Choosing a Mood Stabilizer

What This Guide Is For

The library already has a deep, standalone guide for each of the five mood stabilizers: lithium, valproate, lamotrigine, carbamazepine, and oxcarbazepine. This is not another one of those. This guide does the job those guides deliberately don't: it treats the class as a system and helps you choose among the agents at the bedside.

The individual guides answer "how do I use lithium well?" This one answers the questions that come before that: Is lithium even the right drug for this patient? If not, which one is, and why? What do I reach for in mixed mania versus classic euphoric mania versus a depression-dominant course? What do I do when the patient is pregnant, or won't do blood draws, or is on the pill, or has failing kidneys? When two agents are reasonable, what actually breaks the tie?

So this guide is comparative by design. It leans on tables, on head-to-head reasoning, and on a decision framework organized by clinical scenario rather than by drug. Where it states a dose, a level, or a monitoring interval, those numbers match the individual guides exactly, because those guides are what's published on the site, and the class guide must never contradict them. When you need the full depth on any single agent (the complete titration, the exhaustive side-effect management, the whole toxicity workup) go to that drug's guide. When you need to decide which agent, start here.

Why "Mood Stabilizer" Is a Messy Category

"Mood stabilizer" is a marketing-era term of convenience, not a clean pharmacologic class. There is no shared receptor, no shared mechanism, and no shared chemistry that unites the drugs we file under it. Lithium is an element, a monovalent cation that competes with sodium. The rest are anticonvulsants that wandered over from neurology: valproate, lamotrigine, carbamazepine, and oxcarbazepine all block voltage-gated sodium channels to varying degrees, but they otherwise behave like four different drugs, because they are.

What actually holds the category together is a clinical job description: an agent that treats or prevents the episodes of bipolar disorder without flipping the patient into the opposite pole. That is a functional definition, and it is a loose one. It is why the members of the class have wildly different profiles, and it is exactly why choosing among them is a real clinical skill rather than a coin flip.

Two honest caveats before we go further:

The anticonvulsants are not interchangeable, and most of them are weaker than their reputations. Lithium and valproate and carbamazepine have genuinely "passed" controlled trials in mania. Lamotrigine works for the depressive pole and does essentially nothing for acute mania. Oxcarbazepine sits on a pile of flawed studies and cannot be relied on as a monotherapy for bipolar I at all. Topiramate failed every acute-mania trial it entered and should not be used as a mood stabilizer (its legitimate roles are weight loss and, marginally, migraine; it is not covered as a member of this class here). Grouping these agents under one label obscures how differently they perform.

The atypical antipsychotics compete directly in this space, and for some jobs they win. This is the biggest thing the term "mood stabilizer" hides. Quetiapine, lurasidone, cariprazine, aripiprazole, olanzapine, and lumateperone are, for acute mania and acute bipolar depression, often faster and better-evidenced than the classic mood stabilizers. Any honest account of how to stabilize mood has to acknowledge that the atypicals are frequently the first move, especially acutely. This guide points to them repeatedly because you cannot make good decisions in this class while pretending they don't exist, but it does not cover them in depth. They have their own guides, and their monitoring (metabolic surveillance, EPS, prolactin) belongs there. Here, treat them as the fast-acting neighbors you reach across to constantly, particularly in the acute phase and in pregnancy.

The Five Agents, One Paragraph Each

Before the decision framework, here is the identity of each drug: the one thing it is the drug for. Internalize these five sentences and half the decisions make themselves.

Lithium

Lithium is the drug for classic bipolar I and for suicide prevention. It is the oldest, best-evidenced, and, as monotherapy in bipolar I, the most effective agent in the class, superior to valproate, lamotrigine, olanzapine, and quetiapine head-to-head. It is the only psychiatric drug with replicated evidence that it prevents suicide (about a 60% reduction in completed suicide), an effect independent of its mood-stabilizing action, and it lowers all-cause mortality. It shines in the Kraepelinian patient: euphoric mania, then depression, with genuinely well intervals between. Its price of admission is blood draws, a narrow therapeutic window, and attention to the kidney and thyroid.

