Abstract

This commentary reviews the tricyclic antidepressants (TCAs) and their place in the treatment of depressive illness, updating information from my seminal review paper on their pharmacology (2007) with data from the small number of subsequent publications that have some additional original information.  Several misconceptions and distortions limit the value of what has been written about TCAs. 

First, it must be noted that they are a distinctly heterogeneous group, and that it is necessary to describe and assess them individually, linking the principles of neuroscience-based nomenclature with side-effects (SEs) and clinical effects.  Few sources do that. 

Second, following on from the first point, a reassessment of their disadvantages and side effects must consider them as individual agents, not agglomerated under the meaningless category of ‘the TCAs’.  Furthermore, studies that do not use plasma levels to assure appropriate dosage are of lesser value — that includes most studies included in the many meta-analyses that have been performed.  Plasma levels are often useful to inform rational decisions concerning their benefits and SEs.  New comparative data with other antidepressant drugs (including antipsychotics which are increasingly used as augmentation strategies) are reviewed, especially relating to cardiovascular risk.  That indicates that newer drugs have a similar risk profile to some TCAs.

Third, one must also consider TCAs undoubted superior efficacy in biological melancholic depression, the importance of understanding this in relation to supposed toxicity risk and suicide is discussed in detail.  Some differences in TCA toxicity following overdose and routine therapeutic use, compared to newer drugs, have undergone reassessment, and are less pronounced than hitherto perceived, generally being of minimal or doubtful clinical relevance.

Fourth, such points are central to scientific investigations seeking to establish cause-effect relationships between drugs with different properties and possible differential effects on particular symptom-clusters.    The paucity and deficiency of answers to such questions illustrate the inadequacies of randomised control trials and their impotence for addressing the central question of causes and mechanisms.

Scientific observations made when these drugs were first discovered are highlighted because revising that history reveals significant gaps in subsequent research: first, in relation to establishing the time course for response; second, in establishing the clinical characteristics of likely responders; third in investigating different responses to particular symptom-clusters.  If closer attention was paid to establishing the early time-course for response, there might be a clearer distinction between different AD drugs.

It is concluded that imipramine, amitriptyline, nortriptyline, and (especially) clomipramine are key TCAs with which clinicians should be familiar and that they remain important first-choice treatment considerations for selected cases of major depressive disorder, ahead of newer drugs.  They should also receive serious consideration earlier in the treatment algorithm, as both first and second-choice treatments for selected cases.

Introduction

The tricyclic antidepressants (TCAs) are a group of drugs of similar structure (hence their tag of ‘tricyclic’, which refers to their structural arrangement in three fused rings), but they are markedly heterogeneous (i.e. un-alike) in their pharmacological actions.  Some of the TCAs should be considered to have been mis-classified — some are not appropriately labelled as antidepressants (ADs).  This illustrates that although the structural similarity of drugs may be close, their pharmacological actions can be dissimilar.  For instance, clomipramine (CMI) is a close structural analogue of chlorpromazine, but their effects could hardly be more different.

Serious depression is not the only condition for which TCAs may be appropriate — it is well to remember they have many other uses which this commentary does not consider.  Therefore, if some TCAs become unavailable, partly because of underuse, this will impact many patients.

Any discussion that generalises by saying ‘the tricyclics …’ is certain to be unhelpful and equally certain to be inaccurate (e.g., see the appalling APA textbook chapter on TCAs about which I have commented ).  See also my seminal review paper about the TCAs [1], which is the only paper that I am aware of that analyses these drugs using an approach grounded in the vitally important principle of neuroscience-based nomenclature, which is yet to be adequately adopted [2-7].  My paper is the most highly cited review paper about these drugs in the last 40 years.  There has been little new information about TCAs in the 20 years since my review, with one notable exception and ‘must-read’ paper (discussed below) being by Sicouri et al. [8], which has detailed data and explanation about cardiac ion channels and hERG etc.

Imipramine (IMI), amitriptyline (AMI), and (especially) clomipramine (CMI) are likely more efficacious in the treatment of severe depression (endogenous/biological, melancholic) depression than any of the new drugs [9, 10].  Nortriptyline (NTP, a metabolite of amitriptyline) has some particular advantages for use in combinations, e.g., with sertraline or escitalopram.  Desipramine (DMI) is effectively a selective noradrenaline reuptake inhibitor (NRI) and should not be forgotten, although there are several more recent and more specific NRIs.

When TCAs were introduced in the early 1960s — the first, imipramine, hit the shelves at the beginning of 1958, then amitriptyline in 1961 — these tricyclic derivatives of promethazine and promazine (scil., the antihistamines, the antipsychotics (APs), and the TCAs) were labelled according to how their properties were then (mis)perceived).  Some were mis-classified: doxepin and trimipramine should have been designated as specific H1 antagonists (antihistamines); chlorpheniramine, in contrast, is a serotonin reuptake inhibitor (SRI) rather than an antihistamine — indeed, its bromine analogue, brompheniramine, was the molecule from which Carlsson oversaw the development of the first SSRI, zimelidine, which was synthesised before Prozac [11, 12].

