RESEARCH DOSSIER
Therapeutic Drug Monitoring in Psychiatry
Abstract. Therapeutic drug monitoring (TDM) is the clinically contextualized measurement of medication concentrations in blood, usually serum or plasma, to inform psychiatric pharmacotherapy. This monograph examines how drug exposure is measured, how reference concentrations are established, and what concentration measurements can and cannot establish about therapeutic response, toxicity, adherence, and individual pharmacokinetic differences. The method is a critical narrative synthesis of original clinical studies, pharmacokinetic research, systematic evidence reviews, international consensus guidelines, professional recommendations, and medicine-safety communications available through November 2026. The analysis distinguishes concentration-guided treatment with relatively strong clinical foundations, especially lithium and clozapine, from applications in which serum measurement serves mainly as a problem-solving instrument. It evaluates sampling time, steady-state assumptions, active metabolites, analytical methods, biological variation, pharmacogenetics, interpretive bias, longitudinal monitoring, service implementation, and patient-centered consequences. The principal conclusion is that the interpretive unit of TDM is the concentration in its temporal, pharmacological, clinical, and laboratory context. A numerical laboratory result acquires therapeutic meaning through a reasoned relationship to the patient's symptoms, adverse effects, exposure history, organ function, comedications, and goals of treatment. No new clinical trial, prospective patient dataset, de novo systematic review, or meta-analysis was conducted.
Keywords: therapeutic drug monitoring; psychiatry; psychopharmacology; plasma drug concentration; serum drug concentration; lithium; clozapine; antipsychotics; antidepressants; tricyclic antidepressants; valproate; carbamazepine; pharmacokinetics; pharmacogenetics; therapeutic reference range; laboratory medicine; medication adherence; clinical toxicology; precision psychiatry.
Introduction: The Scientific Problem of Measured Exposure
Psychiatric pharmacotherapy operates through a sequence of events that ordinarily remains invisible in the examination room. A medication is prescribed and administered, absorbed and distributed, transformed or eliminated, and reaches tissue sites whose exposure affects biological processes associated with therapeutic and unwanted responses. The prescribed dose specifies a clinical intervention; it does not establish the circulating exposure achieved by a particular person. A blood concentration offers a partial observation of that exposure, allowing a clinician to test hypotheses that cannot be resolved by dose history and symptoms alone. The difference matters because the same dose can generate substantially different concentrations across people and within the same person over time.
Therapeutic drug monitoring (TDM) is the purposeful measurement and interpretation of medicine concentrations in a biological specimen to support treatment decisions. In psychiatry the specimen is most often blood plasma or serum. The contemporary AGNP consensus update (Hart et al., 2026) Source 10 defines a systematically expanded field of neuropsychopharmacological TDM; the guideline presents therapeutic reference ranges for 160 drugs, introduces 30 agents, revises ranges for 88, and separates strong routine recommendations from selected-situation usefulness. These counts describe the consensus guideline's classification, rather than independent evidence of improved outcomes for every included medication. Its associated practice model integrates pharmacokinetics, pharmacogenetics, and clinical interpretation.
The research question is therefore not simply which concentration is normal. It is under what conditions a measured exposure changes the probability of benefit, reveals a preventable risk, explains apparently ineffective treatment, or improves a decision compared with careful clinical follow-up alone. Different drugs require different answers. Lithium carries a relatively narrow margin between helpful exposure and poisoning; clozapine displays substantial between-patient variability and a documented exposure-response association; tricyclic antidepressants can show clinically consequential concentration effects; many newer antidepressants permit useful measurements without equally robust concentration-response evidence. Hiemke et al. (2018) Source 11 and Schoretsanitis and Paulzen (2026) Source 22 provide historically linked but differently focused evidence syntheses.
This work analyzes the concentration as a biomedical measure and the interpretation as a psychological and clinical activity. A psychiatric patient's response includes symptoms, function, adverse experiences, treatment preferences, sleep, and adherence patterns. Laboratory information improves care only if it enters a reliable interpretation-and-action process. The scientific object is consequently the entire process of sampling, analysis, inference, communication, and longitudinal therapeutic evaluation.
History of Concentration-Guided Psychopharmacology
### From empirical treatment to quantitative exposure
The conceptual foundations of TDM arose from pharmacokinetics: the quantitative relationship between dose, absorption, distribution, metabolism, elimination, and concentrations over time. A drug's circulating concentration can act as a measurable indicator of exposure if sampling methods and kinetic conditions are understood. In psychiatry, development of lithium maintenance treatment and the detection of lithium toxicity made reproducible serum measurement particularly important. The subsequent introduction of tricyclic antidepressants, antiepileptic mood stabilizers, and antipsychotics created additional questions about variability of exposure and the possibility of identifying useful concentration ranges.
The history must be distinguished from retrospective assumptions that modern guidance existed from the beginning. Early clinical experience, controlled serum-level comparisons, laboratory assay improvements, and formal guideline development represent different stages. Gelenberg et al. (1989) Source 9 provided an influential randomized, double-blind comparison of lithium maintenance strategies. The study allocated 94 patients to target serum ranges of 0.8–1.0 or 0.4–0.6 mmol/L, observing more relapses in the lower-target condition together with more adverse effects in the higher condition. That design provides stronger causal evidence about those specific treatment strategies than a cross-sectional association between concentrations and clinical status would provide.
The AGNP guideline sequence in 2004, 2011, 2017, and 2026 demonstrates a second institutional history: clinical laboratories, psychiatrists, and pharmacologists progressively classified medicines according to the strength of evidence for monitoring. The 2017 consensus publication, issued in its final journal volume in 2018 Source 11, established a widely used framework; the 2026 update Source 10 substantially revised how therapeutic and dose-related ranges are established. The evolution illustrates why older ranges require a date, a source, and attention to methods of derivation. It also explains why a single inherited table of concentrations is an incomplete description of current practice.
### Changing research goals
Earlier concentration research often sought one therapeutic window for each medicine. Later work exposed the importance of distributions, overlapping response groups, pharmacokinetic covariates, subgroup characteristics, active metabolites, and longitudinal within-person changes. A single cutoff may raise the chance of response without accurately predicting the outcome of one patient. For clozapine, for instance, individual participant data from nine studies (https://pmc.ncbi.nlm.nih.gov/articles/PMC10201335/) showed an area under the receiver operating characteristic curve of approximately 0.61, indicating only modest discrimination even though a useful group-level exposure-response relationship was present.
The research question has expanded accordingly. TDM can test inadequate exposure, unexpected accumulation, interactions, changes in smoking status, pregnancy-related pharmacokinetic changes, organ impairment, or a discrepancy between reported dosing and measured concentrations. The purpose of contemporary monitoring is to explain a clinical problem and guide the next decision, rather than to optimize a number in isolation.
Pharmacokinetic Foundations and Exposure Models
### Dose, bioavailability, clearance, and concentration
At steady state under a simplified linear model, average concentration is proportional to bioavailability multiplied by the administered dose rate and inversely proportional to clearance. This is an explanatory model, not a bedside dose calculator. The assumptions include approximately stable clearance, dosing, adherence, and relevant physiology. Absorption patterns, saturable metabolism, active transport, enzyme induction, and fluctuations in dosing can produce departures from the simplified relationship.
The observed concentration varies during the dosing interval. Peak concentrations can be more sensitive to absorption rate and formulation; trough concentrations can better standardize serial comparisons for some medicines. Neither is inherently an exact estimate of central nervous system exposure. Distribution into brain, protein binding, tissue partitioning, transporter activity, and temporal delay between plasma concentration and receptor or network effects complicate extrapolation from blood to behavior. A quantitative laboratory measurement is therefore a biomarker of systemic exposure, whose relationship to the therapeutic site must be established drug by drug.
Several pharmacokinetic quantities are especially useful. Half-life describes how quickly concentration changes during elimination when the underlying approximation applies. Clearance describes the body's capacity to remove a substance, while volume of distribution helps relate the amount present to the observed concentration. At steady state with time-invariant kinetics, average input is balanced by elimination; repeated dosing does not remove within-interval peaks and troughs. A metabolite can have its own half-life and activity, so the clinically relevant measurement may include parent compound, active metabolite, or their combined concentration. Parent-to-metabolite ratios can help formulate hypotheses about adherence timing and altered biotransformation, although they rarely identify one mechanism conclusively.
