What follows is not a dictionary but an aide-mémoire—a listing of terms of art that appear in more than one chapter, followed by a reminder of their use in context. In the relevant academic disciplines, many of the concepts have definitions that are more precise or that cover a larger frame of reference. My intent here is only to make it unnecessary to flip back through the book to locate the original explanation of a word or expression. In each entry, the first use of a term that has its own entry elsewhere in the glossary appears in italics.
active placebo A dummy pill with side effects, ideally ones that mimic those of an active treatment under test. The tricyclics caused immediate problems such as dry mouth, and some researchers worried that the telltale symptom would allow patients and observers to break the blind and then attach extra hopeful expectancy to the antidepressant. An otherwise inert dummy pill that also caused dryness of mouth would, so the thought went, allow for a fairer experiment.
active treatment/experimental treatment The psychotherapy or medication (believed likely to benefit patients through its inherent efficacy) being tested in a controlled trial, often in contrast to placebo.
additivity/additive A statistical property of a pair of interventions: two interventions are said to be additive if when given together their effect is the sum of the effects of each given individually.
Where additivity applies, the efficacy of one treatment (such as minimal supportive psychotherapy) can be subtracted from the efficacy of a combined treatment (such as an antidepressant plus minimal supportive psychotherapy) to arrive at a difference that accurately reflects the efficacy of the other treatment in the combination (here, the antidepressant). If additivity does not apply, then the subtraction will yield an underestimate of the efficacy of the treatment (again, here, the antidepressant) whose power is being assessed indirectly.
allegiance bias The tendency for an active treatment to show excess efficacy in outcome trials or meta-analyses conducted by adherents of or researchers with a stake in the intervention under test. For example, trials conducted by the inventor of a psychotherapy tend to show stronger favorable results than trials (of the same therapy) conducted by neutral parties.
antidepressant controversy The long-standing debate about whether antidepressants are inherently effective in the treatment of major depression. The contrary view has it that they act mostly through hopeful expectancy and so are little better than “placebos with side effects.” A more restrained version of the skeptical position holds that antidepressants outperform placebo only in severe or very severe major depression. See also active placebo and severity hypothesis.
antidepressants Medications used primarily in and for the treatment of depression. The names of most antidepressants discussed in the text appear in the SSRIs and tricyclics entries below. See also monoamine oxidase inhibitors, reboxetine, Serzone, and trazodone. I refer to some antidepressants by their brand name and others by their generic name, depending on which term is in more general use.
antipsychotics Medicines such as Thorazine (called Largactil in Roland Kuhn’s Switzerland) used primarily in the treatment of illnesses, such as schizophrenia, in which delusions and hallucinations are prominent symptoms. See also psychotic depression.
anxiolytics Medicines (such as Librium, Miltown, Valium, and Xanax) used primarily to treat anxiety.
arm A group of participants in a controlled trial that receive a particular treatment regimen. (Every controlled trial will have a control arm and at least one active treatment arm; some will also have a comparator arm.) The term also covers the protocol applied to each group.
augmenter A secondary treatment (generally one that, at the given dose, would not elicit marked favorable responses on its own) intended to help a primary treatment to work or work better. For example, when a patient does not respond to a full dose of an antidepressant, doctors may add a low dose of a lithium salt—a half to a third of the amount given in lithium’s most common use, stabilizing patients with bipolar disorder—as an augmenter, to jump-start the antidepressant treatment.
baseline score inflation At the start of an outcome trial, the exaggeration (often by raters pushing to meet a quota) of participants’ depression symptom scores, especially those near the minimum set for inclusion in the study. With inflation, a trial that requires a Hamilton score of 16 will include a suspiciously high number of participants with initial scores of 16 and 17. Among them will be patients who in reality have lesser levels of depression and who, in the face of subsequent accurate ratings, will appear to have improved with any intervention, including dummy pills. Baseline score inflation can act as a confound, causing controlled trials or meta-analyses to show a severity gradient when there is none.
