Skip to content

Lithium

TL;DR — Lithium remains the best-supported single maintenance treatment across manic and depressive polarity, but benefit depends on sustained exposure, serum concentration and active renal, thyroid, calcium and toxicity surveillance (Miura 2014, PMID 26360999). Enriched randomized evidence estimates any-relapse RR 0.52 (95% CI 0.41–0.66; NNT 2.3), while real trial cohorts still show about 40% recurrence over roughly 78 weeks (Oya 2019, PMID 31026388; Kishi 2020, PMID 32701902). Serum-level evidence suggests 0.4–0.8 mmol/L reduces recurrence versus below 0.4, while an older randomized comparison found 0.4–0.6 carried 2.6 times the relapse risk of 0.8–1.0 mmol/L (Hsu 2021, PMID 34227095; Keller 1992, PMID 1586272). Toxicity synthesis finds hypothyroidism, impaired urinary concentration, hyperparathyroidism and weight gain, with a small absolute renal-replacement risk; newer kidney evidence estimates impaired function in about one quarter of long-term users but with substantial heterogeneity (McKnight 2012, PMID 22265699; Schoretsanitis 2022, PMID 34783413). Suicide evidence is genuinely disputed: a 2013 meta-analysis found fewer suicides, whereas a 2022 rare-event reanalysis was inconclusive.

Where lithium’s evidence is strongest

Lithium’s main comparative advantage is breadth. The 2014 maintenance network meta-analysis concluded that it should remain first line because it prevented both manic and depressive recurrence and had higher-quality supporting evidence than alternatives, notwithstanding tolerability (Miura 2014, PMID 26360999).

Setting Quantitative result Interpretation
Enriched maintenance trials Any relapse RR 0.52 (95% CI 0.41–0.66), NNT 2.3 (1.6–4.2) Large effect among people stabilized and randomized in two small trials (Oya 2019, PMID 31026388)
Recently manic/hypomanic Lithium prolonged time to any intervention vs placebo, P=.006; delayed mania/hypomania/mixed, P=.006 Strong elevated-polarity prevention (Bowden 2003, PMID 12695317)
Recently depressed Median intervention-free time 170 d vs 93 d placebo; mania-free at 1 y 86% vs 72% Depression prevention less distinct than lamotrigine; mania prevention clear (Calabrese 2003, PMID 14628976)
BALANCE Primary event 59% lithium vs 69% valproate; HR 0.71 (0.51–1.00) Lithium favored over valproate over up to two years (BALANCE 2010, PMID 20092882)
Acute bipolar depression MADRS −13.6 vs −11.8 placebo, P=.123 Maintenance efficacy should not be misrepresented as rapid acute antidepressant efficacy (Young 2010, PMID 20122369)

The enriched maintenance RR is not a universal treatment-start estimate. Participants first had to stabilize and tolerate treatment, a design that selects likely beneficiaries (Oya 2019, PMID 31026388).

Residual risk despite continuation

A single-group synthesis of 21 randomized trials, 1,415 participants and mean 78.4-week duration estimated the following while lithium was continued (Kishi 2020, PMID 32701902):

Outcome Pooled event rate (95% CI)
Any mood recurrence 39.8% (32.8–47.1)
Depressive recurrence 25.6% (18.8–34.0)
Manic/hypomanic/mixed recurrence 18.5% (13.7–24.7)
All-cause discontinuation 67.0% (57.2–75.5)
Adverse-event discontinuation 8.7% (5.1–14.7)

These rates show why a relapse plan remains necessary during successful pharmacotherapy. They should not be read as untreated counterfactual risks because the underlying trials used different designs and many were short or enriched.

Serum concentration and dose–response

Lithium has a narrow therapeutic window; “taking lithium” is not a binary exposure. A 2021 multicenter cohort and dose-response meta-analysis included 1,406 participants and found lower recurrence at 0.4–0.8 mmol/L than below 0.4, adjusted HR 0.75; the 0.8–1.2 group had adjusted HR 0.77 but did not reach significance in the cohort (Hsu 2021, PMID 34227095).

