Psychology by BogdanovaThe Encyclopedia of Psychology for the Artificial Era

RESEARCH DOSSIER

Psychotropic Tolerance, Withdrawal, and Discontinuation

Abstract. Psychotropic tolerance, withdrawal, and discontinuation are longitudinal processes spanning pharmacological adaptation, the consequences of changing exposure, and the course of underlying psychiatric disorders. This fundamental research monograph synthesizes foundational experiments, randomized discontinuation and maintenance trials, systematic reviews, regulatory warnings, and clinical practice guidance. It distinguishes tolerance, physiological dependence, addictive behavior, withdrawal, rebound, and illness recurrence; analyzes pharmacokinetic and pharmacodynamic hypotheses, the limits of receptor-occupancy models, and psychological contributors to symptom experience; and compares antidepressants, benzodiazepines, antipsychotics, lithium, and other classes. Evidence from ANTLER, RADAR, class-specific withdrawal research, the 2024 antidepressant discontinuation meta-analysis, and the January 2026 network comparison of deprescribing strategies is evaluated for causal interpretation, measurement validity, safety, and generalizability. The synthesis identifies serious risks associated with abrupt benzodiazepine cessation, clinically consequential differences among maintenance and discontinuation strategies, and unresolved questions about long-term outcomes, tapering pace, and differentiation of withdrawal from relapse. This is a critical narrative evidence synthesis, without new clinical data, registered systematic-review protocol, or independently pooled statistical estimates.

Keywords: psychopharmacology; medication tolerance; physical dependence; antidepressant discontinuation syndrome; benzodiazepine withdrawal; antipsychotic tapering; lithium withdrawal; relapse prevention; pharmacodynamics; pharmacokinetics; psychiatric medication; psychotropic deprescribing; rebound; clinical measurement; long-term outcomes; shared decision-making.

Introduction: The Scientific Problem of Stopping Psychotropic Drugs

Psychotropic tolerance, withdrawal, and discontinuation form a single longitudinal research problem: the effects of a drug evolve during exposure, and the physiological and psychological consequences of reducing exposure depend on the history of that exposure. A medication may improve symptoms during treatment, produce adaptations to repeated administration, yield new symptoms when reduced, and still provide protection against recurrence of the original illness. Any comprehensive account must hold all four propositions together. This dossier examines the temporal transition from sustained pharmacological treatment to lower exposure or cessation, drawing on psychopharmacology, learning theory, clinical epidemiology, patient-reported outcomes, and causal inference. The relevant endpoint is not merely whether a person swallows a final tablet; it is the evolution of functioning, symptoms, adverse effects, preferences, and safety after a change in treatment.

The central questions are empirical. Under what conditions does a repeated medication exposure lead to a diminishing effect, and which effects remain stable? What constitutes a withdrawal syndrome, and how reliably can it be distinguished from relapse, rebound, adverse effects, or a new medical condition? Which strategies for discontinuation reduce clinical risk, and which claims arise mainly from pharmacological modeling or expert consensus? Finally, how do trial design and measurement determine apparent rates of relapse and withdrawal? A cross-class synthesis published by Vinkers and colleagues (2024) Source 16 identifies these as linked decisions across antidepressants, antipsychotics, benzodiazepines, mood stabilizers, opioids, and stimulants. The present monograph addresses the psychological and psychiatric research problem, with the most detailed evidence devoted to prescribed antidepressants, benzodiazepines, antipsychotics, and lithium.

This work is a critical narrative research synthesis based on identifiable original trials, published systematic reviews, regulatory documents, and practice guidelines. It does not pool primary datasets, register a systematic-review protocol, estimate new incidence rates, or test a novel intervention. Clinical guidance is analyzed as a research object and interpreted in its evidentiary context. Medication changes require evaluation of diagnosis, current state, exposure, comorbidity, co-prescribed substances, and individual preferences; published generalizations cannot supply those facts for an individual. The aim is an explanatory and methodological reference suitable for scholars, health professionals, and readers who need a rigorous conceptual map.

Terminology and Operational Definitions

Tolerance is a reduction in a specified response to a specified exposure after repeated administration. Its operational meaning requires a measured endpoint, dose, interval, and context: tolerance to hypnotic sedation, for example, need not imply comparable tolerance to every anxiety-related, anticonvulsant, cognitive, or respiratory effect of the same compound. Pharmacodynamic tolerance refers to changes in response to a given concentration at a target or within a downstream circuit; pharmacokinetic tolerance concerns changes in drug disposition that alter exposure at a given dose. Behavioral tolerance, observed in particular settings, may arise through learning and compensatory responding. Sensitization describes an enhanced rather than diminished effect with repeated exposure. A global statement that someone is tolerant to a drug conceals this plurality.

Physical dependence denotes adaptation to sustained exposure such that reduction or cessation can generate a characteristic withdrawal state. Its presence does not establish compulsive consumption, craving, or a substance use disorder. The distinction is explicit in the 2025 Joint Clinical Practice Guideline on Benzodiazepine Tapering Source 2, which describes physical dependence as an expected possible outcome of regular therapeutic exposure and distinguishes it from benzodiazepine use disorder. Addiction-related diagnoses involve additional behavioral criteria, impaired control, continuing despite harm, and functional consequences; tolerance and withdrawal alone are insufficient to infer them in a person taking medications under appropriate medical supervision. Scientific and public discussion should preserve the conceptual boundary.

Discontinuation refers to stopping a drug, while dose reduction and tapering describe pathways to lower doses that may or may not culminate in cessation. Discontinuation symptoms are new or intensified experiences temporally associated with reducing exposure, although temporal succession alone does not prove pharmacological causation. Withdrawal syndrome denotes a recognizable cluster plausibly caused by adaptation to prior exposure and loss of that exposure. Rebound is an acute worsening beyond the preceding symptomatic baseline, sometimes linked to drug withdrawal. Relapse is return of an illness episode before recovery is secure; recurrence is a subsequent episode following recovery. These distinctions require explicit observation windows, baseline assessment, and definitions rather than interchangeable everyday labels.

Historical Development of the Discontinuation Problem

The history of psychotropic discontinuation runs alongside the development of modern psychiatric pharmacotherapy. Lithium maintenance transformed the long-term management of bipolar disorders; tranquilizers, antidepressants, and antipsychotics subsequently expanded both the range of treatment and the duration of exposure. Reports of withdrawal and symptom recurrence accumulated through pharmacovigilance and follow-up observation before contemporary discontinuation trial designs became routine. Earlier debates frequently focused on the apparent return of illness after drug cessation, whereas later research asked whether the act and speed of discontinuation might themselves modify observed outcomes.