Valproate

Valproate is the drug for the messy, non-classic bipolar patient. Where lithium owns the textbook case, valproate earns its keep in mixed features, rapid cycling, irritable or dysphoric mania, and bipolar with comorbid substance use. It loads fast (a real advantage in acute mania), doses flexibly, and doesn't make you watch the kidney. Its defining liability is reproductive: it is by far the most teratogenic drug in the cabinet (roughly 50 to 100 times lithium), and it lowers childhood IQ even beyond the first trimester, which should keep it out of anyone who might become pregnant when an alternative exists.

Lamotrigine

Lamotrigine is the drug for the depressive pole and for bipolar II. It prevents the depression that patients actually spend most of their lives in, it is the best-tolerated agent in the class (weight-neutral, non-sedating, no required labs), and it has an excellent pregnancy profile. Its two great failure modes are using it for acute episodes (it doesn't work fast and doesn't treat mania at all) and titrating it too fast (Stevens-Johnson syndrome). It is a foundation you lay for the long term, not a rescue.

Carbamazepine

Carbamazepine is the specialist's antimanic agent and lithium's complement. It is a legitimate, FDA-approved (as Equetro) antimanic drug for the second line: the lithium-refractory rapid cycler, the manic patient who can't take valproate, and, distinctively, the patient in whom lithium and carbamazepine cancel each other's worst side effects. Its burden is real: two boxed warnings, a genetic screen, a self-lowering drug level (autoinduction), and a metabolic profile (potent CYP3A4 induction) that reaches into nearly every other prescription the patient holds.

Oxcarbazepine

Oxcarbazepine is the anti-irritability, anti-anxiety agent for bipolar II and the soft spectrum. It is carbamazepine's easier-to-live-with cousin (no marrow catastrophe, no autoinduction, no required levels, milder interactions), bought at the cost of weaker efficacy evidence: it cannot be trusted as a bipolar I monotherapy. Its home is the anxious, revved, irritable bipolar II patient, often at surprisingly low doses (300 to 600 mg/day), and its one genuine hazard is hyponatremia.

The Core Decision Framework: Choosing by Clinical Scenario

This is the heart of the guide. Most prescribing decisions in this class are not "which drug is best in the abstract" but "which drug fits this clinical situation." Work through the scenario that matches your patient.

Classic euphoric mania

This is lithium's home turf, and the scenario where it is at its most superior. The Kraepelinian phenotype (euphoric rather than dysphoric mania, followed by depression, with clean euthymic intervals, no rapid cycling, no mixed features, minimal comorbidity, ideally a hyperthymic temperament and early illness onset) predicts the "excellent responder," the roughly 27 to 35% of bipolar patients who have no further episodes once stabilized and adherent. No other agent approaches that.

The practical wrinkle is speed. Lithium takes weeks to work as monotherapy, which is too slow for a floridly manic patient. The standard acute move is to lead with a faster agent (a sedating atypical such as olanzapine or quetiapine, with or without a benzodiazepine for sleep and agitation) and introduce lithium underneath it, then taper the antipsychotic once the patient is stable. Valproate is the other fast, loadable antimanic option and is the better choice if the mania is not classic. Carbamazepine is second-line antimanic, reached for when lithium and valproate fail or can't be used.

Mixed features

Mixed states tip you away from lithium and toward valproate. Lithium is weakest exactly where mania carries dysphoric, depressive, or agitated coloring, and valproate is the anticonvulsant with the best mixed-episode data. Carbamazepine also has genuine mixed-features efficacy and is a reasonable alternative or add-on.

Be honest with the patient that mixed states are a hard patch of the illness that neither lithium nor valproate treats spectacularly (the Calabrese comparisons found both similar and neither great). Atypicals are frequently part of the regimen here. Avoid leaning on lamotrigine as your antimanic tool: it doesn't cover the manic/activated side of a mixed picture.

Rapid cycling

Rapid cycling (four or more episodes a year) is another scenario where lithium underperforms and the anticonvulsants earn their place. Valproate is the workhorse. Carbamazepine has supporting data and is a classic partner to lithium in this population (see combinations, below). Lamotrigine specifically helps the ultra-rapid, cyclothymic end of the spectrum, producing a roughly 50% reduction in daily-to-weekly mood swings in a placebo-controlled trial, so it is a rational add-on when the cycling is depression-weighted.

