Bipolar II maintenance: the evidence base for lithium and lamotrigine is thinner than the guidelines suggest
- Systematic review of 10 randomised trials, total n = 645 adults with bipolar II disorder, comparing lithium, lamotrigine, and placebo across acute depression, hypomania, and maintenance phases.
- Lithium vs placebo for prevention of new affective episodes: RR = 0.86 (95% CI 0.66–1.11; I² = 0%; 4 studies). Confidence interval crosses 1 — no statistically reliable signal of relapse prevention in BD-II.
- Lithium vs lamotrigine on depression severity (HAMD-17): mean difference = 1.27 (95% CI −3.84 to 6.38; 2 studies). The two drugs are statistically indistinguishable here, but the comparison rests on under 200 patients.
- Risk of bias was rated "some concern" for 84% of outcomes; GRADE certainty was "very low" for every outcome examined. A pooled lamotrigine-vs-placebo synthesis was not even possible because the trials were too few and too heterogeneous.
If you treat bipolar II patients, you have probably been telling them that lithium prevents relapse and lamotrigine prevents the depressive pole. The Copenhagen group (Fredskild, Kessing, Munkholm) just put numbers on how thin that claim actually is. After 70 years of lithium and 25 years of lamotrigine, BD-II maintenance evidence still rests on roughly 645 randomised patients — fewer than a single mid-size depression RCT.
What the data shows
The lithium-vs-placebo relapse-prevention estimate (RR 0.86) leans in lithium's favour but the confidence interval (0.66 to 1.11) is wide enough to include no effect at all. Heterogeneity was zero — the trials agreed with each other, they just collectively failed to deliver enough events. That is a classic signature of an under-powered evidence base: consistent direction, not enough signal to reach statistical certainty.
The head-to-head lithium-vs-lamotrigine comparison on HAMD-17 (MD 1.27, CI spanning −3.84 to +6.38) is even more diluted. Two trials. Almost any clinically meaningful difference is compatible with the data. The lamotrigine-vs-placebo arm could not be quantitatively pooled at all — which means the most prescribed BD-II maintenance drug in many countries has, by 2025 standards, no meta-analytic placebo comparison for its core indication.
GRADE "very low" across all outcomes is not a footnote. It is the headline.
For your practice
This does not mean stop using lithium or lamotrigine in BD-II — there is no better evidence-based alternative, and the indirect evidence from BD-I and naturalistic registries continues to support both. What changes is the conversation with the patient. The honest framing is: "We are recommending the best-supported option we have for your subtype, but the trial evidence for BD-II specifically is sparse and uncertain — so we will track your response carefully and adjust early."
That changes three concrete things: 1) you build in earlier review points (8 and 12 weeks, not just 6 months) because you cannot lean on confident expected timelines from population data; 2) you document mood charting more rigorously, because individual response data partly substitutes for missing population-level certainty; 3) you stop framing lamotrigine and lithium as interchangeable on the depressive pole — the head-to-head data are not strong enough to support that equivalence claim either way.
Bipolar II is treated, in many practices, as if it were "BD-I lite" — but the maintenance evidence base for it has barely a tenth of BD-I's statistical mass, and our prescribing confidence should reflect that gap.
All ten trials carried "some concern" or higher risk of bias, dropout was high in maintenance arms, and BD-II diagnostic boundaries shifted across the studies (some pre-DSM-IV criteria) — so even pooled, the synthesis is closer to a careful mapping of uncertainty than a definitive comparative effect estimate.