### Steady state and sampling timing
A specimen drawn shortly after a dose is not directly interchangeable with a specimen obtained at an established trough. The distinction is pronounced when laboratory reference ranges were developed for a specific collection time. A measured value must therefore be accompanied by the clock time of the last dose, the actual blood draw time, formulation, dosing interval, recent missed or extra doses, recent treatment changes, and whether pharmacokinetic steady state was reasonably achieved.
For lithium maintenance, Nolen et al. (2019) Source 18 explicitly describe a standardized morning measurement approximately 12 ± 1 hours after the evening dose. Other compounds have distinct recommended collection procedures. Carbamazepine poses a special challenge because it induces enzymes involved in its own metabolism. DailyMed product labeling Source 17 describes autoinduction over several weeks; an early concentration may not predict later exposure under the same prescribed regimen.
A laboratory result with missing sampling metadata may still alert the clinician to extreme exposure, but its ability to establish whether a routine reference range has been reached is materially reduced. Clinical services need collection procedures that make the reported number reproducibly interpretable. Whenever feasible, comparison of serial concentrations should use comparable dosing intervals, collection times, and assay conditions.
### Between-person and within-person variability
Two patients prescribed the same milligram dose can differ in hepatic enzyme activity, kidney function, age, body composition, smoking exposure, inflammatory state, drug interactions, adherence, and absorption. The same patient can develop a new exposure profile after hospitalization, infection, medication changes, dehydration, altered nutrition, or smoking cessation. TDM can measure the consequence of multiple influences without itself proving which influence caused a given concentration.
Longitudinal results create a personalized comparison. A concentration documented during stable clinical response provides a reference for interpreting a later change. This within-person benchmark may be useful even for a drug with incomplete population-level exposure-response evidence. Its interpretation remains probabilistic: concurrent changes in illness severity, treatment adherence, comorbidity, and measurement conditions can occur without a single identifiable pharmacokinetic cause.
Therapeutic Ranges and the Logic of Evidence
### What a reference range represents
A therapeutic reference range is a population-derived interval intended to describe concentrations associated with useful therapeutic outcomes and acceptable tolerability under defined clinical circumstances. It must be distinguished from a laboratory analytical reference interval, a toxicological threshold, a dose-related reference range, and an individual's previously effective concentration. A range can summarize data from trials, observational studies, pharmacokinetic models, or expert interpretation; the strength and purpose of those inputs change the certainty attached to it.
A dose-related range predicts concentrations expected when a particular drug is taken at a specified regimen under assumptions about patient pharmacokinetics. It can be valuable for detecting unexpected exposure but says less by itself about symptom remission. Conversely, a therapeutic range claims a relationship to benefit or tolerability, yet even well-supported ranges overlap with individual nonresponse and adverse effects. The 2026 AGNP recommendations Source 10 deliberately differentiate these functions and articulate how ranges were derived. This is a substantive methodological advance over reproducing numbers without their evidentiary origin.
### Probability and decision thresholds
A threshold for concentration is a decision aid, not a biological switch. Response is often gradual, heterogeneous, and influenced by symptom duration, diagnosis, comedication, and psychosocial conditions. A higher concentration can increase the probability of response while also increasing adverse-effect burden. A laboratory value can be within a population range despite severe symptoms; a value can fall outside it while an individual experiences sustained benefit. Clinical decision-making must integrate both observations.
One must also consider the clinical consequences of false reassurance and false alarm. In lithium treatment, interpreting an unexpectedly elevated level requires attention to symptoms and renal function because serious toxicity can occur even when a single number inadequately represents tissue burden or temporal change. In clozapine treatment, prematurely classifying a patient as pharmacologically treatment-resistant can lead to unnecessary augmentation if exposure was never adequately established. In a newer antidepressant, imposing a poorly validated concentration cutoff might generate needless changes in otherwise appropriate treatment.
### Levels of evidence
A randomized comparison of concentration-targeting strategies can directly test whether one monitoring approach leads to better outcomes than another under trial conditions. A prospective concentration-response cohort can establish temporal association, but prescribers may alter doses in response to illness severity, introducing reverse causality or confounding by indication. A cross-sectional laboratory series usually supports a narrower claim about achieved exposure and determinants of variability. A pharmacokinetic simulation evaluates behavior of a specified model and requires external validation.
Systematic reviews combine studies of different populations, sampling conventions, diagnoses, and outcome definitions. Pooled estimates are meaningful only to the extent that those studies are sufficiently compatible. Specialist consensus contributes clinical expertise but does not create experimental evidence where trials are missing. Nolen et al. (2019) (https://pmc.ncbi.nlm.nih.gov/articles/PMC6688930/) are explicit: strict eligibility criteria yielded no qualifying trials for their initial lithium synthesis, and a broader set of seven trials supported a Delphi-based recommendation. This transparency is crucial to understanding the basis of clinical reference values.
The Measurement Chain: From Clinical Question to Assay Result
### Clinical indication and test ordering
The first step is a defined question. Routine monitoring after initiation or dose change, unexpected toxicity, lack of response, suspected interaction, deterioration in kidney or liver function, changes in physiology, and uncertainty about medication exposure are distinguishable indications. Their distinctions should shape when sampling occurs and how the result is used. An order for concentration testing without a clinical question risks generating a number that is technically accurate but clinically inert.
The testing request should identify the substance and, when relevant, metabolites or free fractions; the formulation and current schedule; last dose; sample date and time; relevant co-prescribed drugs; and the reason for testing. Repeated monitoring can be important in long-term therapy, but a fixed schedule does not remove the need to respond to acute illness or other changes between scheduled draws.
### Analytical techniques and quality assurance
Psychotropic concentrations may be measured by chromatographic methods, including high-performance liquid chromatography with ultraviolet or tandem mass-spectrometric detection, or by validated immunoassays for particular analytes. Analytical selectivity matters when structurally similar metabolites or other medicines coexist. Method-dependent cross-reactivity, calibration, limits of quantification, extraction recovery, matrix effects, and sample stability influence what a reported value actually represents.
Laboratory validation should establish accuracy, precision, specificity, reportable range, and performance at clinically relevant decision regions. Internal quality control and external proficiency assessment support comparability. A result obtained from a different laboratory or assay platform may not be perfectly interchangeable with a previous result when the methods measure different molecular species or differ in calibration. Clinical chemistry and laboratory medicine therefore participate directly in the interpretation, rather than merely providing a technical service.
### Preanalytical errors and the denominator problem
Collection time, wrong specimen, incorrect tube selection, specimen misidentification, transport conditions, delayed processing, and recording errors can all compromise interpretation. The most dangerous error may be a plausible value accompanied by incorrect dose timing. Before attributing a sudden change to disease or metabolism, one should determine whether the laboratory and clinical records describe the same pharmacological interval.
Serial concentrations are sometimes summarized using a concentration-to-dose ratio. This ratio is not an invariant patient trait. It assumes comparable dose timing and adherence, stable dosing, and an interpretable relationship between dose and exposure. It can change with smoking, comedication, inflammation, renal clearance, autoinduction, and deliberate dose adjustments. Unexpected ratios help generate investigative questions but rarely substitute for mechanistic assessment.
Clinical Interpretation as a Structured Inference
A defensible interpretation brings together measured concentration, preanalytical conditions, exposure history, pharmacological plausibility, and contemporaneous clinical assessment. A low concentration can indicate insufficient prescribed exposure, a late sample, missed doses, an interaction, malabsorption, or elevated clearance. An elevated concentration can reflect reduced clearance, an early draw, altered protein binding, a formulation change, dosing errors, or interactions. The laboratory observation narrows the possibilities; it does not identify the cause automatically.
The clinician's assessment should include symptoms, adverse effects, mental status, physical findings, relevant organ function and comorbid illness, and treatment goals. For suspected toxicity, immediate assessment and established emergency management take precedence over routine interpretation of a reference band. For treatment nonresponse, an adequate concentration can support the conclusion that exposure was present, while leaving open questions about duration, diagnosis, psychosocial influences, and underlying biological response.