Serum comparison Outcome Caveat
Each 0.1 mmol/L increase Dose-response OR 0.85 for recurrence Observational contribution; confounding by indication possible (Hsu 2021, PMID 34227095)
Modeled 0.4 mmol/L Recurrence OR 0.42 relative to modeled zero Model-based, not a randomized threshold
Modeled 0.8 mmol/L Recurrence OR 0.27 relative to modeled zero Benefit must be balanced against concentration-related harms
0.4–0.6 vs 0.8–1.0 mmol/L 2.6-fold major-relapse risk at low range Older randomized 94-person maintenance trial (Keller 1992, PMID 1586272)

In the randomized level trial, low-range participants also had nearly twice the risk of subsyndromal symptoms. The first subsyndromal symptom increased major-relapse risk fourfold; 76% developing subclinical hypomania relapsed versus 39% developing subclinical depression (Keller 1992, PMID 1586272). This supports symptom monitoring alongside concentration monitoring.

Target concentration cannot be chosen from efficacy alone. Age, renal function, interacting medication, acute versus maintenance phase, previous response and adverse effects all change the acceptable balance; exact protocols belong in current guidelines rather than being inferred from these studies.

Suicide: an unresolved evidence conflict

Lithium’s anti-suicide reputation has major clinical weight, but randomized rare-event evidence depends strongly on inclusion and analytic method.

Synthesis Trials / participants Suicide result Other result Reading
Cipriani 2013 48 RCTs / 6,674 OR 0.13 (95% CI 0.03–0.66) vs placebo All-cause death OR 0.38 (0.15–0.95); self-harm OR 0.60 (0.27–1.32) Supports suicide prevention (PMID 23814104)
Nabi 2022 12 eligible modern trials / 2,578 0.2% lithium vs 0.4% control; OR 0.41 (0.03–2.49), P=.45 No significant nonfatal suicidal behavior or attempt effect Inconclusive; does not confirm specific protection (PMID 36111461)

The newer review included all eligible rare-event data and emphasized that previous analyses excluded trials with zero events or difficult reporting. Its exact-method estimate was OR 0.42 (95% CI 0.01–4.5), an interval compatible with large benefit, no effect or harm (Nabi 2022, PMID 36111461). The correct synthesis is uncertainty, not reversal: randomized trials are too sparse in suicide events to settle the magnitude.

Renal, thyroid, parathyroid and weight effects

The 2012 toxicity review screened 5,988 abstracts and included 385 studies. It found a coherent organ-system profile rather than undifferentiated “toxicity” (McKnight 2012, PMID 22265699).

Outcome Quantitative estimate Meaning
Glomerular filtration rate −6.22 mL/min (95% CI −14.65 to 2.20), P=.148 Mean difference imprecise and not significant in that synthesis
Urinary concentrating ability −158.43 mOsm/kg (−229.78 to −87.07) Consistent concentrating defect
Renal replacement therapy 18/3,369 (0.5%) Severe renal outcome uncommon in included data
Clinical hypothyroidism OR 5.78 (2.00–16.67) vs placebo Clear thyroid risk
TSH +4.00 IU/mL (3.90–4.10) Biochemical effect supports surveillance
Serum calcium +0.09 mmol/L (0.02–0.17) Small mean shift, relevant to parathyroid monitoring
Parathyroid hormone +7.32 pg/mL (3.42–11.23) Consistent hyperparathyroid signal
Weight gain vs placebo OR 1.89 (1.27–2.82) Meaningful but less than olanzapine in comparative data

A newer kidney-specific meta-analysis of 20 studies and 25,907 lithium-treated patients estimated impaired kidney function prevalence 25.5% (95% CI 19.8–32.2). Compared with 722,529 non-lithium-treated patients, the pooled OR was 2.09 (1.24–3.51); prevalence rose with longer treatment duration (Schoretsanitis 2022, PMID 34783413).