Two practical notes: check thyroid function, since thyroid dysfunction is over-represented in rapid cycling and correcting it (or supraphysiologic thyroid augmentation) can help; and reassure the patient that rapid cycling is usually a phase that resolves over years rather than a permanent sentence.

Bipolar depression as the dominant pole

When the patient's illness is mostly depression (the typical bipolar II shape, where patients spend roughly half their lives depressed and only a few percent hypomanic), lamotrigine is the anchor. It prevents the depressive pole, it is weight-neutral and well-tolerated for the long haul, and it is often first-line as monotherapy or as the foundation of a regimen.

The crucial caveat is the "happily ever after fallacy": lamotrigine prevents bipolar depression but does not reliably treat an acute depressive episode, and it is slow (4 to 6 weeks just to titrate, then another 1 to 3 weeks to work, plateauing at 10 to 21 weeks). For acute bipolar depression, lead with a faster proven agent (lurasidone, cariprazine, quetiapine, lumateperone, or olanzapine-fluoxetine), add lamotrigine underneath, and taper the antipsychotic after roughly 6 months of recovery, leaving lamotrigine as the maintenance foundation. Lithium is also a legitimate option for the depressive pole, especially where suicide risk is prominent, and lithium plus lamotrigine cleanly covers both poles.

Oxcarbazepine is not an antidepressant, but in bipolar II it can indirectly help by stripping away the anxious, hypomanic churn that drives the instability.

Maintenance and prophylaxis

For long-term prevention, the two best-evidenced monotherapies are lithium and lamotrigine, and they cover different poles: lithium prevents mania (and depression) and carries the mortality and anti-suicide benefit; lamotrigine prevents depression and is the easier drug to stay on for decades. In the pivotal maintenance trials they perform comparably on time-to-relapse, and they combine cleanly, so lithium plus lamotrigine is a rational two-pole maintenance regimen.

Valproate's maintenance evidence is softer than its reputation: the large, unenriched STEP-BD program did not show valproate (or lithium) clearly superior to placebo for long-term maintenance in unselected patients, and it is FDA-approved for acute mania, not maintenance. Use valproate for maintenance with realistic expectations, often in combination rather than as solo prophylaxis. Carbamazepine has some maintenance evidence but less robust than lithium's.

Suicidality

Suicide risk is an indication for lithium, not a reason to avoid it. This is one of the most consequential and most misunderstood points in the class. Lithium is the only agent with strong, replicated evidence for reducing both attempted and completed suicide (roughly 60% reduction in completed suicide), the effect is unique to lithium and not shared by the other mood stabilizers, and it spans both unipolar and bipolar illness because it is independent of mood stabilization.

Many clinicians hesitate to give a drug that is "toxic in overdose" to a suicidal patient. But lithium is disabling far more often than lethal in overdose, and the prevention benefit is enormous. Prescribe it even to patients with overdose histories; if acute risk is high, limit quantities and engage family to dispense. Choosing a different mood stabilizer to avoid the overdose question means forgoing the one drug proven to reduce the very outcome you're worried about.

Pregnancy and women of childbearing potential

This scenario reorders the whole class, and the reordering should happen before conception, not after a positive test. The teratogenicity ranking is the organizing fact:

• Valproate: avoid. Roughly 5% neural tube defect risk (50 to 100 times lithium), plus a dose-dependent hit to childhood IQ that persists beyond the first trimester. Do not use it as a default in anyone who could become pregnant. Folate does not make it safe.
• Lamotrigine: preferred. Overall first-trimester malformation risk around 2.2%, within the general-population baseline; the old cleft-palate signal was largely reassured by later registries. It is one of the strongest arguments for lamotrigine.
• Lithium: acceptable if monitoring is tolerated. The Ebstein's anomaly fear is real but small in absolute terms (0.05 to 0.1%, versus roughly 1 in 20,000 baseline), and far safer than valproate. The burden is pharmacokinetic: clearance rises in pregnancy (levels fall, dose often up), then crashes at delivery (cut the dose promptly postpartum to avoid toxicity). Monitor levels frequently.
• Carbamazepine: not preferred. Roughly 3.3% malformation rate with elevated spina bifida risk, plus the contraceptive-failure hazard of its own enzyme induction. Little used for psychiatric indications in US pregnancy.
• Oxcarbazepine: uncertain, not established safe. Human data are thin; treat it as a drug of uncertain reproductive safety rather than a reassuring choice.

