Lithium, lamotrigine, valproate, carbamazepine, oxcarbazepine. What each one does, what it does not, and why the drug is the line and the work is yours.
Mood stabilizers smooth the peaks and the troughs of bipolar and the bipolar-spectrum conditions without flattening the patient. The point is not to make you feel nothing, the point is to make the cycle stop running your life. If you have been spending six months a year either too high to sleep or too low to get off the couch, the line these drugs hold is the thing that gives you back enough steady ground to do anything else. The drug is leverage, the work is yours, and I want to be clear about that up front because most of the writing on this class either pitches the meds as a fix that absolves the patient of any further responsibility or treats them as a crutch for people who lack discipline, and both framings are nonsense.
The honest version is that bipolar is real biology, the cycle has a momentum of its own, and the medication is what slows the momentum enough that the patient’s choices can start to matter again. Sleep regularity, alcohol, stimulant use, medication adherence, life rhythm… all of that is still yours, all of that still moves the dial, and none of it goes away because you started lithium. What changes is that the biology stops screaming louder than the willpower, which is the trade, and which is what these drugs are for.
The actual medications, what to call them, and what they do
The class is small, five drugs do most of the work, plus a couple of atypical antipsychotics used as adjuncts that end up doing some of the mood-stabilizing job even though they are not technically in the class.
Lithium is the original mood stabilizer and still the gold standard for bipolar I. The mechanism is genuinely not fully understood sixty years in, which tells you something about how slow brain pharmacology has been. What we know is that it modulates intracellular signaling in ways that quiet the cycle, reduces the frequency and severity of manic episodes, has a maintenance-phase effect on depressive recurrence, and is the only psychiatric medication with a real anti-suicide signal in the data. It also requires the most respect, because the therapeutic window is narrow, you need blood draws, and you have to watch the kidneys and thyroid over the long haul… if you commit to lithium, you commit to the labs.
Lamotrigine (Lamictal) is the workhorse on the depression side of bipolar, especially bipolar II. It has a real effect on preventing depressive episodes in maintenance and a weaker effect on the manic side, and it is generally well tolerated once you get to the target dose. The catch is the slow titration up over six weeks, which is non-negotiable, because rushing it is what triggers Stevens-Johnson syndrome, the rare but dangerous skin reaction that puts people in the hospital. Done right, the ramp is uneventful and the patient lands at 200mg and stays there, and most of them report it as the easiest psych med they have ever been on.
Valproate / Valproic Acid (Depakote) is the workhorse on the manic side. It is fast, effective for acute mania, and the trough levels are easy to monitor, which all helps when somebody is in an episode and you need to move quickly. It also comes with weight gain, hair loss, and teratogenicity that makes it a hard no for women of childbearing age who might want kids. Useful drug, with a clear cost. Most adult men who need an antimanic that works fast end up on it at some point, and the trade-offs are usually worth it for the right patient.
Carbamazepine (Tegretol) is older, less commonly used now, and has a quirk most patients have never heard of called autoinduction, meaning the drug induces its own metabolism, so the dose that worked at week two stops working at week six because the liver has revved up to clear it faster. It also has drug interactions with basically everything because the same liver enzymes it induces are the ones that metabolize half the medications anybody is on. Still useful in patients who have failed lithium and valproate, but it is a second-line drug for a reason.
Oxcarbazepine (Trileptal) is the cleaner cousin of carbamazepine, with similar mechanism, much less autoinduction, fewer drug interactions, and generally better tolerability. The evidence base is thinner than the carbamazepine evidence base because oxcarbazepine has been less studied, but in real practice it has mostly displaced carbamazepine for the patients who need that pharmacology and cannot tolerate the carbamazepine side effects.
