It may be worth mentioning the potential dangers of expansion of the term bipolar disorder to its spectrum. Specific evidence does not exist for treatment in "softer" bipolar spectrum. Prescribers may be prescribing beyond the evidence if they do not make a specific diagnosis of bipolar I (and possibly II).

This situation may be reminiscent of the overinclusive use of the term schizophrenia in USA in the 1960s, which led to a tightening of diagnostic criteria in DSM-III, particularly a narrowing of the definition of schizophrenia. Maybe we should now also have concern about the validity of the broadening of the concept of bipolar disorder. Estimates of the proportion of the population with bipolar spectrum are as high as at least 4-5%. The use of the diagnosis bipolar disorder has become variable and unreliable.

The definition of bipolar disorder in the guideline is insufficiently critical. For example, there is considerable variability in studies of the overall morbidity in first degree relatives, which was the basis for defining the disorder. Has there really been an advance in creating the term bipolar disorder, rather than the traditional term manic-depressive illness? If so, could the guideline make clear what it thinks it is?

Thank you for mentioning an important issue. Further discussion of this point has been added (3.1.1 'The bipolar spectrum' p.40 end of paragraph)

We broadly agree with this comment but feel the recommendations regarding diagnosis in this Guideline are evidence based, vigorous and avoid over-inclusiveness.

What is the evidence that biological factors are more primary that psychosocial factors in aetiology? As far as we know there is none. Why therefore does the guideline seem to agree with "the current mainstream view".

There is no suggestion that there is evidence that biological factors are primary, the guideline merely asserts that that is the current view.

Discussion of psychosocial factors in aetiology is included as well as discussion of biological factors.

It may be worth making clear that ICD10 does not recognise bipolar II as a separate category.

This has been added (3.4.1, 'ICD-10', final sentence 2^nd para, p53).

More could be said about the question of the term 'mood stabiliser'. Common understanding of the term may encourage unrealistic expectations of the effects of drugs that are so designated and is almost certainly responsible for the widespread prescription of such drugs to patients with a wide range of diagnoses. When people hear the term they may misunderstand it to mean that the drug provides a kind of "mental tranquillity" or similar state, which may well seem highly desirable if it could be achieved if one's life is in emotional turmoil. The term "mood stabiliser" implies a specificity of action which is unjustified by evidence and obscures the fact that all such drugs are basically sedative, and may well exert their "antimanic" effects in this way.

The evidence in clinical trials relates to prevention of further episodes of mania and depression, in people diagnosed with bipolar disorder (most of the convincing evidence from bipolar I patients only). From this, a possibly unjustified extrapolation is made that these drugs "smooth" or stabilise mood. Moreover, it is somewhat incidental that lithium was first used as a prophylactic agent. Schou et al had tried lithium in several manic patients and were convinced of its effect. It does not necessarily logically follow that a drug which had a therapeutic effect during an ongoing manic episode must also prevent further episodes of mania, but this seems to be what Schou and colleagues unthinkingly assumed.

Why use the term mood stabiliser at all if the definition is really antimanic drugs (plus lamotrogine), as used in the guideline? The guideline does not fully explain why it thinks that the term mood stabiliser is unlikely to be abandoned. It could set a trend by doing so.

After all, antidepressants are supposed to be prophylactic for further episodes of depression. Aren't these as much 'mood stabilisers', if the term really must be used, as much as antimanic drugs? Perhaps because lithium was primarily used in clinical practice as a prophylactic agent, rather than for its antimanic effects, the term mood stabiliser was introduced. Hypothetically, if antidepressants stopped being used for the acute phase and were only used for prophylaxis they may well also be called mood stabilisers.

The term mood stabilisation therefore seems to primarily refer to prophylaxis. Its continued use as a term allows drugs so identified to acquire a market, perhaps particularly in "softer" bipolar spectrum conditions, for which they are not fully justified.

Thank you - we agree that 'mood stabiliser' is not a good term and have changed it to 'anti-manic agent' or 'anti-manic medication' where appropriate in the treatment of acute episodes and 'prophylactic agent/medication' in the long-term section. We do not consider antidepressants to be mood stabilising since, in people with bipolar disorder, they can induce mania.

There seems to be insufficient discussion of the potential bias in trials of acute mania because of restricted selection and high drop out rates.

Thank you - we have added some text on these issues.

There seem to be several instances where the guideline acknowledges the weakness of the evidence but then concludes that specific drugs are effective. For example, the efficacy of valproate is questionable may well be a better wording of the conclusion than "it is reasonable to regard it as effective".

Thank you - in light of this and other comments we have revised the clinical summary as appropriate. In addition, additional data which we had previously missed from an update to the Bowden1994 trial, has been added.

Has this recommendation about benzodiazepines crept in because atypical antipsychotics may not be as sedative as traditional neuroleptics? Where is the discussion about the potential use of traditional neuroleptics, or has the market for treatment of mania completely been taken over by the newer drugs?

BZDs have always had a place in the management of acute behavioural disturbance and insomnia.

Reference to evidence to potential anti-suicidal effect of lithium should be made with more caution. Although it is mentioned that the evidence comes from observational studies, it is not apparent that the implication is that the link is correlational, not necessarily causal. Also, there are several well designed studies that find higher rates of suicide in people on lithium, and the positive ones have methodological flaws such as excluding people lost to follow up. Mentioning that the effect may be "nonspecific" understates the lack of causality. There may well be other explanations for the correlation eg. as mentioned but perhaps not as explicitly stated as it might be, lithium patients in these studies may well be unusual in some ways and tend to be selected, not necessarily representative of lithium patients in general for whom suicide rate may well be as expected in bipolar disorder.