A landmark experimental contribution came from Rosenbaum, Fava, Hoog, Ascroft, and Krebs (1998) Source 13, whose randomized clinical trial investigated emergent symptoms during selective serotonin reuptake inhibitor discontinuation. Its historical significance rests on studying withdrawal-like outcomes under controlled conditions rather than treating all post-treatment symptoms as ordinary relapse. Drug-specific differences helped establish the need to consider half-life, active metabolites, prior exposure, symptom timing, and systematic elicitation of adverse events. Controlled interruption paradigms, however, may not reproduce the consequences of discontinuation after years of therapy or the experiences of people who have previously struggled with tapering.

The early lithium discontinuation literature established another dimension: clinical course may depend on the manner of stopping treatment. Suppes, Baldessarini, Faedda, and Tohen (1991) Source 14 summarized recurrence after stable lithium maintenance, while a prospective observational study by Faedda and colleagues (1993) Source 3 compared outcomes after faster and more gradual stopping. The latter followed 64 outpatients and reported greater early recurrence following rapid discontinuation. The original comparison was not an independently randomized test of taper speed, so it is evidence for a clinically important association with residual confounding possible. Its enduring contribution is to make the speed of withdrawal an explicit exposure of interest.

The Time Course of Drug Exposure

A prescribed dose is an input to a pharmacokinetic system, not a direct measure of brain receptor activity. Absorption, distribution, metabolism, active metabolites, elimination, protein binding, individual physiology, and drug interactions determine concentration over time. The effective decline after the last dose can differ substantially between compounds. A short elimination half-life may permit concentrations to fall quickly and provoke symptoms earlier; a long-lived active metabolite may delay the onset of symptoms and make a seemingly successful short observation window misleading. Vinkers et al. (2024) Source 16 emphasize marked differences among medication classes and individual drugs.

Repeated administration generates a temporal history that a concentration measured on one day cannot fully represent. Receptor regulation, transmitter release, feedback loops, gene expression, neural network dynamics, endocrine modulation, circadian rhythms, and learned expectations all can change during ongoing treatment. Evidence for one adaptive mechanism may be strong in an animal preparation and weak in direct human clinical tests. A mechanistic hypothesis becomes a clinically established withdrawal explanation only when it links exposure history, measured biological change, symptom pattern, and intervention response in an appropriate design. The absence of readily measured biological markers should not be mistaken for absence of withdrawal, just as plausible receptor adaptation should not be treated as a complete explanation for each reported symptom.

For longitudinal inference, treatment duration must be separated from time since a dose change, and dose from receptor engagement. Concomitant medications, missed doses, illness fluctuations, adherence, and life events can produce apparent changes independent of an intended taper. Studies that record only baseline and end-of-trial medication status discard the transitions that often determine tolerability. Repeated exposure and symptom assessments allow researchers to evaluate whether the same change produces different outcomes early versus late in a withdrawal sequence.

Pharmacodynamic Adaptation and Homeostatic Compensation

Organisms preserve regulation through processes that respond to repeated perturbations. Long-term receptor blockade can produce compensatory receptor or downstream signaling changes; sustained agonism or positive allosteric modulation may alter receptor responsiveness and neuronal network balance. The language of homeostasis and allostasis supplies organizing principles, but it does not identify a universal molecular cause of psychiatric withdrawal. Different drugs act through different target profiles, and adaptation varies across tissue, behavior, and duration. A strong model makes drug-class-specific and symptom-specific predictions rather than treating all adverse experiences as consequences of one generic imbalance.

Benzodiazepines enhance GABA-A receptor-mediated inhibition through positive allosteric modulation, and a reduction after sustained exposure can expose altered inhibitory and excitatory regulation. Serotonergic antidepressants act on transporters and other targets; changes in transporter occupancy and downstream regulation have informed proposed taper trajectories. Antipsychotics differ in dopamine receptor antagonism, partial agonism, anticholinergic actions, histamine antagonism, and additional pharmacological effects. Lithium's intracellular and systems-level actions are distinct again. These differences mean that withdrawal cannot be reduced to a single universal sequence or transferable percentage schedule.

Behavioral outcomes depend on both the directly pharmacological change and the person's adaptive environment. Anticipatory anxiety, bodily monitoring, sleep disruption, compensatory routines, and learned avoidance can amplify distress after a reduction without making the initiating symptoms imaginary. Conversely, a reassuring clinical relationship can improve coping without eliminating severe physiological complications. A complete model integrates measured exposure, regulatory adaptation, symptom interpretation, environmental conditions, and underlying disorder trajectory.

Drug-Effect Curves and the Hyperbolic Taper Hypothesis

Pharmacological dose-response relationships are frequently nonlinear. Equal numerical dose reductions do not necessarily produce equal changes in receptor occupancy. In their influential 2019 Personal View Source 5, Horowitz and Taylor examined serotonin transporter occupancy data and argued that sequentially smaller absolute reductions at low doses may yield more even changes in estimated biological engagement. This is a plausible pharmacological rationale for proportionate, often called hyperbolic, reduction. It explains why the final movement from a low dose to zero may be biologically important even when its milligram magnitude is small.

The inference from modeled occupancy to patient outcomes requires restraint. Transporter binding is neither a direct measurement of adaptation nor a validated surrogate for withdrawal severity. Drugs may have multiple receptor effects, individual exposure can vary, and patient symptoms can respond to contextual and psychological influences. The cross-class synthesis by Vinkers et al. (2024) Source 16 identifies limited direct randomized evidence about hyperbolic tapering despite widespread clinical and theoretical interest. A model that improves pharmacological continuity deserves clinical testing; the model does not establish an optimal schedule for every patient.

Contemporary guidelines endorse graded and individualized dose reductions for relevant medication classes, and NICE depression guidance Source 12 discusses progressively smaller reductions and formulation choices when stopping antidepressants. These recommendations are general clinical guidance, not proof that one percentage or interval optimally balances withdrawal, relapse, practical burden, and other treatment goals. Scientific comparisons must specify whether an intervention manipulates reduction size, total duration, patient choice, monitoring frequency, psychological support, or several of these simultaneously.

Withdrawal Phenomenology and Symptom Classification

Withdrawal can manifest through physical, emotional, cognitive, behavioral, autonomic, and sleep-related experiences. The frequently cited FINISH mnemonic groups antidepressant discontinuation phenomena as flu-like symptoms, insomnia, nausea, imbalance, sensory disturbances, and hyperarousal; Bhat and Kennedy (2017) Source 1 illustrate its clinical utility. Dizziness, abnormal sensations, gastrointestinal disturbance, vivid dreams, anxiety, irritability, and impaired concentration may occur in variable combinations. A list of symptoms alone cannot demonstrate causation because numerous symptoms also appear in anxiety, depression, intercurrent medical illness, and adverse medication reactions.

