Letters

Comparison of St John's wort and imipramine

Remission is important outcome

Study design casts doubt on value of St John's wort in treating depression

Finding must be treated with caution

Sensitivity of assay is questionable

Author's reply

Remission is important outcome

EDITORThe authors of either of the two large trials of St John's wort for treating depression published in the BMJ have reported the proportion of patients entering full remission of symptoms after acute treatment. ^{1 2} This is unfortunate as remission of symptoms rather than response to treatment should be the key outcome objective. Partial remission is a common adverse outcome of depression after short term treatment and carries with it a high risk of relapse and continuing disability.³

There is now emerging evidence that although all antidepressants seem to be similarly efficacious in producing a short term response, there are differences in remission rates, particularly in severely ill patients.⁴ This shows the importance of reporting categorical as well as continuous outcomes in clinical trials of treatment with antidepressants.

Peter L Cornwall, senior lecturer in community psychiatry. 
University of Newcastle, Newcastle upon Tyne NE1 4LP p.l.cornwall{at}ncl.ac.uk

Competing interests: None declared.

Study design casts doubt on value of St John's wort in treating depression

EDITORAlthough it is impressive to see a study in which hypericum is equivalent to an antidepressant, suggesting efficacy, the design of this study by Woelk et al makes it difficult for us to know the value of St John's wort for the treatment of depression.¹

Firstly, imipramine is an outdated antidepressant. The current batch of selective serotonin reuptake inhibitors and atypical antidepressants have far fewer side effects and can be tailored to the specific needs and responses of the patient. Will hypericum be coupled with activating herbs (such as Gingko biloba or Siberian ginseng) or sedating ones (such as kava or valerian root) to achieve the same effect? This study strikes me as an example of statistical interest (yes, hypericum may be at least as effective as imipramine), but it has minimal practical clinical relevance.

Secondly, the study ran for only six weeks. Most studies exploring the efficacy of an antidepressant run for a substantially longer period. Various factors can come into play after this initial period. Do subjects discontinue use due to long term side effects they are willing to put up with in the short term? Does an initial placebo effect run its course, and then the two drugs have differential efficacy? What is the impact of either drug on recurrence? A longer trial reveals a tremendous amount, whereas such a brief trial only leaves us wondering as to its veracity.

Thirdly, without a comparison group that received no active drug, we have no way of knowing if either treatment has much benefit beyond natural improvement or placebo. The is because of the natural course of the disease (especially if a significant proportion of the subjects had an adjustment disorder with depressive mood, which tends to resolve on its own). In addition, the placebo effect is even higher when the placebo has some antihistamine effect (similar to the influence of the antidepressant on histamine receptors). In earlier studies imipramine may have been similar to inert placebo, thus making current comparisons and analysis of effect sizes difficult, if not impossible.

Thus, any replication of this study may want to consider hypericum versus a more modern antidepressant, a course of treatment that runs for six months, not six weeks, and a third arm to the study that uses a placebo with some antihistamine to mimic the immediate physiological effects of the active medicines in the other arms of the study.

James L Spira, head. 
Division of Health Psychology, Navy Medical Centre, 34800 Bob Wilson Drive, San Diego CA 92134, USA JimSpira{at}aol.com

Competing interests: None declared.

Finding must be treated with caution

EDITORAlthough the trial reported by Woelk et al was designed to avoid the limitations of other hypericum trials, it still has many limitations.¹ The dose of imipramine was increased from 25 mg twice daily to 75 mg twice daily in one week, which does not represent current clinical practice. This played a part in the higher rates of withdrawal symptoms and side effects in this trial. In a meta-analysis by Linde et al the dropout rates due to side effects for hypericum and different types of tricyclic antidepressants were 2.2 and 6.8, respectively.² The dosage of imipramine should be individually determined according to the requirements of each patient and increased gradually according to the clinical response. A lag in therapeutic response usually occurs at the onset of treatment, lasting from several days to several weeks; increasing the dosage does not normally shorten this latent period and may increase incidence of side effects.³

It would be more appropriate if the sample size was calculated by using a two tailed test, which will increase the sample size by 25 patients in each group. Woelk et al showed that patients who have anxiety associated with depression may derive more benefit from treatment with hypericum than with imipramine, which is an interesting finding, especially in a general practice setting where it might be difficult to draw a distinct line between mild to moderate depression and anxiety disorders. This finding must be interpreted with caution, since imipramine is not indicated for the treatment of anxiety disorders.³

Abdullah Alkhenizan, clinical fellow. 
University of Toronto, Sunnybrook Hospital, Toronto, Ontario, Canada M4N 3M5 alkhenizan{at}sprint.ca

Competing interests: None declared.