Brief mention on the atypical antipsychotics, because they keep showing up in the bipolar conversation. Lurasidone (Latuda) and Lumateperone (Caplyta) are the two I reach for most often as adjuncts in bipolar depression. They are not technically mood stabilizers, but they hold the depressive side of bipolar in a way that SSRIs sometimes do not, because SSRIs in bipolar can flip the patient into mania or hypomania or accelerate cycling, a liability an atypical does not carry. Pair an atypical with a mood stabilizer and you cover both sides of the cycle, which is most of what real bipolar prescribing looks like.
When mood stabilizers fit
Bipolar I is the cleanest indication, and any of the lithium, valproate, lamotrigine, or atypical-augmentation options is reasonable depending on the phase of illness, the patient’s history of response, and the side-effect tradeoffs. Most bipolar I patients end up on a combination, a mood stabilizer plus an atypical, because the cycle has both a manic side and a depressive side and a single agent rarely covers both well over years. Bipolar II, where the highs are hypomanic rather than fully manic but the depressions can be debilitating, is most often lamotrigine first, sometimes with an atypical added for the depressive episodes when they get severe. Lithium also works in bipolar II with a slightly weaker evidence base than in bipolar I, and it is still a reasonable first-line choice when the depressive episodes have been suicidal in character, because the anti-suicide signal is real and you take it where you can find it.
Cyclothymia, which is the mild version of the bipolar spectrum where the highs and lows are not severe enough to meet full bipolar criteria but the cycling is real and disruptive, usually gets lamotrigine, sometimes lithium at a low dose. The argument for treating it at all is that the cycling tends to get worse over time if left alone, and catching it early is easier than catching it ten years in after the patient has built a life around the cycle. Mixed states, which is what happens when manic and depressive features show up at the same time and the patient is agitated, sleepless, irritable, and despairing all at once, almost always need combination therapy, usually valproate plus an atypical, and this is one of the situations where moving fast matters because mixed states carry a higher suicide risk than either pure mania or pure depression.
The “an SSRI flipped me into mania” patient is another classic indication. If somebody has been on an antidepressant by itself and ends up in a manic or hypomanic episode, the diagnosis is bipolar (the SSRI did not cause it, it unmasked it), and the move is to taper the antidepressant, start a mood stabilizer, and figure out whether they need the antidepressant back at all once the cycling is under control. A lot of the time they do not.
One more indication worth naming. Lithium augmentation of an antidepressant in treatment-resistant unipolar depression has real evidence behind it. If somebody has been through two or three antidepressants without remission, adding lithium at a low dose (often 600 to 900mg, lower than the bipolar maintenance dose) is one of the better-evidenced augmentation strategies, ahead of most of the atypicals and the off-label tricks. It does not get used as often as it should because the labs are a barrier and prescribers tend to reach for something easier first, which is a fair criticism of the field.
The drug holds the line. The patient does the work of staying inside that line. The combination is what changes the outcome.
When they do not fit
Pure unipolar depression with zero bipolar features is overkill for a mood stabilizer as monotherapy. If the patient has never had a hypomanic episode, no family history of bipolar, no SSRI-induced activation, the move is to treat the depression with an antidepressant first, and the lithium augmentation case is a different conversation for treatment-resistant depression specifically.
Pregnancy or planning pregnancy is a hard no for valproate, with rare exceptions where nothing else has worked and the patient and obstetrician have walked through the math. Valproate carries a roughly 10% risk of major congenital malformations plus a separate signal for neurodevelopmental delays in exposed kids, the highest teratogenic profile in the psychiatric pharmacopoeia. Lithium has a smaller signal (Ebstein’s anomaly, a heart malformation), real but rare, manageable with careful monitoring, and a lot of women have safely been on lithium through pregnancy. Lamotrigine is the safest of the class in pregnancy and tends to be the default for women who plan to conceive.