Most readers will know than doxepin has recently been ‘rebadged’ as a sleeping tablet — only 50 years late!  That exemplifies psychiatrists’ lack of knowledge of pharmacology.  I have been using it as a hypnotic for many years, ever since Richelson’s seminal work in the early 1980s [13, 14].  I explained the rationale of regarding it as a hypnotic, not an AD, in my TCA review in 2007, table 6: it is a specific potent H1-antagonist with no significant NRI potency, and little ability to modify the tyramine pressor response [15].  It is an questionable argument to claim doxepin is an antidepressant, despite early claims — but it is an effective sedative anxiolytic, which knocks quite a few points of the HDRS score [16, 17].  Knowledge percolates slowly into the corpus of psychiatric writing and teaching.

Selecting a TCA

It is logical to select a few of the TCAs for AD treatment and familiarise oneself with the use of those — now, in most jurisdictions, that is an ersatz choice because not all of them remain on the market.  The notes on receptor affinities, and data in my TCA review, further clarify and substantiate the rationale for my personal preferences.  The activity of different CYP450 enzymes varies as a result of their genetic polymorphism (details about this are in the commentary on clomipramine); because the rate of metabolism of TCAs is highly variable it is frequently helpful, and sometimes necessary, to check plasma levels.

AMI is metabolised into NTP; IMI (and lofepramine, which is a pro-drug) into desipramine (DMI).  CMI is metabolised into desmethyl-clomipramine, which is a potent NRI, thus making CMI a serotonin and noradrenaline reuptake inhibitor (SNRI).

Clomipramine is still the best SNRI.  Many regard it as a more potent AD — in my view there is no doubt that CMI is superior to the SSRIs, SNRIs, and the other TCAs.  See the separate commentary about clomipramine.  The fact that it has a lower FTI (fatal toxicity index) may constitute further evidence indicating it is a more effective AD, both Buckley and Farmer have alluded to this notion [18, 19].

Nortriptyline: has several favourable attributes.  It is potent as an NRI, has a modest but useful sedative effect, low anti-muscarinic effect (approximately 4:1 NAT/muscarinic selectivity compared to amitriptyline’s 1:1), linear pharmacokinetics, no active metabolites, no significant inhibition of CP450 enzymes in the therapeutic dose range, and it is cheap.  It can be safely combined with both an SSRI (top choices being sertraline and escitalopram) and with a monoamine oxidise inhibitor (MAOI): thus, it has good properties as a ‘bridging’ drug.  It is also probably less toxic than other TCAs, and venlafaxine.  That adds up a lot of favourable attributes.

Nortriptyline may be less toxic in overdose than venlafaxine

Amitriptyline: there is persuasive evidence for its superior efficacy vs. SSRIs and SNRIs in severe depression [9].  For many years, this drug has been described as an SNRI, which is not quite correct — see the information in the serotonin toxicity (ST) section and in my TCA review paper.  It can be inferred that only CMI and IMI have significant clinical effects mediated by serotonin, see [20] for a succinct explanation of this issue, (Fig. 2, p 356).

Imipramine: IMI was the first member of this class to be introduced (1958) and does have modest SRI potency — sufficient to make it a danger of precipitating serotonin toxicity (ST) in combination with an MAOI, but only when used in high doses.  Modest SRI potency may therefore underlie part of its efficacy and make it a sort of ersatz SNRI, even if it is not as powerful as CMI, by a long shot.

Desipramine has the best side-effect vs. potency profile (it is the most NRI-selective of the available TCAs), particularly the lowest propensity to cause postural hypotension or anti-muscarinic SEs (note: lofepramine [21] is similar and is metabolised into DMI [22]).  But DMI has high affinity for the NAT (nearly an order of magnitude greater than any other TCA) and the oft-recommended dose range (150-300 mg) is high — 25 mg may be sufficient in combination with sertraline.  It was used in high doses in the 1970s, perhaps because of the factors mentioned above: thus, the high recommended dose will undoubtedly produce, in some people, plasma levels that are in the toxic range.  The toxicity of drugs taken in OD is worked out according to the defined daily dose (DDD), which determines the number of tablets in a packet, and the tablet size, and is usually a month’s supply.  Therefore: when people took an OD of a packet of DMI, they were more likely to come to harm, because the total number of milligrams of the drug in a packet was higher than other TCAs.  However, if it is used in lower doses of 25-75 mg daily, which produces a fully effective blockade of NAT [1], then it is safe and has a low FTI.