A useful methodological standard is to document the original clinical question, specimen characteristics, result and applicable reference evidence, competing explanations, intervention or decision, and subsequent patient outcome. This transforms TDM from a series of isolated laboratory reports into an auditable longitudinal clinical process. It also permits service evaluation: one can examine whether an actionable result altered care appropriately and whether that change was associated with improved safety, symptom control, or patient experience.
Lithium: Concentration-Guided Mood Stabilization
### The exceptional clinical role of lithium levels
Lithium illustrates why therapeutic drug monitoring is embedded in routine psychiatric care rather than reserved for suspected misuse or unusual cases. The drug has a clinically consequential relationship between concentration and both long-term mood stabilization and toxicity. It is principally eliminated through the kidneys, and changes in kidney function or fluid and sodium balance can change its handling. Monitoring must therefore integrate serum concentration with renal function, electrolytes, thyroid status, calcium, symptoms, and interacting drugs. Neither a serum target nor a normal laboratory panel by itself describes the total safety of long-term treatment.
The original randomized study of Gelenberg and colleagues (1989) Source 9 compared 94 patients assigned to two lithium concentration strategies in maintenance treatment. Relapses while participants remained on protocol occurred in 6 of 47 patients assigned to the standard target of 0.8–1.0 mmol/L versus 18 of 47 assigned to the lower target of 0.4–0.6 mmol/L; the higher-concentration condition also produced more adverse effects. These findings establish a trade-off within the studied population and trial design. They do not establish that all patients or all ages should receive the higher target.
The ISBD/IGSLI systematic review and expert consensus (Nolen et al., 2019) Source 18 recommended a standard maintenance serum concentration of 0.60–0.80 mmol/L for adults, with a lower interval of 0.40–0.60 mmol/L considered where response is good but tolerability is poor, and a higher interval of 0.80–1.00 mmol/L in insufficient response with good tolerability. Importantly, the authors acknowledged limitations in the evidence and the absence of comparable consensus for all age groups. These ranges should be read as clinician-facing recommendations with explicit conditions, not universally applicable targets.
### Different practice guidelines and clinical aims
The NICE bipolar disorder recommendations Source 16 propose an initial plasma lithium target of 0.6–0.8 mmol/L, while specifying circumstances in which a trial at 0.8–1.0 mmol/L may be considered. They require levels after initiation and dose changes until stability, then scheduled checks, with more frequent surveillance for specified high-risk conditions. The CANMAT/ISBD 2023 update summarized in 2024 (https://pmc.ncbi.nlm.nih.gov/articles/PMC11058959/) differentiates acute and maintenance treatment and age groups, and also recommends regular serum monitoring for valproate and carbamazepine. Differences between recommendations partly reflect target population, phase of illness, and the purpose of treatment; they cannot be resolved by selecting the highest number from several published tables.
Standardizing the approximately twelve-hour serum sample is particularly important for lithium. The same plasma concentration drawn at different postdose intervals can correspond to different kinetic states. When a value is higher than expected, clinicians must assess both the laboratory number and symptoms such as tremor, gait problems, gastrointestinal disturbance, or altered consciousness, alongside renal and fluid status. Clinical evaluation and the appropriate emergency pathway govern suspected intoxication.
### Long-term longitudinal safety
Lithium monitoring has two interconnected objectives: exposure management and detection of adverse consequences outside the concentration itself. Renal function and thyroid measures describe physiological domains that can change despite a nominally acceptable lithium level. Changes in calcium and parathyroid function may also matter during extended treatment. Dehydration, acute illness, changes in sodium intake, nonsteroidal anti-inflammatory drugs, renin–angiotensin system medicines, and some diuretics can modify lithium exposure. A stable historical regimen does not guarantee a stable current concentration.
There is an important methodological distinction between the effect of a serum target and the effect of a surveillance program. A patient receiving frequent measurement may also benefit from more clinical contact, improved coordination, and earlier identification of problems. Direct estimates of the additional clinical value of particular monitoring frequencies require studies that compare care processes, not just associations between concentrations and response.
Clozapine: Exposure, Response, and Patient-Specific Variability
### Concentration-response evidence
Clozapine is central to psychiatric TDM because considerable variability exists between administered dose and plasma concentration, and treatment resistance is a high-stakes clinical determination. The most influential exposure-response analyses evaluate whether patients achieving higher concentrations are more likely to respond. A 2021 systematic review and meta-analysis (https://pubmed.ncbi.nlm.nih.gov/34374073/) reported that concentrations above 350 ng/mL were associated with higher response rates across contributing studies, while very high concentrations did not show a comparable straightforward improvement.
The individual patient data meta-analysis reported in 2023 (https://pmc.ncbi.nlm.nih.gov/articles/PMC10201335/) assembled 294 individuals from nine studies and estimated 372 ng/mL as the cutoff with the greatest diagnostic benefit in its receiver operating characteristic analysis. The area under the curve was 0.612, with appreciable overlap between responders and nonresponders. The analysis supports concentration optimization as one component of the investigation of apparent nonresponse, while the modest discriminative accuracy precludes treating a threshold as a reliable prediction for each patient.
A separate systematic review and individual-data analysis by Tralongo et al. (2023) Source 24, drawing on 15 eligible studies, estimated higher mean concentrations among responders and a response-associated threshold using a different dataset and statistical approach. Differences between reported thresholds are scientifically informative: a cutoff is partly an outcome of inclusion rules, population, response definition, measurement schedule, and optimization criterion. The common conclusion concerns exposure-response association, with uncertainty in the exact boundary.
### Metabolism, smoking, and inflammation
Clozapine exposure can change rapidly when the metabolic environment changes. Cigarette-smoke constituents induce CYP1A2 activity, whereas a cessation of smoking may substantially increase clozapine exposure despite unchanged prescribed dose. A 2024 case report of smoking cessation Source 21 directly documented concentration changes over time; its clinical relevance is clear, although one patient cannot establish a universally predictable time course.
Inflammation and infection also demand attention. Elevated inflammatory markers can coincide with altered metabolism, while clozapine itself is associated with clinically important inflammatory adverse reactions. A 2023 systematic review of clozapine-associated inflammation (https://pubmed.ncbi.nlm.nih.gov/37842767/) described the literature on C-reactive protein, cardiac complications, and early surveillance. The mechanism and clinical decision depend on the cause and timing of inflammation; a high clozapine level with fever raises questions different from those of a high level in a patient with stable physical health.
Concomitant inhibitors or inducers of metabolism, dosing history, age, sex-associated physiological distributions, and sampling time can alter the concentration-to-dose relation. Parent clozapine and norclozapine measurements may jointly aid interpretation, but neither the parent-to-metabolite ratio nor concentration alone establishes deliberate nonadherence. Documentation of changes in smoking and infection status is indispensable to longitudinal interpretation.
### Blood counts and the distinction between monitoring systems
TDM of clozapine concentrations serves a different purpose from absolute neutrophil count monitoring. The latter concerns the risk of severe neutropenia; a satisfactory plasma clozapine concentration is no substitute for surveillance of this hematological adverse effect. In a consequential regulatory change, the US Food and Drug Administration removed the clozapine REMS program effective June 13, 2025 Source 25, while emphasizing continued absolute neutrophil count monitoring according to prescribing information. The administrative removal of REMS therefore must never be misread as removal of a clinical safety requirement. Other jurisdictions can operate under different monitoring and access systems.
TDM also does not replace assessment of constipation, cardiopulmonary symptoms, sedation, metabolic effects, seizures, or psychosocial functioning. The complete clozapine-monitoring program combines several complementary measures with different biological targets and clinical thresholds.
Antidepressants: From Tricyclic Windows to Individualized Investigations
### Tricyclic antidepressants and stronger concentration evidence
Tricyclic antidepressants have a longstanding role in TDM because pharmacokinetic variability, active metabolites, and concentration-related adverse effects can make fixed-dose assumptions unreliable. Analytical interpretation may involve measuring parent compound and a metabolite together. For some tricyclic agents, the concentration-response evidence and consensus recommendations are more developed than for many newer antidepressants. This does not make every tricyclic share one reference interval or one toxicological risk.
