Published on 16 September 2026
Once-daily or twice-daily lithium: what does brain MRI show at 12 hours?
In brief
Lithium has been prescribed since 1949 by monitoring a blood concentration, for want of anywhere else to look. A German and British team used lithium-7 MRI to measure the brain signal in 41 euthymic patients with bipolar disorder, three times in one day, with a matched blood sample at each scan. The main result: 12 hours after the last dose, the sampling time used in practice for the immediate-release formulation, the brain signal and the serum concentration do not differ between a single daily dose and two doses. Over the course of the day, however, the two profiles diverge, and the brain closely follows serum rather than dampening its variations. The ⁷Li signal is higher in white matter than in grey matter, in signal intensity rather than absolute concentration, and brain-to-serum ratios point towards faster equilibration in grey matter. These are 41 patients split into two groups according to the regimen they were already taking: the abstract reports no randomisation, and the cross-sectional design effectively rules it out. All were White European, aged 18 to 50, recruited in a single city. The outcome is an imaging signal, not a clinical result.
The context
The question comes up at every initiation. One dose in the evening, or two doses a day? The usual argument for once-daily dosing is adherence; the argument for twice-daily dosing is the smoothing of concentration peaks, aimed at peak-related effects such as tremor and gastrointestinal intolerance. On the renal side, the available data point rather the other way and favour a single evening dose (Carter et al., 2013). The two regimens are considered clinically equivalent in efficacy, but they produce very different serum curves over the day.
Hence the question asked here, which is a question not of efficacy but of compartment: does the brain dampen the serum fluctuations of once-daily dosing, or does it reproduce them? The answer determines whether it is legitimate to use the same serum lithium range for both regimens.
The mechanism
What lithium-7 MRI measures, and what it does not
The ⁷Li nucleus can be detected by magnetic resonance, but its signal is weak, which calls for long acquisitions and a spatial resolution far below that of proton MRI. The published abstract speaks of signal intensity, not of absolute concentration expressed in millimoles per litre. The distinction is not cosmetic: it limits what can be claimed.
| What is measured | What it establishes | What it does not tell us |
|---|---|---|
| Whole-brain ⁷Li signal, brain observable lithium threshold (BOLT) approach | Primary outcome. The presence of lithium in the brain and its variation over the ten hours observed | Neither the absolute concentration, nor the intracellular distribution, nor the occupancy of lithium’s molecular targets |
| Tissue-specific profiles, Newcastle lithium image voxel evaluation (N-LIVE) pipeline | Secondary analysis. Lithium is not evenly distributed: higher signal in white matter, faster equilibration in grey matter than in white matter and cerebrospinal fluid | Why this distribution exists, and whether it has any clinical consequence at all |
| Serum lithium matched to the three scans | That the brain signal follows the serum curve specific to each regimen | Nothing about the compartment in which lithium actually exerts its mood-stabilising effect |
| Population pharmacokinetic modelling over 08:00, 14:00 and 18:00 | The shape of the profiles over a ten-hour window | Night-time exposure, which was not observed: the protocol stops at 18:00 whereas the evening dose is taken at 20:00 |
The study at a glance
| Question (PICO) | |
|---|---|
| Population | |
| Adults aged 18 to 50 years with bipolar disorder type I or II, euthymic, on stable lithium carbonate. 44 eligible participants recruited, 3 withdrawals before assessment, 41 assessed. All White European, 21 women (51%) and 20 men (49%), mean age 39.2 years (SD 9.8). Four psychiatric outpatient departments in Dresden, Germany. Recruitment from 19 April 2022 to 9 December 2023 | |
| Intervention | |
| Three consecutive ⁷Li MRI scans over a 10-hour period (08:00, 14:00, 18:00) with matched serum sampling at each scan. In the preceding week, a dose at 20:00 for everyone, plus a dose at 08:00 for the twice-daily group, withheld on assessment day until after the 08:00 sample | |