TCAs vs newer ADs

Selected TCAs remain important and effective first-choice treatments for selected cases of major depressive disorder (biological, endogenous, melancholic) in the opinion of many authors [23-31]; despite that opinion, the overwhelming majority of patients, all over the world, are treated with SSRIs and other newer ADs, not only as the 1st choice, but also as second and third choice medications. 

It is not unusual to encounter patients who have had every single SSRI sequentially, without any other class of AD being tried

The varied threads in this section reflect the fact that TCAs have been around for nearly 70 years now, and our understanding of them has shifted and evolved with changing knowledge about their pharmacology and the classification of illnesses, and much else.  If I seem a little discursive, it is because I am trying to draw together significant historical, and recent, threads that should be considered in reassessing our views.

The evidence for the superior effectiveness of various TCAs (in ‘endogenous, melancholic’ depression), over all other ADs except MAOIs, is robust and dates back to the earliest observations of the effect of these drugs, from around 1960 [32]; it has been reinforced by research studies and clinical observation ever since, with research studies, over many decades, showing a consistent trend indicating that TCAs, especially  AMI and CMI, are more efficacious in melancholic-type depression [9, 10, 28, 33, 34].  There is widespread opinion that the (mis)classification of depression, facilitated by the (rightly) much vilified DSM system [35, 36], is largely responsible for creating confusion and blurring the distinction between melancholic depression and other types of depression [37-41].  Indeed, since anergia and anhedonia are core signs of melancholia, it is astonishing — almost unbelievable — that anhedonia, was not included as a symptom of depression until DSM-IV in 1994, thus there is hardly any wonder that the distinction — between biological and other forms of depression —was blurred.

Most trials on which guidelines are based have been carried out in patients who do not have melancholic depression, hence TCAs have not been shown, except in a few trials, to be superior to SSRIs

Important as the evidence from studies might be, it is also undoubtedly the case that most experienced psycho-pharmacologists would agree that there are many patients for whom selected TCAs are of superior clinical efficacy compared to newer AD drugs.  That necessitates a major re-evaluation of statements about the lesser SEs and greater safety of newer drugs.  Scil., having lesser SEs is of much reduced beneficial impact if TCA treatments are significantly more effective, and not only reduce suicide, but also the associated morbidity, including cardiovascular morbidity — the Standardised Mortality Ratio (SMR) in depression has been estimated to be between 10-30 times elevated [42-50].

Therefore, the use of less-effective drugs with less SEs takes the dubious dictum of ‘Primum non nocere’ to an extreme [51].

This commentary does not address details of the SEs of TCA drugs, their treatment, nor how they differ between individual TCAs.  This is dealt with in other commentaries and in my TCA review.  However, it may be noted that CMI is the most potent at hSERT and has much the same sexual SEs as the SSRIs, albeit that these may be ameliorated somewhat by the inherent 5HT2A antagonism.

Recent reviews

I had not updated my TCA review until now because there was insufficient new original information to justify that effort: the papers by Sicouri [8], and Taylor  [52] have precipitated this update.

Sicouri et al. has extensive detail about how antipsychotic drugs and antidepressant drugs influence ion channels [8], particularly in relation to sudden cardiac death (SCD).  Antipsychotic drugs (APs) are more commonly associated with QT prolongation and Torsades de Pointes (TdP) than ADs (see ‘crediblemeds’), and some of the newer AD & AP drugs appear just as problematic as some of the TCAs.  Wu’s analysis of nearly one million cases [53] found no difference in ventricular arrhythmias/sudden cardiac death (VA/SCD) risk across AD classes and suggested that TCAs were associated with a lower risk than SSRIs.

The focus on QTc has distracted attention from the fact that a prolonged QT interval is a necessary — but not a sufficient — condition for the development of TdP [54].  Most cases of AD-induced TdP occur following ODs or when administered in combination with other QT-prolonging agents or conditions (e.g. long QT syndromes).

TCAs are class one anti-arrhythmic drugs, and as such were once recommended for patients with ventricular arrhythmias.  However, such drugs are now thought to increase the risk of arrhythmia, albeit only in anoxic post-infarction states.

What is clear from the evidence presented about newer drugs in more recent papers is that the idea that new drugs are safe, and the TCAs are not, is a naïve and simplistic oversimplification of complex data: in fact, many newer drugs (both ADs & APs) have just as many problems and risks, and sometimes more, than some of the TCAs.  I highlight the mention of AP drugs (covered in Sicouri’s review) because they are increasingly used in combination for difficult cases of depression. 

What is also clear, and useful to recall, is that all these epidemiology-association studies only show association, not causation, and they are subject to myriad confounding factors, which explains why few of them agree with each other.

