Foreword

Lithium is not an obsolete drug, although some wish that it were. Nowadays lithium is used primarily for the prevention of recurrences of bipolar disorder, and prophylactic lithium treatment has been marvelously helpful for very many very ill patients. Studies from the last decade have demonstrated a significant association between long-term lithium treatment, on the one hand, and lowered mortality and suicidal behaviour, on the other [1, 2]. No such association has been demonstrated for other moodstabilisers. When administered to bipolar women during pregnancy and lactation, lithium has a lower risk/benefit ratio than do other mood stabilisers [3].

These positive experiences should nevertheless not keep us from searching for something that is still better, but it must be realised that to compare the prophylactic efficacy of contending drugs with the efficacy of lithium in an open-minded and balanced manner makes demands on the discriminatory power of the trial designs and the integrity of the investigators.

The search for prophylactic drugs other than lithium has faced substantial obstacles [4]. Since lithium prophylaxis is so effective, there has been difficulty about obtaining permission from ethics committees and informed consent from patients in order to compare new drugs of unproved prophylactic efficacy with a placebo or with lithium. The investigators have employed two approaches to justify the testing of new drugs. One was to emphasise the shortcomings of lithium treatment. The other was to treat patients who were refractory to lithium treatment or had troublesome side-effects.

Shortcomings of lithium treatment

Lithium has many side-effects, and sometimes they are so troublesome that the treatment has to be discontinued. However, the frequency and severity of the side-effects are different at different serum lithium concentrations. In 1979 at the psychiatric hospital in Risskov, Denmark, the range of serum lithium concentrations was lowered from 0.8 to 1.0 mmol/L, to 0.5 to 0.8 mmol/L. The average serum lithium concentration fell by approximately 30% and the proportion of patients not having any side-effects rose from 10 to 40%. Lithium-induced tremors and diarrhoea were markedly less pronounced during treatment with low serum concentrations than with high; so too were lithium-induced increases of urine volume and reduction of renal concentrating ability [5, 6]. The prophylactic efficacy of the treatment before and after the change was not determined quantitatively, but it did not seem to fall markedly.

It is worthy to note that the claim that lithium has many and severe side-effects is made primarily in American texts. In 1989 a group from Boston compared doses leading to what they called ‘standard’ serum lithium levels, 0.8–1.0 mmol/L, with doses leading to low levels, 0.4–0.6 mmol/L [7]. The trial showed that the former levels were more effective but also associated with a higher incidence of side-effects than the latter, and the authors chose to recommend the high levels. They did not examine effects and side-effects of the levels in between, that is, 0.6–0.8 mmol/L. In Europe the serum lithium range recommended for optimum balance between effects and side-effects has for several years been 0.4–0.8 mmol/L, to be exceeded at either end in particularly sensitive and particularly resistant patients [8].

Change from lithium to a new drug

Switching to another drug when lithium is not effective or not tolerated is a sound and clinically responsible practice. One must, however, realise that observations made under these circumstances are biased in favour of the new drug; they cannot show whether it is prophylactically efficacious when given as a first treatment, let alone more efficacious than lithium. When a patient is switched from drug A to drug B because the former was ineffective or not tolerated, then drug B must necessarily come out as good as or better than drug A, for it cannot come out worse. Since many studies have had this design, the evidence of a prophylactic effect of new drugs must be assessed with caution.

Treatment with anticonvulsant drugs

A prophylactic action in bipolar disorder has been claimed primarily for the anticonvulsant drugs carbamazepine and valproate, but as far as typical bipolar disorder is concerned, the evidence is weak or missing [4]. Post et al. listed a number of studies that in their opinion demonstrated a prophylactic action of carbamazepine [9]. Many of the trials were of the drug A/drug B type, and the methodological reservation mentioned above was not taken into account. Most of the trials did not distinguish between typical and atypical patients, and such a distinction makes a difference. Greil et al. [10] carried out a prospective, double-blind study on 171 bipolar patients allocated randomly to carbamazepine and lithium and followed for 30 months. In patients with typical bipolar disorder, those with bipolar I disorder and mood-congruent symptoms, lithium prevented recurrences more efficaciously than did carbamazepine (P = 0.005). In patients with atypical bipolar disorder, those with bipolar II disorder, mood-incongruent symptoms, rapid cycling, dysphoric mania, and comorbidity, there was a non-significant trend for carbamazepine to be better than lithium (P = 0.075).