In their 2026 review of antidepressant TDM Source 22, Schoretsanitis and Paulzen compare evidence across agents, identifying stronger monitoring recommendations for selected tricyclic medicines and lower-confidence recommendations for others. The authors describe use cases involving inadequate response, relapse, adverse reactions, comorbidity, interacting treatments, and populations with altered pharmacokinetics. Their account should be read with the latest AGNP 2026 consensus Source 10, because numerical tables and evidence grades may have been revised.
For a medicine with active metabolites, a single parent concentration can underestimate or misrepresent clinically relevant exposure. Pharmacological interpretation must ask which molecular species the assay reports, whether a combined range is used, and which exposure was linked to efficacy or toxicity in the relevant research. Anticholinergic, cardiovascular, sedative, and neurological adverse effects also require clinical surveillance independent of assay values.
### Newer antidepressants and selective application
Selective serotonin reuptake inhibitors and serotonin–norepinephrine reuptake inhibitors differ in elimination routes, active metabolite formation, exposure variability, and available concentration-response evidence. For many drugs, serum measurement is valuable chiefly in selected circumstances: apparent nonadherence, unexpected pharmacokinetics, pronounced side effects, interacting medicines, or persistent nonresponse despite a plausible trial. It is scientifically unwarranted to infer that routine measurement improves outcomes for all patients treated with every antidepressant.
An undetectable concentration can raise a substantial question about exposure. Yet interpretation depends on assay sensitivity, sample timing, expected trough concentration, time since discontinuation, and the molecule's pharmacokinetics. A detectable concentration confirms some exposure within the assay's capability; it does not by itself prove continuous adherence over weeks. Likewise, an out-of-range level is an indication to investigate, rather than an automatic dose-change command.
### Pharmacogenetic relevance
Some antidepressants exhibit large inherited variation in enzyme-mediated metabolism. The 2023 CPIC guideline led by Bousman and colleagues Source 3 reviews CYP2D6, CYP2C19, and CYP2B6 evidence for serotonin reuptake inhibitor prescribing. Genotypes can inform anticipated metabolism and selection, but they do not directly measure a patient's present concentration. TDM observes the composite consequence of genotype, environment, dose, adherence, physiology, and comedication. Combining these sources of evidence can sharpen an explanation of unusually low or high exposure.
Pharmacodynamic genotype claims deserve their own evidentiary assessment. CPIC concluded that available SLC6A4 and HTR2A evidence did not support their clinical use in antidepressant prescribing. A clinical laboratory should not promote a broad multigene score as an equivalent of verified metabolism-based evidence. The distinction between genotype-driven predictions and measured exposure is fundamental to responsible precision psychiatry.
Valproate and Divalproex: Total Concentration, Free Drug, and Safety
Valproate, including divalproex formulations, is used in some bipolar-disorder treatment regimens, and serum concentration monitoring may help characterize exposure after initiation or changes in treatment. The CANMAT/ISBD summary (https://pmc.ncbi.nlm.nih.gov/articles/PMC11058959/) identifies trough monitoring as part of ongoing use. Interpretation needs to consider the specific product, clinical indication, collection timing, concurrent medication, and response or adverse effects. A conventional total concentration reference band is an imperfect measure of the biologically active fraction.
Valproate is extensively protein bound, and binding can become nonlinear or change with albumin concentration, illness, and other medicines. In an investigation by Doré et al. (2017) Source 7, high unbound concentrations could occur despite total levels that appeared normal or low in patients with hypoalbuminemia, and unbound exposure was relevant to neurological adverse manifestations. This illustrates a measurement problem: the total concentration may be an inaccurate proxy for free, pharmacologically available drug when protein-binding conditions change.
An unexpected clinical presentation with apparently acceptable total concentration therefore warrants consideration of free concentration, albumin status, and other causes of illness. Clinical laboratories must distinguish measured free fractions from calculated estimates, particularly when prediction models are not reliable in medically complex patients. A single conversion ratio cannot reconstruct altered binding across all patients.
Concentration-guided evaluation also exists within a broader safety regime. Potential hepatotoxicity, hematological effects, neurological complications, and important reproductive and developmental risks require investigation and monitoring outside drug-level testing. The European Medicines Agency's valproate risk-minimization measures Source 8 address pregnancy and reproductive risk through regulatory measures separate from routine concentration measurement. A reassuring serum number never cancels these established pharmacological risks.
Carbamazepine: Autoinduction, Interactions, and Clinical Limits
Carbamazepine creates a distinctive monitoring problem through substantial drug-interaction potential and autoinduction of metabolism. The relationship between a fixed dose and concentration evolves after treatment initiation, making early measurement a provisional snapshot. Current DailyMed carbamazepine labeling Source 17 reports a conventional adult therapeutic concentration range of 4–12 micrograms/mL and describes autoinduction over approximately three to five weeks. This labeling is an authoritative product statement, but the clinical interpretation of ranges derived across indications and settings requires particular care when the psychiatric objective is mood stabilization.
Carbamazepine-10,11-epoxide is a pharmacologically active metabolite. Enzyme-inducing and inhibiting comedications can affect carbamazepine and metabolite exposure in different ways, so a parent-compound level alone may not fully explain unexpected neurological symptoms. The relevant questions include assay species, time since regimen change, concomitant antiepileptics or psychotropic agents, clinical response, and adverse effects.
Hematological, hepatic, dermatological, and electrolyte-related complications are monitored through other clinical and laboratory pathways. Serious cutaneous reactions, infection symptoms, and other safety concerns require immediate clinical assessment rather than reassurance from a concentration that lies within a typical reference band. Concentration monitoring is an informative component of medication management, not a comprehensive safety certificate.
Controlled and Observational Valproate Concentration Studies
### Early inpatient concentration-response research
The clinical literature on valproate in acute mania provides a particularly instructive comparison between population thresholds and intervention-specific results. Bowden and colleagues (1996) Source 4 studied 65 hospitalized patients receiving divalproex in a 21-day treatment study with an initially structured dose-escalation design. Their analysis associated day-five concentrations at or above 45 micrograms/mL with a higher probability of at least 20% improvement in several mania symptom domains than concentrations below that threshold. Characteristic adverse experiences occurred disproportionately in those reaching at least 125 micrograms/mL. These values describe observed associations in a defined inpatient population and clinical outcome definition. The statistical relationship cannot establish that 45 micrograms/mL is a universally sufficient concentration for maintenance stabilization or that a particular concentration guarantees remission.
This finding is frequently compressed into a broad valproate “therapeutic range,” obscuring the population and the outcome used to define it. The study examined hospitalized individuals during acute mania, where severity, speed of symptom control, supervised medication administration, and short-term dose adjustment differ from outpatient maintenance treatment. The initial dose structure reduced some variability associated with uncontrolled prescribing, but concentration groups were not themselves independently randomized. Observed response differences may still be influenced by pharmacokinetic and clinical features correlated with concentration.
A different pooled analysis by Allen and colleagues (2006) Source 1 used individual data from three randomized placebo-controlled divalproex studies of acute mania. The secondary concentration-response analysis included 374 participants categorized into six serum concentration groups and a placebo group. It reported greater efficacy than placebo beginning in the 71.4–85.0 micrograms/mL interval, with higher concentration categories showing stronger associations under the authors' analytic model. This result creates a more demanding exposure estimate than the minimum signal described in the 1996 study. The discrepancy is not a reason to erase either study; it calls for close comparison of definitions of improvement, follow-up times, distributions of baseline severity, statistical methods, and whether the objective is demonstrating some benefit or identifying a concentration associated with larger average effects.
### Interpretation of inconsistent thresholds
A population threshold selected because it separates responders from nonresponders is not equivalent to the concentration at which expected treatment benefit is greatest. Lower threshold estimates may identify the onset of detectable efficacy, whereas higher estimates may seek a larger average effect. Methods can also differ in whether they include placebo comparators, how they account for treatment discontinuation, and whether concentration is treated as an ordered category or a continuous exposure. For an individual patient, total versus unbound valproate creates further divergence between the reported value and pharmacologically active exposure.
The valproate literature demonstrates why a monitoring report needs to identify the scientific purpose of its range. The evidence is clinically informative and deserves serious attention, yet it does not establish a single universal value across acute mania, maintenance treatment, antiepileptic indications, pregnancy, hypoalbuminemia, or interacting regimens. Such distinctions are necessary before transferring results across clinical contexts.