1.	Woelk H, for the Remotiv/Imipramine Study Group. Comparison of St John's wort and imipramine for treating depression: randomized controlled trial. BMJ 2000; 321: 536-539. (2 September.)
2.	Linde K, Ramirez G, Mulrow CD, Pauls A, Weidenhammer W, Melchart D. St John's wort for depressionan overview and meta-analysis of randomised clinical trials. BMJ 1996; 313: 253-258[Abstract/Free Full Text].
3.	Gillis MC, Welbanks L, Bergeron D, Cormier-Boyd M, Hachborn F, Jovaisas B, et al. Compendium of pharmaceuticals and specialties. 34th ed. Canadian Pharmacists Association, 1999:1805-1807.

Sensitivity of assay is questionable

EDITORWoelk et al have undertaken to confront the criticism of design and methods of some studies evaluating the efficacy of hypericum by attempting to conduct a trial that meets current methodological standards.¹ Their report does, however, also seem to have methodological shortcomings.

Firstly, owing to the absence of a placebo control group, the magnitude and relevance of the `treatment effect' cannot be assessed and therefore the trial's assay sensitivity cannot be demonstrated.² Woelk et al report 43% of patients in the hypericum group and 40% in the imipramine group whose scores on the Hamilton depression scale decreased by at least 50% against baseline, Linde et al found rates of "50% responders" of up to 54% in the placebo groups of the trials included in their review.³ Therefore the treatment effect in this trial could alternatively be explained by a placebo effect, since there is no proof that the trial was appropriate to discriminate between pharmacologically active and inactive treatments. Although the absence of a placebo control group is an issue in most active control trials, a proof of assay sensitivity seems to be particularly essential in depression, owing to the very large extent and variability of the placebo effect.

Secondly, the non-inferiority margin used in the trial is debatable. Woelk et al claim that a difference in improvement 3.5 points between hypericum and imipramine on the 17 item Hamilton depression scale is clinically irrelevant. But taking into account that an effect size of 3 points is accepted as clinically relevant to distinguish between an active drug and placebo,⁴ this non-inferiority margin seems much too large. It was larger than the significant differences between hypericum and placebo in several of the trials reviewed by Linde et al,³ and it would permit to accept non-inferiority even if hypericum were no more efficacious than placebo.²

Thirdly, the actual standard deviation of the primary variable turned out to be only about half of the value assumed during sample size calculation. This strongly indicates that there are essential differences between this study and earlier trials, causing doubts on its external validity.

Fourthly, although the trial was designed to show the non-inferiority of hypericum extract in comparison to imipramine by a one sided non-inferiority test, Woelk et al report the result of a (non-significant) two sided superiority test for the primary variable.

Andreas Völp, independent biostatistician. 
Fuchstanzstrasse 107, D-60489 Frankfurt, Germany Andreas.Voelp{at}psy-consult.de

Competing interests: None declared.

1.	Woelk H, for the Remotiv/Imipramine Study Group. Comparison of St John's wort and imipramine for treating depression: randomized controlled trial. BMJ 2000; 321: 536-539. (2 September.)
2.	International Conference on Harmonization. Note for guidance on choice of control group in clinical trials. Draft consensus guideline London: European Agency for the Evaluation of Medical Products, 1999.
3.	Linde K, Ramirez G, Mulrow CD, Pauls A, Weidenhammer W, Melchart D. St. John's wort for depressionan overview and meta-analysis of randomised clinical trials. BMJ 1996; 313: 253-258
4.	Montgomery SA. Clinically relevant effect sizes in depression. Eur Neuropsychopharmacol 1994; 4: 283-284[CrossRef].

Author's reply

EDITORThe proportion of patients having full remission of symptoms after acute treatment is one way of defining the key outcome objective. Yet, clinically, partial remission is defined as at least a 50% reduction on the Hamilton depression rating scale, which corresponds to very much or much improved on the clinical global impression scale. Such an improvement is clinically relevant, and most of the conducted trials of antidepressants have used it.¹

Six weeks is a reasonable time for a short term trial looking at the efficacy and tolerability of antidepressants.² It is not ethical to treat patients with a depressive disorder for longer than this in a placebo controlled study.

The Ze 117 extract of St John's wort has also been tested against fluoxetine, a more modern antidepressant. This trial showed that the efficacy of Ze 117 was equivalent to that of fluoxetine 20 mg daily.³ An open multicentre long term study in 440 patients concluded that Ze 117 250 mg twice daily for up to 12 months is a safe and effective treatment in patients with mild to moderate depression.⁴

The dose of imipramine was indeed increased to its therapeutic concentration comparatively quickly. Yet the number of participants who dropped out because of adverse events was similar to that in other trials.⁵

Of course any trial can always be improved, especially by having an additional arm that includes placebo, but the large response to our study, both negative and positive, shows fair criticism of our study of a herbal product.

H Woelk, professor. 
Klinik für Psychiatrie und Psychotherapie, Akademisches Lehrkrankenhaus der Universität Giessen, Lichterstrasse 106, D-35394 Giessen, Germany