Notably, in Taylor’s [52]) recent review of many studies on TCA cardiovascular effects, using various types of control groups, they found no clinically significant increased incidence of problems with arrhythmia, increased QTC, myocardial infarction, stroke, or heart failure.

Ergolou’s study of 11,000 cases of out-of-hospital cardiac death (OHCD) over 15 years [55] found increased rate of OHCD for citalopram (>20 mg) and escitalopram (>10 mg) among SSRIs, and for high-dose mirtazapine; but  none of the TCAs were associated with increased OHCD rate.

Untreated depression more than doubles the risk of OHCD [56], and the SMR is greatly elevated

For those interested, in addition to Sicouri — which is a recommended source for those wishing to know more about this complex subject — there are other recent papers on the details of the neurophysiology of these receptors and cardiac issues [57-60].

Interactions, metabolism, genetics

There are few new data concerning CYP450 interactions with TCAs since my review.  Recent examples usefully elaborating on such interactions are in Preskorn et al.’s papers [61-63] and in the following references on therapeutic drug monitoring (TDM)  [64-66].

One may note that CYP2D6-status can be reliably inferred from the metabolic ratio of NTP and its principal metabolite [67], as explained in detail by Preskorn [61], in figure 2.  That is helpful to know, especially if genotyping of CYP450s is unavailable or unaffordable.

More recent general aspects of CYP450 genetics are contained in other references [68, 69], with some cautionary comments about the current trend for genotyping to predict drug responses [70].  There is further detailed explanation of this in my commentary on clomipramine.

Some general comments

Review papers that consider the TCAs as a group are certain to contain misleading generalisations, and this illustrates the vital importance of adhering to neuroscience-based nomenclature — readers are urged to familiarize themselves with this essential classification system [2-7].

Some papers require careful [re]interpretation [52, 71], as do most papers commenting on potential drug-related increase or decrease in cardiovascular events, or depressive illness subsequent to cardiovascular disease — a complex two-way relationship about which it is difficult to say much of clinical use, although there have been many pronouncements on those subjects over the years.  Few of those pronouncements have stood the test of time.  Over the decades, some pronouncements about newer drugs being less risky than the TCAs have originated from less-than-objective sources and have been fuelled by the need to promote newer drugs and justify their higher price.  Now that a more informed and balanced view of the SEs and problems of newer ADs is emerging, as the ‘new-is-better’ hyperbole subsides, it is possible to make a useful reassessment of the pros and cons.

An interpolatory note is relevant here: Desipramine — which is a selective NRI — is not an effective ‘antidepressant’ by itself (as was found to be the case with other NRIs such as reboxetine and atomoxetine).  Unfortunately, as is so often the case in psychiatry and psycho-pharmacology, words are used with insufficient definition and precision.  I agree with the pioneers in the field, Arvid Carlsson, Ronald Kuhn, and Paul Kielholz, who when comparing DMI with IMI in the early 1960s observed that DMI had a different profile of activity in depressive illness, being more activating than mood enhancing.  As Carlsson said to David Healy in interview

It was clomipramine that made us so excited and also we felt that, on the basis of Kielholz’s scheme, imipramine and amitriptyline were perhaps more mood elevating than desipramine. … [He also said when commenting about dirty drugs and multiple pathways for effects] ‘However, I’m not sure about the anticholinergic action, whether that could also contribute … this is generally assumed to be just a side effect.  I’ m not so sure.’  and …but the [immediate] effects of 5HT drugs on sex prove that this can’t be the case.  How can we now explain the 2 or 3 or 4 week delay in response of depression to ADs?

Kielholz’ [72, 73] view, based on his clinical impression, despite the so-called ‘solid data’ showing no difference, was:

desipramine, possess a predominantly psychomotor stimulating effect. The actual mood-elevating effect is thus somewhat less marked in both these groups.

I am interested in Carlsson’s observations (and he included comment on 5HT2A antagonism) because I gained the clear impression that when I switched patients from the sertraline and NTP combination (sometimes because of anti-muscarinic effects) onto even low doses of DMI, many of them (~80%) came back and said it was less satisfactory because they felt ‘wired’.  Perhaps I was observing the same ‘activation’ as Kielholz, Kuhn, and Carlsson were commenting on.

The comments about the Hamilton depression rating scale (HDRS) scores, from some of these early researchers, were that scores should not be analysed aggregated, and that improvement was evident in the first week or so: those are interesting because relatively little attention has been paid to those issues in the ensuing 60 years, and few better observations have been made subsequently.  Much research has gone off at all sorts of tangents, without concentrating sufficiently on these fundamental findings — findings which generate hypotheses that can be tested experimentally by cause-effect driven research.  Thus, there has been a relative paucity of research tracking the rate of improvement in agglomerated symptoms, or different psychopathological domains of symptoms, [74-76], or the different profiles of symptom change, with different drugs.  If the right instruments are not used to elucidate this question, as generally they have not been, then no useful answer will emerge; this is especially so because most assessments have been insufficiently frequent [77, 78].  Such research would constitute useful scientific investigation of cause, mechanism, and effect.