The newer estimate does not contradict the low renal-replacement rate. “Impaired kidney function” spans much milder endpoints than end-stage failure. Cross-sectional studies gave 14.5% prevalence while retrospective studies gave 29.5%, demonstrating design sensitivity (Schoretsanitis 2022, PMID 34783413).

Monitoring logic

This page does not prescribe a jurisdiction-specific schedule. It maps each evidence-supported risk to the variable that can detect it.

Risk Monitoring domain Evidence basis
Underexposure / recurrence Trough serum lithium plus symptom course Recurrence fell across concentration range; low randomized range relapsed more (Hsu 2021, PMID 34227095; Keller 1992, PMID 1586272)
Reduced filtration Creatinine/eGFR trajectory, not a single value Impaired kidney function prevalence 25.5%, duration-associated (Schoretsanitis 2022, PMID 34783413)
Concentrating defect Thirst, polyuria and urinary concentrating symptoms Mean concentrating capacity fell by 158 mOsm/kg (McKnight 2012, PMID 22265699)
Hypothyroidism TSH and clinical thyroid status Hypothyroidism OR 5.78 (McKnight 2012, PMID 22265699)
Hyperparathyroid physiology Calcium, with PTH evaluation when indicated Calcium and PTH increased in pooled evidence (McKnight 2012, PMID 22265699)
Weight/metabolic burden Weight trajectory and relevant metabolic assessment Weight-gain OR 1.89 vs placebo (McKnight 2012, PMID 22265699)
Pregnancy exposure Preconception/perinatal risk review and dose context Cardiac-malformation risk was dose-associated (Patorno 2017, PMID 28591541)

Monitoring is useful only if abnormal trends lead to a reasoned response. A single threshold should not erase competing risks: abrupt cessation can raise recurrence, while continuing unchanged can worsen organ injury.

Pregnancy signal

In a Medicaid cohort of 1,325,563 pregnancies, cardiac malformations occurred in 16/663 lithium-exposed infants (2.41%) versus 15,251/1,322,955 unexposed infants (1.15%); adjusted RR was 1.65 (95% CI 1.02–2.68) (Patorno 2017, PMID 28591541).

First-trimester daily dose Adjusted RR for cardiac malformation (95% CI)
600 mg or less 1.11 (0.46–2.64)
601–900 mg 1.60 (0.67–3.80)
More than 900 mg 3.22 (1.47–7.02)

Right-ventricular outflow obstruction defects occurred in 0.60% exposed versus 0.18% unexposed, adjusted RR 2.66 (1.00–7.06) (Patorno 2017, PMID 28591541). This supports a dose-sensitive shared decision, not automatic continuation or automatic withdrawal; maternal relapse risk is part of the same equation. See pregnancy and reproductive health.

Discontinuation and reintroduction

Across maintenance drugs, discontinuation increases recurrence, and abrupt stopping may compress risk into the early months (Kishi 2021, PMID 33046156). A narrower question is whether lithium becomes pharmacologically less effective after it is restarted.

Five studies addressed this; three entered meta-analysis with 212 cases. The pooled odds of at least one relapse after interruption/restart versus uninterrupted treatment were 1.40 (95% CI 0.85–2.31), P=.19 (de Vries 2013, PMID 23911110). Evidence was too sparse to confirm discontinuation-induced refractoriness.

This should not be distorted into reassurance about abrupt cessation. Recurrence risk from being off maintenance and uncertain response after restart are separate questions.

Toxicity as a systems problem

Serum lithium can rise when clearance or volume status changes. The evidence above establishes a narrow efficacy–harm balance, but these abstracts do not provide a complete interaction protocol. The red-flags and safety page should hold current interaction and urgent-toxicity guidance.