Thank you - we have redrafted this section.

There does not seem to be any discussion of the bias in discontinuation studies due to the potential confounding because of discontinuation effects. Why do more recent placebo controlled lithium discontinuation studies not find as great an effect of lithium as previous reviews (and in one case more or less no effect) - is it worth discussing this? Non-specific factors do seem to be important in these studies as evidenced by the finding that rapid withdrawal is associated with higher relapse than gradual withdrawal.

Thank you, we have amended the guideline in light of your comments. We should point out that this section was meant to provide a brief overview of study design rather adding it to the summaryof study characteristics. We have also amended our recommendations concerning lithium.

The guideline about when to start prophylactic treatment seems to be a "rule of thumb" without evidence - this should be stated as such.

Thank you - it is a central responsibility of the guideline development group to use consensus methods or offer expert opinion in areas were high quality evidence does not exist, as is made clear in our methods section. We would be interested on your views on when to start prophylactic treatment.

Where is the evidence about the need for 5 years of prophylactic treatment? Why for example should this recommendation not be for a much shorter period? Where is the advice about helping patients discontinue medication if they wish? Where is there any discussion about the potential advantages of dealing with personal and social problems in bipolar disorder without reliance on medication and weighing up the pros and cons of prophylactic medication?

Thank you for this comment we believe that this period is sufficient to allow for an evaluation of the long term benefits of treatment and a review of whether or not treatment should be continued if, as many people do, there is a desire to stop medication. We suggest that you read the guideline carefully to review the evidence for prophylactic medication. We would be grateful for any evidence you have on the advantages of dealing with bipolar disorder without medication.

Why is a minimum of lithium level of 0.6 recommended (and 0.8 for those that have relapsed)? This is contrary to evidence that a lower level is effective and may be based on a trial which demonstrated discontinuation effects of lowering dose.

This is incorrect - please see our review in chapter 9 (first draft numbering)

There may be more flexibility required than this guideline suggests about lithium monitoring. For example, where is the evidence that 6 monthly testing is less effective than 3 monthly? Should there be some discussion about whether blood monitoring is necessary at all in fit and well patients?

Thank you but we believe that routine monitoring in an important precaution given that some side effects of high lithium levels may not always be clinically apparent.

Why is there this apparently specifically zealous caution about discontinuing lithium? If it is related to the increased risk of relapse after lithium discontinuation, then the guideline should be very hesitant about recommending lithium treatment in the first place. Does the minimum recommendation of 2 years conflict with 9.4.1.4? It seems to. If a particularly zealous paragraph about the need for continuation treatment with lithium comes up with a figure of 2 years, it could well suggest that the 5 year figure in 9.4.1.4 is too long. Certainly it seems to be without evidence.

The recommendations are based on the best available evidence and on the expert opinion of the group.

There seems to be a suggestion that the antidepressant discontinuation syndrome is different for tricyclic antidepressants and SSRIs. What is the evidence for this? The evidence for a half-life effect is not strong, and the warning about fluoxetine discontinuation needs to be more cautious.

The evidence for discontinuation symptoms is currently stronger for some of the SSRls and other drugs such as venlafaxine -we accept that this may not be simply related to the half-life of a drug or that problems may not exist with the TCAs. Nevertheless we believe that we have developed recommendations in this area on the best available evidence.

The scope makes clear that "When referring to pharmacological treatments, normally guidelines will recommend only within the licensed indications. However, where the evidence clearly supports it, recommendations for use outside the licensed indications may be made in exceptional circumstances."

This principle does not appear to have been followed in the guideline. Reference to licensed indication is made in relation to the treatment of acute mania and for prescribing in children. However, there needs to be more discussion of the use of medication for prophylactic treatment outside licensed indications - beyond reference to the US approval of lamotrogine and the lack of approval of risperidal consta.

Is such prophylactic treatment not fairly commonly outside licensed indications? It is difficult to see how this can be resolved with the scope that such recommendations should be "in exceptional circumstances". Where is the guidance about what to do if a patient is started on a licensed drug for acute treatment and requires prophylactic treatment for which there may not be a license eg. depakote and quetiapine?

There may need to be reference to the fact that drugs such as valproate have been strongly marketed by its manufacturer through educational activities. Hence maybe the reason they do not need to obtain a license. Not that we are suggesting that drug companies should continue to be allowed to control clinical testing of their own drugs - this function should be taken over by an independent body. The considerable off-license prescribing in bipolar disorder may be a consequence of the fact that lithium is the exception to the pharmaceutical industry-led introduction of psychotropic medication.

The licence status of individual drugs mentioned in the guideline has been made clearer.

The pharmaceutical industry is clearly prepared to exploit its position by using NICE recommendations to promote its products eg. Lilly awarding itself a NICE recommendation for olanzapine. NICE needs to be seen as taking a sceptical view of the evidence, and not merely going through the motions of endorsing data which may have little clinical significance. We are not convinced it has been sufficiently critical in this respect in this guideline.

Thank you - we try to maintain a critical perspective and have reviewed all our recommendations for antipsychotics. In addition we have expanded our comments about the weaknesses of the existing literature. Where appropriate we have highlighted key areas for further research.