Symptom interpretation requires onset latency, duration, progression, course after an adjustment in exposure, and a plausible association with the drug's pharmacology. New sensory disturbances emerging soon after a reduction may support a withdrawal explanation, whereas a sustained recurrence of a person's characteristic depressive syndrome following a longer symptom-free interval may support relapse. These patterns are tendencies, not infallible diagnostic algorithms. Withdrawal and relapse can coexist; sleep disruption and distress during withdrawal may precipitate worsening of an underlying illness. Severe symptoms should prompt clinical assessment rather than conceptual overconfidence.

Research must also attend to unusual or prolonged courses. Some patients report symptoms persisting beyond the brief windows traditionally emphasized in trials; their experiences deserve careful measurement and evaluation. Attributing every prolonged presentation to a medication can neglect alternative explanations, while ruling out withdrawal solely because symptoms persist can prematurely exclude an important hypothesis. Duration distributions, patient-level trajectories, and comparisons against appropriate control conditions are needed to determine the prevalence and causes of prolonged states.

Expectation, Conditioning, and the Psychology of Discontinuation

Expectations influence symptom attention, distress, adherence, and the interpretation of bodily signals. Participants who know they are stopping a medication may monitor for anticipated effects; those who anticipate improvement may experience the same transition differently. Placebo-controlled discontinuation studies make these processes visible because some participants assigned to placebo also report discontinuation-like symptoms. This observation is important for causal estimation, but it cannot justify dismissing an individual's symptom report. It shows that observed symptom frequency combines pharmacological, contextual, and background components.

Conditioning can operate in both directions. A patient who repeatedly experiences dizziness after missed doses may learn to associate changes in medication timing with threat, while another who has undergone successful shared planning may develop more confidence in interpreting transient discomfort. Clinical conversations shape the opportunity to distinguish symptom categories and to report them accurately. A symptom scale administered only after an adverse event will capture different information from structured daily assessment that includes anticipatory expectations, mood, sleep, exposure, and functional consequences.

A psychologically informed discontinuation study therefore measures beliefs about medication, previous experiences, anxiety sensitivity, therapeutic alliance, and reasons for stopping alongside pharmacological variables. These measures can test hypotheses about moderation and mediation rather than treating expectations as explanations introduced after conflicting results appear. Importantly, psychological treatment may help people tolerate change or prevent relapse independently of any effect on the neurobiology of withdrawal; intervention designs must distinguish those pathways.

Physical Dependence and Substance Use Disorders

Tolerance and physical dependence arise in both therapeutic and nontherapeutic settings. The diagnosis of a substance use disorder involves a wider behavioral pattern and functional impairment, not the mere fact that dose reduction causes unpleasant physiological effects. Brunner and colleagues' 2025 joint guideline Source 2 makes this distinction central to benzodiazepine care. It matters scientifically because conflating dependence with addiction distorts prevalence estimates, treatment populations, risk assessment, and the moral interpretation of medication-related harm.

Withdrawal severity may nevertheless be elevated in settings that also involve substance use disorder, repeated escalation, multiple sedating drugs, or unstable supply. Alcohol and opioids can compound risks associated with benzodiazepines and other central nervous system depressants. Clinical trials often exclude precisely these complex populations, creating limitations for external validity. A general psychiatric outpatient estimate cannot be transferred to an individual with concurrent substance use, seizure vulnerability, severe medical comorbidity, or multiple medication changes.

The study of craving adds an additional distinction. Craving is an appetitive or aversive motivational state that can arise through reward learning, negative reinforcement, and contextual cues. A person may wish to resume a prescribed drug to relieve a withdrawal syndrome without meeting criteria for addictive behavior. Meanwhile an individual with substance use disorder may experience craving, impaired control, withdrawal, or combinations that change over time. Accurate measurement preserves each construct's independence and permits study of their interaction.

Withdrawal, Relapse, Rebound, and Recurrence as Causal Hypotheses

The differential diagnosis after medication reduction is a problem in causal inference. A new symptom may reflect withdrawal, loss of a preventive drug effect, recurrence of an underlying disorder, rebound, a nocebo or expectation effect, medication interactions, a separate medical illness, or several pathways operating together. A clinician's judgment and a researcher's causal model must both consider the complete timeline. Reinstatement followed by rapid symptom relief can support withdrawal, but response to a medication is not in itself definitive because drugs can also relieve symptoms of the underlying disorder and because symptom courses vary.

Randomized maintenance-withdrawal trials estimate differences between treatment strategies in a defined trial population, under an actual discontinuation protocol. Such an estimate can include effects of loss of maintenance protection, withdrawal-induced symptoms classified as relapse, differential unblinding, and downstream events caused by tapering. It answers a legitimate pragmatic question about a strategy; it does not by itself identify how much of a measured difference is attributable to the illness returning independently of the transition. The ANTLER trial Source 8 is an important example of rigorous randomized strategy comparison that still needs interpretation in light of reported withdrawal symptoms.

A satisfactory design measures several competing outcomes and explicitly models their timing. Withdrawal-specific sensory and autonomic scales, disorder-specific symptom measures, structured clinical episode criteria, functional outcomes, and adverse-event records should be collected together. Investigators can report early post-reduction outcomes separately from later recurrence, compare patterns by half-life, and conduct sensitivity analyses of alternative relapse definitions. These methods make explanatory claims more precise while preserving the primary randomized comparison.

Antidepressant Discontinuation: Incidence and Measurement

Antidepressant discontinuation is the most extensively quantified part of the contemporary withdrawal literature, but estimates depend strongly on study design and how symptoms are counted. Henssler and colleagues (2024) Source 4 synthesized randomized, controlled, and observational studies and estimated that around 31% of participants discontinuing an antidepressant reported at least one discontinuation symptom, compared with approximately 17% of those discontinuing placebo. The authors interpreted the difference as suggesting that symptoms attributable specifically to medication discontinuation affect about 15% of people. Severe symptoms were estimated at around 3% after antidepressant cessation. Substantial heterogeneity prevents these summary percentages from being treated as individual probabilities or as the incidence for every medication.

The subtraction of event proportions across different groups is a useful approximation rather than a perfect decomposition of causation. Nocebo processes, background symptoms, incomplete blinding, diagnosis, study selection, and nonstandard definitions can influence both rates. Observational surveys may overrepresent individuals with difficult withdrawal experiences, while trials may underrepresent long-term users or people who have experienced problems during previous dose changes. The 2024 meta-analysis Source 4 therefore provides a stronger quantitative foundation than isolated anecdotes but leaves important questions about duration, severity, and representativeness unresolved.