Bech scale/HAM-D6 A slimmed-down antidepressant rating instrument, developed by the Danish psychiatrist Per Bech in the 1970s, that employs only six of the seventeen Hamilton scale factors: depressed mood, guilt, work functioning, psychomotor retardation, psychic anxiety, and general bodily symptoms, such as low energy. (Bech calls these six factors “core symptoms” of depression.) The shortened scale was designed to reflect clinicians’ assessments of patients and to give a more coherent picture than the full scale does of the course of major depression.
bias A systematic flaw in an experiment or body of evidence that will tend to push results in a certain direction and so lead to spurious findings. See also allegiance bias, dropout bias, file-drawer bias, and susceptibility bias.
blind/double-blind/break the blind Unaware of the treatment assignment. An outcome trial is double-blind if neither the patient nor the treating doctors and raters know who is in which arm. If staff or patients can tell who is taking what (drug or placebo), they have “broken the blind.” Blinding is a protection against bias—in particular against a tendency to report favorable results for subjects known to be receiving the active treatment.
candidate-drug trial An outcome trial testing a new medication, the “candidate,” being proposed for approval by the Food and Drug Administration.
categorical analysis/category/categorical A research summary that counts people (in an outcome trial’s treatment and control arms) who achieve a defined level of improvement. Response, remission, and recovery are categories. See also number needed to treat. (I do not use the expression in the text, but comparisons built around average results in groups—effect size is an example—are called numerical analyses.)
CBT/cognitive behavioral therapy A psychotherapy focused on correcting patients’ maladaptive cognitions (“Because I failed at this task, I am worthless”), often through direct didactic means.
classic placebo effect Improvement of illness, in response to an inert intervention, that arises from the impact of pill taking, doctorly authority, the clinical setting, and so on. The term also covers similarly based progress occurring in patients receiving potentially effective treatments. For well over a century, the term hopeful expectancy has been used to describe the psychological state that mediates the classic placebo effect.
Cochrane Collaboration An international professional organization dedicated to evidence-based medicine that, among its other tasks, sponsors, approves, and publishes overviews of treatment outcome research. The group is named after Archie Cochrane, a Scottish physician and public health researcher who, in the aftermath of World War II, advanced the use of randomized trials in the evaluation of medical treatments.
comparator A treatment of established efficacy used, in its own arm of an outcome trial, to “validate the sample” of patients. The idea is that if the group of participants is typical of those with the target illness, those in the comparator arm should improve to the expected degree, a result that will indicate that the remainder of the experiment (say, candidate drug versus dummy pill) is trustworthy. A trial in which the comparator does not outperform placebo is said to be “failed,” meaning that the rest of the experiment cannot be expected to yield valid results.
completer analysis A statistical summary (of results of an outcome trial) that ignores patients who drop out and bases calculations on scores from participants who show up for rating sessions through the end of the experiment. If patients in a placebo arm, more than those in a drug arm, quit for “inefficacy”—that is, failure to improve—completer analyses will overstate the benefits of dummy pills. One alternative is an intention-to-treat analysis, which includes results from all patients who attend the first rating session of a trial. See also differential dropout.
confound An extraneous factor in an experiment that is liable to distort the results. Most commonly, in our discussions, we encounter confounds such as differential dropout that leave patients in one arm of a trial relatively susceptible to ongoing illness while patients in the other arm remain, on average, more resilient—after which point the trial will suffer from susceptibility bias. Confounds can lead to spurious conclusions.
Confound, as a noun, is a term of art in statistics; in recent years, some authors have begun replacing it with the less felicitous (to my ear) noun confounder.
control arm See control condition.
control condition An intervention (such as administration of a placebo or comparator) included in an outcome trial for purposes of comparison. When the control condition involves a placebo, the control arm of a trial is meant to duplicate the circumstances in the treatment arm—the same passage of time, the same contact with raters, and so on—in the absence of the inherently effective element of the intervention under test. In this case, the control arm results represent the hypothetical counterfactual, what would have happened to patients in the treatment arm had they not received the benefits (and side effects) of the active treatment, via its inherent efficacy.
controlled trial A trial in which an active treatment is compared to a control condition.
core symptoms of depression See Bech scale.
curse of Roland Kuhn/Kuhn’s curse My shorthand for the historical tendency, which Kuhn recognized, for patient samples in drug trials to become progressively less representative of depressed people in general as (with effective antidepressants widely available) run-of-the-mill cases are treated in doctors’ offices.
