Old Salt, New Molecules: An Elenchus on Bipolar Medication
A Personal Ellenchus — by Dr. Chris
Greetings all.
An elenchus is the Socratic method of cross-examination. You take a belief everyone treats as settled, you ask it questions until it either holds or falls, and you do not spare your own beliefs in the process. So before I cross-examine the standard bipolar prescription pad, I will admit up front that I walked into this post carrying two premises of my own that did not survive the checking. I will show you exactly where they broke, because the argument gets stronger without them.
What follows are the questions I think every member of this page should bring to their next appointment.
Why is gabapentin still here
Two randomized, placebo-controlled trials published in 2000 found no advantage for gabapentin over placebo in any phase of bipolar disorder, and one of them found it worse than placebo for mania (Fullerton, Busch and Frank, Medical Care). The same paper shows that gabapentin’s spread into bipolar prescribing between 1997 and 2000 closely tracked marketing dollars aimed at psychiatrists, and in 2004 Warner-Lambert paid $430 million to resolve criminal and civil charges over precisely that kind of off-label promotion (U.S. Department of Justice). A quarter century later the evidence still does not support gabapentin for bipolar disorder (Psychiatric Uses of Gabapentin, PMC), and the CANMAT/ISBD guidelines rate it negative for mania (CANMAT/ISBD 2018 guidelines).
If gabapentin is on your list “for bipolar,” the question is simple: what is it actually treating? Anxiety, sleep and nerve pain are honest answers. “Mood stabilizer” is not one.
Gold standard, or just old
Here is where my first premise broke. I came in ready to write that lithium is prescribed off-label. It is not. The FDA label lists lithium as monotherapy for acute manic and mixed episodes and for maintenance treatment of bipolar I (FDA lithium label), and it has held that approval since April 1970 (FDA safety review).
I also came in ready to call lithium a heavy metal, like copper. It is the opposite. Lithium is the lightest metal on the periodic table, with a density about half that of water (Chemistry LibreTexts). So the chemistry insult is wrong. The fair charge is different, and better: lithium is an element, not a designed molecule. It was not built for a receptor. It has a narrow window between the dose that works and the dose that poisons, and it asks for blood draws for the rest of your life.
The receipts on both sides are real. Lithium cuts one-year relapse from 61 percent on placebo to 36 percent (Cochrane data summarized in PMC), and it is the bipolar medication with the strongest evidence for reducing suicide (BMJ 2013 meta-analysis). It also nearly doubles the risk of stage 3 chronic kidney disease and more than doubles the risk of hypothyroidism (The Lancet), and the kidney risk roughly triples once you pass ten years on it (population-based study, PMC).
So the elenchus question is not “is lithium bad.” It is: “Doctor, is lithium the best drug for me, or is it the default because it is the one you were trained on?”
Why is valproate a reflex
Depakote is on-label too, approved for manic episodes of bipolar disorder (FDA Depakote label). The problem is not the label. It is what the drug does to a pregnancy that has not happened yet. As many as 11 percent of exposed children are born with major malformations and up to 40 percent show neurodevelopmental delay (MHRA). The United Kingdom now bars valproate in anyone who could become pregnant unless they are enrolled in a formal Pregnancy Prevention Programme (Valproate and the Pregnancy Prevention Programme, PMC), and its teratology service says flatly that valproate should not be used to treat bipolar disorder in pregnancy (UKTIS).
Question: if you are a person who could become pregnant, why was this the drug reached for first?
Half a stabilizer
Lamotrigine is on-label as well, approved in June 2003 for maintenance treatment of bipolar I (FDA). Read the fine print, though: maintenance. For acute mania, CANMAT/ISBD rates lamotrigine Level 1 negative, which is the strongest grade of evidence that a drug does not work (CANMAT/ISBD 2018 guidelines). It is half a stabilizer: a decent umbrella against depression, useless in a fire. If you are manic and someone hands you a titration schedule that takes six weeks to reach a working dose, ask what is supposed to hold you in the meantime.
