Published on 23 September 2026

Analysis · Bipolar disorder · Psychopharmacology

Lamotrigine and depression: real prevention in bipolar disorder, a marginal acute effect, no evidence in unipolar depression

◆ Collection
Pharmaceuticals · 2025; 18(10) · Arnone et al.
DOI 10.3390/ph18101590
PMID 41155702
Scientific 72
Editorial 79

Editorial collections Myth or Reality? A Study Under the Microscope

The essentials

This synthesis screened 3521 references, retained 32 eligible randomized trials, and pooled the 24 double-blind trials into the meta-analysis, for 2257 patients on lamotrigine and 2320 on a comparator cumulated across all analyses, unpublished trials included. The message is double and asymmetric. In bipolar maintenance treatment, lamotrigine reduces the emergence of depressive symptoms (RR 0.78, 95% CI 0.63 to 0.98) and prolongs the time spent free of depressive symptoms (RR 1.59, 95% CI 1.19 to 2.11). In the acute phase of bipolar depression, an effect exists but stays marginal (SMD 0.155, 95% CI 0.005 to 0.305), an effect size the authors themselves describe as small. In unipolar depression, no benefit is demonstrated, neither in monotherapy (SMD 0.10, 95% CI −0.20 to 0.41) nor as an add-on to an antidepressant (RR 0.95, 95% CI 0.73 to 1.23): this is an absence of demonstration, which is not the same as a demonstration of inefficacy. Major limitation: the acute-phase add-on analysis pools trials that are incompatible with one another (I² 96.7%) and cannot be meaningfully read.

Context

Lamotrigine is an anticonvulsant that has become, in routine psychiatric practice, a maintenance treatment for bipolar disorder. The question that comes up in consultation is not whether it works, but when it works: does it prevent depressive relapse in a stabilized patient, does it relieve the episode currently underway, and does it have a place in resistant unipolar depression, where it is sometimes added to an antidepressant? These three situations are routinely conflated in clinical discussion even though they do not rest on the same level of evidence.

Regulatory context, worth checking wherever you practice. Marketing authorization, commercial availability, and reimbursement are three separate questions, and a meta-analysis answers none of them directly: each jurisdiction sets its own rules, and a finding of efficacy in trials does not automatically settle any of the three. The authors situate this in their own introduction: the United States Food and Drug Administration currently supports lamotrigine for maintenance treatment of bipolar I disorder, while in Europe it is approved for preventing the emergence of depressive symptoms in bipolar depression, not for treating an episode already underway. The French regulatory file offers a concrete, checkable illustration of the same structural point. Lamotrigine’s psychiatric indication in France, as worded in the product information published by the national medicines agency and checked on August 13, 2026, reads as prevention of depressive episodes in adult patients with bipolar I disorder who have a predominance of depressive episodes, and the same document states explicitly that the product is not indicated for the acute treatment of manic or depressive episodes. Unipolar depression does not appear among the approved indications there: using lamotrigine in that situation is an off-label prescription, with the information and documentation duties that this carries under French law. Reimbursement status, a further and distinct question, was not checked for this article. Readers practicing outside France should verify the equivalent authorization, availability, and reimbursement rules in their own jurisdiction before transposing any of this: they will very plausibly differ, even though the trial evidence discussed below is the same everywhere.

Safety reminder, drawn from the prescribing framework rather than from this meta-analysis. Lamotrigine carries a risk of severe cutaneous reactions, including Stevens-Johnson syndrome and toxic epidermal necrolysis, and of drug reaction with eosinophilia and systemic symptoms. These reactions cluster overwhelmingly in the first eight weeks of treatment, and the risk rises with an initial dose that is too high and with concomitant valproate, which calls for the lamotrigine dose to be reduced substantially. The slow titration schedule set out in prescribing information is not an optional precaution: it is the principal measure that lowers this risk, and it is one reason acute-phase use sits awkwardly in practice, since a therapeutic dose is reached only after several weeks. Any rash occurring on lamotrigine warrants prompt evaluation and, in most circumstances, discontinuation, unless another cause is clearly established. Exact titration schedules and dose adjustments vary from one regulator to another, so readers should consult the prescribing information in force in their own country rather than assume the French schedule applies unchanged. The meta-analysis under review was not designed to study tolerability, but it is not silent on the point: it reports a skin-rash frequency ranging from 2% to 9% depending on the analysis, and no cases of Stevens-Johnson syndrome across the pooled trials.