Smoking Cessation as a Natural Experiment in Clozapine Pharmacokinetics
### Within-person exposure changes
An abrupt change in tobacco smoking can generate a measurable perturbation in clozapine metabolism. Meyer (2001) Source 15 examined 11 patients receiving stable clozapine doses before and after implementation of a hospital-wide smoking prohibition. Mean concentrations increased by 71.9% after cessation in the observed cohort, but dispersion was substantial, and the author explored a model predicting the nonsmoking level from the earlier smoking level. The scientific importance of this small study lies in its within-person measurements under a clear environmental change. Its numerical model is not an externally validated universal rule for calculating a new regimen.
Cormac and colleagues (2010) Source 5 retrospectively evaluated inpatient clozapine concentrations around a complete smoking ban. The proportion of patients who had smoked and showed a plasma concentration of at least 1,000 micrograms/L increased from 4.2% before the ban to 41.7% after the ban over the observation period despite reported dose reductions. This observation documents a clinically consequential distributional change in a particular inpatient service. It does not separate all potential effects of concurrent illness, medication management, diet, or admission circumstances, and denominators at each measurement point are important when interpreting percentages.
The later case observation by Qi and colleagues (2024) Source 21 illustrates how concentration trajectories can evolve after cessation rather than following a single permanent step change. Taken together, the studies support active attention to smoking status and its temporal relationship with sampling, especially at transitions into smoke-free care. Their different research designs and patient populations preclude translating the largest observed proportional change into a routine assumption for all patients.
### Why the mechanism matters
The relevant exposure in this research is to constituents of tobacco smoke that induce drug-metabolizing enzymes, especially CYP1A2. Nicotine replacement and cigarette smoke are different pharmacological exposures. This distinction affects the explanation a patient receives when a nicotine replacement product is started during a smoke-free hospitalization. The smoking history must describe actual combustible tobacco use, cessation, resumption, and the timing of changes in relation to doses and blood samples.
The clinical research value of serial clozapine monitoring is thus mechanistic as well as practical. Concentration change can corroborate altered exposure after a meaningful external change. The result still requires assessment of infection, inflammation, adherence, interacting drugs, clinical status, and assay timing. Precisely because concentration changes can be large, their interpretation deserves an explicit and consistently recorded exposure history.
The Tricyclic Antidepressant Laboratory Tradition
### Early large clinical laboratory series
A historical laboratory series can show variability that a short drug-efficacy trial is poorly positioned to capture. Dawling (1982) (https://doi.org/10.1016/S0009-9120(82)90511-2) reported 2,141 plasma samples from 1,055 patients treated with amitriptyline or nortriptyline during a three-year period. The laboratory observed extensive between-person variation, including after consideration of dose, and found that subjective adverse-effect reporting did not reliably predict elevated measured levels. This is strong descriptive evidence that administered dose and symptom reports alone may fail to identify unexpected systemic exposure; the retrospective laboratory sample was not a randomized evaluation of concentration-guided treatment.
An American Psychiatric Association Task Force report (1985) Source 2 synthesized the early laboratory and clinical research. It concluded that plasma measurements for imipramine, desmethylimipramine, and nortriptyline were clearly clinically useful in specified circumstances. The report is significant as an institutional historical document and a judgment about clinical utility, but its conclusions must be understood against the measurement techniques, pharmacotherapy landscape, and studies available at that time.
### The challenge of free and total concentrations
Even within the same drug class, competing measurement hypotheses require direct empirical examination. Perry and colleagues (1985) Source 20 measured the unbound fraction of nortriptyline in 18 depressed inpatients whose steady-state total concentrations were maintained within a defined interval. Their statistical analysis did not establish a significant relationship between free concentration and the measured clinical response over 21 days. This small negative study matters because it illustrates the risk of inferring that a mechanistically attractive biomarker necessarily improves prediction in a specific clinical population.
The tricyclic research history therefore includes both positive laboratory observations and uncertain relationships between alternative exposure metrics and outcomes. Class-wide monitoring enthusiasm should be tempered by drug-specific evidence, variation across psychiatric diagnoses and treatment periods, assay differences, and the small samples that characterized many early concentration-response experiments. The question for current research is whether carefully selected concentration measures improve decisions over the contemporary clinical comparator, not whether blood measurement was historically influential.
Lithium Safety and Reproductive Pharmacokinetics: Beyond a Single Therapeutic Interval
### Long-term toxicity and parallel monitoring
The adverse consequences of lithium can occur in physiological systems that serum lithium measurement observes only indirectly. McKnight and colleagues (2012) (https://doi.org/10.1016/S0140-6736(11)61516-X) systematically reviewed 385 studies addressing renal, thyroid, parathyroid, and other potential adverse effects. The synthesis documented clinically meaningful adverse-effect domains, while the design and heterogeneous source studies limited simple prediction of risk for one patient. Their findings support combining serum concentration monitoring with organ-function and symptom assessments. They do not show that a universally selected lithium concentration eliminates long-term renal or endocrine risks.
This distinction is important to causal reasoning. Long-term adverse effects may depend on cumulative exposure, episodes of acute toxicity, duration, comorbid illness, vulnerability, and treatment decisions. A cross-sectional measurement may be entirely correct while offering little information about a patient’s historic exposure profile. Research evaluating long-term lithium harm should account for treatment duration, cumulative burden, clinical indications, frequency and severity of intoxication episodes, and other determinants of organ function. Better longitudinal models are needed to test whether optimizing measured exposure changes long-term outcomes beyond standard care.
### Pregnancy as a longitudinal pharmacokinetic experiment
Wesseloo and colleagues (2017) Source 26 investigated 1,101 lithium measurements across 113 pregnancies in 85 women. This retrospective observational study measured concentrations, doses, creatinine, and temporal changes across pregnancy and the postpartum period. Dose-normalized lithium concentrations fell during early and middle pregnancy, with the largest reported reduction in the second trimester, and then changed toward prepregnancy values as physiology evolved. Their data show how the same treatment can produce shifting systemic exposure across closely spaced periods in one person's life.
The study also helps distinguish a biological signal from a prescribing instruction. A retrospective cohort describes trajectories in a particular group under the treatment decisions made by clinicians. It cannot itself establish the optimal serum concentration for every stage of pregnancy or resolve competing maternal and fetal risks. Individualized specialist decisions require a broader clinical and reproductive safety assessment, including kidney function, risk of mood relapse, dose changes, and maternal preference.
The methodological lesson extends beyond pregnancy. For a physiological transition with a predictable time axis, fixed-interval monitoring may miss clinically important nonlinear changes. The appropriate comparison is an individual's longitudinal exposure and health trajectory, not a series of isolated laboratory values treated as independent observations.
Interpreting Unexpected Concentrations: A Clinical Research Taxonomy
### Unexpectedly low or absent concentrations
Low concentration raises at least four distinct hypotheses: insufficient exposure despite taking the intended dose, incorrect or unstable timing of sampling, a dosing or adherence discrepancy, and an analytical problem. Excessive clearance may result from enzyme induction or particular physiological conditions. A collection performed after a missed dose may be indistinguishable from pharmacokinetic acceleration without an accurate history. Similarly, a specimen collected substantially later than the reference trough can underestimate the intended comparison value. The interpretive question is whether the result documents genuinely inadequate exposure under defined conditions.
Adherence should be assessed with care. Psychiatric treatment may be interrupted because of adverse effects, cost, disrupted routines, personal beliefs, difficulty obtaining prescriptions, concerns about identity and autonomy, or uncertainty about benefit. A laboratory result provides information about exposure and supports a collaborative discussion; it cannot determine intention or moral responsibility. Repeated unexpected results can motivate closer investigation, but medication possession histories, clinician reports, and electronic records are not equivalent to direct observation of ingestion.
### Unexpectedly elevated concentrations
Elevated concentrations invite a review of recent dose changes, accidental duplicate administration, reduced hepatic or renal elimination, drug interactions, inflammation, smoking cessation, altered formulation, and altered specimen timing. In a patient with concerning symptoms, clinical urgency is determined by the whole presentation. Some medicines have toxic effects whose severity is poorly represented by an isolated plasma concentration, particularly when exposure is rapidly changing or tissue distribution lags behind blood values.