Where the illness actually lives
In the landmark long-term study of bipolar I, patients were symptomatic 47 percent of the weeks across nearly thirteen years of follow-up, and depressive symptoms (32 percent of weeks) outnumbered manic and hypomanic symptoms (9 percent) by more than three to one (Judd et al., Archives of General Psychiatry). Depression, not mania, is what tracks most closely with lost jobs and lost function in this illness (PMC review). So the sharpest question of the whole elenchus is this one:
Why are we still starting with drugs from 1970 and 1995 when there are molecules that were tested specifically for the phase we actually spend our lives in?
Cariprazine (Vraylar) was approved for bipolar I mania in September 2015 and for bipolar I depression in May 2019 (FDA review). Alongside quetiapine, it is one of only two monotherapies approved for both poles of the illness (PMC). The 2023 CANMAT/ISBD evidence update moved it to first-line for bipolar I depression, on the same shelf as lithium, quetiapine, lurasidone and lamotrigine (CANMAT/ISBD 2023 update). Lumateperone (Caplyta) was approved in December 2021 for depressive episodes of bipolar I or II, as monotherapy or as an add-on (Intra-Cellular Therapies); the same guideline update lists it as second-line for now (CANMAT/ISBD 2023 update). In pooled short-term trials, lumateperone’s weight and metabolic changes were indistinguishable from placebo, and not a single patient had clinically significant weight gain (metabolic analysis).
These are not obscure drugs. They are on-label, guideline-listed and built for bipolar depression. Ask why you had to read about them on Facebook.
What the new drugs cost
An elenchus that spares its own favorites is just advertising, so let me turn it on my side of the table.
Cariprazine’s akathisia, that maddening inner restlessness, hit 20 percent of patients in the mania trials versus 5 percent on placebo, and 6 to 10 percent in the depression trials, by the manufacturer’s own numbers (Allergan and Gedeon Richter). Every antipsychotic, new or old, carries a tardive dyskinesia risk; the annualized incidence across the second-generation drugs is about 2.6 percent, lower than the old generation’s 6.5 percent but not zero (meta-analysis, PMC). Then there is the price. Vraylar lists at $1,594.82 for a 30-day supply as of January 2026 (Vraylar), and Caplyta launched at $1,320 a month (Reuters). Lithium costs pennies. That gap is why your insurer makes you “fail” the old drugs first, and it is a legitimate thing to be angry about. But it is the insurer’s failing, not the lithium’s.
One more honest tension. Caplyta’s own bipolar approval includes use as an add-on to lithium or valproate (FDA Caplyta label). The newest molecule was tested standing on the shoulders of the oldest salt. The future is probably not “new instead of old.” It is “the right one for the phase you are in.”
What about magnesium
I wanted to tell you that if lithium works for you, magnesium threonate would work better because it crosses the blood-brain barrier. I went looking for the evidence. It is not there. The systematic review of magnesium in mood disorders calls the evidence inconclusive (magnesium and mood disorders meta-analysis, PMC), and I could not find a single controlled trial of magnesium threonate in bipolar disorder. Take magnesium if you like it. Do not swap it for a mood stabilizer.
The one rule
Whatever you decide to ask, do not stop anything cold. When lithium is stopped rapidly, the median time to the next episode is four months; tapered slowly, it is twenty (Baldessarini et al.). Bring questions to the appointment, not an ultimatum to the medicine cabinet.
Off-label was never the problem. Off-evidence is. Gabapentin is both. Lithium, valproate and lamotrigine are on-label and real, with real costs the profession has been too casual about for too long. Cariprazine and lumateperone are on-label, built for the phase we actually live in, and too often the last thing offered rather than the first thing discussed.
Ask why.
Love,
Dr. Chris
This post is personal opinion and lived experience, not medical advice. Do not change or stop any medication without talking to your prescriber.