Mechanism

Why a preventive effect might exceed a curative one

Element	Status
Pharmacological target	Textbook pharmacology, outside the scope of the trials
FindingThe publication describes lamotrigine as an inhibitor of voltage-gated sodium channels, licensed for partial seizures, primary generalized tonic-clonic seizures, and Lennox-Gastaut syndrome. The resulting reduction in presynaptic glutamate release is standard pharmacology textbook material. None of the trials pooled in this synthesis measured this mechanism directly.
Clinical hypothesis	Consistent with the results, not tested
FindingThe idea that a drug stabilizing neuronal excitability would prevent recurrence more effectively than it resolves an episode already underway fits the observed profile. This synthesis does not test that idea: there is no mediation analysis and no intermediate biomarker.
Dose and exposure	Wide range, dose-response relationship unexplored
FindingTarget doses across the pooled trials range from 100 to 500 mg per day, with 200 to 400 mg per day in the acute bipolar phase, 200 to 300 mg per day as an add-on, and up to 400 to 500 mg per day in maintenance. The meta-regression covered publication year, age, number of women, number of rapid-cycling patients, age at diagnosis, episode duration, depression severity, number of suicide attempts, and study duration. Neither dose nor plasma concentration appears among these covariates, and the authors state that they could not run sub-analyses because data across the studies were inconsistent and limited. No conclusion about an optimal dose is therefore possible, in either direction.

The study at a glance

Population, intervention, comparator, outcomes
Population
Adults with unipolar depression or bipolar I or II disorder, diagnosed according to DSM-IV, DSM-IV-TR, or ICD-10. Studies with fewer than ten participants were excluded.
Intervention
Lamotrigine, target doses from 100 to 500 mg per day depending on the trial, given as monotherapy or as an add-on, in both the acute phase and maintenance.
Comparators
Placebo, lithium, olanzapine combined with fluoxetine, quetiapine, lithium combined with divalproate, desipramine, selective serotonin reuptake inhibitors.
Outcomes
In the acute phase, continuous change on depression rating scales, preferred over response rates to preserve statistical power. In maintenance, the rate of emergence of depressive symptoms and the duration of time free of symptoms. The add-on analysis in unipolar depression instead relies on response rates.
Design
Systematic review and meta-analysis: 3521 references screened, search conducted up to June 2024, 32 eligible trials, 24 double-blind trials retained for the meta-analysis, including unpublished trials from the manufacturer’s registry. Random-effects model in STATA 18.0, leave-one-out sensitivity analysis, meta-regression, Egger’s test.
Sample sizes
2257 on lamotrigine and 2320 on a comparator overall, but this total adds together each analysis’s own subtotal: two datasets contribute to two separate analyses, so their participants are counted twice. By analysis: acute bipolar monotherapy 531 versus 516; add-on 188 versus 187; maintenance versus placebo 365 versus 276; maintenance versus lithium 350 versus 242; unipolar monotherapy 395 versus 394; unipolar add-on 81 versus 78. The publication does not consistently distinguish randomized from analyzed samples; the acute bipolar trials are described as analyzed by intention to treat with last observation carried forward.

