Four weeks, three months, two years: NICE CG185 times every taper and sets no threshold for planned withdrawal
- NICE clinical guideline CG185, "Bipolar disorder: assessment and management", was published on 24 September 2014 and last amended on 2 September 2025. It replaced CG38 of July 2006, and the guideline states its own scope of change in the Context chapter: "All areas of CG38 have been updated." Its stated purpose is on the overview page: "It aims to improve access to treatment and quality of life in people with bipolar disorder." Reading costs nothing and no account is asked for: recommendations, the plain-language version for patients, the full guideline and the appendices all download without registration. Reuse is a separate question, since the files carry the notice "All rights reserved. Subject to Notice of rights."
- The document carries 136 numbered recommendations in eleven sections, from care across all phases (1.1) to children and young people (1.11). Behind them sit a full guideline of 390 pages, 36 appendices distributed across 27 files (1 to 10 in one PDF, 11 to 36 separately, number 35 in Excel), and 23 numbered review questions in the published review protocols, grouped in five topic areas. GRADE was applied to questions about the effectiveness of interventions; for questions about the experience of care and about the organisation and delivery of care, methodology checklists were used instead. Strength is carried by the wording: "The strength of each recommendation is reflected in the wording of the recommendation, rather than by using ratings, labels or symbols."
- Planned stopping is timed in detail. Lithium is to be reduced gradually over at least 4 weeks and preferably up to 3 months, even if another antimanic drug has been started (1.10.25), with close monitoring for early signs of mania and depression during reduction and for 3 months after (1.10.26). An antipsychotic and lamotrigine are each to be reduced over at least 4 weeks (1.10.13, 1.10.39), and valproate over at least 4 weeks when the purpose is "to change to an alternative medication" (1.10.35). When long-term treatment is being stopped, early signs of relapse are to be discussed and the person monitored (1.7.11), and symptoms, mood and mental state are to be monitored for 2 years after medication has stopped entirely (1.7.12).
- A threshold for offering planned withdrawal is not set in sections 1.7 and 1.10 of the recommendations page: no count of episodes, no months of remission, no other measured quantity. Recommendation 1.7.1 requires a discussion after each episode covering, among other things, the risk of relapse after reducing or stopping medication "for an acute episode", the potential benefits and risks of long-term medication, the potential benefits and risks of stopping medication "including for women who may wish to become pregnant", and "possible duration of treatment, and when and how often this should be reviewed". Recommendation 1.7.11 begins after the decision has been made.
- Stopping on safety grounds is a different matter and is specified. Valproate is to be stopped immediately if abnormal liver function or blood dyscrasia is detected, with the accompanying note that hepatic enzymes persistently elevated to over 3 times the upper normal limit, rising, or accompanied by clinical symptoms, mean the suspected drug should be withdrawn (1.10.30); a woman or girl of childbearing potential already taking valproate is to be advised to stop it gradually under medical supervision (1.10.27). The nearest thing to a rule about who decides on planned stopping is 1.2.9: where bipolar disorder is managed solely in primary care, a person "considering stopping any medication after a period of relatively stable mood" is listed as a reason to re-refer to secondary care.
- The base text is 2014. Database searches began in July 2012 and, in the words of the full guideline, "the final re-runs carried out in January 2014 ahead of the guideline consultation"; after that point only studies judged exceptional were admitted. The amendment of 2 September 2025 concerns valproate in men and updates links to technology appraisal guidance in two sections; it is not a re-examination of the evidence on maintenance treatment. Development was commissioned by NICE and carried out within the National Collaborating Centre for Mental Health, which is funded by NICE; the guideline development group met 13 times. Appendix 2 holds 22 declaration blocks, three of which record a personal pecuniary interest; the field "Action taken" is filled 14 times, 13 of them reading "None" and one reading "Trustee for Bipolar UK".
What the document is, and what it costs
CG185 is the current NICE guideline for bipolar disorder. NICE commissioned it; the National Collaborating Centre for Mental Health, a partnership led by the Royal College of Psychiatrists and the British Psychological Society, developed it. The eleven sections run from care across all phases of the disorder to children and young people, taking in primary care, assessment, crisis, mania, bipolar depression, longer-term management, physical health monitoring and how to use medication. Recommendations on carers and on physical health were not developed here: the full guideline presents them in a table of recommendations "incorporated or adapted from another NICE guideline", the psychosis and schizophrenia guideline, where their evidence base lies. Physical health is handled by transfer, with an annual check in primary care (1.2.12).