| Comparator | |
| Once-daily dosing (20 patients) versus twice-daily dosing (21 patients). Between-subject comparison, on the regimen already in use. The abstract reports no randomisation, and the cross-sectional design effectively rules it out | |
| Primary outcome | |
| Whole-brain lithium measurements across timepoints, compared by regimen using the BOLT region-of-interest approach | |
| Design | |
| Repeated-measures cross-sectional study, imaging surrogate outcome, no clinical outcome · repeated-measures ANOVA, linear mixed-effects models, population pharmacokinetics · patient and public involvement and engagement (PPIE) in the design |
The quality check
| Criterion | Status |
|---|---|
| Repeated measures and serum-brain matching | Sound |
| FindingThree scans and three simultaneous samples, a standardised dosing protocol in the preceding week, morning dose withheld on test day so that the 08:00 timepoint is comparable | |
| Participant flow | Sound |
| Finding44 recruited, 3 withdrawals before assessment, reasons not reported, 41 assessed | |
| Patient involvement | Sound |
| FindingDesign informed by people living with bipolar disorder and on long-term lithium treatment | |
| Funding | Reservation |
| FindingThe only funders declared in the abstract are the Baszucki Brain Research Foundation and the Deutsche Forschungsgemeinschaft, both non-industry. The PubMed record does, however, carry a declaration of interests: one author declares a relationship with Philips Healthcare, an MRI manufacturer, within a research agreement that provides access to advanced MRI scanner research functionality and research tools, which bears directly on the technique studied. The other declarations concern academic funding, including the European R-LiNK programme that funded the network of lithium-7 MRI centres, and roles in learned societies | |
| Allocation of regimen | Reservation |
| FindingBetween-subject comparison of regimens chosen in routine practice. The abstract reports no randomisation, and the cross-sectional design effectively rules it out. The gold standard would be a within-patient crossover protocol | |
| Sample size | Reservation |
| Finding20 versus 21 patients. No power calculation and no equivalence margin declared in the abstract | |
| Nature of the outcome | Reservation |
| FindingSurrogate outcome. No clinical outcome measured: no relapse, no renal tolerability, no thyroid function | |
| Protocol registration | Not verifiable |
| FindingThe record contains no registration field. The point cannot be verified as things stand, so it is not counted against the study | |
The findings
| Outcome | Published result |
|---|---|
| 12 hours post dose | Brain ⁷Li signal and serum lithium equivalent between regimens |
| PEB readingNo equivalence margin no confidence interval and no p value appear in the abstract: the study has not demonstrated equivalence, it has found no difference | |
| Serum profile over the day | Once-daily: steady decline over the day. Twice-daily: rise after the morning dose, then decline |
| PEB readingConsistent with the expected pharmacology | |
| Brain and serum concordance | Secondary analysis. The brain signal closely follows the serum profile specific to each regimen |
| PEB readingThe brain does not dampen fluctuations, it reproduces them the primary outcome is the only whole-brain comparison between regimens; everything else comes from secondary analyses | |
| Tissue distribution | Secondary analysis. ⁷Li signal higher in white matter than in grey matter, in both regimens |
| PEB readingReplicates earlier work by the same team | |
| Speed of equilibration | Brain-to-serum ratios indicating faster equilibration in grey matter than in cerebrospinal fluid and white matter |
| PEB readingPhysiological description, no established clinical translation | |
| Clinical outcomes | None |
| PEB readingNo efficacy, tolerability or renal function data the study says nothing about which regimen to prefer for a given patient | |
What the study establishes fits in one sentence: over the window observed, the brain compartment does not behave as a buffer. The idea of cerebral dampening of peaks, unresolved until now, finds no support here. That does not make once-daily dosing preferable; it removes a theoretical argument from one of the two options.