Quality control

Point checked	Verdict
Protocol and review reporting	PROSPERO CRD42025633709, PRISMA
FindingRegistration under number CRD42025633709, reporting compliant with PRISMA, its PRISMA-S extension for the search strategy, and the Cochrane Handbook. The publication does not state the date the protocol was filed.
Search for unpublished trials	Manufacturer’s registry searched
FindingThe molecule’s historic manufacturer’s trial registry was searched on May 21, 2025, in addition to six bibliographic databases. Seven of the datasets in Table 1 come from that registry, four in acute bipolar depression and three in unipolar depression, the latter three explicitly described as unpublished. This is the principal strength of the work: the published literature alone did not give the same picture.
Risk of bias and blinding	One high-risk trial, three without double blinding
FindingThe RoB 2 tool was applied by two independent assessors. The overall quality of the studies is described as generally high, with a low or medium risk of bias, with one exception: the Schindler and Anghelescu trial, rated high risk. More importantly, inclusion in the meta-analysis was meant to be restricted to double-blind trials, and three studies in Table 1 do not meet that bar: the Suppes trial, single-blind, and the Licht and Schindler-Anghelescu trials, open-label. The Licht trial enters the pooled lamotrigine-versus-lithium maintenance comparison; the other two are reported only in the narrative text.
Heterogeneity between trials	I² at 96.7% and 73.7% on two analyses
FindingHeterogeneity is low on the maintenance analyses (I² 25.6% and 0%) and on acute bipolar monotherapy (I² 33.6%). It reaches 96.7% on the acute bipolar add-on analysis and 73.7% on unipolar monotherapy, and no meta-regression covariate explains either. At 96.7%, the pooled average no longer describes anything common to the trials it aggregates.
Publication bias	Egger’s test not significant
FindingNo asymmetry detected in any analysis, p values ranging from 0.152 to 0.56. This result is not very informative when the number of trials per analysis is small, and it does not substitute for the inclusion of unpublished data, which remains the real safeguard used here.
Internal consistency of the report	Three mismatches between text and table
FindingThe Suppes trial’s target dose is given as 400 mg per day in Table 1 and as 200 mg per day in the text. The number of women in the unipolar monotherapy trials is 236 in the text and 342 in the table, the latter crediting the Santos trial with 171 women out of 34 participants in total. The forest plot for acute bipolar monotherapy displays a rounded estimate of 0.16, interval 0.00 to 0.31, where the text gives 0.155, interval 0.005 to 0.305. None of this overturns the conclusions, but the copyediting was not as careful as the analysis.
Funding and conflicts of interest	No external funding, ties declared
FindingThe publication declares no external funding and no new data. The first author declares sponsorships from Janssen-Cilag, Servier, Lundbeck, and Viatris; the last author declares paid lectures, advisory board roles, and principal-investigator roles across a large number of companies, plus editorial roles and numerous institutional and industry research grants; the other authors declare no conflicts. The authors state they have no ties to the trials included in this work. The molecule evaluated is an old generic, which limits the direct commercial stake without erasing the question.

Results

0.78
Relative risk of emergence of depressive symptoms under lamotrigine in bipolar maintenance treatment, 95% CI 0.63 to 0.98, a 22% risk reduction. On the same molecule, in the acute phase, the effect shrinks to a standardized difference of 0.155, and in unipolar depression it is not demonstrated at all.