The atypical antipsychotics are also reasonable in pregnancy and are often preferred first-line for reproductive-age women, since they lack the teratogenic signatures of valproate and act fast.

Set all of this against the danger of untreated illness: bipolar relapses in roughly 71% of pregnancies, and stopping medication roughly doubles that risk, with abrupt cessation triggering relapse far faster (about 2 weeks) than gradual (about 22 weeks). The lesson is not "keep the teratogen." It is: switch reproductive-age women to a safer agent (lamotrigine, or an atypical, or lithium) deliberately and ahead of time, rather than stopping cold or discovering the problem in the first trimester.

Renal disease

Lithium is the agent the kidney cares about; the anticonvulsants are mostly hepatically handled. This makes renal disease a strong steer away from lithium and toward the others.

• Lithium: eGFR above 60, use normally with annual monitoring. eGFR 30 to 60, do not reflexively stop; many patients hold steady for years on a lower dose with frequent monitoring and nephrology co-management, and this can be the right call for an excellent responder with severe illness. eGFR below 30, generally contraindicated. Protect the kidney by keeping levels at or below 0.8, dosing once nightly, avoiding toxic spikes, and keeping the patient hydrated.
• Lamotrigine: hepatically glucuronidated, no major adjustment for ordinary renal impairment, a real advantage.
• Valproate: hepatically metabolized; reduce dose in moderate-to-severe impairment; rare AKI reported but low.
• Carbamazepine: hepatically metabolized, so no eGFR-driven scaling, but watch sodium (hyponatremia).
• Oxcarbazepine: the active metabolite (MHD) is renally excreted and accumulates; in severe impairment (CrCl below 30) cut the starting dose (often by half) and titrate slowly.

Hepatic disease

The mirror image: here lithium becomes attractive precisely because it has no hepatic metabolism, while the anticonvulsants demand caution.

• Lithium: not liver-dependent; safe in hepatic impairment.
• Valproate: contraindicated in active hepatic disease; rare but potentially fatal hepatotoxicity (highest in the first 6 months and concentrated in the very young, polypharmacy, and mitochondrial or urea-cycle disorders); reduce dose in moderate-to-severe impairment.
• Carbamazepine: accumulates in liver disease; use caution, monitor LFTs and levels, avoid in severe impairment.
• Oxcarbazepine: hepatically converted; conservative dosing in advanced disease.
• Lamotrigine: hepatic clearance means significant liver disease can raise levels; reduce dose in moderate-to-severe impairment.

The elderly

Age shifts everything toward lower doses, slower titration, and more vigilant monitoring, and it sharpens specific hazards for each drug.

• Lithium: reduced renal clearance and a more permeable blood-brain barrier mean you target roughly 30% lower serum levels and monitor more often. The upside is real: older patients with treatment-resistant depression respond better to lithium augmentation (68.2% versus 46.9%), and mortality is lower on lithium. Get a baseline ECG.
• Valproate: watch thrombocytopenia and free-level traps (lower albumin, polypharmacy) more closely; check a free level when a patient looks toxic but the total is "normal."
• Carbamazepine: more CNS sensitivity (sedation, unsteadiness, falls), more hyponatremia, more conduction risk; slower titration, lower doses, baseline ECG, and careful med reconciliation given the interaction burden.
• Oxcarbazepine: hyponatremia risk is highest here, compounded by the near-universal thiazides and SSRIs; check sodium more often and keep a low threshold when an older patient feels "off."
• Lamotrigine: among the gentler choices, but titrate slowly and watch for the (uncommon) hyponatremia.