Drug-specific differences also matter. Shorter effective elimination and the drug's broader pharmacology may affect risk, while the patient’s prior experience and the method of reduction shape recorded outcomes. Rosenbaum et al. (1998) Source 13 established experimentally that SSRIs differ in the frequency of symptoms during temporary interruption. The relevance of that finding extends beyond a simple long-versus-short half-life distinction: effects can depend on active metabolites, prior use, nonserotonergic action, and how long a participant is observed after drug interruption.

Antidepressant Maintenance Trials and Relapse Risk

The randomized ANTLER trial by Lewis and colleagues (2021) Source 8 compared maintenance with a protocol involving tapering and discontinuation among people in UK primary care who felt sufficiently well to consider stopping long-term antidepressants. The trial randomized 478 people, with relapse by 52 weeks recorded in 92 of 238 maintenance participants (39%) and 135 of 240 discontinuation participants (56%). The reported hazard ratio was 2.06 (95% confidence interval, 1.56–2.70). Participants assigned to discontinue also reported more withdrawal symptoms and higher symptom burden. These findings support continued treatment as an effective relapse-prevention strategy for the studied population under the studied stopping protocol.

Interpretation requires attention to the intervention itself. The comparison was between whole treatment strategies, not an isolated drug molecule and an otherwise identical physiological state. Reduction after prolonged exposure can cause early symptoms that overlap with depression and anxiety measures, and a trial's finite follow-up cannot determine the lifetime benefit of either strategy. ANTLER does not demonstrate that all people should remain on antidepressants, nor does it show that slow individualized tapering with structured psychological support would yield the same results as its specific discontinuation procedure.

The latest comparative evidence further refines the question. Zaccoletti and colleagues (2026) Source 17 analyzed 76 randomized trials involving 17,379 participants with remitted depression or anxiety. In their network meta-analysis, slow tapering accompanied by psychological support reduced relapse relative to abrupt discontinuation (reported risk ratio 0.52, 95% confidence interval 0.38–0.72), while slow tapering without the additional support did not show a statistically clear advantage (0.81, 0.56–1.18). The authors concluded that slow tapering with psychological support performed comparably to continuation for relapse prevention in remitted depression, while the anxiety evidence was less extensive. Network comparisons depend on transitivity, study similarity, available comparisons, and the distinction between direct and indirect evidence.

Antidepressant Tapering as an Intervention Bundle

The expression slow tapering is often used as though it defined one reproducible procedure. Across studies and settings it can refer to different total durations, numbers of reduction steps, final doses, available formulations, follow-up schedules, and thresholds for pausing. It may be paired with psychotherapy, psychoeducation, clinician contact, relapse-prevention planning, or self-monitoring. Therefore the relevant experimental question is which components of a discontinuation strategy improve outcomes for which patients, rather than which unspecified pace is intrinsically superior.

NICE's 2022 depression guidance Source 12 asks that antidepressant reductions reflect the medication's pharmacokinetics, period of use, and a pace acceptable to the person; it recognizes the role of progressively smaller reductions and appropriate formulations. This guidance supports individualized practice, although exact optimal reductions and durations are incompletely established by head-to-head randomized trials. A widely cited proposal by Horowitz and Taylor (2019) Source 5 emphasizes the nonlinear relationship between dose and serotonin transporter occupancy. The difference between a plausible receptor-engagement model and validated patient outcomes remains an important research distinction.

Psychological support plausibly acts by supporting coping, identifying early mood changes, increasing confidence in treatment decisions, improving monitoring, and establishing rapid access to care when problems arise. Yet researchers must test whether support reduces genuinely pharmacological withdrawal manifestations, prevents relapse through behavioral mechanisms, increases adherence to a planned taper, or affects outcome reporting. The 2026 network analysis Source 17 provides comparative outcome evidence; mechanistic attribution needs more direct factorial experiments and prospective exposure data.

Benzodiazepine Tolerance and Dependence

Benzodiazepines occupy a special place because severe withdrawal from established physical dependence can be medically dangerous. Positive allosteric modulation of GABA-A receptors contributes to therapeutic sedative, anxiolytic, anticonvulsant, and muscle-relaxant effects; repeated exposure may produce tolerance to some effects and dependence even when patients take medications exactly as prescribed. These changes are heterogeneous and cannot be inferred solely from dosage or duration. Sleep, anxiety, age, medical vulnerability, comedications, and dependence history determine the assessment context.

The US FDA's September 23, 2020, class-wide safety communication Source 15 strengthened boxed warnings concerning misuse, addiction, physical dependence, and withdrawal. It is a regulatory communication in a specific jurisdiction; it supplies safety evidence and prescribing-information context, not a universal cross-national clinical protocol. Its significance for the scientific literature is recognition that serious withdrawal can occur after therapeutic exposure and that abrupt change may create major hazards.

The 2025 multisociety guideline led by the American Society of Addiction Medicine Source 2 distinguishes physiological dependence from benzodiazepine use disorder and emphasizes individualized risk-benefit reassessment, patient collaboration, clinical supervision, and adjustment of tapering to observed response. The guideline explicitly cautions against abrupt cessation in patients likely to be physically dependent. Its recommendations are partly grounded in limited comparative evidence and partly in a transparent clinical consensus process. The strength of the clinical warning and the scarcity of randomized trials testing precise taper trajectories are compatible facts.

Benzodiazepine Withdrawal Safety and Clinical Risk

Withdrawal manifestations can include anxiety, insomnia, tremor, perceptual change, autonomic activation, agitation, and—at the severe end—seizures or delirium. Their onset and duration vary with drug kinetics, exposure, coexisting conditions, and the actual pace of reduction. When benzodiazepines are used with opioids or alcohol, the risks of ongoing treatment, intoxication, respiratory depression, and the transition away from treatment must all enter the assessment. Brunner et al. (2025) Source 2 treat selection of level of care and management of complicated withdrawal as central clinical decisions.

A safety-oriented study must identify high-risk contexts rather than extrapolate from uncomplicated outpatient cohorts. Previous withdrawal seizures, concurrent sedative use, seizure disorders, marked instability, older age, pregnancy, and coexisting medical conditions may affect risk. Attributing severe symptoms simply to anxiety disorders can miss a drug-related emergency. At the same time, interpreting every episode of anxiety as life-threatening withdrawal can misrepresent ordinary symptom fluctuation. Assessments integrate history, timing, examination, current medications, and clinical judgment.

