6.1. Introduction

Drug treatments currently have an important place in the management of PTSD. This is supported by reviews, which suggest that they are effective (e.g. Van Etten & Taylor, 1998; Friedman et al, 2000). Drug treatments have been shown to achieve statistically significant (positive) effects on each of the three main elements of PTSD (re-experiencing, avoidance and hyperarousal). However, other reviews, for example Stein et al (2004), have suggested that the efficacy of drug treatments may be less strong (they estimated an SMD for drug treatments against placebo of −0.46 (k=4; n=327; 95% CI −0.71 to −0.2). This more modest view was based on a rigorous systematic review of relevant studies, including only randomised controlled trials and applying more strenuous inclusion criteria (in particular those for trial quality). As with this review, Stein et al (2004) used between-group rather than within-group effect sizes, thereby reducing the likelihood of artificially inflating the effect size (as is acknowledged by Van Etten & Taylor, 1998).

Given issues concerning the lack of wide-scale availability of psychological interventions and the desire to provide increased patient choice, it is important to establish the relative efficacy of drug treatments in PTSD. If there are drugs with comparable efficacies to the psychological treatments, for which there is currently the strongest evidence base (see Chapter 5), these would allow rapid access to an effective intervention, especially in primary care settings. They would be available as adjunctive or alternative treatments in case of treatment failure and would allow individuals with PTSD to make a choice as to the approach they favour. Even if the efficacy were less good, these drugs would still be available as a second-line intervention.

In the UK, only two drugs are currently licensed for the treatment of PTSD, paroxetine and sertraline (the latter being licensed only for women). However, other drugs that are not licensed for use in the UK have been subjected to randomised clinical trial for the treatment of PTSD and are considered within this review.

6.2. Current clinical practice

Robust evidence on the pattern of drug usage for the treatment of PTSD in the UK is not available. However, it is widely accepted that many patients in the UK are treated with drugs: predominantly selective serotonin reuptake inhibitors (SSRIs) and newer antidepressants, although to a lesser extent tricyclic antidepressants and atypical antipsychotic agents are also prescribed. The major uses of these drugs are for the treatment of PTSD symptoms, related sleep disturbance and agitation and comorbid depression. Low-dose amitriptyline may also be used in the treatment of chronic pain syndromes. In one major UK centre for the treatment of PTSD it was estimated that around 30% of patients at point of assessment are taking some form of psychotropic medication and a further 30% are prescribed medication during their attendance at the clinic (C. Freeman, 2004, personal communication).

Mellman et al (2003) have reported on a survey of drug treatments in the USA in a community-based sample of mental health clinic attenders with PTSD. They compared drug prescribing for patients with PTSD alone, depression alone and PTSD comorbid with depression. Of the PTSD alone group, 77% were prescribed medication, compared with 89% of the comorbid group and 82% of the depression alone group. Patients with PTSD alone were prescribed a range of antidepressant medication, but 17% were taking atypical antipsychotics and 41% were taking benzodiazepines and related drugs. Mellman et al (2003) point out that the use of the latter two drug types is likely not to conform to international guidelines (Friedman et al, 2000).

6.3. Limitations of the literature: comparing RCTs of pharmacological and psychological treatment

There are some important differences between drug treatment and psychological treatment clinical trials. Drug trials use a double-blind method, often with a placebo arm. What this means is that the effect of the drug treatment can, in principle, be separated from the non-specific effects of being in a trial. There are a number of non-specific effects that may occur. The most prominent of these is the placebo effect, which can account for a significant component of a drug treatment effect, and in the case of mild depression this may be as much as 80% (Kirsch, 2000). In addition, in any trial – and in particular in drug trials – much attention is paid to measurement, and the assessment interviews are likely to be mutative (i.e. they may in themselves have a therapeutic effect). It is therefore no surprise to find that the total effect of drug intervention is greater than the chemical effect of the drug treatment alone.

In trials of active drug versus placebo, the control for non-specific attentional effects is greater than is the case in a psychological therapy trial when a waiting list control is used (see intervention v. waiting list comparisons in Chapter 5), as opposed to a trial in which some form of attentional control has been used (see intervention v. intervention comparisons in Chapter 5). Further, in drug trials, researchers are masked to the treatments being evaluated. In psychological therapy trials there is greater potential for the degree of enthusiasm of the researcher for a particular treatment to affect the outcomes. There may therefore be both a less effective control for the placebo effect and also the potential for inflation of the estimated active treatment effect. What this means is that when comparing the results of meta-analyses of drug treatment and psychological treatment trials (particularly where these are intervention v. waiting list comparisons) caution is required, especially in drawing conclusions from simple comparisons of effect sizes. The design of the drug trial may lead to a lower effect size for the active drug than for an equivalent psychological therapy when a comparison is made against no other active intervention. Unfortunately, direct comparisons between pharmacological and psychological treatments are largely lacking.

