Counted by the decade: the lithium renal algorithm and the patient who pays over forty years
- Genre, stated plainly. This is an expert opinion with a management algorithm, published on 25 April 2026 in the International Journal of Bipolar Disorders, volume 14, article 22, open access under CC BY-NC-ND 4.0. There are 42 authors across twelve countries, led from the department of nephrology at Aalborg University Hospital and the department of clinical medicine at Aalborg University in Denmark. The authors state their motive in the conclusion: concerns regarding renal side effects have contributed to declining use of lithium in clinical practice. It is not a trial. The paper generated and analysed no data of its own, and its section on evidence and scope states that no formal systematic database search, no predefined inclusion or exclusion criteria and no structured quality scoring procedures were applied. There is no registered primary outcome, no comparison group, no blinding and no outcome assessor, and no prospective evaluation of the algorithm is reported in the publication. Confidence intervals are absent from the tables, in Table 1 and in the magnitudes of Tables 4 and 5 alike, while the body of the paper carries intervals for the same estimates.
- What the authors put forward as their own headline conclusion, in the Conclusion section: lithium treatment should neither be withheld nor continued reflexively solely on the basis of renal concerns, and what replaces the reflex is individualized, trajectory-based assessment together with shared decision-making. The same section says nephrology referral should be considered when eGFR declines below 60 mL/min/1.73 m2, with close co-management warranted once eGFR approaches 30, particularly where there is rapid decline, recurrent intoxication or significant somatic comorbidity. This trajectory-based formulation does not appear in the structured abstract, where a more general sentence stands in its place: that lithium's benefits often outweigh the risks when managed appropriately.
- The exposure arithmetic, from Table 5, whose column of magnitudes is headed "Approximate quantitative signal". Each decade of therapy, or a comparable cumulative dose increment, is given as a 20 to 40% relative risk increase. An average serum concentration of 0.60 to 0.79 mmol/L is given as a 2.9-fold increase and 0.80 to 0.99 mmol/L as a 4.3-fold increase; the body carries the same two estimates with their intervals, as hazard ratios of 2.93 (95% CI 1.97 to 4.36) and 4.31 (2.66 to 6.99) in Gislason's Icelandic cohort of 2,695 lithium users against 1,615 mood-disorder controls, in which levels of 0.30 to 0.59 mmol/L were not associated with excess chronic kidney disease. The endpoint behind those two figures is labelled stage 3 in Table 5 and stages 3 to 5 in Table 1 and in the body; I followed the body and Table 1. Any level above 1.0 mmol/L is given as approximately a twofold hazard for a first toxicity episode, with additive risk on each recurrence. Every additional 500 defined-daily-dose equivalents, roughly 18 months at one defined daily dose per day, is given as about a 30% increase. Table 4 adds that a baseline eGFR below 60 doubles the later risk of stage 3 or higher disease and that an early fall of more than 5 mL/min in the first 12 to 18 months adds a further approximately twofold risk. The note under Figure 2 states that target concentrations are individualized and that while maintenance commonly aims for 0.6 to 0.8 mmol/L, higher or lower targets may be appropriate on clinical efficacy and tolerability, which places the common target band inside the range to which Table 5 attaches the 2.9-fold signal. The paper draws that collision itself: since kidney disease risk rose with serum concentration, it states, even 0.6 mmol/L may be excessive in some patients, and it cites recommended maintenance targets of 0.4 to 0.8 mmol/L for ages 60 to 79 and 0.4 to 0.7 mmol/L from 80 years.
- Two places where the abstract does not match what stands behind it. The abstract attributes the lower incidence of end-stage kidney disease in newer studies to appropriate monitoring, whereas the Interpretation column of Table 1 attributes the same difference to the choice of comparator and to how completely cases were detected: not comparator-adjusted, likely influenced by ascertainment, surveillance bias likely, comparator-controlled. No row of Table 1 names monitoring. The abstract also states that management strategies can significantly improve outcomes, whereas the body of the paper states that no intervention has convincingly halted progression except early discontinuation of lithium. The same body text also states that once eGFR has declined to approximately 30 to 40 mL/min/1.73 m2 deterioration frequently continues despite cessation. I followed the table in the first case and the body in the second.