The relevant quantitative research deficit concerns which gradual strategies minimize clinically important outcomes across different patient groups. A schedule that is well tolerated by one person may be intolerable for another; selected observational success stories do not establish comparative efficacy. Clinical consensus supports flexible, collaborative changes, but controlled evidence is needed for treatment setting, psychological adjuncts, differences between long- and short-acting agents, and sustained functioning after cessation. Because complicated benzodiazepine withdrawal can be dangerous, this dossier does not provide individual medication schedules or replacement prescriptions.

Antipsychotic Reduction and the Relapse-Protection Tradeoff

Antipsychotic maintenance provides significant protection against relapse in schizophrenia and related recurrent psychoses. Leucht and colleagues' 2012 systematic review and meta-analysis Source 7 synthesized randomized maintenance comparisons and established the protective effect of continued medication over follow-up. Randomized withdrawal designs nevertheless have interpretive limitations: recurrent psychotic episodes, physiological rebound, and potential effects of rapid transitions may be difficult to disentangle if withdrawal-specific outcomes are inadequately measured.

The 2023 RADAR randomized trial by Moncrieff and colleagues Source 9 directly tested a supported reduction strategy against maintenance treatment in people with schizophrenia or other recurrent psychotic disorders. Among 253 randomized participants, the two groups showed no statistically significant difference in social functioning at two years, while serious adverse events and severe relapses were more frequent in the reduction group. The randomized result constrains claims that gradual dose reduction necessarily improves social functioning while preserving clinical stability. At the same time, the trial's selection process, achievable reduction, participants' different medication histories, and incomplete outcome follow-up limit universal generalization.

The 2026 report of the broader RADAR mixed-methods program Source 10 adds qualitative and longer-form interpretation of patients' and clinicians' experiences. Some participants identified positive effects of greater agency or reduced burden, while others experienced symptom worsening and relapse. A meaningful outcome framework should register both. The right clinical comparison is not defined by a single preferred ideology of medication continuation or discontinuation, but by observed effects on episodes, hospitalization, adverse effects, quality of life, cognition, and personally valued participation.

Antipsychotic Receptor Adaptation and Tapering Models

The theoretical rationale for especially gradual antipsychotic tapering often invokes dopamine receptor adaptation and nonlinear occupancy. Horowitz, Jauhar, Natesan, Murray, and Taylor (2021) Source 6 synthesized pharmacological and clinical arguments supporting a gradual reduction model. Their work contributes mechanistic hypotheses about why step size and time between steps may matter. It should be identified as an explanatory proposal supported by mixed lines of evidence rather than a directly confirmed universal intervention regimen.

Clinical heterogeneity complicates this model. Antipsychotics differ in receptor profile, dosing form, active metabolites, duration of action, prior exposure, and withdrawal-associated phenomena such as sleep disruption, cholinergic rebound, and movement symptoms. Long-acting injectable preparations introduce distinct exposure curves, while partial agonists cannot be assumed to behave identically to full antagonists. A single dopamine-occupancy curve does not explain the entire post-discontinuation phenotype.

A rigorous trial of proposed taper trajectories would compare well-described strategies over adequate follow-up while preserving rescue treatment, measuring drug exposure, repeated psychiatric states, motor effects, functional outcomes, and patient preferences. Researchers should differentiate relapse that occurs during dose change from recurrence after sustained stability on a lower dose. The RADAR trial Source 9 shows that even an extended, supported protocol can carry meaningful relapse risks, underscoring the importance of direct empirical testing rather than inference from receptor theory alone.

Lithium Maintenance and Discontinuation

Lithium occupies a distinctive position as a mood-stabilizing treatment for bipolar disorders with evidence for long-term prevention of mood episodes. The phenomenon of early recurrence after withdrawal has been studied for decades. Suppes et al. (1991) Source 14 summarized recurrence patterns after lithium discontinuation, while Faedda and colleagues (1993) Source 3 followed a cohort of 64 stable outpatients after rapid or gradual cessation. Their findings highlighted greater early recurrence after more rapid discontinuation. These associations are clinically consequential, but treatment-choice confounding and small cohorts limit simple causal interpretations.

The current NICE bipolar guidance Source 11 recommends gradual lithium reduction and monitoring for early mood changes during and after cessation. Its recommendations combine available evidence and expert judgment for clinical risk management. They should not be recast as proof of a precise biologically optimized dose-reduction curve. Lithium also has an unusually important physical health monitoring context, including renal, thyroid, and serum concentration considerations during therapy; a decision to continue or stop is grounded in the complete benefit-harm balance.

Withdrawal-related mood destabilization and loss of prophylactic protection are not readily separable in every study. Rapid recurrence soon after stopping may be compatible with withdrawal-related destabilization, recurrence of an underlying episodic illness, or both. A stronger study would randomize among ethically acceptable discontinuation speeds, ascertain baseline recurrence history, and follow participants for an extended period after changes in exposure. The history of lithium withdrawal demonstrates why a clinical choice can be both medically individualized and of general scientific importance.

Stimulants and Other Psychotropic Classes

The broad psychotropic category includes medications with very different indications and discontinuation profiles. Prescribed stimulants used for attention-deficit/hyperactivity disorder have characteristic changes in alertness, fatigue, sleep, appetite, mood, and symptom control when altered; the recurrence of attention-related difficulties must be separated from a physiological transition. The research literature is less extensive for long-term stimulant deprescribing than for major antidepressant and antipsychotic maintenance questions. Vinkers and colleagues (2024) Source 16 review stimulants alongside other major psychiatric treatment classes and identify evidence gaps.

Other prescribed psychotropics include sedating antihistaminergic and anticholinergic agents, anticonvulsant medications used as mood stabilizers, and medications acting on opioid pathways. Abrupt interruption of an anticonvulsant may have seizure implications in an individual whose underlying neurological illness requires continued treatment. Opioid-related withdrawal has a well-described physiological syndrome and a separate, extensive addiction-treatment literature. The mechanisms, approved indications, and safety concerns differ, so an all-purpose psychotropic withdrawal schedule would be scientifically indefensible.

Discontinuation research should build a medication-specific mapping of receptor and transport actions, exposure trajectories, withdrawal phenomenology, relapse protection, baseline severity, and safety risk. The relevant unit of evidence is the drug-condition-population-strategy combination. Although higher-level cross-class principles improve conceptual clarity, reliable recommendations must descend to the studies that match the population and clinical question.

Trial Architectures for Discontinuation Research

Discontinuation trials answer different causal questions according to how participants enter treatment, how remission is established, and how stopping is implemented. In an enrichment design, all eligible people first receive active treatment and only those responding are randomized to maintenance or withdrawal. Such a design is informative for patients who already benefit from treatment, but it does not estimate average treatment effects in everyone with the disorder. In a direct randomized taper trial, established users may be allocated to continuing treatment or to alternative reduction strategies. Both designs must specify what counts as a relapse, whether rescue treatment is allowed, how subsequent treatment is analyzed, and what follow-up is sufficiently long.