demand characteristic A confound arising from the tendency of patients to offer symptom reports that reflect what they believe their caregivers prefer or implicitly “demand.” The “hello effect” (exaggerating the level of suffering early on) and “goodbye effect” (in the late going, reporting more progress than actually occurred) are demand characteristics.
differential dropout A confound that can arise when patients exit different arms (of an outcome trial) for different reasons. If, in the extreme case, patients drop out of the placebo arm because they are floundering while patients leave the treatment arm when they feel all better, then the placebo arm will be left with naturally resilient patients while the treatment arm will be left with partly responsive (that is, somewhat treatment-resistant) patients. Thereafter, the trial will suffer from susceptibility bias. See also dropout bias.
differential sieve A cause of confounds that arise in a controlled trial when the interventions in the contrasting arms leave them with patient populations whose risk of further illness differs. For instance, over time, treatment with dummy pills may, through its lack of efficacy, act as a narrow-gauge sieve, causing vulnerable patients to exit a trial so that only naturally resilient participants pass through and remain available for further study. In the same interval, antidepressant treatment may act as a broad-gauge sieve, sustaining patients with chronic and recurrent depression and allowing them to pass through to the next phase of the research. From that point on, the experiment will suffer from susceptibility bias.
double-blind See blind.
dropout bias Susceptibility bias arising from differential dropout. If differential dropout leaves patients in one arm of an outcome trial more liable to continued illness than patients in another arm, thereafter the trial results are likely to be distorted.
dysthymia Chronic low-level depression. The official definition in use for much of the last half century required depressed mood on most days for at least two years with at least two accompanying depressive symptoms, such as feelings of hopelessness. Other requirements concerned the relationship to major depression (episodes in the distant past and episodes occurring late in the course of dysthymia were permitted) and intervals (brief ones were allowed) without depressive symptoms. Some of the meta-analyses we have encountered used looser criteria.
effect size A measure, developed by the statistician Jacob Cohen, of the average advantage of an active treatment over a placebo condition. Cohen provided a guide whose overuse he then warned against: an effect size of 0.2 is small; 0.5 is medium; and 0.8 is large. An effect size of 0.5 says that the average treated subject will do better than 69 percent of untreated subjects. The effect size goes beyond statistical significance to attempt to answer our How much? question—how well a treatment works. Like the number needed to treat, effect size provides a means of comparing the efficacy of treatments in disparate medical specialties.
Ekselius Repeatedly I reference a study by the Swedish researcher Lisa Ekselius that showed high response rates to SSRIs in depressed primary-care patients—substantial improvement in over three-quarters of those entering the trial and 90 percent of those who stuck with it. The point is that in ordinary use (when the curse of Roland Kuhn is not in effect) the medicines work well.
elaboration In the theory of Philip Cowen, the social and psychological processes through which, over time, a person builds on the immediate change in feeling and perception provided by antidepressants to achieve substantial relief from mood disorder.
evidence-based medicine/EBM A movement announced in the 1990s that aims to ground clinical decision making less in doctorly experience and more in objective research results. The extreme or narrow version of EBM demands near-exclusive reliance on the findings of randomized, controlled, double-blind outcome trials.
experimental treatment See active treatment.
file-drawer bias/publication bias The tendency for a scientific literature to show exaggerated efficacy for treatments as a result of the preference (on the part of researchers and editors) for publishing favorable outcomes and shelving unfavorable ones. Erick Turner’s analysis of data from the FDA files (largely involving candidate-drug trials) demonstrated that, compared to the published research, the entire collection of evidence showed lower antidepressant efficacy. The published literature had suffered from file-drawer bias.
grab bag (of placebo effects) My term for the jumble of influences (including baseline score inflation, classic placebo effects, demand characteristics, minimal supportive psychotherapy, and natural healing that occurs with the passage of time) that act in the control arms of antidepressant outcome trials.
Hamilton scale/Hamilton score A rating instrument developed by Max Hamilton in the late 1950s and used ever since in outcome trials of depression treatment. The scale recognizes seventeen symptoms of depression, called factors: depressed mood, guilt, suicidality, early (in the night) insomnia, middle insomnia, late insomnia, inhibition of work and other interests, psychomotor retardation, agitation, psychic anxiety, somatic anxiety, gastrointestinal symptoms, general bodily symptoms, genital symptoms, hypochondriasis, loss of insight, and weight loss.