Analysis	Result
Bipolar maintenance, emergence of depressive symptoms	RR 0.78, 95% CI 0.63 to 0.98
FindingThree trials, 365 patients versus 276, about fourteen months of average follow-up. The authors translate this into a 22% risk reduction. The signal is consistent with the molecule’s prophylactic use, and its precision stays modest: the upper bound nearly touches 1.
Bipolar maintenance, duration free of depressive symptoms	RR 1.59, 95% CI 1.19 to 2.11
FindingA result convergent with the previous one, on a duration outcome, without heterogeneity (I² 0%). It rests on the same three trials: this is two views of one signal, not two independent demonstrations. Worth noting, this time-based outcome is expressed as a relative risk rather than a hazard ratio, which limits how far it can be read in terms of time to relapse.
Acute bipolar depression, monotherapy	SMD 0.155, 95% CI 0.005 to 0.305
FindingFive trials, 531 patients versus 516, about 8.2 weeks. Statistically significant, clinically trivial: the authors themselves note that 0.2 is conventionally regarded as a small effect, conclude that efficacy is low, and describe a large overlap between the distributions. The lower bound nearly touches zero.
Acute bipolar depression, add-on	SMD 0.88, 95% CI −0.41 to 2.18, I² 96.7%
FindingThis estimate cannot be meaningfully interpreted, and it is not statistically significant. It aggregates three trials whose individual effect sizes range from 2.13 (95% CI 1.79 to 2.48) to −0.11 (95% CI −0.67 to 0.45), with the third at 0.58 (95% CI 0.22 to 0.94), at nearly equal weights. The authors do not rely on this figure, and it should never be cited on its own.
Unipolar depression, monotherapy	SMD 0.10, 95% CI −0.20 to 0.41
FindingFour trials, 395 patients versus 394, three of them unpublished. No benefit demonstrated, with unexplained heterogeneity (I² 73.7%, p = 0.01) that calls for caution in both directions. The confidence interval does not formally exclude a small effect: the precise statement is that no efficacy is established, not that inefficacy is proven.
Unipolar depression, added to an antidepressant	RR 0.95, 95% CI 0.73 to 1.23
FindingThree add-on trials to fluoxetine or paroxetine, 81 patients versus 78, analyzed on response rates, with no heterogeneity. The augmentation strategy, the most common off-label use in practice, does not separate from the comparator. Same cautious reading applies: absence of demonstration, not demonstration of absence, on small samples.
Comparison with lithium	RR 0.82 and 0.90, not significant
FindingIn maintenance, 350 patients on lamotrigine versus 242 on lithium, about 35 months: lamotrigine distinguishes itself from lithium neither on emergence of depressive symptoms (RR 0.82, 95% CI 0.63 to 1.06) nor on duration free of symptoms (RR 0.90, 95% CI 0.68 to 1.18). In the acute phase, the comparison rests on a single, single-blind trial in bipolar II depression. This absence of difference does not amount to equivalence.

Critical appraisal

Domain	Judgment
Integration of unpublished data	Main strength of the work
FindingThe negative result in unipolar depression rests in part on three trials the published literature did not contain. This is exactly the situation in which a synthesis built on publications alone would have concluded differently, and wrongly.
Heterogeneity of the add-on analysis	Uninterpretable estimate
FindingPooling three trials whose effect sizes run in opposite directions produces a number with no clinical referent. That the authors do not claim it is to their credit; that it appears in a table exposes it to misuse by a third party quoting it out of context.
Clinical relevance of the acute effect	Significant but trivial
FindingSignificance obtained on large cumulative samples says nothing about whether the patient perceives the benefit. At 0.155, the question of a clinically meaningful threshold arises, and the synthesis does not settle it.
Scope of the unipolar result	Not demonstrated, not disproven
FindingThe rigorous formulation is this: after including the unpublished trials, no benefit is shown, either in monotherapy or as an add-on. Concluding that inefficacy is established would go beyond what these data allow, all the more so since the add-on samples are small and the monotherapy heterogeneity remains unexplained.
Maintenance outcomes	Two distinct measures, not to be merged
FindingEmergence of symptoms and duration free of symptoms do not measure the same thing. The publication states that the maintenance outcomes used were the rate of emergence of depressive symptoms and the difference in survival time, without giving a trial-by-trial definition; that has to be checked in the original publications, which remain three trials of six to eighteen months. Their convergence strengthens the reading, it does not duplicate it.
Journal and disclosures	Editorial model worth flagging
FindingThe work appears in an open-access, author-pays journal, under a Creative Commons license. This model invalidates nothing by itself, and a synthesis is judged on its methods, but it justifies reading the report with the same rigor applied to a traditional peer-reviewed journal, particularly on funding and conflicts of interest, here declared in detail.