For research and documentation, any suspected toxicity episode should distinguish:

Question Why it matters
Was exposure acute, chronic, or acute-on-chronic? Serum concentration and tissue burden can diverge
Did kidney function or hydration change? Clearance is central to accumulation
Was a new interacting medicine introduced? Preventable systems failure may be present
Were neurologic, gastrointestinal or cardiac features documented? Severity cannot be inferred from a number alone
What was the post-event maintenance plan? Avoids trading toxicity for unmanaged recurrence

Evidence limitations

  • Maintenance trials commonly enrich for responders and tolerators.
  • Kidney studies use heterogeneous impairment definitions and designs (Schoretsanitis 2022, PMID 34783413).
  • Suicide is rare enough that randomized estimates remain extremely imprecise (Nabi 2022, PMID 36111461).
  • Serum concentration studies mix randomized and observational evidence.
  • Pregnancy cohorts capture live births and residual confounding, not every pregnancy outcome.
  • Monitoring schedules are guideline-dependent and should not be invented from meta-analysis alone.

Dose–response is polarity dependent

Across six randomized trials (n=975), modeled recurrence odds fell as serum lithium increased: OR 0.50 at 0.60 mmol/L and 0.15 at 1.20 mmol/L for any episode. Depression prevention required a modeled 1.13 mmol/L to reach the same 56% risk reduction achieved for mania at 0.60 mmol/L (Hsu 2022, PMID 35158229). This does not make 1.13 mmol/L a universal target; toxicity, age, kidney function and non-random allocation to achieved concentration constrain the curve.

Kidney effects: measurable decline, uncertain event risk

A lithium-versus-non-lithium meta-analysis found mean eGFR 11.14 mL/min/1.73 m² lower with lithium (95% CI −16.61 to −5.68; n=1,622; I²=86%) and annual decline 0.13 mL/min/1.73 m² faster (0.06–0.20; n=13,280). New/progressive chronic kidney disease did not differ significantly (OR 2.16, 0.59–7.94; n=17,740; I²=99%), an interval too imprecise to establish equivalence (Macaron 2026, PMID 41727809).

Evidence in older adults remains extrapolative. A focused review found sparse trials and recommends the same broad first-line agents as in working-age adults, but with greater attention to somatic comorbidity, interactions and adverse effects; lithium can remain an option with close monitoring rather than being excluded by age alone (Arnold 2021, PMID 34201098).

New interaction and prediction uncertainties

Three cases described rising lithium after semaglutide initiation; two developed toxicity despite stable renal function, while pre-emptive dose reduction mitigated toxicity in the third. Proposed mechanisms include reduced intake, gastrointestinal fluid loss and delayed gastric emptying, but a case series cannot establish a pharmacokinetic class interaction (Al-Soleiti 2025, PMID 40999647).

A lithium-response GWAS of 2,563 people identified four linked chromosome-21 variants at genome-wide significance. In a prospective 73-person cohort, carriers of response-associated alleles had lower relapse (reported HR 3.8 for alternate versus associated alleles, 95% CI 1.1–13.0), but the validation interval was wide and no clinical decision rule followed (Hou 2016, PMID 26806518). Pharmacogenomic selection therefore remains investigational.

Open questions

  • What individualized serum target maximizes relapse prevention while minimizing kidney and endocrine burden (Hsu 2021, PMID 34227095)?
  • Can kidney-risk trajectories distinguish reversible hemodynamic change from progressive tubulointerstitial injury before clinically important decline (Schoretsanitis 2022, PMID 34783413)?
  • Which rare-event analytic framework best estimates lithium’s suicide effect, and what trial or target-trial design could narrow the interval (Cipriani 2013, PMID 23814104; Nabi 2022, PMID 36111461)?
  • Can structured monitoring reduce the 67% all-cause discontinuation observed in lithium trials without sacrificing efficacy (Kishi 2020, PMID 32701902)?
  • What is the safest taper strategy after long stability, and does restart response depend on taper speed or time off lithium (de Vries 2013, PMID 23911110)?