Each factor receives a weighting—from 1 to (for some factors) as many as 4 points—and the sum is the Hamilton score. A score of 50 points is possible, but totals above the low 30s are rare. In simple terms, a Hamilton of 30 represents severe depression; 20, moderate; and 10, mild. Although they may still consider themselves depressed, in most formal research people with Hamilton scores below 7 or 8 are judged to be in remission.
hopeful expectancy See classic placebo effect.
How much? question My term for the debate over (or our general interest in) the level of efficacy of antidepressants. Almost all overviews show that antidepressants work. The issue is, how well?
hypothetical counterfactual See control condition.
inherent efficacy/inherently effective Inherent efficacy refers both to the contribution that an intervention makes through its special properties and to the basis for its making that contribution. With streptomycin, its inherent efficacy arises from its power to inhibit the growth of bacteria. With antidepressants, their ability to alter neurotransmission in the brain is the presumed source of their inherent efficacy. “Pharmacologic potency” is an earlier term for roughly the same concept. The implicit contrast is to the grab bag of placebo effects: drugs’ efficacy based on the fact of pill taking, the passage of time, and so on.
instruction effect In an outcome trial, the influence on results exerted by what the tester says. For example, telling experimental subjects given caffeinated coffee that they are drinking decaf will result in decreased reports of tension. See also lessebo effect.
IPT/interpersonal psychotherapy A brief, manualized psychotherapy for depression, developed in the 1960s by Gerald Klerman and others, intended as a stand-in (in outcome trials) for psychoanalysis and related psychotherapies.
lessebo effect An instruction effect whereby patients given active medication and told that they may be on placebo show lower drug efficacy than patients told, correctly, that they are taking a real drug. The lessebo effect may arise in part from patients’ failure, once they have been made uncertain about their treatment, to engage in elaboration of the initial improvement they experience.
maintenance The extended use of medication over time. A typical maintenance trial begins with research subjects who have done well—say, on an antidepressant—in a short-term outcome trial. In the maintenance phase, these patients will be randomly assigned to take a dummy pill or to continue taking the antidepressant, and the time to relapse or recurrence will be measured. Other designs are possible.
major depression For most of the last half century, the official definition of major depression required impairment of functioning based on at least a two-week interval of depressed mood or inability to experience pleasure, accompanied by symptoms (for the diagnosis, five symptoms were required in all) from a group that included low energy, guilt, impaired concentration, suicidality, psychomotor retardation or agitation, disturbed or excessive sleep, and change in appetite. Before the formulation of this definition (and oftentimes, later, in parallel with its use), clinicians diagnosed depression experientially, based on their empathic assessment of a patient’s level of despondency and their past observations of similar presentations of mood disorder. See also mild major depression and minor depression.
meta-analysis A study that uses statistical methods to amalgamate the results of many small outcome trials. Gene Glass developed the approach, incorporating Jacob Cohen’s concept effect size.
mild major depression Major depression—entailing substantial impairment and at least five depressive symptoms—but with a modest score, perhaps in the 8-to-15 range, on the Hamilton scale. Mild major depression is distinct from minor depression, which shows scores in a similar range but involves fewer symptoms.
minimal supportive psychotherapy The aspect of participation in an outcome trial—based on frequent contacts with interested, helpful staff members (and perhaps fellow participants as well)—that acts like a supportive psychotherapy. (Supportive psychotherapy works via encouragement more than insight.) The term was coined by the designers of the NIMH collaborative study, who required psychiatrists monitoring the placebo and medication arms of the trial to provide “advice and encouragement” to prevent suicides or substantial worsening in research subjects.
minor depression A mood disorder involving fewer symptoms than the number required for major depression. In the example we consider, the “Dartmouth study,” the researchers required three or four depressive symptoms and a Hamilton scale score greater than 10. The symptoms had to include depressed mood or the inability to experience pleasure. Minor depression is not mild major depression; in the Dartmouth study, patients with a fifth symptom were excluded. So were patients with dysthymia—which means that (in this instance) minor depression was nonchronic low-level depression.
monoamine oxidase inhibitors/MAOIs Medications that affect the way that the brain handles a broad class of neurotransmitters, the monoamines: dopamine, norepinephrine, and serotonin. MAOIs were the first modern antidepressants. Because they are difficult to use—they have life-threatening side effects—the MAOIs became less popular with time, in the United States especially.
neuroticism Hans Eysenck’s term for a personality trait characterized by negative thinking, uncomfortable self-consciousness, and emotional vulnerability and instability.
neurotransmitter A chemical that carries signals from one nerve cell to another. Serotonin, norepinephrine, dopamine, and glutamate are neurotransmitters.