Professor Merton Sandler (he and Pare proposed the monoamine hypothesis) asked:

Why did the three major breakthroughs in drug treatment all occur in the 1950s, and then nothing since then?’  Shorter’s answer was, …these breakthroughs occurred precisely because FDA-style trials using DSM diagnoses were not in force in the 1950s [79].

In a separate commentary about the epistemological difficulties with RCTs, I report in detail how the major players who initiated the FDA trials system, like Louis Lasagna (head of the FDA in the 1960s), soon expressed the view that they were relying too much on RCTs, at the expense of other methods and clinical experience [80, 81]; and yet the FDA have continued on the path of almost exclusive reliance on RCTs despite continued criticism over decades by a host of other eminent clinical researchers [82].

My selection of TCAs

General considerations

For treating depression, as a specialist, where most people came to me (by the 1990s) already on an SSRI, I would usually start with adding NTP.  If patients were drug-free at the time of consultation (a rare occurrence) I would start NTP at night first, and then add sertraline or citalopram (usually after a short stabilisation period on the NTP).  But there is a place for these other TCAs mentioned.  However, in the third millennium, I advocate attaining an SNRI effect, which I regard as indisputably superior to a specific agent (SSRI/SNRI), by using the above NTP/sertraline combination, which is more flexible, especially because it facilitates easier transition if a different class of antidepressant, especially an MAOI, becomes indicated — for comparative data on the reuptake affinity of SSRIs, SNRIs, and TCAs, and their effect on the tyramine pressor response see table 4 in my TCA review [1]. 

My view is that Arvid Carlsson was correct when he speculated that other ‘dirtier’ effects of TCAs were relevant.  I found that using reboxetine (or DMI) instead of NTP was sometimes excessively activating, which suggests, (as Carlsson hinted) that the other effects, H1 antagonism, anti-muscarinic effects, and possibly 5HT2A antagonism, might indeed contribute to the overall effect — no wonder he got a Nobel prize.  The concept of ‘dirty’ drugs started with Carlsson and colleagues and has been commented on intermittently ever since.  This contrasts with the recent trend for developing highly specific drugs, like SSRIs, which appear to be less effective.

The spurious dangerousness proposition

Many common objections to TCA usage apply only in specific and delineated circumstances:

1. To patients in primary care
2. When TCAs are used by less-expert practitioners and without therapeutic drug monitoring (TDM)

The most common of these objections are, 1) a more burdensome level of SEs, and 2) the danger of fatal toxicity if an OD is ingested.  I argue here that both of those objections are considerably overstated and are seldom relevant for specialist practice, or for any patients who have already failed treatment with the usual first-line drugs.

The evidence on toxicity does not logically support the widely advocated, but simplistic, approach of avoiding TCAs because they have higher toxicity if someone takes an OD, for the reasons enumerated herein.

Although the statistics vary from time to time, and country to country, the overall picture is clear.  Of all deaths by suicide about 90% do not involve drugs at all [83-85], only about 10% of them involve drugs of any sort.  Of this 10% that do involve drugs, only about 10% of those involve prescribed drugs [86, 87], and of the prescribed drugs only a fraction of those are ADs. [88-91].

Thus, using a marginally less toxic AD drug is going to influence overall death by suicide by about 0.1% of the total number of all people who commit suicide (less than 10% of 10% of 10%).

The other side of the coin is that if, as we have good reason to conclude, TCAs are more effective than the alternatives, then potentially most suicides could be more advantageously treated — that constitutes an enormously greater total of potentially saved lives, than the fraction of 0.1% that might not have survived their OD, because their prescribed AD was marginally more toxic.  The principal behind this argument is sound even if it is simplified; there are clearly other factors and influences at play relevant to this equation.

One more key consideration in the risk equation that has not been discussed to my knowledge is this: it is analogous to the ‘area under the curve’.  The total period of risk-exposure, if AD treatment is effective, is only a few weeks.  The total period of risk-exposure for ineffective treatment is potentially ‘lifelong’, no irony intended there.  It could be considered as the inverse of the quality-adjusted life year (QALY) measure.

Ego requiem meam causa

A recent review paper recommended various measures to reduce fatal suicides, none of which involved AD drugs [92]. Those, to add a further dimension of perspective, included better restriction of opiates, restriction of gun availability, and better alcohol control policies.