Renal disease is a hard no for lithium, because the drug clears through the kidneys and a marginal kidney cannot keep up with the load, which puts the patient at risk of toxicity, and anybody with a GFR under 60 needs a different stabilizer. The drugs to reach for in renal disease are lamotrigine and the atypicals, both of which clear primarily through the liver. The last group who should not be on lithium is the patients who will not commit to blood draws and follow-up, not as a moral judgment but as pharmacology… lithium without monitoring is genuinely dangerous, the therapeutic window is narrow, the consequences of an out-of-range level can be severe, and starting somebody on lithium and hoping they will come around to the labs is not a plan.
The lithium conversation
Lithium has been the gold standard for bipolar I since the 1970s, and most of the patients who finally get stable after years of bouncing between meds get stable on lithium. The strongest evidence base in the class for preventing manic recurrence, the strongest for preventing depressive recurrence, and the only real anti-suicide signal in psychiatric medication… the number you’ll hear every clinician repeat is that lithium cuts completed suicide by sixty or seventy percent, and honestly that one’s lore, it’s what we all got told in school and nobody can quite tell you which study it came from. The cleanest data we actually have, the Cipriani 2013 meta-analysis in the BMJ, pooled forty-eight randomized trials in people with mood disorders and found lithium cut the odds of completed suicide to roughly an eighth of placebo (odds ratio 0.13) and lowered death from any cause, which is a signal no other psychiatric medication has produced, so the tidy percentage is fuzzy but the direction’s real and, anecdotally, it does seem to hold water.
The maintenance target for blood levels is 0.6 to 1.0 mEq/L drawn twelve hours after the last dose, acute mania often needs higher levels (0.8 to 1.2), and the maintenance target gets dialed back once the patient is stable. Below 0.6 the prophylactic effect drops off, above 1.2 side effects climb, above 1.5 is early toxicity, above 2.0 is a medical emergency. The reason the labs matter is that lithium levels can climb without warning if the patient gets dehydrated, starts an NSAID like ibuprofen, starts an ACE inhibitor for blood pressure, develops a kidney problem, or has a stomach bug that knocks out their hydration for a day. Any of those can push a stable level into the toxic range.
Toxicity signs in rough order of severity are coarse tremor (different from the fine tremor a lot of patients have at baseline), nausea, vomiting, diarrhea, confusion, sedation, slurred speech, ataxia (unsteady walking), and at the high end seizures and cardiac arrhythmias. If any of these show up, the move is to hold the dose, get hydrated, and call the prescriber that day, not the next morning. Levels that climbed fast can keep climbing, and any prescriber who waves off a tremor-and-nausea call as nothing is a damn liar.
Thyroid and kidney monitoring is the long-haul piece. Lithium can suppress thyroid function over years (TSH every six to twelve months is standard), and roughly a quarter of long-term patients develop hypothyroidism, which is treatable with levothyroxine. Kidney function (BUN, creatinine, eGFR) goes the same way, with a small percentage of long-term patients developing chronic kidney disease, the rate climbing after twenty years on the drug. Most patients can stay on lithium for decades without serious renal damage if the monitoring is done, the damage tends to happen in the patients who were never tracked, and the deal is quarterly draws for the first year, semiannually after that, lifetime.
The lamotrigine conversation
Lamotrigine is the drug I reach for first in bipolar II and in any patient where the depressive side of the cycle has been the bigger problem. It works on the depression side, has a real maintenance effect, and is generally well tolerated. The catch is the titration… 25mg daily for two weeks, then 50mg daily for two weeks, then 100mg daily for one week, then 200mg daily as the target. Six weeks of climbing from a dose that does nothing to a dose that does the work, annoying for everybody involved, and the price of doing this safely.
The reason the ramp is non-negotiable is Stevens-Johnson syndrome and toxic epidermal necrolysis, the two skin reactions that are rare, dose-rate dependent, and can be fatal. Faster ramps multiply the risk, slow ramps drop it into the territory of one in several thousand, which is roughly the rate for most drug allergies in general. If somebody is also on valproate, the lamotrigine titration has to go even slower, because valproate inhibits lamotrigine clearance and effectively doubles the lamotrigine level at any given dose. That is one of the most important drug interactions in the class and it bites prescribers who do not pay attention to the combination.