Adolescents

Data are thinner across the board, but two points steer the choice. First, early lithium initiation predicts better lifelong prophylaxis: an excellent young responder may be spared decades of episodes, so lithium is worth serious consideration in the classic phenotype. Second, the pediatric antimanic evidence for carbamazepine and oxcarbazepine is weak (both failed to separate from placebo in controlled pediatric mania trials), so neither is a well-supported choice for youth mania. Lamotrigine carries a higher baseline SJS risk in children, making slow titration even more critical. And any adolescent girl started on valproate inherits its full teratogenic risk as she reaches reproductive age, which should factor into long-term planning from the start.

The patient who won't do blood draws

Monitoring burden is a legitimate deciding factor, and it points hard toward lamotrigine, with oxcarbazepine as a secondary option. Lamotrigine requires no routine serum levels, no CBC, no LFTs, no metabolic panel, and no ECG in low-risk patients; its monitoring is event-driven (ask about rash) rather than schedule-driven. Oxcarbazepine needs no drug levels either, only a baseline sodium and a couple of early rechecks.

The agents this rules out (or makes hard) are lithium, valproate, and carbamazepine, all of which depend on serum levels and periodic labs. Be honest with yourself: a patient who genuinely will not do any monitoring probably cannot be on lithium safely, and forcing it is worse than choosing a drug that fits their reality. But don't over-apply this: most patients, given a concrete tool (a shared lab tracker), clear instructions, and a standing lab order, comply far better than expected. Reserve the "no monitoring" pathway for patients who have actually demonstrated they won't, not for your assumption that they might not.

The patient on hormonal contraception

Two drugs interact with the pill in opposite, both-important directions.

• Lamotrigine and estrogen: estrogen induces lamotrigine's metabolism and can drop its level by 40 to 50%. You may need a higher dose while she's on the pill, and, critically, you must lower it when she stops the pill (or during the pill-free week), or she can climb toward toxicity. The same dynamic drives the pregnancy (dose up) and postpartum (dose promptly back down) swings.
• Carbamazepine and oxcarbazepine (at higher doses) induce the pill's metabolism, causing contraceptive failure and unintended pregnancy. For carbamazepine this is a can't-miss counseling point at any dose: advise a non-hormonal or supplemented method. For oxcarbazepine the risk is dose-dependent and becomes clinically meaningful toward and above 1,200 mg/day, another reason to keep bipolar II patients low.

Lithium and valproate do not meaningfully interact with hormonal contraception, which can itself be a tiebreaker in a woman who wants to stay on the pill.

Head-to-Head Comparison Tables

These tables compress the class into the dimensions that actually drive choice. All figures are consistent with the individual guides; consult those for the reasoning behind each.

Efficacy by pole and phase

Monitoring burden

Teratogenicity ranking (safest to most dangerous)

Weight and metabolic impact

Cognitive burden

Drug-interaction liability (lowest to highest)

Lethality in overdose

Combination Strategy

Bipolar disorder is frequently a combination-therapy illness, and knowing which pairings are rational (and which are dangerous or pointless) is a core class skill.

Rational, evidence-supported combinations

Lithium plus lamotrigine

The cleanest two-pole regimen in the class. Lithium covers mania (and carries the anti-suicide benefit), lamotrigine covers depression, and there is no meaningful pharmacokinetic interaction between them. This is the go-to when a single agent leaves one pole uncovered.

Lithium plus valproate

A mainstream, well-tolerated pairing for treatment-resistant bipolar I and for the rapid cycler or mixed-features patient who needs more antimanic power than either alone provides. Adding valproate to lithium is a standard move for the resistant bipolar I patient.

Lithium plus carbamazepine

The underappreciated complementary pair. The two drugs partly cancel each other's worst side effects: lithium raises the white count and can reverse carbamazepine-induced neutropenia, and lithium raises sodium while carbamazepine lowers it, blunting the hyponatremia. It has open-label support in rapid-cycling and treatment-resistant illness. The tradeoffs to watch are additive hypothyroidism (both suppress thyroid; check TSH) and rare neurotoxicity (confusion, ataxia, coarse tremor) even when both levels read normal.

A mood stabilizer plus an atypical antipsychotic

The standard acute strategy: the atypical delivers speed (mania or bipolar depression), the mood stabilizer delivers durable prophylaxis, and you taper the antipsychotic after roughly 6 months of recovery to shed its metabolic burden. If the patient is gaining weight on the combination, trim the antipsychotic first: it's usually the bigger culprit.