Lithium is followed on two records across the same years: the episodes that did not occur, and glomerular filtration. This document sets out the second. A joint task force of the International Group for the Study of Lithium Treated Patients and the International Society for Bipolar Disorders states what to measure before lithium begins, how often to repeat it, and at which numbers a nephrologist enters the case, and it states its reason: concerns regarding renal side effects have contributed to declining use of lithium in clinical practice. It is an expert opinion carrying an algorithm, not a trial, and the authors say so themselves: no systematic search, no predefined inclusion or exclusion criteria, no structured quality scoring.
What the algorithm consists of
Before lithium is started, and once or twice yearly afterwards, the paper asks for serum creatinine with an estimated glomerular filtration rate, urea nitrogen, sodium, potassium and calcium, thyroid and parathyroid function, and an electrocardiogram. Frequency rises with age, a reduced baseline eGFR, a fast fall, somatic comorbidity, and any episode of intoxication. Once-daily dosing is preferred. Table 3 maps stages onto management: at an eGFR of 45 to 59 monitoring intensifies and nephrology input is considered, at 30 to 44 referral is recommended with dose reduction, at 15 to 29 co-management is mandatory, below 15 lithium is usually discontinued. Table 6 works in rates: a fall slower than 3 mL/min per year with eGFR above 40 favours continuation, a fall faster than 5 mL/min per year, or an eGFR in the 30 to 40 zone, favours reduction or cessation. Where lithium is tapered, the taper runs over no less than three months, with another mood stabiliser introduced in parallel and a relapse action plan agreed.
The arithmetic of exposure
Table 5 turns duration into a risk figure rather than counting events. Its four rows stand in the third key finding above; the column of magnitudes is headed "Approximate quantitative signal", and the asterisk has no accompanying footnote in any of the three representations checked: the publisher PDF, the structured full text and the PMC page. The tables carry no intervals, but the body does, for the same estimates: the two largest, 2.93 and 4.31, come with 95% confidence intervals of 1.97 to 4.36 and 2.66 to 6.99, measured against mood-disorder controls in a cohort where mean levels of 0.30 to 0.59 mmol/L showed no excess kidney disease.
Two lines meet on the same patient. The note under Figure 2 gives 0.6 to 0.8 mmol/L as the common maintenance target while stating that targets are individualized, and Table 5 attaches the 2.9-fold signal to an average level of 0.60 to 0.79. The paper does not leave that unremarked: it states that since risk rose with serum concentration, even 0.6 mmol/L may be excessive in some patients, and it cites targets of 0.4 to 0.8 mmol/L for ages 60 to 79 and 0.4 to 0.7 mmol/L from 80 years. Table 6 records that in an older or frail patient the competing risk of death from other causes may exceed the likelihood of reaching end-stage disease. What accumulates, accumulates in whoever is exposed longest at the common target: the patient started at twenty with four decades ahead.
Where the abstract parts company with the tables
The abstract credits appropriate monitoring with the lower rate of end-stage disease in newer studies, while the Interpretation column of Table 1 explains that contrast by comparator choice and case ascertainment and names monitoring in no row. The abstract also states that management strategies can significantly improve outcomes; no table compares monitored with unmonitored care.