How stopping is timed
The withdrawal instructions are among the most concrete in the document. Lithium is reduced over at least 4 weeks and preferably up to 3 months, and the instruction holds even when another antimanic drug has already been started (1.10.25); monitoring for early signs of mania and depression runs through the reduction and for 3 months beyond it (1.10.26). An antipsychotic and lamotrigine each carry a minimum of 4 weeks (1.10.13, 1.10.39), and valproate the same minimum where the stated purpose is to change to an alternative medication (1.10.35). In each case the reason given is to minimise relapse risk. Above these sit 1.7.11, which requires that early warning signs be discussed and the person monitored, and 1.7.12, which extends monitoring of symptoms, mood and mental state to 2 years after medication has stopped entirely.
Where the threshold is not written, and where it is
What is timed is the method; for planned withdrawal the trigger sits inside a conversation: 1.7.1 requires a discussion after each episode covering, among other items, the benefits and risks of stopping medication "including for women who may wish to become pregnant" and "possible duration of treatment, and when and how often this should be reviewed"; 1.7.6 repeats it for each drug. Neither section 1.7 nor section 1.10 sets a threshold in episodes or in months of remission at which planned withdrawal should be offered. Stopping for safety is written differently: valproate is to be stopped immediately on abnormal liver function or blood dyscrasia, the marker given being enzymes over 3 times the upper normal limit (1.10.30), and gradual withdrawal under supervision is prescribed for women and girls of childbearing potential already taking it (1.10.27). The patient version addresses the reader directly: "If you decide to stop taking medication, your doctor should discuss with you how to tell if you are becoming unwell again and what to do if that happens", and lists "How long will I have to take medication?" among questions to ask. NICE's own research recommendation records that the relative effects of lithium, an antipsychotic or their combination, regarding efficacy, tolerability, cost effectiveness and quality of life, are unknown.
What changed, and what the change covered
The guideline names its own headline change in the Context chapter: "All areas of CG38 have been updated." The update log records amendments between 2015 and 2025, most tracking MHRA safety advice on valproate, plus a separate list of minor changes for July 2024, October 2024 and March 2025. Of the year tags, 119 read [2014] and one reads [2025]; the two untagged items are 1.5.13 and 1.11.9, both pointers to technology appraisals, and those are the two sections whose links the September 2025 entry says were updated. The July 2024 change removed recommendations covered elsewhere; NICE published no list, but archived copies of the page show 150 numbered items in June 2024 against 139 in September 2024, the eleven being 1.1.10 and 1.1.11 on advance statements and lasting power of attorney, the carer support block 1.1.12 to 1.1.19, and 1.4.3. The 2023 change has its own 46-page evidence review, calling itself an urgent update that "did not follow the processes described in the Guidelines Manual" and states that "no quality assessment of the included studies was undertaken as recency was prioritised over quality".
The guideline times every taper to the week and specifies immediate withdrawal when the liver or the blood count says so; for planned stopping of maintenance treatment, sections 1.7 and 1.10 name no threshold in episodes or months of remission and leave the moment inside a required conversation.
The base text is 2014; searches began in July 2012 and the final re-runs were carried out in January 2014 ahead of consultation, after which only studies judged exceptional were admitted. The 2 September 2025 amendment concerns valproate in men and two sets of updated links, not the evidence on maintenance treatment. The statement that no threshold is set applies to planned withdrawal and to sections 1.7 and 1.10 of the recommendations page; withdrawal on safety grounds is specified in 1.10.30 and 1.10.27, and 1.2.9 makes a patient's consideration of stopping a reason to re-refer from primary to secondary care. Seventeen documents were read: the overview, recommendations, context, research recommendations, committee details and update information pages, the Evidence, History and Tools and resources tabs, three chapters of the information for the public, the full guideline, appendices 1 to 10, the review protocols, the 2023 valproate evidence review, quality standard QS95, and two archived versions of the recommendations page from June and September 2024. Six sets of files were listed and not read: appendices 11 to 36, the 2017 surveillance report with its two appendices, the 2021 exceptional surveillance on diabetes testing with its single appendix, individual GDG meeting minutes, quality standard QS102, and the endorsed competence framework. The word England does not appear on the CG185 overview, recommendations or context pages; the statement about territorial scope appears in the disclaimer of the 2023 evidence review, which says NICE guidelines cover health and care in England with decisions for other UK countries taken by their governments, and on QS95, which says quality standards apply in England and Wales. The individual declarations of the 2023 working group are not printed in the 46-page review, which states only that the declaration of interests policy was applied. Other national guidelines were revised in the same window, among them the practice guidelines of the Japanese Society of Mood Disorders published in 2024. Access is free to read; the files are not open-licensed and carry "All rights reserved. Subject to Notice of rights."