Some TCAs are probably equally toxic, or less toxic, than some newer ADs (e.g., NTP is probably less toxic than venlafaxine).  Also, since most suicides are not caused by the AD drug prescribed, giving a less toxic AD does not address the main problem which is depressive illness itself causing death by suicide.  The illness generates the risk of suicide, not the treatment — using a drug that is safer in overdose, but less efficacious, is not the answer.  Indeed, the evidence that some old drugs are more effective than most of the new drugs remains strong — especially for AMI, IMI and CMI, and also for the MAOIs, tranylcypromine, phenelzine, selegiline, and isocarboxazid [93]; see other notes for elaboration.

The risk of suicide is further minimised by good clinical management, e.g., among other measures, patients should have frequent appointments (or be in hospital), only a small (sub-lethal) supply may be issued at each visit, and a responsible person can be designated to supervise the supply.

The illness generates the risk of death by suicide — not treatment with AD drugs that might facilitate fatal overdoses

One of the most specious and disingenuous contributions to the debate about the danger of TCAs was published by Beerworth [94], who suggested that the very act of prescribing a TCA was de facto negligence.  Perhaps it would be truer to say that the expression of Beerworth’s opinion (she is a barrister) was closer to de facto negligence, by culpable ignorance; altogether it remains an ill-considered and hubristic comment.

Caution is required when interpreting the relative toxicity of TCAs (cf. Taylor) because the various studies span many decades of treatment experience, during which time the approach to the treatment of serious overdoses and the prevention of morbidity and mortality has advanced and improved.  For instance: it is notable that there was only one death from OD of a TCA in the series of 300 reported by the expert toxicologists in the HATS data; this obliged them to analyse the data differently, and they stated:

Death after TCA poisoning is rare nowadays [95], so we used intermediate outcome measures.

Studies over many decades are not easily comparable because treatment of ODs has improved

Likewise, the fatality rate from ODs of MAOIs dropped near the end of the last millennium when it was realised that intubation and paralysis, if the temperature was elevated beyond 39°, was the definitive course of action; indeed, between 1985 and 1991 the fatality rate per hundred thousand scripts dropped from 30 to 0 [96].

Dosulepin (aka dothiepin) toxicity

The evidence that Dosulepin is significantly more toxic in OD than other TCAs is supported by the high-quality prospectively gathered series of cases from Professor Whyte’s HATS toxicology group [95].  Although this drug was never approved in the USA it was one of the most frequently prescribed TCAs in the UK, and Australia, over couple of decades [18, 95].  It is a rather weak NRI and may, like doxepin, have been popular because of its prominent sedative effect, as suggested by Montgomery (below).

There are several reasons (see above) for supposing that the data from Professor Whyte’s HATs toxicology group are more reliable, contrary to the view of Taylor et al.

One might note that Professor Montgomery [97] offered the opinion that:

In the UK, dothiepin is generally perceived as being a less effective antidepressant and is used in doses where the sedative properties are more appreciated than any antidepressant properties.

All these mis-informed views about SEs and toxicity have cultivated apprehension and caused most doctors to stick to the guidelines and make SSRIs their ‘default’ choice — follow the guidelines, then you do not have to think for yourself.

Overdoses, toxicity (FTI), and death by suicide

The relatively higher toxicity of some of the TCAs in OD (particularly dothiepin, likely the most toxic, see below), and other SEs, persuade some authorities to over-generalise, over-simplify and say TCAs are no longer suitable as a first-choice treatment, especially in primary care — I argue here that is a mistaken and unbalanced view of the evidence.

It is notable that Buckley [18], a toxicologist and clinical pharmacologist, and author of several key reviews on serotonin toxicity and AD toxicity, felt the need to state clearly:

Toxicity should be based on data for each individual drug and not on the therapeutic class …

That statement does not just apply to toxicology, but to all other aspects of comparing TCAs to any other treatment, in any other way. The FDA ‘class labelling’ requirement [98] requires that all drugs in a class have, with certain criteria and provisos, all SEs ascribed to any members of that class listed for each individual drug, irrespective of its individual pharmacological properties: this FDA-endorsed lumping-together perpetuates the myth that TCAs are all the same in SEs and toxicity.  The FDA’s rationale (lame excuse?) may be that the TCAs are such a long-established class of drug that there has been no impetus to reconsider or re-classify their properties, since they are out of patent.  It is clear that they need to be re-classified; thus, the current situation is that the FDA have created a situation in which their system is misleading and is misinforming the medical profession.

The FDA ‘class-labelling’ requirement perpetuates the falsehood that all TCAs are the same

The discontinuation rates are not meaningfully different for TCAs vs SSRIs, due to treatment-failure dropouts and SEs, [27]— and note, that conclusion comes from multiple studies, most of which disadvantaged TCAs either because they did not, or could not, compare individual agents, or did not employ therapeutic drug monitoring. 

The DUAG studies demonstrated clear superiority for CMI [9, 100-102].  Those studies were more recently summarised and reviewed by Kragh-Sørensen [10].