What to do if a rash shows up. If it is on the trunk, blistering, involving the mucous membranes (mouth, eyes, genitals), or accompanied by fever, that is an ER trip and the lamotrigine stops immediately. If it is a small benign-looking patch with no other symptoms, call the prescriber and get eyes on it before deciding. Most rashes in the first eight weeks are not serious, but you do not get to know that on your own, and the cost of a missed Stevens-Johnson is high enough that the conservative call is the right call.
The 200mg target dose is where most patients land, though some go higher for inadequate response (up to 400mg), and the drug is generally well tolerated at the upper end with no weight gain, no sedation worth mentioning, no sexual side effects, and no cognitive blunting. Headaches early on fade with time, and the patient experience on lamotrigine is usually that they are not aware they are taking a medication, which is exactly the kind of mood stabilizer experience patients want.
The valproate conversation
Valproate is what gets used when the manic side of the cycle is the bigger problem. It works fast for acute mania, faster than lithium in most cases, and is easier to titrate because the therapeutic window is wider. The trough level target is 50 to 100 mcg/mL drawn before the morning dose, and most patients land in that range on 1000 to 1500mg daily. The mechanism is mostly GABA enhancement, which is also why it works as an anti-seizure medication, which is what it was originally developed for before anybody realized it had a second life as a mood stabilizer.
The costs of valproate are real. Weight gain often substantial, often ten to twenty pounds in the first year. Hair loss, which sometimes regrows if the patient supplements with zinc and selenium and sometimes does not. Tremor at higher levels, sedation, polycystic ovary syndrome in women (its own metabolic problem and one of the reasons valproate is a tough call in any premenopausal woman), and the teratogenicity, the big one. Valproate exposure in the first trimester carries roughly a 10% risk of major malformations (neural tube defects being the classic one) and a separate signal for lower IQ and autism in exposed kids, the highest teratogenic risk in psychiatry, the reason valproate is essentially off-limits in women of childbearing age who have not actively decided they are done having kids.
The patients valproate fits well are adult men with mania-dominant bipolar I who have not done well on lithium or who need something faster acting for an acute episode. The trade-offs in that population are usually manageable, the response is reliable, the maintenance is straightforward. Liver enzymes and platelets need to be monitored (valproate can suppress both at the rare end), and the trough levels need to be checked a few times a year. The story that keeps repeating is the patient who quits valproate because of the weight gain, the weight starts coming off, they feel like they made the right call, and ninety days later they are manic in the ER… the lesson is that the move is to talk to the prescriber about switching, not to taper off into nothing.
The carbamazepine and oxcarbazepine side effects
Lithium, lamotrigine, and valproate got covered in their own sections, but the carbamazepine pair did not. Carbamazepine brings sedation that is often pronounced early, GI upset, hyponatremia (low sodium) that can present as confusion and needs a basic metabolic panel to catch, aplastic anemia and agranulocytosis as rare but serious bone marrow effects (monitored with a CBC), Stevens-Johnson at a higher rate than lamotrigine especially in patients with certain HLA markers more common in Asian populations (HLA-B*1502 is the gene to know about), drug interactions with most of what anybody is on because of liver enzyme induction, and autoinduction that climbs the dose over the first few months. Oxcarbazepine is similar but milder across the board, with one counterintuitive piece… hyponatremia is actually more common with oxcarbazepine than with carbamazepine, which is the kind of thing prescribers forget. Less sedation, fewer drug interactions, less autoinduction, generally easier to live with. The lithium side effect worth a footnote here, separately, is the cognitive slowing some patients call “lithium fog,” which usually goes away if the level is brought down.
