) are observed in a study sample, subsequent or repeated measurements are likely to be less extreme and closer to the true average value simply due to chance variation. Regression to the mean is particularly operative when comparing two variables that are poorly correlated. The risk is especially true in the case of small sample sizes, which increases the chance of Type II (false negative) errors. It is more likely to occur if treatment allocation is not by randomization (so, for example, regression to the mean is of greater concern in observational studies such as the comparison of suicide risk in bipolar patients nonrandomly assigned to lithium or divalproex, as described in Chapter 3, Box 3.1). Risk of regression is greater for disease states that are inherently variable in course, where symptoms are more prone to come and go, or wax and wane (such as rapid cycling bipolar disorder, or panic disorder, or relapsing binge drinking) than is the case for more persistent, nonepisodic, and unvarying conditions (such as persistent/chronic depression, obsessive-compulsive disorder, or generalized anxiety disorder).
A baseball player at the start of the season who gets three hits in his first four at-bats has an impressive 0.750 average. Regression to the mean will nearly always bring that number back down to earth over more at-bat opportunities.
It can be hard to know with certainty if a high placebo response rate truly reflects a beneficial effect from the placebo as opposed to random or natural variation in the course of illness. The only way to make such a distinction is to compare placebo to no treatment. Such a meta-analysis was undertaken by a group from the University of Copenhagen, involving 114 RCTs encompassing 40 different general medical or psychiatric conditions (including depression, anxiety, ADHD, schizophrenia, phobias, and insomnia, among others) (Hrøbjartsson and Gøtzsche, 2001). No significant differences were found between placebo and no treatment in any of the psychiatric disorders studied, and the lack of significant differences was not an artifact of higher dropout rates in one group versus another, or whether investigators were unblinded to study conditions.
The strength of the therapeutic alliance may influence treatment outcomes more strongly during treatment with a placebo than with an active pharmacotherapy.
However – knowing what we know from Chapter 3 – significant heterogeneity was found across trials, and the meta-analysis may not have been adequately powered or sufficiently sensitive to identify small yet meaningful effects from placebo in at least some clinical settings. The meta-analysis also did not account for qualitative aspects of the therapeutic alliance which, as they state, “may be largely independent of any placebo intervention” (p. 1599). Elsewhere, at least in the case of depression, the strength of the therapeutic alliance has been shown to exert a greater mediating effect on outcomes with placebo than with active drug therapy (Zilcha-Mano et al., 2015). Indeed, “placebo effects” for psychiatric disorders, treated within a mental health setting, may very well differ fundamentally from all other clinical situations involving placebo.
In physics, the “observer effect” means that the act of measuring a phenomenon, in itself, unavoidably changes it.
In Chapter 1, we examined potential obstacles to drawing accurate causal inferences in the course of pharmacotherapy. Coupling that with the information on Bayesian analysis (i.e., how iterative experiences cause us to revise our estimates of the probability that an event will happen), we can now make things even more complicated when trying to infer cause and effect if we bring in the concept of regression to the mean. If two variables are poorly correlated (say, a treatment and a change in a symptom severity rating), then poor baseline scores can only improve (move closer to the population mean) over time simply due to random variation. Similarly, if an adverse event (especially a rare or unexpected adverse event) temporally follows exposure either to a drug or a placebo, one must consider the possibility that the occurrence is random with no evidence for causality. When successful outcomes occur with unproven compounds (and, especially, when they are extolled in case reports), the same risk surrounding causal inference is unavoidable. However, unlike an RCT, in real-world practice this methodological dilemma is often irrelevant to the goals of treatment; unlike the clinical investigator, both the practitioner and the patient for better or for worse generally care less about why improvement occurs, so long as it does. This is a fundamental difference in the basic approach of research versus non-research-based treatment delivery.
Particularly in psychiatry, unless the prescriber is an unplugged vending machine, the dispensing of inert substances is never devoid of a meaningful placebo effect. Skilled clinicians knowingly or unknowingly capitalize on what psychotherapy researchers sometimes call the “nonspecific” factors that operate in psychotherapy; they actively listen, reassure, validate the legitimacy of concerns, reframe questions, and (intentionally or not) often incorporate psychotherapeutic techniques such as abreaction, clarification, and suggestion – consequently imposing a sense of order and containment over the distress typically associated with most forms of psychopathology. In psychotherapy research, the existence of a positive therapeutic alliance has been found to have an effect size of 0.26 (Horvath and Symonds, 1991). The act of simply asking structured interview questions about a patient’s problems has been shown to diminish subjective distress, anxiety and depression in more than half of depressed patients (Scarvalone et al., 1996). So impactful is the sheer act of conducting an assessment interview that some study protocols deliberately restrict time durations for conducting assessments and prohibit empathic statements (e.g., “that must have been very difficult for you…”), lest inadvertent supportive psychotherapeutic exchanges contaminate the assessment à la the “observer effect” (just assessing the patient may alter their symptom presentation) plus a smattering of the Hawthorne effect (that is, just knowing that they are being assessed may also alter how patients present themselves).