In my experience, patients are less likely to need to cease appropriately dosed NTP than any of the SSRIs.  When longer-term SEs (and risks, cf. gastric bleeds below) are factored in, we can confidently predict that the balance swings more towards TCAs.

In summary, a failing of reviews and studies is that, beyond lumping TCAs together as a group, they also fail to adjust the therapeutic dose according to plasma levels — a practice that is not only useful for proper management but sometimes essential for therapeutic success [100].

SSRI problems vs TCAs

This commentary is not a review of the relative frequency and severity of side-effects with these drugs: it is intended to be a reminder of some less appreciated SEs that help to balance the relative merits between different drugs.  Some problems with SSRIs are not assessed adequately by looking at RCT evidence because some SEs were either not assessed, or are only manifest after treatment of longer duration.

A report on a large meta-analysis (of 67 trials) found an SSRI discontinuation rate in short-term studies of 20% due to any side-effect [103, 104], including tremors, headaches and dizziness, anxiety, restlessness and agitation.  These were all short-term studies (<6 weeks) and it is established that discontinuation increases significantly over longer times [27].  It is therefore reasonable to suppose that in a real-world scenario around one third of all SSRI treated patients drop out of treatment, and possibly as many as one half.

Bleeding tendency

This is an example of a missed risk that has taken a long time to be recognised, that of bleeding (of all sorts), and gastric disorders — there have been an increasing number of papers on this subject over the last 20 years, and evidence continues to accumulate [105, 106].  Gastric bleeds still have a 5-10% mortality rate, so even a small increase in bleeding tendency is a serious issue. This is probably for two reasons; SSRIs are ‘pro-kinetic’ in the GI tract (hence the side effect of diarrhoea); they furthermore increase the bleeding time (by depleting platelet 5-HT).  It has been forgotten that depletion of platelet 5-HT was the assay initially used to indicate the potency of a drug as an SRI [107].

GI tract overstimulation

Serotonin is pro-kinetic in the GI tract, accounting for the commonly encountered GI problems, nausea, reflux, dyspepsia, indigestion, and loose bowel motions [108-110].

Sexual dysfunction

SSRIs cause frequent sexual dysfunction [111-113], which is particularly relevant for younger female patients where anorgasmia occurs in at least 50% of cases.  Sexual dysfunction was ignored and minimised in a cavalier manner, that smacked of deceitfulness, in most of the SSRI trials.  The frequency of self-reported persistent post-treatment genital hypo-aesthesia is uncertain because few studies have been carried out, although it is well documented in Internet discussion groups [114, 115].

A point about ‘SEs’: it is a slightly mis-used term.  The main adverse drug reactions (usually referred to as ‘SEs’) of many drugs are an inevitable consequence of their main intended mechanism of action, referred to as ‘type A’ [116], as illustrated by increased bleeding time caused by depletion of platelet 5-HT, gastric upset, and sexual dysfunction.  Hence, these effects tend to occur in all subjects, dose-dependently, to a greater or lesser degree.  Note that clomipramine’s potent SRI affinity gives it the same sexual SEs as SSRIs, although these are possibly ameliorated by its inherent 5HT2A antagonism, the same may apply to IMI.  There is little accurate information about these problems with other types of AD [117].

CYP450 inhibition

Interactions caused by CYP450 inhibition are significant and can lead to fatalities [118, 119].  Further detailed information about this is tabulated in my TCA review and is explicated in many papers by Sheldon Preskorn, to whose work interested readers are recommended, e.g. a series of papers Drug-Drug Interactions (DDIs) in the journal Psychiatric Practice [61, 62, 120-124].

Sleep

SSRIs are the only AD drugs that worsen sleep continuity, which can impair executive function and cognitive ability [125, 126].  Wichniak et al. opine [125] that sedative TCAs are a valuable treatment option of insomnia if cognitive behavioural therapy fails, and patients still require sleep-promoting drugs more than 3-4 times per week.  Since sleep disturbance is such an integral part of depression, SSRIs inability to improve this is a disadvantage.

Impaired thinking and concentration

In addition to possible cognitive worsening as a result of impaired sleep continuity, SSRIs also probably lower frontal dopamine, which is probably linked with reduced mental concentration and energy [127],  however both positive and negative effects have been suggested [128].  Emotional blunting is also common in association with all monoaminergic ADs, not just SSRIs [129].  These effects may not be as consequential as the memory impairing effects of some TCAs, resulting from their antimuscarinic activity.

In summary, it behoves us to be sceptical about SEs and to be alert to the emergence of long-term problems — such problems are still emerging with SSRIs after 30 years! E.g., see [130-132], post-marketing surveillance in most countries is a relatively neglected area; we must not expect that SEs will always be obvious.