The antidepressants-make-my-bipolar-worse conversation
Straight SSRIs in undiagnosed or untreated bipolar can flip the patient into a manic or hypomanic episode, accelerate the cycling so the highs and lows come faster, or induce a mixed state, which is the worst version of the cycle. This is real, it is well-documented, and it is the most common medication accident in bipolar care. A guy shows up with what looks like depression, gets started on sertraline or escitalopram or whichever, feels fine for two weeks, then ends up in the ER with a frank manic episode that nobody saw coming, or more commonly ends up with rapid cycling where instead of one or two episodes a year he is now cycling every couple of months and feels worse on the antidepressant than he did off it.
The lesson is not that SSRIs are bad, the lesson is that SSRIs in bipolar need to be paired with a mood stabilizer or replaced entirely. The two most common moves are either to drop the SSRI and put the patient on lamotrigine plus an atypical (which often covers the depressive side without the activation risk), or to keep a lower-dose SSRI alongside a mood stabilizer that covers the manic side, with the understanding that the patient needs to be monitored for activation. Wellbutrin (bupropion) and Trintellix (vortioxetine) have somewhat less of the activation signal than the standard SSRIs in bipolar populations, but they still carry some risk and are still not first-line monotherapy for bipolar depression.
The diagnostic version of this conversation matters too. If somebody comes in with depression that has not responded to two or three antidepressants, or where every trial has produced a brief “wow this is great” period followed by a crash, the question of whether this is actually bipolar is worth asking carefully. A lot of bipolar II goes undiagnosed for years because the hypomanic episodes felt good and never got reported as a problem, and asking specifically about periods of decreased sleep with high productivity, irritability without external cause, racing thoughts, or impulsive spending or sexual behavior is how the diagnosis gets caught.
What the patient has to do
The biggest non-medication lever in bipolar is sleep regularity. Bipolar disrupts sleep, sleep disruption triggers episodes, and you control your sleep more than you think. Going to bed at the same time every night and getting up at the same time every morning is one of the most powerful tools in the bipolar toolkit, and it is the one most patients ignore because it sounds too boring to be real medicine. It is real medicine, the patients who hold the line on sleep have a meaningfully different course than the patients who do not, and that is true even on the best medication regimen.
Alcohol use is a direct trigger, disrupting sleep architecture, kicking the mood stabilizers around metabolically, and correlating with cycling and relapse in basically every study that has looked at it. The patient who drinks four nights a week is not running the same protocol as the patient who does not, and the medication has to fight harder to hold the line. This is not a moralistic point, it is pharmacology… the drug is leverage, and alcohol is pulling on the other end of the lever.
Stimulant use, recreational and prescribed, is a complicated piece. Some bipolar patients also have ADHD and genuinely need a stimulant, in which case the move is to make sure the mood stabilizer is on board first and add the stimulant at a conservative dose with close monitoring for activation. Recreational stimulants (cocaine, methamphetamine, high-dose caffeine binges) are direct triggers for mania and are not negotiable, and if we are being honest the patient who is binging on caffeine to make it through the workday is also probably underslept, which is the same pattern wearing two costumes.
Medication adherence is the one most patients struggle with at some point, and it is worth talking about plainly. The single most common reason patients end up in the ER with another episode is that they stopped taking their medication, not that the medication failed, just that they stopped. The reasons vary, side effects, feeling better and deciding they did not need it anymore, running out and not refilling, frustration with the labs, all real and all human. The honest version is that missing doses is what reliably brings patients back to a crisis, and the conversation worth having with your prescriber is not “should I be on this” but “what is the lowest reliable version of this regimen that I can actually hold to.”
The Naming Method is the framework I use for catching prodromal shifts early. Prodromal means the early, subtle stage before a full episode, the version where your sleep has started slipping or your thoughts are getting busier or your spending is creeping up and you have not yet broken into either pole. Naming what is happening, plainly, out loud, to a person who can hold you accountable, is the single biggest predictor of catching a shift before it becomes an episode. The patients who can say “I think I am sliding” while they are still functional get the dose adjusted and ride through, and the patients who cannot name it until they are already in the episode end up in the hospital.