Level of evidence

Scientific72

Editorial79

Systematic review with meta-analysis of double-blind randomized controlled trials, level 1a on the Oxford Centre for Evidence-Based Medicine scale. Confidence is high on two points: the direction of the effect in bipolar maintenance treatment, and the small size of the acute-phase effect. It is moderate on the exact magnitude of these effects, given the limited precision of the intervals and the small number of trials per analysis, three to five. It is low to absent on the acute-phase add-on analysis, which heterogeneity renders uninterpretable, and on any claim of equivalence with lithium, which was never tested as such. A meta-analysis’s level of evidence does not automatically transfer to each of its sub-analyses: here, the same publication houses a solid result, a marginal result, and an unusable one.

The colleague test

What an experienced colleague might say about this study in two minutes, between two consultations.

“Lamotrigine, I keep it for stopping a bipolar patient’s depressive relapse, and there it’s genuinely something. In the acute episode, it’s paper-thin, I honestly don’t expect much from it. In unipolar depression, nothing has ever been shown, and this time they went and dug up the trials nobody had published, that’s what interests me. What I don’t even look at is their add-on analysis, with heterogeneity at 96%.”

Translated for practice: the drug is prescribed as a prophylactic, not as a fast-acting antidepressant, and its use in unipolar depression finds no support in these data.

What you can do with this

In bipolar maintenance treatment, these data support the prophylactic use of lamotrigine on the depressive side, with a modest magnitude (a 22% reduction in the risk of emerging depressive symptoms) and no demonstrated superiority over lithium. The choice between the two remains a clinical one, not one the evidence imposes.
In the acute phase, what can be told to a patient becomes simpler: this treatment mainly serves to space out and prevent episodes, not to resolve the one currently underway. The delay imposed by slow titration also makes an expectation of rapid effect unrealistic.
In unipolar depression, no data support adding lamotrigine to an antidepressant, including in resistant situations where the reflex to do so exists. The accurate statement to carry into a team meeting is: no benefit demonstrated, which is not the same as inefficacy proven. In most jurisdictions, including France, this use falls outside the approved indication, and clinicians should check the rule that applies where they practice.
What to monitor regardless of context: the gradual titration set out in prescribing information, the substantially reduced dose required with concomitant valproate, vigilance for any rash, especially during the first eight weeks, and re-evaluation of the titration schedule after any treatment interruption. This does not come from the results of this study, it comes from the prescribing framework, and it takes priority over any consideration of efficacy.
What to remember about the method: a synthesis that goes looking for unpublished trials can overturn an established therapeutic belief. The question to ask of any meta-analysis from now on is: were the negative trials recovered, and how?

Frequently asked questions

Can lamotrigine be used to treat a bipolar depressive episode that is currently underway?

An effect exists, and it is small (SMD 0.155, lower bound 0.005). It does not place the drug among the fast-acting options, and the slow titration required by its safety profile makes it poorly suited to an acute situation. Internationally, marketing authorizations for lamotrigine in bipolar depression center on prevention rather than acute treatment; the prescribing information in force in your own country will typically state this explicitly, so it is worth checking rather than assuming.

Does this meta-analysis prove that lamotrigine is ineffective in unipolar depression?

No, and the nuance matters. It shows that, after including three unpublished trials, no efficacy is demonstrated, neither alone (SMD 0.10, 95% CI −0.20 to 0.41) nor as an add-on (RR 0.95, 95% CI 0.73 to 1.23). A small effect is not formally excluded, and the heterogeneity in the monotherapy analysis remains unexplained. In practice, nothing supports prescribing in this indication, and doing so falls outside the approved use almost everywhere.

Does lamotrigine outperform lithium?

No superiority is demonstrated: in maintenance, RR 0.82 (95% CI 0.63 to 1.06) on emergence of depressive symptoms and RR 0.90 (95% CI 0.68 to 1.18) on duration free of symptoms, in 350 patients versus 242. This does not mean the two drugs are equivalent: a non-significant comparison across three datasets, one of them an open-label trial, does not prove equivalence. Lithium also carries properties this synthesis does not evaluate, notably on the manic side of the illness.