Critical appraisal
| Domain | Judgement |
|---|---|
| Selection and comparability of groups | Reservation |
| FindingThe abstract reports no randomisation, and the cross-sectional design effectively rules it out. A patient is on once-daily dosing for reasons of their own, adherence, gastrointestinal tolerability, total dose, and those reasons are potentially associated with what is being measured. Residual confounding that cannot be controlled with 41 patients | |
| Characterisation of exposure | Reservation |
| FindingThe abstract reports neither the daily doses by group nor the formulation used, immediate or prolonged release. Transposition to settings where once-daily dosing relies on a prolonged-release formulation is therefore not direct | |
| Outcome measurement | Reservation |
| FindingSignal intensity on ⁷Li MRI, a research technique available in a very small number of centres, with a detection threshold that itself defines the region analysed | |
| Observation window | Reservation |
| FindingTen daytime hours. The hours following the 20:00 dose, when peaks are expected in the once-daily group, are not observed | |
| Analysis and reporting | Reservation |
| FindingAppropriate statistical methods, but the abstract states equivalence without providing the elements needed to judge it | |
| External validity | Reservation |
| Finding18 to 50 years, all White European, four departments in one city. The population in whom the choice of regimen matters most, older people with declining renal function, is excluded by the age criteria | |
| Fit between claim and evidence | Reservation |
| FindingThe authors do not confine themselves to description. Their abstract suggests that these differences in profile “might inform regimen selection and therapeutic monitoring strategies in clinical practice”. The opening is phrased conditionally, but it projects a clinical use that none of the study’s outcomes supports | |
One point deserves to be put the opposite way from how it will be put elsewhere. This study validates no serum lithium range. It reports no target value, and the question of ranges appears only in the interpretation, as a discussion these data feed into. In France, for example, the summary of product characteristics for lithium carbonate sets an effective serum lithium level of 0.5 to 0.8 mEq/L, whatever the pharmaceutical form. What changes with the form, in that document, is not the target but the sampling time: 12 hours after the dose for the immediate-release form, 24 hours after the dose for the prolonged-release form. The value of 0.8 to 1.2 mEq/L often quoted is not a different target: it is an intermediate marker measured at the twelfth hour on the prolonged-release form, corresponding to the same exposure. A German study that does not specify the formulation therefore does not tell us which sampling reference its measurements relate to.
Level of evidence
PEB appraisal: high confidence in the description, low confidence in the comparison between regimens. That brain lithium follows serum lithium, that the signal is higher in white matter, that grey matter equilibrates faster: these observations are measured directly, replicate earlier work and do not depend on the comparison between groups. Confidence falls as soon as the claim is that the two regimens are equivalent, because the abstract reports no randomisation and the cross-sectional design effectively rules it out, because the sample is 20 versus 21, and because no equivalence bound is published.
The colleague test
What an experienced colleague would say if you put this study to them in two minutes, between two consultations.
“ What interests me is that the brain buffers nothing: that buries the theoretical argument for twice-daily dosing, and that I will keep. The rest, no. They compare twenty patients with twenty-one, with no randomisation reported, on an image, in people under fifty whose renal function is not reported. My patients at risk are precisely the others. ”
What this means in practice: we gain a pathophysiological argument, we do not gain a prescribing rule. The choice of regimen is still decided on the adherence and tolerability of the patient in front of you.
What you can do with this on Monday morning. No prescription changes, but four things become usable straight away.
- Put the time of the last dose on the laboratory request, not just a request for a lithium level. This study shows that the brain signal moves with serum over the day: a sample taken at 10:00 in a patient who took their tablet at 8:00 does not measure the same thing as a sample taken 12 hours after the dose, and the two regimens diverge precisely at these intermediate times.
- Answer the patient who asks why a single evening dose, staying within the limits of what was measured. Over the window observed, from 8:00 to 18:00, the brain signal follows serum: nothing supports the idea of a cerebral buffer that would dampen variations. The window of the evening peak, after the 20:00 dose, was not measured. The adherence argument, for its part, stands intact.
- After a change of regimen, recheck the serum lithium level after five to seven days, once steady state is reached, rather than relying on the previous figure, and realign the sampling time with the new dosing time.
- Go back over the safety reminder, which this study does not cover and which remains the true determinant of prognosis on lithium. The contraindications listed in the French summary of product characteristics are hypersensitivity, breastfeeding, renal failure where strict and regular monitoring of serum lithium and creatinine is impossible, heart failure, Addison’s disease, Brugada syndrome or a family history of Brugada syndrome. A low-sodium diet is not a contraindication but a toxicity factor, as is dehydration. Vigilance over non-steroidal anti-inflammatory drugs, angiotensin-converting enzyme inhibitors, angiotensin receptor blockers and thiazide diuretics, over dehydration, fever and periods of intense heat; monitoring of creatinine and glomerular filtration rate, TSH and serum calcium; information for every woman of childbearing potential about the risk of cardiac malformation.
