I am concerned that Palaniyappan et al's review of combining antidepressants (Reference Palaniyappan, Insole and FerrierPalaniyappan 2009) contains imprecise genera2lisations, inaccuracies and misquotation of references.

The introduction states ‘we review the nature and extent of the side-effect burden and potential risks of these combinations’. That should entail a clear exposition of pharmacodynamic and pharmacokinetic drug–drug interactions, which are the heart of the issue. Yet these two key words (and discussion of the important concepts associated with them) do not appear in Palaniyappan et al's text.

Combinations of different antidepressants are either implicitly or explicitly proscribed by various reviews and ‘authorities’, often the same ones who, as in this instance, get important basic facts wrong and cite references inappropriately (including mine: Reference GillmanGillman 2006). This is a subject about which there is already more misinformation in psychiatry texts (and in the British National Formulary) than you can shake a stick at: Palaniyappan et al's review adds to it.

I will begin by suggesting some resources that readers may access, because, without understanding the pharmacodynamics and pharmacokinetics of interactions, they will never become confident about what to do and what not to do. My own work includes a review detailing the properties, receptor potencies, cytochrome P450 (CYP450) enzyme inhibition potencies and pharmacokinetic interactions of tricyclic antidepressants (TCAs), selective serotonin reuptake inhibitors (SSRIs) and serotonin and noradrenaline reuptake inhibitors (SNRIs) (Reference GillmanGillman 2007); and a detailed analysis of the pharmacodynamic interaction of serotonin toxicity, explaining which MAOIs and other drugs can precipitate it (Reference GillmanGillman 2006). Preskorn's major review of interactions (Reference Preskorn and FlockhartPreskorn 2006) is highly recommended, as are his website (www.preskorn.com) and that of Flockhart (http://medicine.iupui.edu/clinpharm/ddis) and the Allele Nomenclature Committee (www.cypalleles.ki.se). Most other web sources and software programs are inaccurate. Other key papers on CYP450 enzymes are: Reference BrosenBrosen 2004; Reference Ingelman-SundbergIngelman-Sundberg 2005a,Reference Ingelman-Sundberg and Rodriguez-Antonab; Reference Sim and Ingelman-SundbergSim 2006.

It is also notable that there are great differences between tranylcypromine (TCP) and phenelzine (PLZ), especially with regard to CYP450 interactions (Reference Holt, Berry and BoultonHolt 2004): briefly, PLZ is a mechanism-based (i.e. irreversible) inhibitor of most CYP450 enzymes, whereas TCP is not, except for nano-molar (reversible) potency for CYP2A6 inhibition (Reference Draper, Madan and ParkinsonDraper 1997). These differences affect potential interactions.

Palaniyappan et al cite my review of the TCAs (Reference GillmanGillman 2007) but seem to disregard most of its content and produce an unreferenced interaction table out of step with that (and other) sources. They state, concerning my comments on dual action strategies: ‘The combination of a predominantly noradrenergic TCA such as nortriptyline and an SSRI may overcome this ceiling effect and produce a different sodium:5-HT reuptake blockade ratio. However, there is no evidence that this ratio is related in any way to clinical effectiveness.’

Aside from the rendering of ‘NA:5-HT’ as ‘sodium:5-HT’, the quote indicates a misunderstanding of what I discussed, which had nothing to do with ceiling effects. Quite the opposite in fact: I argued that most SNRIs, but particularly venlafaxine, exhibit a subtherapeutic noradrenergic effect, as evidenced by low affinity at the human cloned noradrenaline transporter, and no substantial effect on the tyramine pressor response (Reference Blier, Saint-Andre and HebertBlier 2007; Reference Debonnel, Saint-André and HébertDebonnel 2007) – a ‘floor’ effect perhaps? Therefore, if it were possible to attain full noradrenaline reuptake inhibition, the dose required would be about 10 times the maximum dose of 300 mg, i.e. a toxic level (Reference Whyte, Dawson and BuckleyWhyte 2003). In fact, this extremely weak NRI is the rationale for adding reboxetine, to attain a more balanced SNRI effect. So their comment (regarding venlafaxine + reboxetine) is wrong: ‘Any synergism of such a combination is doubtful, as both drugs act via the same mechanism; the same effects could be achieved by a higher dose of venlafaxine alone, with more predictable pharmacokinetics’ – indeed, predictably toxic.