What stopping does not buy back
What accumulates is stage 3 chronic kidney disease, not dialysis: Table 1 records Bosi 2023 at a hazard ratio of 1.11, 95% CI 0.86 to 1.45 in the body, with a ten-year risk of 8.4% against 8.2%, no increase in severe endpoints for Kessing 2015 and Hayes 2016, and Chan 2025 at 1.35, 1.15 to 1.60, for stage 3 and above with no end-stage association. The renal counterweight is stated directly: once eGFR has declined to approximately 30 to 40 mL/min/1.73 m2, deterioration frequently continues despite lithium cessation, which the paper calls a pragmatic renal point of no return. Table 7 puts numbers to it: in Bocchetta 2017, 42% gained 5 to 10 mL/min within 12 months but only 8% regained their pre-lithium level; in van Alphen 2021 the slope stabilised only where eGFR at stopping was at least 40; in Fransson 2025 reversal was confined to a baseline above 35. The second counterweight is in Table 6: recurrent attempts or persistent ideation are listed as arguments for continuation, with an approximately 60% suicide risk reduction against comparators given in the cell without an interval and referenced to earlier literature, and the same row records that the rebound in suicidal behaviour peaks in the first 6 to 12 months after lithium is stopped. Stopping late is not the symmetrical remedy for what the exposure tables count, and the document declines to resolve any of this into a rule.
The sentence the authors placed in their conclusion and not in their abstract: lithium treatment should neither be withheld nor continued reflexively solely on the basis of renal concerns, and what replaces the reflex is an individualized, trajectory-based assessment that balances substantial therapeutic benefit against renal risk, preserving lithium's role as a first-line treatment while safeguarding long-term patient health.
This is an expert opinion with a management algorithm, not a study. The paper generated and analysed no data of its own, there is no registered primary outcome, no comparison group, no blinding and no outcome assessor, and no prospective evaluation of the algorithm is reported in the publication. The authors state in their section on evidence and scope that no formal systematic database search, predefined inclusion or exclusion criteria, or structured quality scoring procedures were applied. Confidence intervals are absent from the tables: the hazard ratios in Table 1, including Shine 2015 at 1.93, Bosi 2023 at 1.11, Gislason 2024 at 1.90 and Chan 2025 at 1.35, are given there as point values, and the magnitudes in Tables 4 and 5 sit under a column headed "Approximate quantitative signal" whose asterisk has no accompanying footnote in the publisher PDF, in the structured full text or on the PMC page. The body of the paper does report intervals for those same estimates, among them 1.76 to 2.12 for Shine, 0.86 to 1.45 for Bosi, 1.32 to 2.75 for Gislason, 1.15 to 1.60 for Chan, and 1.97 to 4.36 and 2.66 to 6.99 for the two serum-concentration figures of Table 5, so a reader working from the tables alone sees less precision than the paper holds. Registration of the work is not stated in the full text, the declarations, the ethics section or the data availability statement, and the article carries no supplementary material. The funding section states that no specific funding was achieved for this study, while the acknowledgements list named grants to individual authors, among them an NHMRC Leadership 3 Investigator grant GNT2017131, CTSA grant KL2 TR002379 from the National Center for Advancing Translational Science, the Berlin University Alliance award 611_Nachwuchsgruppe, the German Federal Ministry of Education and Research project LiNaKre 13GW0484B, the European Union Horizon programme 101057454 PsychSTRATA, the German Federal Ministry for Economic Affairs and Climate Action project TransCare KK5680901AP4, the German Association for Psychiatry, Psychotherapy and Psychosomatics, the Spanish Ministry of Science and Innovation grants PI22/00840, PI18/01055 and PI21/00713, the La Marató-TV3 Foundation grant 202230-31, CIBERSAM, the Basque Government grant 2022111054 and the University of the Basque Country grant IT1631-22. The competing interests section reads in full that the authors declare no competing interests, and it covers all 42 authors. The document's origin in the joint task force "Role of Lithium In Bipolar Disorder" of the International Group for the Study of Lithium Treated Patients and the International Society for Bipolar Disorders is stated in a footnote and not in the competing interests section. An author contributions section is present: three authors wrote the first draft, the remaining authors critically revised the manuscript, and all authors read and approved the final version. The renal outcome studies harmonised in Table 1 and the discontinuation series in Table 7 are observational, and the paper itself names confounding by indication and differences in surveillance intensity as limitations of that literature.