An elevated value obtained at an unexpected postdose time should not automatically be categorized as sustained excessive trough exposure. Repeating a correctly timed specimen can clarify ambiguity when medically appropriate, while severe symptoms require immediate care. The laboratory's critical-value policies, drug-specific alerts, and the local emergency escalation procedure should be explicit. Reference concentrations intended for maintenance therapy cannot be reused as a universal toxicity algorithm.
### Apparently adequate exposure with inadequate response
A concentration within an accepted therapeutic range can strengthen confidence that medication reached the circulation over the sampled interval, but it does not verify receptor-level exposure or sufficient treatment duration. Apparent treatment resistance may involve diagnostic heterogeneity, pharmacodynamic insensitivity, psychiatric comorbidity, substance use, psychosocial stress, inadequate duration, or clinically meaningful adverse effects that limit effective treatment. In clozapine, concentration assessment is especially consequential before making irreversible conclusions about persistent treatment resistance. For drugs lacking strong exposure-response evidence, a laboratory value contributes more modestly to the differential explanation.
The correct scientific interpretation is a posterior assessment: given the recorded dose and clinical context, what explanations remain plausible after observing this concentration? That inferential framing makes explicit the value of the test, its dependence on pretest assumptions, and the uncertainty in translating a group-level reference into an individual outcome.
Special Populations and Changing Biological Conditions
### Older adults and organ impairment
Age may coincide with altered renal clearance, hepatic function, body composition, protein binding, polypharmacy, and vulnerability to adverse effects. Chronological age alone is a limited predictor of exposure; specific physiological measures carry more information. In lithium therapy, renal function and comedications are central determinants of whether a previously stable regimen remains safe. Older adults may require different individualized therapeutic goals, but changes should follow evidence from age-relevant populations rather than applying adult cutoffs without examination.
Kidney impairment can increase exposure to renally cleared medicines or active metabolites. Liver impairment can alter metabolism, protein synthesis, or clearance; the effect is drug specific. Low albumin is particularly relevant when a measured total concentration is used to infer the active free fraction. These considerations make drug-level testing a dynamic measure of physiology and pharmacotherapy together.
### Pregnancy, postpartum changes, and reproductive safety
Pregnancy can alter distribution volume, renal filtration, hepatic metabolism, and protein binding. The clinical implications differ across drugs and gestational stages, and postpartum physiology can change rapidly. Monitoring may inform dose decisions in specialist care, but a concentration cannot evaluate embryonic, fetal, or developmental safety. For valproate, reproductive risk prevention and legal or regulatory requirements must be considered in their own right. The EMA safety framework Source 8 illustrates why some safety decisions depend on risks that cannot be reduced to drug concentration.
### Children and adolescents
Developmental differences in metabolism, growth, organ function, and indications complicate extrapolation from adult reference ranges. Ethical constraints and less extensive trial evidence can make pediatric concentration-response conclusions uncertain. A blood concentration may still answer a clinically important question about unusually high or low exposure, but pediatric use requires developmentally appropriate evidence and specialist interpretation. Any normative claim should identify whether its supporting dataset actually included the relevant age group.
### Acute medical illness and inpatient transitions
Hospital admission can change tobacco use, medication adherence, diet, hydration, renal status, co-prescribing, and inflammatory physiology simultaneously. The resulting laboratory findings are frequently multicausal. Acute infection may elevate clozapine exposure, while cessation of cigarette smoking can further modify metabolism; separating these effects may require serial measures and careful documentation. Transitions between inpatient and outpatient care are especially vulnerable to lost information about the dose, formulation, specimen time, and recent treatment changes.
Pharmacogenetics, Quantitative Models, and Precision Approaches
Pharmacogenetics estimates one subset of the causes of exposure variability. Genotypes relevant to drug-metabolizing enzymes can be used before treatment to predict a metabolizer phenotype, whereas TDM observes net exposure after the patient has received medication. These methods offer complementary evidence and are particularly informative when an exposure result conflicts with the phenotype predicted from genotype alone. Enzyme inhibition can create phenoconversion, in which observed metabolism differs from genetically anticipated activity.
The CPIC antidepressant guideline (Bousman et al., 2023) Source 3 demonstrates the necessity of specific gene–drug relationships. CYP2D6 and CYP2C19 information may help interpret some antidepressant choices, while evidence for selected pharmacodynamic candidate genes remains insufficient for routine clinical prescribing. The AGNP 2026 update Source 10 incorporates pharmacogenetic guidance alongside its reworked drug concentration tables. Together the sources support a combined exposure-and-context approach without warranting claims that every genomic panel reliably predicts treatment outcomes.
Population pharmacokinetic models can estimate the distribution of likely concentrations under specified covariates. Bayesian therapeutic monitoring can use prior population distributions updated by an individual's observed concentration and dosing information. Its performance depends on parameter identification, sampling fidelity, model validation, relevant populations, and communication of predictive uncertainty. A model built on samples from stable adult outpatients may not predict an adolescent or acutely ill patient accurately. Dosing software deserves prospective outcome evaluation rather than validation through internal fit alone.
Artificial computational tools may assist with temporal reconstruction of doses and samples, detection of data inconsistencies, or presentation of concentration trajectories. Under the Artificial Era research horizon, the psychologically important issue is how decision-making responsibility and explanatory understanding are distributed between clinician, patient, laboratory, and artificial systems. A computational recommendation should record its inputs, model version, uncertainty, and limits of generalization. The clinical basis for acceptance remains a verifiable patient-specific interpretation, rather than an automated numerical instruction.
Clinical Decision Research and the Problem of Utility
### Analytical validity, clinical validity, and clinical utility
These three levels must be examined separately. Analytical validity asks whether the laboratory reports the intended molecular concentration reliably. Clinical validity asks whether that concentration predicts or helps discriminate an outcome in the population of interest. Clinical utility asks whether acting on the result improves outcomes compared with a reasonable alternative process of care. A laboratory may achieve superb analytical performance for a compound with weak response prediction. A response-associated concentration range may exist without a trial demonstrating that routine monitoring improves long-term health.
Evidence of utility also depends on implementation. A laboratory result returned too late to influence an acute decision offers limited benefit. A clinician unfamiliar with sampling assumptions may misinterpret a reliable result. A health system without follow-up processes may repeatedly order assays without learning from them. Evaluating TDM therefore requires both drug-specific evidence and service-level evaluation.
### Study designs and causal inference
Randomized comparison of monitoring strategies can establish whether concentration-guided care improves predefined outcomes, provided testing and treatment differences are adequately specified. Outcomes should include adverse drug events, symptom response, relapse, hospitalization, treatment discontinuation, functioning, and patient-reported burden. Blinding and allocation concealment may be challenging because result-driven decisions are visible; transparent outcome assessment and preregistration become important.
Cohort designs can explore exposure-response relations under real-world treatment. However, dose escalation often follows inadequate response; higher concentrations may therefore appear among sicker patients because of prescriber behavior. Conditioning on continued treatment can also introduce selection bias. Longitudinal analysis must account for changes in dosing, illness severity, and comedication across time. Cross-sectional concentration data are particularly vulnerable to reverse causality and unmeasured adherence differences.
The lithium randomized comparison by Gelenberg et al. (1989) Source 9 remains informative because of its assigned concentration strategy. Yet its specific protocol, era, population, and competing tolerability outcomes constrain generalization. The Nolen et al. (2019) synthesis Source 18 clarifies why extrapolation to contemporary practice should acknowledge limitations rather than present one historical trial as final resolution.
### Predictive performance of plasma thresholds
In the clozapine individual-data meta-analysis, an area under the curve near 0.61 means observed concentrations discriminate responders from nonresponders only modestly. This fact belongs at the center of interpretation, alongside the existence of a meaningful association. A threshold can support a sensible intervention while leaving many people on either side whose outcomes diverge from the model's binary prediction. Continuous response probabilities and calibrated individualized models may be more informative than treating an empirical threshold as a clinical law.