We have attempted to respond to this problem in comparability by assuming that the placebo arm of a drug trial, with its significant clinical input and enhanced measurement strategies, constitutes an active intervention and therefore we have adopted the lower of the two thresholds for effect size used in the psychological treatment comparisons. This means that we have regarded, on a priori grounds, a standardised mean difference (SMD) of −0.5 as indicative of a clinically meaningful difference between active drug and placebo. It may be helpful at this point to explain this process in more detail. What we set out to achieve was a recommendation based not on statistical difference alone but on the sort of change in symptom score likely to be experienced as beneficial by clinicians and PTSD sufferers. Selecting an SMD of −0.5 was a decision taken before the statistical analyses were undertaken. The evidence statements derive not only from the SMD but also from the confidence intervals in the meta-analysis (see Chapter 4).

We have been faced with data from smaller, older trials and larger, more recent trials. There are some limitations that should be pointed out in any comparison of these studies, because the more recent trials have tended to be more robust, using intention-to-treat analyses. They may also have included more patients in primary care, who would have been less severely affected. These differences might have influenced the effect sizes found in the different studies, to the disadvantage of some of the newer drugs.

Before leaving the general issue of differences between study designs, there is another factor to take into account. In drug treatment trials it is common to offer a flexible dosage regimen. This allows the clinician (masked to the allocation of placebo and active drug) to increase the number of tablets taken within an approved range. Trial designs may therefore encourage early increases in dosage in the absence of a marked treatment response. This may lead to somewhat elevated mean dosages in drug treatment trials as opposed to clinical practice, where experience suggests a more cautious use of dosage escalation.

6.4. Issues and topics covered by this review

In broad terms we have tried to address the issues of comparative efficacy, acceptability and tolerability of pharmacological treatments most commonly prescribed in the UK both for acute phase treatment and continuation/relapse prevention treatments in adults. (There is a separate review of treatments of all types in children and young people.)

For the purposes of this review the following drug treatments were considered: paroxetine and sertraline (the only two drugs licensed in the UK for PTSD), including a small comparison trial of paroxetine versus trauma-focused CBT, fluoxetine (another SSRI antidepressant), amitriptyline, imipramine (a tricyclic antidepressant), phenelzine and brofaromine (monoamine oxidase inhibitors), mirtazapine and venlafaxine (other antidepressants), olanzapine and risperidone (atypical antipsychotics) and combined drug and therapy interventions (phenelzine and psychotherapy, and imipramine and psychotherapy). There are a number of other drugs that are not licensed for the treatment of PTSD in the UK but may be prescribed with some frequency for this condition (based on advice from the Guideline Development Group). These include the SSRIs citalopram, escitalopram and fluvoxamine, the atypical antipsychotic quetiapine and the reversible monoamine oxidase inhibitor moclobemide. However, for these latter drugs no study met the inclusion criteria, and so these drugs could not be assessed within the meta-analysis.

Finally, we have included one non-pharmacological but biological intervention (repetitive transcranial magnetic stimulation of the right dorsolateral prefrontal cortex) in this chapter.

6.5. Studies included

Full details of the search strategy for this and other reviews in the guideline are given in Appendix 6. From the main search for RCTs, 26 separate studies were identified by the Guideline Development Group as meeting the inclusion criteria. These include 23 studies of drug treatments against placebo (BRADY 2000A, BUTTERFIELD 2001, CONNOR 1999B, DAVIDSON 1990, DAVIDSON 2001, DAVIDSON 2001C, DAVIDSON 2003, DAVIDSON 2004, DAVIDSON, ELI LILLY, HERTZBERG 2000, KATZ 1994, KOSTEN 1991, MARSHALL 2001, MARTENYI 2002, MARTENYI 2002A, PFIZER 588, PFIZER 589, SKB 627, SKB 650, STEIN 2002, TUCKER 2001, ZOHAR 2002) and three studies of combined drug and or therapy interventions (FROMMBERGER 2004, HAMNER 2003A, KOSTEN 1992). (References given in shortened format are listed in Appendix 14.) In addition, one study of repeated transcranial magnetic stimulation (rTMS) against placebo (sham treatment) was identified (COHEN 2004B). The report by SAYGIN 2002 is not considered further in this review, because one of the drugs involved has now been withdrawn in the UK. (Details of the search for RCT studies are given in Appendix 6 and summary characteristics of individual included trials are given in Appendix 14.)

In the meta-analysis, a number of crossover studies were excluded owing to the non-availability of pre-crossover point data. (In crossover trials all participants receive all interventions (or control or other non-active intervention) in sequence and hence participants act as their own control. At present, methods for interpreting and using crossover data are not sufficiently developed to allow integration of complete (post-crossover point) data from such trials with standard RCT trials within the same meta-analysis.)