One open question, which the study does not settle and which matters more than the one it asks: does once-daily dosing protect the kidney over twenty years? Brain kinetics say nothing about long-term renal tolerability, which remains one of the main reasons for stopping.
Frequently asked questions
Is once-daily lithium as good as twice-daily dosing?
This study cannot say so. It finds no difference at 12 hours, in 41 patients for whom the abstract reports no randomisation, with no published equivalence margin. An absence of a significant difference is not evidence of an absence of difference. It simply makes the idea that the brain dampens peaks less plausible.
Should the timing of serum lithium sampling change?
No. The sampling reference depends on the pharmaceutical form: in the French summary of product characteristics cited above, 12 hours after the dose for the immediate-release form, 24 hours for the prolonged-release form, with the same effective serum lithium level of 0.5 to 0.8 mEq/L. The study measured the twelfth-hour timepoint, the one at which the two regimens converge. That is an argument for keeping existing references, not for changing them.
Do these results apply to prolonged-release lithium?
The published abstract does not specify the formulation. A prolonged-release form flattens the serum profile by design, which shifts the question without removing it. Transposition remains a reasonable hypothesis, not a result.
Is higher lithium signal in white matter a cause for concern?
Nothing allows us to say so. The result replicates earlier observations by the same team. No link with efficacy, with adverse effects or with any clinical marker was tested here.
Will lithium MRI reach routine clinical practice?
Not in the short term. It requires equipment and methodological development available in a few research centres. Its current value lies in answering pharmacological questions, not in following patients.
Annotated bibliography
Ritter P, Edelmann K, Thelwall PE, Boumezbeur F, Bauer M, Werner A, Hall G, Kavanagh O, Cousins DA (2026). Brain lithium temporospatial kinetics in bipolar disorder: a repeated-measures ⁷Li MRI study of dosing regimens in Germany. The Lancet Psychiatry, 13(8), 647-656. DOI 10.1016/S2215-0366(26)00153-7 · PMID 42365853. The source study analysed here. Funding: Baszucki Brain Research Foundation and Deutsche Forschungsgemeinschaft.
Smith FE, Thelwall PE, Necus J, Flowers CJ, Blamire AM, Cousins DA (2018). 3D ⁷Li magnetic resonance imaging of brain lithium distribution in bipolar disorder. Molecular Psychiatry, 23(11), 2184-2191. DOI 10.1038/s41380-018-0016-6 · PMID 29426954. The work that established the feasibility of three-dimensional mapping of brain lithium and described its heterogeneous distribution. A team sharing several authors with the study analysed here, which explains the methodological continuity but limits the independence of the replication.
Necus J, Sinha N, Smith FE, Thelwall PE, Flowers CJ, Taylor PN, Blamire AM, Cousins DA, Wang Y (2019). White matter microstructural properties in bipolar disorder in relationship to the spatial distribution of lithium in the brain. Journal of Affective Disorders, 253, 224-231. DOI 10.1016/j.jad.2019.04.075 · PMID 31054448. Explores the link between the spatial distribution of lithium and white matter microstructure. Useful for placing the white matter predominance found here, without any causal link being established in either direction.
Neal MA, Strawbridge R, Wing VC, Cousins DA, Thelwall PE (2024). Human brain ⁷Li-MRI following low-dose lithium dietary supplementation in healthy participants. Journal of Affective Disorders, 360, 139-145. DOI 10.1016/j.jad.2024.05.128 · PMID 38810780. Shows the sensitivity of the technique to very low intakes in healthy participants. Sheds light on the detection limit of ⁷Li MRI, which underpins the threshold approach used in the source study.
Carter L, Zolezzi M, Lewczyk A (2013). An updated review of the optimal lithium dosage regimen for renal protection. Canadian Journal of Psychiatry, 58(10), 595-600. DOI 10.1177/070674371305801009 · PMID 24165107. A synthesis of the data comparing once-daily and divided dosing from the standpoint of renal function, whose conclusion favours a single evening dose. It serves here to correct the renal argument often attributed to divided dosing. A narrative review without meta-analysis: it points a direction, it does not demonstrate.