References

  1. Miura T, et al. Comparative efficacy and tolerability of pharmacological treatments in maintenance treatment of bipolar disorder. Lancet Psychiatry. 2014. PMID 26360999
  2. Oya K, et al. Efficacy and safety of lithium and lamotrigine for maintenance treatment: a systematic review and meta-analysis. Neuropsychopharmacology Reports. 2019. PMID 31026388
  3. Kishi T, et al. Recurrence of mania or depression among adult bipolar patients who continued using lithium. Journal of Clinical Psychopharmacology. 2020. PMID 32701902
  4. Hsu CW, et al. Lithium concentration and recurrence risk during maintenance treatment: multicenter cohort and meta-analysis. Acta Psychiatrica Scandinavica. 2021. PMID 34227095
  5. Keller MB, et al. Subsyndromal symptoms in bipolar disorder: comparison of standard and low serum levels of lithium. Archives of General Psychiatry. 1992. PMID 1586272
  6. Bowden CL, et al. Placebo-controlled 18-month trial of lamotrigine and lithium in recently manic or hypomanic patients. Archives of General Psychiatry. 2003. PMID 12695317
  7. Calabrese JR, et al. Placebo-controlled 18-month trial of lamotrigine and lithium in recently depressed patients. Journal of Clinical Psychiatry. 2003. PMID 14628976
  8. BALANCE investigators and collaborators. Lithium plus valproate combination therapy versus monotherapy for relapse prevention (BALANCE). Lancet. 2010. PMID 20092882
  9. Young AH, et al. Quetiapine and lithium monotherapy in acute bipolar depression (EMBOLDEN I). Journal of Clinical Psychiatry. 2010. PMID 20122369
  10. Cipriani A, et al. Lithium in the prevention of suicide in mood disorders: updated systematic review and meta-analysis. BMJ. 2013. PMID 23814104
  11. Nabi Z, et al. Effects of lithium on suicide and suicidal behaviour: systematic review and meta-analysis of randomised trials. Epidemiology and Psychiatric Sciences. 2022. PMID 36111461
  12. McKnight RF, et al. Lithium toxicity profile: a systematic review and meta-analysis. Lancet. 2012. PMID 22265699
  13. Schoretsanitis G, et al. Prevalence of impaired kidney function in patients with long-term lithium treatment. Bipolar Disorders. 2022. PMID 34783413
  14. Patorno E, et al. Lithium use in pregnancy and the risk of cardiac malformations. New England Journal of Medicine. 2017. PMID 28591541
  15. Kishi T, et al. Recurrence rates after drug discontinuation versus maintenance: systematic review and meta-analysis. Psychological Medicine. 2021. PMID 33046156
  16. de Vries C, et al. Effectiveness of restarted lithium treatment after discontinuation. Bipolar Disorders. 2013. PMID 23911110
  17. Macaron MM, et al. Lithium nephrotoxicity: a systematic review and meta-analysis of lithium versus non-lithium control studies in affective disorders. Ther Adv Psychopharmacol. 2026;16:20451253261419633. PMID 41727809
  18. Hsu CW, et al. Differences in the prophylactic effect of serum lithium levels on depression and mania in bipolar disorder: a dose-response meta-analysis. Eur Neuropsychopharmacol. 2022;58:20–29. PMID 35158229
  19. Al-Soleiti M, et al. Lithium toxicity and altered clearance following initiation of semaglutide in patients with bipolar disorder: a case series and literature review. J Clin Psychopharmacol. 2025;45:613–618. PMID 40999647
  20. Arnold I, et al. Old age bipolar disorder—epidemiology, aetiology and treatment. Medicina (Kaunas). 2021;57:587. PMID 34201098
  21. Hou L, et al. Genetic variants associated with response to lithium treatment in bipolar disorder: a genome-wide association study. Lancet. 2016;387:1085–1093. PMID 26806518