Statistical and clinical significance also differ. A small mean exposure difference could be statistically robust yet have limited value for a particular patient. Conversely, identifying an extreme concentration or serious interaction may change a patient's safety management even without a randomized trial covering that exact scenario. Evidence grading needs to reflect the decision at hand.
Behavioral, Psychological, and Relational Dimensions of Monitoring
### Adherence, motivation, and therapeutic relationships
The effectiveness of a monitoring program depends partly on how it is explained and experienced. Frequent blood testing can provide reassurance and a sense of active partnership, but it can also create burden, anxiety, expense, and avoidance, especially when the person sees no connection between tests and well-being. Clinician communication should explain which question the test answers, why the sampling time matters, how results will be discussed, and whether any follow-up is expected.
Concentration measurement can clarify an exposure discrepancy without resolving the reasons behind it. Assessing preferences, cognitive obstacles, adverse experiences, financial access, and shared goals is scientifically relevant to the interpretation of an apparent adherence problem. A nonjudgmental discussion can reveal deliberate dose reduction due to side effects that would otherwise be concealed, while concentration alone would offer a misleading appearance of pharmacokinetic failure.
### Communication of risk and uncertainty
Therapeutic ranges frequently look precise because they are printed as exact numbers. Their evidentiary meaning is usually more conditional. A patient deserves an explanation that a therapeutic interval summarizes research across populations and that the personal response may fall outside it. When a result is concerning, clinicians should discuss immediate clinical priorities and follow-up plainly; when the result is reassuring, they should explain which dangers are still assessed by other forms of monitoring.
The quality of a service can be investigated through patient understanding of the purpose of tests, continuity of results across clinics, missed appointments, treatment retention, and reported trust. These are empirical research questions. They should not be presented as demonstrated effects of TDM without studies that actually measured them.
Implementation Across Laboratories, Clinics, and Health Systems
A complete monitoring service requires an agreed set of analytes, validated analytical methods, sample timing instructions, clinical indication fields, interpretive reports, critical-result escalation, and a documented recipient responsible for acting on results. Outpatient and inpatient systems must exchange enough treatment history to make concentrations comparable. The quality of the underlying record can matter as much as instrument precision.
Reference tables should be dated, drug-specific, and traceable to their original evidence. They should state when ranges refer to parent compounds, total active moieties, or particular metabolites, and when samples are expected to be collected. A report should identify units and avoid ambiguous conversions. Where measurement is calibrated and standardized, service audit can evaluate inappropriate or mistimed orders, uninterpretable samples, delayed reports, critical results, documented follow-up, and clinically meaningful outcomes.
International variation in clinical infrastructure and regulation affects practice. A consensus recommendation may be widely influential without being legally binding everywhere. Availability and reimbursement of assays, specialist laboratory access, and locally approved indications determine feasible implementation. An international scientific encyclopedia can describe these differences without imposing one jurisdiction's prescribing framework on its global readership.
Critical Scientific Discussion
The greatest strength of TDM lies in directly observing exposure that prescription records only approximate. Its principal weakness arises when the observation is mistaken for the whole treatment process. Between concentration and outcome lie complex and heterogeneous pathways: receptor binding, active metabolites, dynamic adaptation, brain distribution, inflammatory changes, behavior, symptom measurement, comorbidity, and personal meaning. For medicines with strong exposure-response foundations, levels can considerably improve the quality of decisions; for others, they are selective tools for clinical investigation.
The available evidence contains several distinct forms of uncertainty. Concentration-response research may not control prescribing behavior and illness severity. Published ranges can originate in mixed diagnostic populations with inconsistent specimen timing. Analytical techniques may differ across eras and sites. Outcome definitions vary from symptom reduction to remission or prevention of relapse, while toxicity analyses often use differing severity thresholds. A consensus table aggregates such information into useful practical guidance, but it does not remove these methodological differences.
There is also a distinction between measuring a medicine and evaluating the clinical act of measurement. Benefits of TDM may depend on the speed of reporting, expertise in interpretation, accessibility of laboratory services, and responsiveness to clinical change. A drug with a well-established pharmacokinetic profile can be managed poorly when results are ordered at the wrong time. A carefully designed monitoring program can provide valuable evidence even when the concentration is not sufficiently predictive to prescribe a precise universal range.
No synthesis of the field should promise that optimization of serum concentrations ensures remission or eliminates toxicity. Empirical improvements in prediction, monitoring intervals, risk identification, and follow-up require prospective studies with suitable comparative designs. Particular attention is warranted for older adults, pregnancy, medically complex patients, populations underrepresented in development studies, and non-Western health systems. Research that makes the monitoring workflow itself observable would strengthen the field beyond another catalogue of drug ranges.
Research Priorities and Testable Questions
The most important unresolved question is when measuring and acting on a concentration improves patient-centered outcomes compared with appropriately supported clinical care. Trials should compare explicit monitoring algorithms, include safety and symptom outcomes, and account for cost, patient burden, and clinical context. Analyses of subgroup effects should be specified in advance and avoid assuming that extreme concentration values alone prove one mechanistic diagnosis.
A second priority is the construction of reproducible, externally validated exposure-response models. These should report continuous risks, predictive uncertainty, assay characteristics, and calibration in independent datasets. Clozapine presents a strong test case because the association with response is meaningful yet individual discriminatory performance remains limited.
A third priority concerns transitions and dynamic states: infection, smoking changes, organ impairment, drug interactions, formulation switches, and pregnancy. Repeated concentration measurements, carefully recorded doses, and longitudinal clinical outcomes can help distinguish pharmacokinetic change from confounding and disease fluctuation. Research should explicitly separate clinical monitoring biomarkers from safety measures that assess other physiological systems.
Finally, interdisciplinary investigation should examine how laboratory explanations are translated into actual shared decisions. Studies of communication, patient comprehension, access inequalities, adherence mechanisms, and digital decision support would connect the pharmacological and psychological components of psychiatric treatment. Transparent computational assistance may improve timeliness and consistency when supported by prospective evaluation, auditability, and clinician oversight.
Conclusion
Therapeutic drug monitoring in psychiatry is a systematic process for transforming measured blood concentrations into clinically contextualized evidence about drug exposure. Its importance is clearest where concentration provides information materially relevant to benefit, toxicity, or an otherwise unresolved treatment question. Lithium, clozapine, selected tricyclic antidepressants, and several mood-stabilizing medicines illustrate different pathways through which an assay can contribute to treatment. Their evidentiary foundations and safety requirements cannot be collapsed into one universal therapeutic window.
The available clinical literature supports a graduated, drug-specific approach. Reliable sampling and analytical measurement form the foundation; reference ranges require transparent derivation; interpretation incorporates dose timing, active metabolites, physiological variation, interactions, and longitudinal clinical outcomes. Genotype and pharmacokinetic models can contribute explanatory information without replacing measured exposure or the clinical assessment. A psychiatric monitoring program is successful to the extent that its test results support better, safer, and more understandable treatment decisions.
The scientific horizon of the field extends toward externally validated individualized prediction, equitable implementation, and collaborative interpretation in increasingly computational systems of care. The central finding remains stable across those developments: a serum concentration is an observation of pharmacological exposure whose therapeutic meaning emerges through the evidence and conditions connecting it to a particular person.