Withdrawal can partially unblind a nominally double-blind trial because characteristic sensations inform participants or raters about treatment allocation. If symptoms are expected in the discontinuation arm, beliefs can modify reporting and clinical contact. Conversely, fully masking a taper may be difficult when patients need to adjust formulations or communicate problems. Rosenbaum and colleagues (1998) Source 13 illustrate the contribution of controlled interruption, while Lewis and colleagues (2021) Source 8 show the value of longer randomized follow-up. These designs are complementary rather than interchangeable.

A trial should prespecify estimands: the effect of assignment to a stopping strategy, the effect of actually attaining cessation, the effect of time at a lower dose, and consequences of deviations from the assigned plan are distinct quantities. Intention-to-treat preserves randomization for the assignment effect; per-protocol or as-treated comparisons require additional assumptions and methods to address confounding introduced after randomization. Reporting should make these distinctions visible to readers who may otherwise interpret one relapse statistic as an immutable biological consequence of discontinuation.

Measurement Instruments and Outcome Validity

Withdrawal research requires more than a single general symptom checklist. A candidate instrument should define its construct, time window, domains, scoring procedures, and clinical threshold. The Discontinuation-Emergent Signs and Symptoms checklist and related assessment approaches helped formalize measurement in antidepressant trials, yet there is no universally validated measure covering every psychotropic class. Instruments can be sensitive to the symptoms they were originally designed to detect and may miss changes in functioning, mood episodes, prolonged physical complaints, or withdrawal patterns specific to another medication.

Construct validity is especially difficult where withdrawal and underlying illness share manifestations. Anxiety, insomnia, depressed mood, concentration problems, and somatic symptoms can enter both withdrawal and disorder-rating scales. A study that defines relapse from a threshold on such a scale may label some transition symptoms as illness recurrence. Additional structured interviews, onset timelines, neurological and physical assessments, and persistence criteria can improve classification without making a perfect diagnostic separation possible. Predefined adjudication rules reduce the risk that outcome labels are chosen according to investigator expectations.

Patient-reported outcomes are essential because individuals directly experience symptom burden, but their validity depends on understandable wording, accessible response formats, recall duration, and selection into follow-up. Clinician-rated measures provide complementary observations yet may underestimate experiences that patients cannot easily communicate in an appointment. Repeated measurement, recording of the actual dose taken, functional outcomes, emergency care, work and caregiving participation, and sleep data make a multidimensional account more reproducible.

Heterogeneity and Individual Differences

Average withdrawal incidence hides wide variation in exposure history, biological vulnerability, psychiatric history, life circumstances, and social support. A meaningful moderator can alter the size or direction of an intervention effect; a prognostic factor merely predicts an outcome regardless of treatment. Distinguishing the two requires formally tested interactions, adequate sample sizes, and independent replication. A correlation between severe withdrawal and long treatment duration in a selected survey may be confounded by prior illness severity, treatment indication, or the reason the person responded to the survey.

Age, pregnancy, metabolic differences, liver and kidney function, comorbid physical illness, and polypharmacy may modify concentration changes and safety, although evidence is stronger for particular pharmacokinetic mechanisms than for reliable general predictions of psychological withdrawal. The 2025 multisociety benzodiazepine guideline Source 2 devotes substantial attention to older adults, pregnancy, substance use disorders, and higher-risk care settings. These recommendations identify clinically relevant variation while also illustrating the limitations of comparative evidence for many subpopulations.

Social context also matters. People may seek reduction because of adverse effects, cost, access problems, pregnancy planning, subjective loss of agency, improvement, or perceived lack of benefit. They may stop inadvertently because of supply interruptions, insurance changes, displacement, or inability to contact a prescriber. These are different populations and exposures. Research should avoid treating a monitored elective taper as equivalent to abrupt involuntary loss of medication access, especially where the latter is associated with major disruption and heightened vulnerability.

The Role of Psychological and Psychosocial Support

Psychological interventions during discontinuation may include collaborative planning, coping strategies for insomnia and anxiety, relapse prevention, behavioral activation, and help with interpreting fluctuating symptoms. Their rationale is partly independent of the pharmacology: improved self-management and timely response to recurrence can matter whether physiological withdrawal is present or absent. The evidence has become more informative through Zaccoletti and colleagues' 2026 network meta-analysis Source 17, where slow tapering with psychological support showed better relapse-prevention outcomes than abrupt discontinuation.

The label psychological support encompasses different interventions and doses of contact. A structured psychotherapy program cannot be treated as equivalent to one brief reassurance appointment; outcomes may depend on therapist expertise, delivery setting, accessibility, and participant engagement. Moreover, a reduction program with frequent monitoring may detect more mild symptoms while preventing escalation of severe ones. Such patterns require careful interpretation rather than assuming that more recorded complaints necessarily signal worse care.

An ethically sound study should report whether participants could pause reduction, whether they were offered rescue treatment, and what happened when symptoms arose. The availability of support is a feature of the treatment strategy, not mere background context. For medication classes capable of severe withdrawal, clinical oversight and emergency response are safety requirements. The research program should assess long-term autonomy and quality of life while maintaining protection against serious adverse outcomes.

Duration, Protracted Symptoms, and Follow-Up Windows

A substantial research gap concerns the full distribution of symptoms over months and years. Short interruption experiments can detect acute phenomena but rarely illuminate prolonged symptom trajectories. Withdrawal that emerges after a delay can be missed when observation ends early; conversely, lengthy follow-up without a comparison group may classify new, unrelated symptoms as persistent medication effects. Researchers should measure both the onset and offset of symptoms and preserve uncertainty about etiology where evidence cannot distinguish mechanisms.

Protracted withdrawal is an important patient-reported phenomenon deserving direct empirical evaluation. Studies require prospectively defined duration criteria, exposure histories, appropriate comparison groups, and repeated functional measures. Self-selected online samples can document the existence and character of reported experiences, generate hypotheses, and reveal underrecognized burdens; they cannot determine population prevalence without sampling assumptions. A sufficiently rigorous program would include people with varied outcomes, including uneventful discontinuation, persistent symptoms, and relapse without withdrawal.

Long-term outcomes matter because an early increase in symptoms may resolve while later functional gains emerge, or an initially easy discontinuation may be followed by a new illness episode. Survival analyses, recurrent-event models, longitudinal mixed models, and competing-risk approaches can characterize different trajectories. The selection of an observation horizon should follow the question: acute pharmacological withdrawal, medium-term episode prevention, and multi-year quality of life need different periods of surveillance.

