There is accumulating evidence that the SSRIs are less efficacious than TCAs and MAOIs for the treatment of major depression (see summary of Anderson’s meta-analysis). This doubt seems to apply to citalopram.

There is substantial evidence from trials that the average degree of improvement ‘effect size (ES)’ from citalopram is less than from TCAs, particularly amitriptyline. There is also some evidence that those on citalopram are more likely to die by suicide than those on amitriptyline; that also suggests it is less effective.

The following recent paper is thus noteworthy.

Double-blind comparison of citalopram and placebo in depressed outpatients with melancholia (Mendels, J, 1999)

This large and recent paper (although parts of it have been presented before) seems to me to exemplify the situation with many ‘new’ drugs like citalopram; it is just the sort of paper a drug representative may put on your desk; what can one make of it? This is good material for a ‘journal club’ discussion.

This randomized, double-blind study compared citalopram (20-80 mg per day; mean 52 mg) with placebo in 180 psychiatric outpatients with major depression who also met DSM-III criteria for melancholia.

We have here a large number of severely depressed patients; so, if this drug does work it should be very clear indeed.

Duration 4 weeks only, limited, the authors state ‘because of severity of illness in the study population’. That is slightly topsy-turvy logic and simplistic methodology. Was the treatment not really working then?

There is no data on suicide attempts or admissions to hospital, this seems odd; it is highly likely such events occurred even though the trial lasted only four weeks. (They do report one death by suicide in the placebo group).

Also, we would like to know what happened to the citalopram patients subsequently; ie after the 4 weeks was over. We also need information on the large (almost 50%) portion who dropped out; perhaps they did better, that is a possibility, especially in view of the modest improvement in the active treatment group.

They present as a main point ‘Patients treated with citalopram showed significantly greater improvement at endpoint than placebo patients’. This is being economical with their full results, almost to the point of being disingenuous.

Statistically significant, maybe, meaningfully significant; well, read on and judge for yourself.

They say the results of the study indicate that citalopram is safe and effective in the treatment of depressed patients with melancholia, and is associated with a favorable side effect profile and a potentially rapid (a strange oxymoron) onset of action (if only I had a dollar for every time I have heard that).

Yes but; from a starting score of 34 (rating scale is the HAM D-24; but the goal posts shift and we get a HAM D-17 score later):–

The placebo patients improved by 14 points at 4 weeks (vs 18 points for citalopram)

The results were even worse for the ‘endpoint’ scores, viz 7 point improvement for placebo vs 9 point for citalopram.

At their greatest improvement a small proportion of the 89 patients who started treatment with citalopram had a best HAM-D score of about 16 (ie start score of 34, minus improvement of 18 = 16). That was only 4 points better than the placebo group who scored HAM-D of 20.

A typical degree of improvement, as expressed by the widely used concept of ‘effect size (ES)’, is 0.4 for the treatment difference between TCAs and placebo. This study yields an ES difference between citalopram and placebo of less than 0.1

The ‘intent to treat’ (ITT) analysis of effect size is risible; that is if the whole matter was not so serious with the danger of patients committing suicide as a result of incompletely treated illness.

Let’s recap on the ‘bottom line’

1. They set out to treat 89 severely depressed patients with citalopram (placebo–91)
2. After 4 weeks only 46 were still taking citalopram (placebo–51), an overall dropout rate of 50%
3. Of these remaining 46 (out of 89) their average benefit was HAM D-24 of 18 points vs 14 points for the 51 still on placebo; a very small difference indeed. It could easily be accounted for by variations in Inter-rater reliability.
4. This difference measured as ‘effect size’ is not even 0.1 (note that 0.4 is the typical ES difference between ADs vs placebo).
5. The placebo patients’ degree of improvement was just as good after 4 weeks as the citalopram patients’ was after 3 weeks.
6. The average practitioner would not have been able to discern the difference between those on placebo vs citalopram at ‘endpoint’.
7. After 4 weeks treatment with citalopram many of these patients were still sufficiently ill to qualify for entry into a new antidepressant drug trial.

Verdict

This does not constitute successful treatment. It seems clear that none of these patients came anywhere near getting better (ie ‘remission’). Whatever the reason for not reporting data on the subsequent course of these patients is, without it we have to be even more sceptical than we already are.

There is no evidence presented relating to ‘objective’ improvement criteria such as social or work functioning, suicide attempts, admission to inpatient facilities during treatment etc.

The conclusions the authors draw can only be said to represent a triumph of hope over experience. It is surprising the referees allowed these conclusions to be thus stated:– ‘citalopram is safe and effective in the treatment of depressed patients with melancholia, and is associated with a favorable side effect profile and a potentially rapid onset of action’.

These results go a long way to persuade me that most SSRIs, perhaps citalopram in particular, are not an appropriate treatment for severe depression.