Why do the unpublished trials change anything?

Because negative trials are published less often, and later, than positive ones. Here, three unpublished trials in unipolar depression weigh on the result. The statistical asymmetry test detects nothing, but it is the actual retrieval of data from the manufacturer’s registry, not the test, that protects against publication bias.

Should the journal this appeared in factor into how it is read?

It is worth knowing, without turning it into an indictment. The work is published in an open-access, author-pays journal. A synthesis is judged on its protocol, its search strategy, its handling of heterogeneity, and how well-calibrated its conclusions are. On these four points, this work holds up, with the exception of one analysis the authors themselves do not rely on and a handful of discrepancies between the text and the tables.

Annotated bibliography

Source study. Arnone D, Östlundh L, Mosa M, MacDonald B, Oldershaw J, Qassem T, Young AH. Efficacy of Lamotrigine in the Treatment of Unipolar and Bipolar Depression: Meta-Analysis of Acute and Maintenance Randomised Controlled Trials. Pharmaceuticals 2025; 18(10): article 1590. DOI 10.3390/ph18101590. PMID 41155702. PROSPERO registration CRD42025633709. Contribution: pools 24 double-blind randomized trials, unpublished trials included, and explicitly separates the acute phase, maintenance, and unipolar depression. Limitations: one analysis rendered uninterpretable by 96.7% heterogeneity, an unexplained 73.7% heterogeneity in unipolar monotherapy, three trials without double blinding inside a set presented as double-blind, and several discrepancies between the text and the tables.

The two conflicting add-on trials. Geddes JR, Gardiner A, Rendell J, et al. Comparative evaluation of quetiapine plus lamotrigine combination versus quetiapine monotherapy (and folic acid versus placebo) in bipolar depression (CEQUEL): a 2 × 2 factorial randomised trial. Lancet Psychiatry 2016; 3: 31 to 39. Kemp DE, Gao K, Fein EB, et al. Lamotrigine as add-on treatment to lithium and divalproex: lessons learned from a double-blind, placebo-controlled trial in rapid-cycling bipolar disorder. Bipolar Disorders 2012; 14: 780 to 789. Contribution: the first, in 101 patients versus 101 over twelve weeks as an add-on to quetiapine, gives an effect size of 2.13 (95% CI 1.79 to 2.48); the second, in 23 patients versus 26 over twelve weeks as an add-on to lithium and divalproate in rapid-cycling forms, gives −0.11 (95% CI −0.67 to 0.45). Together they show concretely why a pooled average can mean nothing at all. Limitations: populations, comparators, and scales differ between them, a gap the meta-regression did not manage to formalize.

The molecule’s historic manufacturer’s trial registry. Public study register, gsk-studyregister.com, searched by the authors on May 21, 2025. Contribution: source of the seven datasets in Table 1 that carry no associated scientific publication, including the three unipolar depression trials that shift the reading of the whole synthesis. Makes visible the gap between the published literature and the full body of data actually produced. Limitations: the completeness and level of detail of these entries vary, which limits an individual assessment of their risk of bias.

Regulatory sources cited by the review’s authors. European Medicines Agency referral procedure for lamotrigine (Lamictal), and United States Food and Drug Administration labeling position, both accessed by the review’s authors on May 21, 2025 per their reference list. Contribution: situates the authorized indications the authors describe in their introduction, prevention of symptom emergence in bipolar depression in Europe, maintenance treatment of bipolar I disorder in the United States. Limitations: these are labeling documents, not efficacy data, and they say nothing about comparative efficacy between drugs; the exact wording of an indication, and whether unipolar depression is covered anywhere, should be checked against the current prescribing information in the reader’s own country before any of it is quoted.

Editorial collections

Myth or Reality? A Study Under the Microscope