Also, their claim that ‘there is no evidence that this ratio is related in any way to clinical effectiveness’ denies the established reality, logic and basic pharmacology of the dose–response curve. Since venlafaxine has approximately a 200:1 differential between 5-HT:NA transporter affinity it is impossible to have an optimum therapeutic: side-effect ratio for both systems simultaneously, or even any meaningful balanced effect on both systems simultaneously. The tacit acceptance by psychiatrists of the notion that venlafaxine has an SNRI effect is astonishing, because it is closer to myth than reality. Contrary to Palaniyappan et al's assertion, there is both theoretical and clinical evidence that dual action (a true SNRI effect) is more effective. Nelson et al's randomised controlled trial (RCT) of fluoxetine and desipramine indicated much higher remission rates at 6 weeks for combined treatment (54%) than for the SSRI (7%) or desipramine (0%) alone (Reference Nelson, Mazure and JatlowNelson 2004). Palaniyappan et al also misrepresent that trial by reporting: ‘In any event, this speed of onset effect could not be replicated in a later RCT (Reference Nelson, Mazure and JatlowNelson 2004)’, without mentioning the remission rate. They also use this same reference a second time, implying that it is subsequent work.

Their material and references on my area of expertise, serotonin toxicity, are poor for a review article. The section headed ‘SSRI with moclobemide’ shows a lack of familiarity with the literature on reversible inhibitors of monoamine oxidase A (RIMAs) and SSRIs (Reference GillmanGillman 2006). One of several examples of this is: ‘The SSRI–moclobemide combination has been tried with the same rationale as the SSRI–MAOI combination. Three small open-label trials (total n = 46) found moclobemide to be effective in combination with SSRIs (Reference Dodd, Horgan and MalhiDodd 2005).’

Dodd, their most frequently cited reference, is another review. It is not the original source of any data. There is limited profit in one review (mis)quoting another review. Palaniyappan et al misrepresent the RIMA/SSRI area of research. They misreport Dodd (who cites two open-labelled studies (Reference Joffe and BakishJoffe 1994; Reference Hawley, Quick and RatnamHawley 1996), not three). Hawley's original report (Reference Hawley, Quick and RatnamHawley 1996) had 50 cases (cited by Dodd, correctly), taking the total beyond their stated 46 (Table 2). They might have noted that Hawley stated, concerning adverse drug reactions, ‘Many events were rated as severe. The high rate of adverse events suggests that there are clinically significant pharmacodynamic interactions between moclobemide and SSRIs’ (i.e. serotonin toxicity). Note that there are two other moclobemide/SSRI papers (Reference Bakish, Hooper and WestBakish 1995; Reference Ebert, Albert and MayEbert 1995), although Ebert was not cited by Dodd.

Why, one might ask, despite the great enthusiasm expressed by all of these authors, has nothing else been published in the subsequent decade? I sought information on this point, repeatedly, from all those authors, but only Dr Hawley ever replied (see below).

My detailed analysis of data on animal and human serotonin toxicity, and of that regarding moclobemide in particular (Reference GillmanGillman 2004, Reference Gillman2006), provides a more substantive basis for understanding the risks of combining moclobemide and SSRIs. Among other things, my review states that ‘Hawley decided to stop his research because of high levels of moderately severe serotonergic side effects’, i.e. incipient serotonin toxicity. Readers of Palaniyappan et al's article are likely to be left with the impression that SSRI/moclobemide combinations are ‘relatively safe’. That notion is seriously contentious and the contrary evidence should have been referenced and discussed. I am supposed to be an expert in serotonin toxicity and I would not be game to try it. The answer to the third MCQ question is wrong: moclobemide + SSRI definitely could cause fatal serotonin toxicity.

The format of Table 1 is unsuitable for conveying that type of data and will probably mislead and confuse many readers. It, and the associated text, has some odd material that does not promote a good understanding of pharmacology, interactions, or CYP450 enzymes, and it is less useful than other pre-existing sources and references given here. It is, inexcusably, unreferenced. The authors state (under the heading, and in the context of, combinations of an SSRI with a TCA) ‘Tricyclic toxicity … is a particular risk for the 7% of White people who lack sufficient CYP2D6 to metabolise TCAs (Reference Albers, Reist and HelmesteAlbers 1996)’. This appears to be muddled thinking because that is precisely the group in whom an interaction is less of a risk, because they are already genotypic poor metabolisers and further diminution of CYP450 activity (by SSRIs) will not make them worse; such people are at increased risk of toxicity with monotherapy with TCAs. Albers is an outdated and inappropriate reference. In my opinion readers would do better to consult the following more helpful and accurate sources: Reference BrosenBrosen 2004; Reference Ingelman-SundbergIngelman-Sundberg 2005a; Reference Preskorn and FlockhartPreskorn 2006; Reference Sim and Ingelman-SundbergSim 2006; Reference GillmanGillman 2007; Reference FlockhartFlockhart 2009; Preskorn's extensive review contains a wealth of good information.

There are various other errors in the article, which is characterised by imprecision, unhelpful generalisations and uncritical repetition of other reviews. In my opinion anyone contemplating using combinations needs to be rather better informed than they would become by reading Palaniyappan et al.