6.6. Study characteristics

In contrast to the reports of psychological interventions, the studies included did not typically provide data on remission of PTSD diagnosis, but instead reported response rate in terms of a percentage decrease in symptoms from baseline score. Response rate data were not used within the meta-analysis because of the inconsistency in reporting (thresholds for reported response rates typically vary from 30% to 50%). This decision was taken because it is known that relatively small differences in mean scores (which are not clinically significant) between two comparison groups can produce statistically significant differences when presented as response rates (Kirsch et al, 2002). Remission rates have the advantage of being clinically determined in advance (diagnosis v. no diagnosis). Recent research in depression suggests that remission is a more reliable indicator of a stable return to normal mood states than response rates (Keller, 2003). The most consistent evidence reported for tolerability was the number of participants leaving the treatment early and this is reported within the review.

6.7. General issues arising in the management of antidepressant medication

This guideline has been developed in parallel with a set of guidelines applicable to the treatment of depressive disorders. There is a certain amount of common ground arising from the use of antidepressant drugs in both settings. This section draws heavily on the forthcoming NICE guideline on the treatment of depression (National Collaborating Centre for Mental Health, 2005).

Common concerns in PTSD sufferers about taking medication, such as fears of addiction or that taking medication will be seen as a weakness, should be addressed in an early discussion about prescribing options. Discontinuation/withdrawal symptoms can occur after stopping many drugs (that are not drugs of dependence), including antidepressants, and may be explained in the context of ‘receptor rebound’: for example, an antidepressant with potent anticholinergic side-effects may be associated with diarrhoea on withdrawal. Discontinuation/withdrawal symptoms may be new or hard to distinguish from some of the original symptoms of the underlying illness. They are experienced by at least a third of patients (Lejoyeux et al, 1996).

All patients who are prescribed antidepressants should be informed, at the time that treatment is initiated, of potential side-effects and of the risk of discontinuation/withdrawal symptoms (particularly with paroxetine and venlafaxine). Patients started on antidepressants should be informed about the delay in onset of effect, the time course of treatment and the need to take medication as prescribed. Written information appropriate to the patient’s needs should be made available.

The onset of discontinuation/withdrawal symptoms is usually within 5 days of stopping treatment, or occasionally during taper or after missed doses (Rosenbaum et al, 1998; Michelson et al, 2000). Symptoms can vary in form and intensity and occur in any combination.

Discontinuation/withdrawal symptoms may be mistaken for a relapse of illness or the emergence of a new physical illness (Haddad, 2001), leading to unnecessary investigations or reintroduction of the antidepressant. It is important to counsel patients before, during and after antidepressant therapy about the nature of this syndrome.

Generally, the antidepressant drugs recommended for use in PTSD should be discontinued over at least a 4-week period (Rosenbaum et al, 1998). (A shorter period may be appropriate for fluoxetine because of its long half-life.) The end of the taper may need to be slower as symptoms may not appear until the reduction in the total daily dosage of the antidepressant is substantial.

If discontinuation/withdrawal symptoms do emerge and are mild, the clinician may reassure the patient that these symptoms are not uncommon after discontinuing an antidepressant and will pass in a few days. If symptoms are severe, reintroduction of the original antidepressant (or another with a longer half-life from the same class) and gradual tapering are advised (Lejoyeux & Ades, 1997; Haddad, 2001).

During treatment with antidepressant medication, it is important to consider the risk of self-harm. Adult PTSD sufferers started on antidepressants who are considered to present an increased suicide risk, and all those aged 18–29 years (because of the potentially increased risk of suicidal thoughts associated with the use of antidepressants in this age group), should normally be seen after 1 week and frequently thereafter until the risk is no longer considered significant.

For PTSD sufferers at high risk of suicide, consideration should be given to prescribing an appropriate quantity of antidepressants and providing additional support for administering medication. Toxicity in overdose should also be considered when choosing an antidepressant for patients at significant risk of suicide. Practitioners should be aware that SSRIs and mirtazapine are safer in overdose than other tricyclic antidepressants.

Patients with PTSD started on antidepressants who are not considered to be at increased risk of suicide should normally be seen after 2 weeks and thereafter on an appropriate and regular basis: for example, at intervals of 2–4 weeks in the first 3 months, and at greater intervals thereafter if response is good.

There are specific issues relating to the use of SSRI medication. Particularly in the initial stages of SSRI treatment, healthcare professionals should actively seek out signs of akathisia, suicidal ideation and increased anxiety and agitation. They should also advise PTSD sufferers of the risk of these symptoms in the early stages of treatment and advise them to seek help promptly if these are at all distressing. If a PTSD sufferer develops marked and/or prolonged akathisia while taking an antidepressant, the use of the drug should be reviewed. In the treatment of PTSD, where this is not an uncommon issue anyway, it is also important to consider the impact of sexual side-effects.