Risk of Bias, Publication Bias, and Confounding

All bodies of discontinuation evidence contain design-dependent vulnerabilities. In open observational studies, people with greater distress may be more likely to seek assessment, participate in surveys, or report outcomes. In randomized trials, high dropout, adherence deviations, rescue treatment, crossover, selective adverse-event measurement, and differential unblinding can complicate the apparent effect of assignment. Analyses of completers alone may exclude exactly those participants with the most difficult clinical transitions.

Meta-analysis increases precision only insofar as the included studies ask sufficiently similar questions and use defensible measurements. Henssler et al. (2024) Source 4 reported substantial heterogeneity across antidepressant discontinuation studies, requiring interpretation of pooled incidence as a broad summary. Zaccoletti et al. (2026) Source 17 examined a network of discontinuation strategies, where transitivity assumptions and treatment classification become central. A pooled estimate should be discussed alongside risk of bias, study populations, confidence intervals, and certainty rather than promoted as a universal personal forecast.

Conflicts of interest and sponsorship should be transparently assessed, particularly in drug-class comparisons, manufacturer-funded trials, and studies evaluating commercially available formulations. Funding alone does not invalidate a result; outcomes and methods determine evidential weight. Prespecified endpoints, publication of protocols, access to trial reports, availability of data definitions, and independent replications improve confidence in both positive and negative findings. Narrative syntheses such as this dossier must explicitly state that they are not protocol-driven exhaustive reviews.

Cross-Class Comparison and the Limits of General Rules

The major classes differ in the balance of withdrawal hazard and recurrence prevention. Benzodiazepine withdrawal can include seizures and delirium, emphasizing immediate safety and supervision. Antidepressants have substantial evidence for discontinuation symptoms and relapse differences, with emerging evidence about psychological support and reduction strategies. Antipsychotics have strong relapse-prevention evidence in recurrent psychosis and randomized evidence that supported reduction may raise relapse risk without improving mean social functioning. Lithium has a distinctive history of early mood recurrence after cessation. These differences appear in the cross-class synthesis by Vinkers et al. (2024) Source 16 and the respective class-specific trials and guidelines.

Three general research principles survive the comparison. First, the history of exposure and the kinetics of drug elimination matter when interpreting symptoms. Second, symptom categories must be separated from causal attributions by observation and appropriate comparisons. Third, dose-reduction strategies have multiple possible outcomes—withdrawal burden, relapse, safety, functioning, and patient preference—and optimization requires a joint rather than single-outcome assessment. These principles do not imply that one drug, condition, or patient subgroup provides a ready-made protocol for the others.

The distinction between an explanatory framework and a normative clinical choice should remain explicit. Scientific research can estimate effects of strategies and identify risks, while the preferred choice also depends on the individual's goals, available alternatives, and real-world safety. Robust evidence allows disagreements to become questions about populations, outcomes, and values instead of disputes built around broad slogans about psychiatric medication.

Artificial Era: Medication Information and Human–Artificial Decision Environments

In the Artificial Era, medication information is increasingly encountered through search systems, automated summaries, conversational agents, and digital symptom trackers. The psychological problem becomes partly epistemic: an individual may receive a convincing statement about withdrawal incidence, a tapering schedule, or receptor mechanisms stripped of its population and evidentiary limitations. An artificial system that cites a trial without its treatment duration, inclusion criteria, or measured outcomes can produce an apparently authoritative but clinically distorted account.

Research on such systems should evaluate factual retrieval, correct mapping of a claim to its source, explicit differentiation of randomized evidence from narrative theory, recognition of safety-relevant symptoms, and preservation of meaningful uncertainty. A valid assessment would use independently curated cases and blinded adjudication of the system's responses by qualified experts. Operational metrics could include source fidelity, medication-class specificity, inappropriate personal scheduling suggestions, clinical-hazard omission, and calibrated uncertainty. This is a proposed research direction, not evidence that artificial information systems currently produce particular clinical outcomes.

Human–artificial collaboration can also assist longitudinal observation when implemented under appropriate clinical governance, privacy protections, and opportunities for human review. Such tools might organize patient-reported symptom timelines or remind users to bring questions to clinicians, but measured benefits must be established in prospective studies. The central distinction is between faithfully organizing evidence and assuming responsibility for diagnosis or dosing. The latter involves clinical accountability that cannot be inferred from the linguistic sophistication of a system.

Unresolved Questions and Research Priorities

The strongest immediate research priority is discriminating withdrawal from relapse with longitudinal, multi-domain measures that capture the same symptom in both diagnostic frameworks. Blinded adjudication, pharmacokinetic exposure records, and observation of early and late outcomes should be combined rather than studied separately. More comparative randomized evidence is needed on taper duration, proportional step size, psychological support, drug formulations, and monitoring frequency. The 2026 evidence for slow tapering with support is promising for remitted depression, but it does not answer all class-specific questions or establish which components of care are causal.

A second priority concerns external validity. Trials must better represent long-term users, people with previous withdrawal difficulty, multiple disorders, polypharmacy, older adults, and populations experiencing interrupted medication access. Trials also need to include participants who decline to discontinue and report how this choice affects generalizability. Transparent measurement of treatment exposure, adverse-event severity, satisfaction, occupational functioning, and healthcare utilization can make the results useful beyond remission scales.

A third priority is mechanistic triangulation. Receptor occupancy studies, pharmacokinetic models, animal and human neurobiological findings, and prospective symptom studies should be connected through preregistered hypotheses. Such convergence could reveal why some drugs cause particular syndromes and why some individuals experience disproportionate effects at low doses. Until that evidence matures, pharmacological plausibility and evidence-based clinical comparison must retain their separate roles.

Critical Synthesis

The contemporary literature supports a differentiated model of psychotropic discontinuation. Repeated drug exposure can generate tolerance to particular effects and physical dependence; reducing exposure can produce distinct withdrawal syndromes; stopping maintenance treatment can alter recurrence risk; and each process is modified by the drug, illness, history, environment, and discontinuation strategy. The empirical evidence is strongest where controlled studies and guidelines can be compared directly, notably for antidepressant discontinuation symptoms, benzodiazepine safety, antipsychotic maintenance and reduction, and lithium recurrence after cessation.

The literature also demonstrates the limits of reductive explanations. Receptor adaptation helps explain biological possibilities but cannot alone determine an individual's outcome. A randomized relapse advantage of maintenance is a robust comparative finding under a study protocol, while explaining how much of that advantage reflects underlying illness versus withdrawal requires additional measurements. Similarly, evidence that a combined reduction-and-support strategy improves outcomes does not prove the effect of its pharmacological and psychosocial components separately. These distinctions improve scientific precision and point toward more informative interventions.