NICE/National Institute for Clinical Excellence A semiautonomous British governmental agency, founded in 1999, tasked with improving care in the National Health Service, largely through a focus on treatment standards. In 2005, the agency (still known as NICE) changed its name to the National Institute for Health and Clinical Excellence.
NIMH collaborative study/TDCRP Supervised by the National Institute of Mental Health in the 1980s, the Treatment of Depression Collaborative Research Program was the first large-scale clinical trial of psychotherapy. The study used imipramine as a comparator.
nuisance variables/noise In an outcome trial, circumstantial factors that can impede the effort to measure the intrinsic efficacy of a treatment. For example, an especially upbeat set of raters might pump up response rates in all arms of a study and so serve as “noise” or a nuisance factor. A major function of the control arm is to track nuisance variables.
number needed to treat/NNT A means of summarizing controlled trial results as measured by categorical outcomes such as response and remission. The NNT estimates the number of patients who would need to take an active treatment to achieve one more favorable result than would occur with placebo.
Consider the case in which four patients are offered an antidepressant. Two respond—but one would have responded on dummy pills. The doctor needed to treat four patients to achieve one additional good outcome. Equally, if all four patients respond to medication but three respond to dummy pills, the number needed to treat would be 4.
Reporting results via the NNT is a hallmark of evidence-based medicine. Because its use is widespread, the measure allows comparisons of the efficacy of treatments in different medical specialties.
outcome trial/outcome study An experiment testing a clinical intervention, such as antidepressants or psychotherapy, to see how well it works in people.
placebo In an outcome trial, the element in a control condition meant to resemble the active treatment but lacking its inherent efficacy. In antidepressant trials, generally the placebo will take the form of a tablet or capsule otherwise identical to the one delivering the drug but containing only the fillers used to stabilize or add bulk to the drug’s active ingredient. There are many variations. In the TDCRP, a dummy pill supplemented by minimal supportive psychotherapy served as a placebo for the two psychotherapies being evaluated, CBT and IPT.
placebo arm In a controlled trial, an arm in which patients receive placebo as the primary intervention.
placebo effect See classic placebo effect and grab bag (of placebo effects).
psychomotor retardation A symptom of depression. Psychomotor retardation includes the slowing of speech, thought, and muscular motion.
psychosomatics The study of the interface between mental illness and medical illness generally treated by nonpsychiatrists.
psychotic depression An extreme form of mood disorder characterized by depression accompanied by marked delusions—that is, fixed and highly unrealistic beliefs, and/or hallucinations. Psychotic depression generally does not respond to antidepressants alone but requires combined treatment that includes antipsychotics.
publication bias See file-drawer bias.
quality-of-life research Studies whose outcome measures track work and social functioning along with overall health.
randomized controlled trial See randomized trial and controlled trial.
randomized trial An outcome trial in which participants are assigned to the different arms randomly. In our image of randomizing, patients approach a triage nurse who, without regard to any of their personal characteristics, such as age or health, hands them a sealed envelope consigning them to an arm. Randomization is meant to protect trials from susceptibility bias that can arise from other methods of allocating participants to treatment arms.
R&R My mnemonic, because of its focus on remission and recovery, for a long-term, multisite outcome trial (conducted by researchers at the University of Pennsylvania and elsewhere) testing the benefits of vigorous prescribing, with or without cognitive therapy, for depression.
reboxetine An antidepressant, not marketed in the United States, that is thought to work via its effect on the brain’s use of the neurotransmitter norepinephrine. In experiments, often responses to reboxetine are contrasted to responses to SSRIs.
recovery See remission.