Valproate is today used extensively for prophylaxis in bipolar disorder, but only one relevant study has been published, namely that by Lambert and Venaud [11]. The effects of valproate and lithium were compared in a prospective, parallel-group trial involving 150 patients followed for an average of 18 months. The two drugs were found prophylactically ‘comparable’ with a non-significant trend in favour of valproate; no distinction was made between typical and atypical bipolar disorder. This study has not been replicated. A publication from 1997 by Bowden et al. described the protocol of a recently completed randomised, double-blind, placebo-controlled, parallel-group comparison of valproate, lithium, and placebo, but the outcome of the trial itself has not been published [12]. In atypical bipolar disorder the prophylactic efficacy of valproate has not been compared with that of lithium.

Future comparative trials

When a distinctly more efficacious drug than lithium comes onto the scene, how can we then recognise it? Prophylactic trials take a long time and in order to justify the consumption of resources and personnel, and the long-lasting exposure of patients to a drug of unproved prophylactic efficacy, the trial design must be such that it provides valid information. Anything else would be unethical. Grof analysed the advantages and disadvantages of different trial designs [13].

The important points are the following. Firstly, because experience has shown that it is difficult for trials with a mirror-design to win credence, prophylactic trials should preferably be prospective, randomised, and double-blind. If the standard treatment is found significantly less efficacious than placebo, the study is invalid and should be discarded.

Secondly, during the trial, adjunctive medication should be restricted to the clinically necessary minimum, and any extra treatment must be recorded. Thirdly, results can be generalised only if a distinction is made between typical and atypical bipolar disorder. Fourthly, prophylactic trials should be carried out on patients with moderate to severe illness intensity. Fifthly, trials where patients are switched from drug A to drug B because drug A did not work or had troublesome side-effects cannot show whether drug B is more efficacious than drug A.

Finally, before the start of a trial it must be established that the patients suffer from a recurrent mood disorder and are at considerable risk of further recurrences. Trials based on patients without such risk cannot be used to estimate prophylactic efficacy, and the patients should not be given long-term medication for they may not need it. Angst suggested that individual patients should have had at least two to three episodes within the last 5 years [14], and Grof [13] emphasised that a certain average number of episodes between the debut of the disease and the start of prophylactic treatment should be exceeded for the whole group.

When no statistically significant difference has been found between the efficacy of two drugs, for example, between a new drug and lithium, this may mean that their prophylactic efficacy is in fact the same. However, it may also mean that not enough patients were followed for a sufficiently long time to disclose a possible difference. A check of the statistical power helps to distinguish between these two possibilities, and authors deprive their readers of relevant information if they do not tell them whether such a test has or has not been carried out.

Choice of prophylactic drug

The proposals outlined above may serve to ascertain whether the prophylactic efficacy of a new drug is higher than that of lithium. But in order to oust lithium a new drug must in addition be as good as or better than lithium with regard to long-term tolerability, interactions, ease of management, and risk/benefit ratio during pregnancy and lactation. Even if a new drug replaces lithium as a first-choice prophylactic agent in bipolar disorder, there is for some patients a further consideration. Until a drug with an equally close association between long-term treatment and a reduction of suicidal behaviour has been found, the use of lithium should remain mandatory in manic-depressive patients who are at high risk of committing suicide; that is, patients with severe depressions or depressions associated with persistent suicidal ideas or with suicide attempts in the past.

Curiosity, daring, and compassion for the patients were required by John Cade when in 1949 he disclosed the antimanic action of lithium. Integrity, skill, and responsibility for the patients are required by us today to ensure that manic-depressive patients are at any time given the best prophylactic treatment, be it with lithium or with another drug.