Kleine-Levin Syndrome: Lithium Prophylaxis Without a Randomized Trial
A single adolescent patient, six recurrent hypersomnia episodes into a two-year course of Kleine-Levin syndrome. The question isn't whether lithium can help — cohort data suggest it can — it's whether starting it now, on evidence that has never cleared a randomized trial, is the right trade for a still-growing sixteen-year-old to make first.
Milo H., a 16-year-old boy, spent most of last spring building an electric go-kart with his high school's engineering club, working toward a regional competition he has now missed two years running — not for lack of a working kart, but because both competition weeks landed inside one of his sleep episodes. He was previously healthy until a mild sore throat and low-grade fever in the spring of freshman year gave way three days later to nineteen or twenty hours of sleep a day, confused and short-tempered when roused. That first episode lasted nine days and ended as abruptly as it began. Five more have followed, spaced three to five months apart and lasting six to twelve days each — close to the ten-day median reported across large KLS series, and nowhere near the multi-week episodes that mark the severe end of the disease. Between them he returns completely to baseline, itself part of what makes KLS rather than something progressive the working diagnosis. Workup — brain MRI, EEG, thyroid panel, toxicology, sleep study, psychiatric evaluation — has found nothing. His most recent episode ended three weeks ago.
There is no approved treatment and, per Oliveira and colleagues' Cochrane review updated in 2016, no completed randomized trial of any kind — a gap the disease's rarity, on the order of one to five cases per million, makes unlikely to close. The largest treatment data available, Leu-Semenescu and colleagues' single-center cohort of 130 patients published in Neurology in 2015, set 71 who took lithium against 49 who took nothing and found episodes reduced in roughly a third, evidence the authors themselves grade Class IV. Forty of those 71 were children, so on age he sits inside the one cohort that exists. What complicates reading it as a clean drug effect is that cohort's baseline: before treatment the lithium patients tended toward more episodes per year (3.8 against 2.9) and had spent measurably longer incapacitated (57 days against 37), meaning the sicker end of the group was the end selected for the drug. Against that sits KLS's own tendency to remit over one to two decades regardless of treatment — so six episodes in two years reads as a statement about his present burden, not about where his course was already heading. The reactive alternative fits him less well than it first appears. Léotard and colleagues' 2018 Neurology cohort of IV methylprednisolone was scoped to long episodes, and counted a response as an episode running at least a week shorter than the one before, averaging twelve days saved — arithmetic his six-to-twelve-day episodes cannot produce. He is inside the studied population for the drug being held back, and outside it for the drug about to be given.
Interepisode clinic visit, three weeks out from episode six
Start lithium now. Six episodes in two years, costing him around forty school days between them and, twice, something he cared about specifically — that's not a mild course, and there's no version of watching and waiting that doesn't mean at least a few more episodes like the last six before anyone revisits this. The largest cohort we have, a hundred-thirty patients followed at a single center by Leu-Semenescu and colleagues and published in Neurology in 2015, found episodes reduced in roughly a third of the seventy-one lithium-treated patients — and forty of those seventy-one were children, which is more than can be said for most of what gets cited at him. That's Class IV evidence by the authors' own grading, not proof, but it's the best evidence that exists for anything in this disease, and "nothing better is coming" is itself a real consideration when the condition is rare enough that a randomized trial may never get built.
If this were his first or second episode, I wouldn't be making this argument — the case for watching a new diagnosis settle before committing to a chronic drug is a real one. Six episodes in, with a clear pattern established, is a different moment.
You're right that it's the best evidence that exists for anything in this disease — I'm not disputing that. What I'd push back on is treating that cohort's response rate as if it were a clean drug effect.
Leu-Semenescu's own paper reports that before treatment started, the patients who went on to receive lithium trended toward more episodes per year — 3.8 against 2.9 — and had spent substantially longer incapacitated, 57 days against 37. The sicker end of the population was the end selected for the drug, which is exactly the setup that can produce apparent improvement through regression toward the mean, independent of what the lithium itself is doing. And Oliveira and colleagues' Cochrane review, updated in 2016, found zero completed randomized trials of any pharmacologic treatment for this disease — that's not a hedge, that's the literal state of the evidence.
Layer on that lithium requires years of renal and thyroid monitoring in a patient who's still growing, with a narrow enough therapeutic window that under- and over-dosing are both real risks, and I think the honest answer is that we don't know the true size of the benefit we'd be buying — only that it's probably not zero.
There's a version of this that doesn't require either of you to be fully right yet. IV methylprednisolone, given within the first ten days of an episode, showed a real effect on shortening episode length in a comparative cohort led by Léotard and colleagues, published in Neurology in 2018 — roughly two-thirds of patients responded when it was given that early, against essentially no spontaneous shortening in the untreated comparison group. That's a tool deployed only when something is actually happening, not a drug he takes every day on the chance it prevents the next one — a materially smaller commitment for a sixteen-year-old who's going to be managing this through at least the next several years of his life, however this goes. If episodes keep coming at anything like this pace despite treating them early, lithium is exactly where this heads next — nothing about starting here forecloses it.
Then let me hold that to the same standard I just used on lithium, because it doesn't survive it either. Léotard's cohort is twenty-six patients, retrospective and open-label — a weaker design than the one I spent the last five minutes taking apart — and it was assembled around patients having long episodes. Their threshold for calling something a response was an episode at least a week shorter than the one before it, and the early-infusion figure you're quoting averaged twelve days saved. His episodes run six to twelve days.
There is no version of that arithmetic that lands on him. And the 2021 American Academy of Sleep Medicine guideline goes the other way from where we're heading: it suggests lithium for Kleine-Levin syndrome and makes no recommendation on methylprednisolone at all — both statements written for adults, which he also isn't.
I'm not arguing us out of the abortive plan. I think it's the right call for a sixteen-year-old, because it's reversible and a daily drug for years isn't. But that's a reasoning about commitment, not about evidence, and I'd rather we write it down that way than let it go in the chart looking like the data pointed here. They don't. They point, weakly, somewhere else.
Agreed: an abortive protocol is now in place for the next episode — IV methylprednisolone started as early as it's recognized, ideally within the first several days. Baseline renal and thyroid labs were drawn today specifically so that starting lithium later, if it comes to that, isn't delayed by paperwork. Modafinil remains available as a symptomatic bridge during any future episode, understood explicitly as not addressing the frequency question. Recorded in the note at the clinical pharmacologist's insistence: the abortive-first sequence was chosen because it is reversible in a sixteen-year-old, not because the evidence favors it over lithium — which, on the guideline and on his age fit with the cohorts, it does not.
Not agreed, and left explicit rather than smoothed over:
The sleep medicine specialist's threshold for starting lithium as a matter of course, not a fresh discussion each time — given how much has already been lost waiting.
The pediatric neurologist's threshold, weighing an already-heavy junior year and genuine uncertainty whether steroids will do for six-to-twelve-day episodes what they did for the much longer ones in Léotard's cohort.
The clinical pharmacologist didn't take either number as evidence-derived and said so directly without picking a side: neither the one-episode nor the one-year threshold comes from a study of this specific decision — both are reasonable clinical judgment, not something the literature actually specifies.