Patterns that keep repeating
Picture a patient who’s had three episodes, the third one bad enough to be hospitalized, finally agrees to commit to lithium and the labs, holds the line on sleep and alcohol, and does not have another full episode for two years. Two years of being able to plan a life, hold a job, be in a relationship, not be afraid of his own head. The cycle is still there in the background, but the medication and the choices together are holding it under the surface, and that is the version that happens when the math works.
Say you’ve got a patient who’s been on valproate for eighteen months, has put on twenty-five pounds, gets frustrated about the weight, decides to quit cold turkey without telling anybody, feels fine for six weeks, then ends up in the ER manic with property damage and a hospital stay and a job he is not going to get back. The weight gain was a real problem, the way to solve it was a conversation with the prescriber and a cross-taper, not a unilateral stop. Bipolar that has been treated and comes off treatment has a momentum that does not show up for a few weeks, which is why the patient thought he was fine and the relapse hit harder than the original episode.
Or imagine a patient with bipolar II who started lamotrigine, titrated up cleanly over six weeks, paired it with a hard rule on sleep (in bed by 10:30, up at 6:30, every day including weekends), cut alcohol to one drink at most twice a week, and has not had another full depressive episode in eighteen months. The medication is real, the sleep is real, the alcohol piece is real, taking any one of those out changes the picture, and together they hold the line over years.
Mood stabilizers do not cure the cycle. They hold the line. The patient does the work of staying inside that line.
Where this lands
Mood stabilizers are the most underrated class in psychiatry, in the sense that they do not get the marketing or the headlines the SSRIs and the GLP-1s and the stimulants do, and they quietly do some of the most important work in the field. Lithium is sixty years old and is still better than most of what has come after it, lamotrigine is twenty years old and one of the easiest psych meds to live on, valproate works fast when fast matters, and the class as a whole earns its keep.
If we are being honest, this is the conversation I would rather have with every bipolar patient at the start of treatment, instead of the one where we hand them a prescription and tell them it will take care of things. The drug is doing real work, the drug is also not the whole job… your sleep, your alcohol, your stimulant use, your adherence, your ability to name a shift early, your willingness to stay in regular contact with the prescriber when something feels off, all of that is yours, and all of that is what decides whether the medication holds the line over years or just over months. The medication is leverage on a set of choices that have been hard to make on willpower alone, the minute the choices stop happening the leverage stops working, and the combination is the only version that holds up two years out when the only thing left is whether the patient slept regularly, drank less, took the medication, watched the prodromes, and called when something felt wrong. None of that is set in stone at the start, all of it is decided every day.
Sources
Burgess SSA et al., Cochrane review CD003013, lithium for maintenance treatment of mood disorders. Hashimoto Y et al., Cochrane review CD013575 (2021), lamotrigine in the maintenance treatment of bipolar disorder. FDA prescribing information for lithium carbonate, lamotrigine (Lamictal), valproic acid / divalproex (Depakote), carbamazepine (Tegretol), and oxcarbazepine (Trileptal). American Psychiatric Association Practice Guideline for the Treatment of Patients with Bipolar Disorder (most recent revision). BALANCE trial (Geddes JR et al., Lancet 2010, 375:385) on lithium plus valproate versus monotherapy for relapse prevention in bipolar I. Cipriani A et al., BMJ 2013, 346:f3646, meta-analysis of lithium in the prevention of suicide in mood disorders. Crossley NA, Bauer M, J Clin Psychiatry 2007, 68:935, two meta-analyses of lithium augmentation of antidepressants in depression. Calabrese JR et al., J Clin Psychiatry 2003, 64:1013, lamotrigine and lithium maintenance in recently depressed bipolar I patients.