References
- Allen, M. H., Hirschfeld, R. M., Wozniak, P. J., Baker, J. D., & Bowden, C. L. (2006). Linear relationship of valproate serum concentration to response and optimal serum levels for acute mania. *The American Journal of Psychiatry, 163*(2), 272–275. https://doi.org/10.1176/appi.ajp.163.2.272
- American Psychiatric Association, Task Force on the Use of Laboratory Tests in Psychiatry. (1985). Tricyclic antidepressants—blood level measurements and clinical outcome: An APA Task Force report. *The American Journal of Psychiatry, 142*(2), 155–162. https://doi.org/10.1176/ajp.142.2.155
- Bousman, C. A., Stevenson, J. M., Ramsey, L. B., Sangkuhl, K., Hicks, J. K., Strawn, J. R., Singh, A. B., Ruaño, G., Mueller, D. J., Tsermpini, E. E., Brown, J. T., Bell, G. C., Leeder, J. S., Gaedigk, A., Scott, S. A., Klein, T. E., Caudle, K. E., & Bishop, J. R. (2023). Clinical Pharmacogenetics Implementation Consortium (CPIC) guideline for CYP2D6, CYP2C19, CYP2B6, SLC6A4, and HTR2A genotypes and serotonin reuptake inhibitor antidepressants. *Clinical Pharmacology & Therapeutics, 114*(1), 51–68. https://doi.org/10.1002/cpt.2903
- Bowden, C. L., Janicak, P. G., Orsulak, P., Swann, A. C., Davis, J. M., Calabrese, J. R., Goodnick, P., Small, J. G., Rush, A. J., Kimmel, S. E., Risch, S. C., & Morris, D. D. (1996). Relation of serum valproate concentration to response in mania. *The American Journal of Psychiatry, 153*(6), 765–770. https://doi.org/10.1176/ajp.153.6.765
- Cormac, I., Brown, A., Creasey, S., Ferriter, M., & Huckstep, B. (2010). A retrospective evaluation of the impact of total smoking cessation on psychiatric inpatients taking clozapine. *Acta Psychiatrica Scandinavica, 121*(5), 393–397. https://doi.org/10.1111/j.1600-0447.2009.01482.x
- Dawling, S. (1982). Monitoring of tricyclic antidepressant therapy. *Clinical Biochemistry, 15*(1), 56–61. https://doi.org/10.1016/S0009-9120(82)90511-2
- Doré, M., San Juan, A. E., Frenette, A.-J., & Williamson, D. (2017). Clinical importance of monitoring unbound valproic acid concentration in patients with hypoalbuminemia. *Pharmacotherapy, 37*(8), 900–907. https://doi.org/10.1002/phar.1965
- European Medicines Agency. (2018). *Valproate and related substances: Referral and safety measures*. https://www.ema.europa.eu/en/medicines/human/referrals/valproate-related-substances-0
- Gelenberg, A. J., Kane, J. M., Keller, M. B., Lavori, P., Rosenbaum, J. F., Cole, K., & Lavelle, J. (1989). Comparison of standard and low serum levels of lithium for maintenance treatment of bipolar disorder. *The New England Journal of Medicine, 321*(22), 1489–1493. https://doi.org/10.1056/NEJM198911303212201
- Hart, X. M., Gründer, G., Hiemke, C., Adamovic, I., Amann, F., Baumann, P., Bergemann, N., Brandt, C., Clement, H.-W., Conca, A., Deckert, J., Dietmaier, O., Eckermann, G., Egberts, K., Endres, K., Erlacher, N., Fekete, S., Ferrea, S., Florio, V., … Schoretsanitis, G. (2026). Consensus guidelines for therapeutic drug monitoring in neuropsychopharmacology: Update 2026—Pharmacokinetic, pharmacogenetic and clinical aspects. *Pharmacopsychiatry, 59*(4), 166–295. https://doi.org/10.1055/a-2860-7861
- Hiemke, C., Bergemann, N., Clement, H.-W., Conca, A., Deckert, J., Domschke, K., Eckermann, G., Egberts, K., Gerlach, M., Greiner, C., Gründer, G., Haen, E., Havemann-Reinecke, U., Hefner, G., Helmer, R., Janssen, G., Jaquenoud, E., Laux, G., Messer, T., … Baumann, P. (2018). Consensus guidelines for therapeutic drug monitoring in neuropsychopharmacology: Update 2017. *Pharmacopsychiatry, 51*(1–2), 9–62. https://doi.org/10.1055/s-0043-116492
- Keramatian, K., Chithra, N. K., & Yatham, L. N. (2023). The CANMAT and ISBD guidelines for the treatment of bipolar disorder: Summary and a 2023 update of evidence. *Focus, 21*(4), 344–353. https://doi.org/10.1176/appi.focus.20230009
- Leung, J. G., Zhang, L., Markota, M., Ellingrod, V. L., Gerberi, D. J., & Bishop, J. R. (2023). A systematic review of clozapine-associated inflammation and related monitoring. *Pharmacotherapy, 43*(12), 1364–1396. https://doi.org/10.1002/phar.2887
- McKnight, R. F., Adida, M., Budge, K., Stockton, S., Goodwin, G. M., & Geddes, J. R. (2012). Lithium toxicity profile: A systematic review and meta-analysis. *The Lancet, 379*(9817), 721–728. https://doi.org/10.1016/S0140-6736(11)61516-X
- Meyer, J. M. (2001). Individual changes in clozapine levels after smoking cessation: Results and a predictive model. *Journal of Clinical Psychopharmacology, 21*(6), 569–574. https://doi.org/10.1097/00004714-200112000-00005
- National Institute for Health and Care Excellence. (2014, updated subsequently). *Bipolar disorder: Assessment and management (Clinical guideline CG185).* https://www.nice.org.uk/guidance/cg185/chapter/recommendations
- National Library of Medicine. (n.d.). *Carbamazepine extended-release tablets: Prescribing information*. DailyMed. https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=d3cd36c4-b25e-4716-be1d-0b2c83ea6a03
- Nolen, W. A., Licht, R. W., Young, A. H., Malhi, G. S., Tohen, M., Vieta, E., Kupka, R. W., Zarate, C. A., Nielsen, R. E., Baldessarini, R. J., Severus, E., & ISBD/IGSLI Task Force on the Treatment with Lithium. (2019). What is the optimal serum level for lithium in the maintenance treatment of bipolar disorder? A systematic review and recommendations from the ISBD/IGSLI Task Force on treatment with lithium. *Bipolar Disorders, 21*(5), 394–409. https://doi.org/10.1111/bdi.12805
- Northwood, K., Pearson, E., Arnautovska, U., Kisely, S., Pawar, M., Sharma, M., Vitangcol, K., Wagner, E., Warren, N., & Siskind, D. (2023). Optimising plasma clozapine levels to improve treatment response: An individual patient data meta-analysis and receiver operating characteristic curve analysis. *The British Journal of Psychiatry, 222*(6), 241–245. https://doi.org/10.1192/bjp.2023.27
- Perry, P. J., Browne, J. L., Alexander, B., Pfohl, B. M., Dunner, F. J., Sherman, A. D., & Tsuang, M. T. (1985). Relationship of free nortriptyline levels to therapeutic response. *Acta Psychiatrica Scandinavica, 72*(2), 120–125. https://doi.org/10.1111/j.1600-0447.1985.tb02582.x
- Qi, L., Ma, B., Fan, H., Qi, S., Yang, F., & An, H. (2024). Case report: Time response of plasma clozapine concentrations on cessation of heavy smoking. *Frontiers in Pharmacology, 15*, 1408915. https://doi.org/10.3389/fphar.2024.1408915
- Schoretsanitis, G., & Paulzen, M. (2026). Utility of therapeutic drug monitoring in the treatment of major depressive disorder. *Current Opinion in Psychiatry, 39*(1), 1–7. https://doi.org/10.1097/YCO.0000000000001046
- Siskind, D., Sharma, M., Pawar, M., Pearson, E., Wagner, E., Warren, N., & Kisely, S. (2021). Clozapine levels as a predictor for therapeutic response: A systematic review and meta-analysis. *Acta Psychiatrica Scandinavica, 144*(5), 422–432. https://doi.org/10.1111/acps.13361
- Tralongo, F., Konecki, C., Feliu, C., Kaladjian, A., & Djerada, Z. (2023). Association between clozapine plasma concentrations and treatment response: A systematic review, meta-analysis and individual participant data meta-analysis. *Clinical Pharmacokinetics, 62*(6), 807–818. https://doi.org/10.1007/s40262-023-01247-1
- U.S. Food and Drug Administration. (2025). *FDA removes Risk Evaluation and Mitigation Strategy (REMS) program for the antipsychotic drug clozapine*. https://www.fda.gov/drugs/drug-safety-and-availability/fda-removes-risk-evaluation-and-mitigation-strategy-rems-program-antipsychotic-drug-clozapine
- Wesseloo, R., Wierdsma, A. I., van Kamp, I. L., Munk-Olsen, T., Hoogendijk, W. J. G., Kushner, S. A., & Bergink, V. (2017). Lithium dosing strategies during pregnancy and the postpartum period. *The British Journal of Psychiatry, 211*(1), 31–36. https://doi.org/10.1192/bjp.bp.116.192799