Scientific claims should be graded by their methods. Regulatory warnings establish recognized hazards; practice guidelines integrate evidence and professional judgment; prospective observational work detects trajectories and associations; randomized trials compare strategies; systematic reviews test reproducibility and heterogeneity; and mechanistic models generate testable hypotheses. A complete psychological account preserves all of these contributions without giving them identical epistemic weight.

Conclusion

Psychotropic tolerance, withdrawal, and discontinuation are dynamic phenomena of adaptation and clinical transition. Their study requires precise separation of tolerance, physical dependence, addiction, withdrawal, rebound, relapse, and recurrence, followed by explicit tests of the relationships among them. The strongest evidence indicates that abrupt or poorly supported change can have important consequences, but those consequences differ substantially across drugs and conditions, and continuation also carries burdens that must be evaluated. Treatment strategies are best understood as multidimensional interventions over time.

The next phase of research must combine pharmacokinetic and pharmacodynamic measurement, patient-reported trajectories, clinical episode adjudication, randomized strategy comparisons, and sustained functional follow-up. Recent comparative trials and syntheses materially advance this work, while important questions persist about individualized tapering, prolonged symptoms, representative populations, and mechanistic attribution. The scientific object is the entire transition in a person's functioning and risk after sustained exposure, not the isolated event of taking a final dose.

References

  • Bhat, V., & Kennedy, S. H. (2017). Recognition and management of antidepressant discontinuation syndrome. Journal of Psychiatry & Neuroscience, 42(4), E7–E8. https://doi.org/10.1503/jpn.170022
  • Brunner, E., Chen, C.-Y. A., Klein, T., et al. (2025). Joint clinical practice guideline on benzodiazepine tapering: Considerations when risks outweigh benefits. Journal of General Internal Medicine, 40, 2814–2859. https://doi.org/10.1007/s11606-025-09499-2
  • Faedda, G. L., Tondo, L., Baldessarini, R. J., Suppes, T., & Tohen, M. (1993). Outcome after rapid vs gradual discontinuation of lithium treatment in bipolar disorders. Archives of General Psychiatry, 50(6), 448–455. https://doi.org/10.1001/archpsyc.1993.01820180046005
  • Henssler, J., Schmidt, Y., Schmidt, U., Schwarzer, G., Bschor, T., & Baethge, C. (2024). Incidence of antidepressant discontinuation symptoms: A systematic review and meta-analysis. The Lancet Psychiatry, 11(7), 526–535. https://doi.org/10.1016/S2215-0366(24)00133-0
  • Horowitz, M. A., & Taylor, D. (2019). Tapering of SSRI treatment to mitigate withdrawal symptoms. The Lancet Psychiatry, 6(6), 538–546. https://doi.org/10.1016/S2215-0366(19)30032-X
  • Horowitz, M. A., Jauhar, S., Natesan, S., Murray, R. M., & Taylor, D. (2021). A method for tapering antipsychotic treatment that may minimize the risk of relapse. Schizophrenia Bulletin, 47(4), 1116–1129. https://doi.org/10.1093/schbul/sbab017
  • Leucht, S., Tardy, M., Komossa, K., Heres, S., Kissling, W., Salanti, G., & Davis, J. M. (2012). Antipsychotic drugs versus placebo for relapse prevention in schizophrenia: A systematic review and meta-analysis. The Lancet, 379(9831), 2063–2071. https://doi.org/10.1016/S0140-6736(12)60239-6
  • Lewis, G., Marston, L., Duffy, L., et al. (2021). Maintenance or discontinuation of antidepressants in primary care. New England Journal of Medicine, 385(14), 1257–1267. https://doi.org/10.1056/NEJMoa2106356
  • Moncrieff, J., Crellin, N., Stansfeld, J., et al. (2023). Antipsychotic dose reduction and discontinuation versus maintenance treatment in people with schizophrenia and other recurrent psychotic disorders in England (the RADAR trial): An open, parallel-group, randomised controlled trial. The Lancet Psychiatry, 10(11), 848–859. https://doi.org/10.1016/S2215-0366(23)00258-4
  • Moncrieff, J., Long, M., Priebe, S., et al. (2026). Antipsychotic discontinuation and reduction in people with schizophrenia and multiple-episode psychosis: The RADAR mixed-methods research programme including RCT. Programme Grants for Applied Research. National Institute for Health and Care Research. https://doi.org/10.3310/GJJM0506
  • National Institute for Health and Care Excellence. (2014, updated). Bipolar disorder: Assessment and management (CG185). https://www.nice.org.uk/guidance/cg185/chapter/recommendations
  • National Institute for Health and Care Excellence. (2022, updated). Depression in adults: Treatment and management (NG222). https://www.nice.org.uk/guidance/ng222/chapter/Recommendations
  • Rosenbaum, J. F., Fava, M., Hoog, S. L., Ascroft, R. C., & Krebs, W. B. (1998). Selective serotonin reuptake inhibitor discontinuation syndrome: A randomized clinical trial. Biological Psychiatry, 44(2), 77–87. https://doi.org/10.1016/S0006-3223(98)00126-7
  • Suppes, T., Baldessarini, R. J., Faedda, G. L., & Tohen, M. (1991). Risk of recurrence following discontinuation of lithium treatment in bipolar disorder. Archives of General Psychiatry, 48(12), 1082–1088. https://doi.org/10.1001/archpsyc.1991.01810360046007
  • U.S. Food and Drug Administration. (2020, September 23). FDA requiring Boxed Warning updated to improve safe use of benzodiazepine drug class. https://www.fda.gov/drugs/drug-safety-and-availability/fda-requiring-boxed-warning-updated-improve-safe-use-benzodiazepine-drug-class
  • Vinkers, C. H., et al. (2024). Discontinuation of psychotropic medication: A synthesis of evidence across medication classes. Molecular Psychiatry, 29, 2575–2586. https://doi.org/10.1038/s41380-024-02445-4
  • Zaccoletti, D., Mosconi, C., Gastaldon, C., Benedetti, L., Gottardi, C., Papola, D., Ponzi, O., Purgato, M., Naudet, F., Cristea, I. A., Barbui, C., & Ostuzzi, G. (2026). Comparison of antidepressant deprescribing strategies in individuals with clinically remitted depression: A systematic review and network meta-analysis. The Lancet Psychiatry, 13(1), 24–36. https://doi.org/10.1016/S2215-0366(25)00330-X