recurrence/relapse The return of depression in a patient previously in remission from a prior episode. I use the terms indiscriminately, as do some authors I cite. (Other authorities reserve relapse for a quick reemergence of depression and contrast it to recurrence, appearance of a new episode after a substantial interval of good health. An associated argument—often marred by analyses that fall afoul of the differential sieve—has it that medication is better at preventing the former than the latter.) Definitions of relapse and recurrence vary from study to study, a problem I allude to in passing in the discussion of tachyphylaxis.
remission Emergence from depression, and then the subsequent interval of freedom from depression. When patients’ Hamilton scores fall below a certain number, generally 8, their depression is said to have remitted. (Some trials require continued low scores for a period of weeks.) Early studies used the term recovery to comparable effect, but the mental health professions later embraced remission, in acknowledgment of the likelihood of recurrence. In the R&R research, the term recovery was reserved for remissions lasting six months.
rescue The prompt administration of antidepressants as an episode of depression emerges. In circumstances that might otherwise call for the prophylactic use of antidepressants, clinicians may, as an alternative strategy, follow patients closely and initiate rescue when the need arises.
response A favorable reaction to treatment. From the earliest analyses, response has been defined as a 50 percent decrease in the Hamilton score.
Serzone An antidepressant believed to work via its effect on serotonin transmission in the brain but that because of its particular mechanism of action does not qualify as an SSRI. Serzone is related to trazodone, an earlier medicine used to treat depression, anxiety, and insomnia.
severity gradient A pattern of outcomes in which a treatment has greater efficacy for severe than for mild illness.
severity hypothesis The theory that antidepressants work best for severe or very severe major depression and may have no superiority over placebo in the treatment of mild (or mild-to-moderate) major depression. See also antidepressant controversy and severity gradient.
SSRIs Antidepressants (such as Prozac, Zoloft, Paxil, Celexa, and Lexapro) thought to work via their influence on the brain’s handling of the neurotransmitter serotonin. The SSRIs are “second-generation” antidepressants, ones with a better-tolerated set of side effects than the tricyclics, which they have largely replaced.
STAR*D A multisite outcome trial (Sequenced Treatment Alternatives to Relieve Depression), conducted in the first decade of this century, meant to test treatments, principally antidepressants, on a broad range of depressed patients. Most participants had chronic depression complicated by other ailments. The intent was to see how patients fare when treated as they might be in clinical practice.
statistical significance, significant A statistical standard meant to indicate whether a research result—a correlation or a treatment outcome—is real, that is, unlikely to be due to chance alone. (The usual criterion for the label significant requires a less than one-in-twenty chance of a random association.) Findings can be statistically significant, that is, not due to chance, without being clinically significant, that is, important to patients. Statistical significance says nothing about the How much? question.
stimulants Medicines, such as amphetamines or Ritalin, now used principally in the treatment of attention deficits and hyperactivity—but which on occasion prove of use in the treatment of major depression.
susceptibility bias A distortion (due to a failure of randomization or an equivalent confound such as differential dropout) that arises when participants in one arm of an outcome trial are more susceptible to further illness than participants in another arm. The arm with the less vulnerable patients may show favorable outcomes even when the treatment it tests is ineffectual.
tachyphylaxis A catastrophic failure of efficacy arising during treatment with a medicine that had previously been stabilizing. Lately, the definition has been broadened to include, in the most liberal usage, any worsening of depressive symptoms (during a maintenance phase of treatment) that requires an adjustment in a drug regimen.
TDCRP See NIMH collaborative study/TDCRP.
trajectory A distinctive, coherent path of response to treatment, generally one shared by a cluster of patients. For example, depressed patients who enjoy an early, abrupt improvement and then stay well or improve further in each subsequent observation would be said to be on a particular (favorable) trajectory.
trazodone A sedating drug with antidepressant and anxiolytic properties. See Serzone.
treatment arm In a controlled trial, the arm in which patients receive the experimental or active treatment.
tricyclics The class of antidepressants (named for their characteristic three-ring chemical structure) that includes imipramine, Elavil, and desipramine. Generally, tricyclics are thought to work via their effects on the brain’s use of the neurotransmitter norepinephrine, with varying additional effects on serotonin pathways. The tricyclics were among the first modern antidepressants, dating back to Roland Kuhn’s observations of imipramine in the 1950s. Effexor and (to a lesser degree) Cymbalta are recent antidepressants thought to resemble tricyclics in their mechanism of action.