When to consider avoiding SRIs as first choice

We can predict which patients may find a TCA to be a more satisfactory option as a first-choice treatment — indeed, suggestions and recommendations follow naturally from the data and analysis in this review, scil:

1. Typical ‘endogenous’ symptom profile, or more severe, depressive illness.
2. Patients with previous episodes of severe depression unresponsive, or only partially responsive, to an SSRI.
3. Those whose current episode is severe or who show psychomotor retardation or lack of energy (optimal choices are imipramine, amitriptyline, and clomipramine; my preference is for clomipramine because it is an ‘SNRI’).
4. Those with a pre-existing history of, or present symptoms of, marked insomnia or restless leg syndrome (as the latter may be precipitated and exacerbated by SRIs).
5. Those with previous or present symptoms of gastrointestinal (GI) disturbance, reflux, dyspepsia, GI ‘overactivity’, GI bleed, or any bleeding risk or tendency or propensity for bruising.
6. Those with anorexia and weight loss
7. Those who exhibit or have in the past exhibited significant suicidal features or attempts (if such patients are not in hospital, then arrange for a restricted or supervised supply of medication. Such patients may be made worse by mono-therapy with SSRIs
8. Those with high levels anxiety or agitation
9. If concern over cytochrome P450 enzyme drug interactions is relevant. Selected TCAs cause less problems in this area than some new drugs, especially fluoxetine and fluvoxamine.
10. Those with a history of hyponatremia.
11. The elderly, because of increased risk of SEs with SSRIs (nausea, vomiting, vertigo) compared to TCAs [46]. Therapeutic drug monitoring is advised.
12. Any concern regarding discontinuation or withdrawal syndromes should be a cause for caution and careful consideration of alternative strategies.

Inferences and conclusions

Progress, far from consisting in change, depends on retentiveness. … Those who cannot remember the past are condemned to repeat it

Santayana, G The Life of Reason 1905

Remarkable reflections emerge powerfully from this review.

First, the observations about the original compounds discovered, IMI and DMI, and those ‘copied’ subsequently, like AMI (that was the first of the ‘me-too’ exercises, with which we are now so familiar), are that there is a difference in their symptom-profile of activity — this was described by pioneering researchers, including the Nobel prize winning Carlsson, and Kielholz, and Khun, when they first tested IMI around 1960.  They noted then that DMI, which was introduced soon after, had a more activating and less mood-enhancing profile (no RCT was required, just clinical observation, without rating-scales).

Second, their clinical observation was that people who remitted showed improvement in less than 1-2 weeks; indeed, Kuhn said five days — this has subsequently been substantiated.  That observation has implications for establishing cause-mechanism-effect relationships which is the essential central method of science [133-135], and something that RCTs are unable to address.

Third, they were clear about the fact that the clinically observed (no rating-scale needed) melancholic symptoms were the ones that best-predicted response — something subsequent research has been slow to substantiate, because most RCTs do not have melancholic, suicidal, or psychotic patients included.

It is apparent from this review, especially from the above three points, that we have not succeeded well in learning from history.  The recognition of the superior efficacy of these drugs for melancholic depression [33], compared to newer ADs, has been lost in the tsunami of pharmaceutical company sponsored RCTs, which are unable to, or inadequate for, investigating cause and effect. 

TCA pharmacology remains poorly taught to the younger generation of doctors, and poorly explicated in standard texts.  The central and important concept of neuroscience-based nomenclature has not been applied adequately to create a clearer understanding of TCAs’ individual properties.  The fact that doxepin is still miscategorised as an AD illustrates that point.

The endless comparisons of one drug with another, via RCTs, to try to establish non-specific ratings-score differences has generated endless fruitless activity, and at great cost to both patients and research funds — one prominent exhibition of this is represented by the weak conclusions and recommendations from the enormous Cipriani meta-analysis [136].  Perhaps, worst of all, it can be argued that this has diverted effort and funds from potentially more interesting and scientifically orientated investigations of causes, mechanisms, and more nuanced effects.

The supposed problems with the SEs of TCAs and their toxicity have been considerably overstated, partly, perhaps largely, driven by the commercial need to make newer drugs appear more advantageous.

Calculated by the number of patients treated, and who could yet be treated, TCAs are substantially under-utilised.  IMI, AMI, and CMI represent archetypal examples of excellent drugs that are indicated as first-choice treatment for some patients and are of superior therapeutic efficacy to newer ADs for many more severely ill patients; yet they are uncomfortably close to being relegated to the dustbin of history.

All doctors, especially early-career doctors, would do their patients and themselves a service by learning how and when to use them, and more often abjure automatic use of newer drugs.

I gratefully acknowledge the corrections and suggestions, made by Professors Shalom Feinberg and Elliott Richelson that have greatly improved this manuscript.