Combinations that require doing the math

Valproate plus lamotrigine

Not forbidden, but the single most dangerous combination to get wrong. Valproate roughly doubles lamotrigine levels by inhibiting its glucuronidation, which directly raises Stevens-Johnson risk. If you combine them, use the reduced lamotrigine titration (halve every starting dose and every step, target roughly 100 mg/day) and go especially slowly. And remember the reverse: if you later stop the valproate, the lamotrigine level roughly halves, so anticipate loss of efficacy and raise the lamotrigine dose as you withdraw the valproate.

Combinations that are dangerous or pointless

Carbamazepine plus oxcarbazepine

Pointless and additive in the wrong ways. They are close chemical cousins with overlapping sodium-channel mechanisms, overlapping HLA-B*1502-linked rash risk, and overlapping hyponatremia liability; combining them stacks the toxicities without adding a distinct benefit. Pick one.

Carbamazepine plus a CYP3A4-substrate antipsychotic (without planning)

Carbamazepine can gut the antipsychotic's level (quetiapine can drop about 80%, and carbamazepine also shifts quetiapine toward its norquetiapine metabolite, which is linked to manic switching). If you need an antipsychotic alongside carbamazepine, asenapine is the one atypical with essentially no significant carbamazepine interaction (note asenapine is itself a CYP2D6 inhibitor).

Carbamazepine plus hormonal contraception (as the sole method)

Carbamazepine induces the pill's metabolism and can cause contraceptive failure. This is a can't-miss counseling point, not a combination to leave unmanaged.

Switching Between Agents: Cross-Taper Principles

Switching mood stabilizers is usually a cross-taper, not a stop-and-start, because abrupt discontinuation of any of these agents risks destabilization, and lithium in particular carries a well-documented rebound (a roughly 7-fold relapse risk within 7 weeks of abrupt stop, plus the possibility of losing future response).

• Overlap and cross-taper. Introduce the new agent and titrate it toward a therapeutic dose before withdrawing the old one, so the patient is never uncovered. Taper the outgoing drug gradually as the incoming drug takes hold.
• Taper lithium slowly. A minimum 30-day taper (roughly 300 mg per month), and 6 to 12 months for long-term users, tapering more slowly near the end. Never stop lithium abruptly except for a genuine emergency.
• Mind the interaction as you switch. Some cross-tapers change drug levels. Switching off valproate onto or alongside lamotrigine will roughly halve the lamotrigine level (raise the lamotrigine as the valproate comes down). Switching off carbamazepine unmasks its induction: over 1 to 2 weeks, every co-medication it was suppressing (antipsychotics, antidepressants, methadone, warfarin, contraceptives) rises toward "overdosed," so plan dose reductions of the co-meds and reconcile the whole list before you start the taper.
• Respect lamotrigine's re-titration rule. If lamotrigine has been off for more than about 5 to 7 days for any reason (a switch, a lapse, a hospitalization), you must re-titrate it from the very bottom; the rash risk resets. Never resume a lapsed patient at their old maintenance dose.
• Watch the pole you're moving toward. Switching a bipolar II patient from an antimanic drug onto lamotrigine covers depression but not mania; switching the other way covers mania but not depression. Make sure the destination agent covers the pole that actually threatens the patient.
• Cross-taper timing by agent speed. Lamotrigine's slow titration means a switch onto it takes weeks and often warrants keeping the old agent (or an interim atypical) on board until lamotrigine reaches a therapeutic dose. Valproate and carbamazepine can be titrated faster; lithium needs its level chased to steady state (≥5 days after each change).

The Monitoring Master Table

All five agents' baseline workup and ongoing monitoring in one view. Details and action thresholds live in the individual guides.

Two screens apply across the anticonvulsants: HLA-B*1502 before starting carbamazepine or oxcarbazepine in patients of Asian ancestry (positive means avoid the drug), and a rash-counseling conversation for lamotrigine, carbamazepine, and oxcarbazepine (any rash with fever, blistering, or mucosal involvement means stop and be evaluated).

The Class-Level Bedside Cheat Sheet