Similarly, there are specific issues relating to the use of the monoamine oxidase inhibitor phenelzine. All patients receiving phenelzine require careful monitoring (including blood pressure measurement) and advice on interactions with other medicines and foodstuffs, and should have their attention drawn to the product information leaflet.

Administration of some drugs to nursing mothers may lead to effects in the infant if the drug passes into breast milk. It is therefore important to exercise additional care. In any individual case, a decision to prescribe should only be made after full and open discussion with the PTSD sufferer, reference to the appendix on breast-feeding in the British National Formulary and full consideration of the risks and benefit of this action. In general, this is likely to be an additional factor pointing strongly to the advantages of a psychological treatment.

It is likely that guidance on the use of antidepressant drugs will be further updated during the life of this guideline and practitioners should maintain an awareness of current guidance set out by the Committee on Safety of Medicines and available on the Department of Health website (http://www.info.doh.gov.uk/doh/embroadcast.nsf/vwDiscussionAll/9AA9EC56B07B3B4F80256F61004BAA88).

6.8. Repetitive transcranial magnetic stimulation

Repetitive transcranial magnetic stimulation (rTMS) is a new technique under investigation in a range of conditions, such as depression and stroke. It involves placing an electromagnetic coil on the scalp and rapidly turning on and off a high-intensity current through the discharge of a capacitor. The magnetic pulse induces electrical effects in the underlying brain cortex. These pulses vary in frequency. If the stimulation occurs faster than once per second (1 Hz), it is referred to as fast or high-frequency rTMS and may result in excitatory physiological changes. On the other hand, low-frequency or slow rTMS may have an inhibitory effect on brain excitability. There is a low risk of seizure and certain exclusion criteria apply. In the study by Cohen (COHEN 2004) investigators administered rTMS over the right dorsolateral prefrontal cortex of the brain for 20 min per day over 10 working days. The sham treatment group also underwent the same procedure but the positioning of the coil was such that it did not have an effect (it was held vertically to the scalp rather than being placed on the scalp). There were two treatment groups: one receiving fast rTMS (10 Hz) and the other slow rTMS (1 Hz).

6.9. Clinical summary

We have drawn attention to the important difficulties of comparing drug treatment trials with psychological therapy trials. The tough design of the drug trial is likely to produce a lower effect size than a comparable psychological treatment trial.

However, using an a priori criterion for a clinically important effect, the drug treatments were disappointing. For paroxetine, there is a reliable, positive but small effect, which (although statistically significant) fell short of the target effect size of 0.5 for a clinically important intervention. Once we included additional unpublished data, we were able to demonstrate neither clinically important nor statistically significant effects in the meta-analysis for sertraline, the other drug licensed in the UK.

There are a number of other randomised controlled trials that met the inclusion criteria, although these tended to be based on relatively small samples and therefore need to be interpreted with some caution. These suggest that there may be a clinically important effect for mirtazapine, amitriptyline and the MAOI antidepressants phenelzine and brofaromine.

The difficulties arising from simple comparisons of recent large trials and older small trials have already been identified. It would be incorrect to conclude that an older drug with an apparently larger effect size in a small trial is preferable. It is likely that none of the drug treatments has a very large effect size using current robust trial methodology. Current policy is to favour the use of licensed drugs in preference to unlicensed drugs. However, prescribers should be aware of other positive trial data in reaching a clinical decision. The use of unlicensed drugs is the responsibility of the prescriber.

This review includes one RCT using a non-pharmacological treatment (rTMS). This report is encouraging (for high-frequency rTMS) but further research is required to determine the place of interventions like this in the management of PTSD.

In the event that few trials for a review had effect sizes that met the thresholds for clinical significance, as was the case for this review of drug treatments, statistical significance in the meta-analysis was also taken into account in reaching a recommendation (see Chapter 4).

A specific drug is recommended here if it meets the threshold for clinical effect as outlined above and is currently available in the UK. Thus, mirtazapine, amitriptyline and phenelzine have been included (brofaromine not being currently available), although we do recognise that in each case, this recommendation is made on the basis of data from single trials.

We have concluded that we should recommend paroxetine on the basis of its robust statistically significant effect based on large-trial data, even though it did not meet the threshold for a clinically significant effect as we have defined it in the list of recommended drugs. This is the only drug in the list of recommendations with a current UK product licence for PTSD.

We have also made a recommendation about the use of adjunctive olanzapine in people who are non-responsive to initial drug treatment. In this case, the recommendation is based on limited evidence of a clinically important effect on comorbid depression symptoms. Although the meta-analysis shows a limited evidence of effect on PTSD symptoms as well, this does not reach statistical significance and needs to be interpreted with caution.

We have also examined tolerability data and side-effect profiles. On this basis, we recommend that paroxetine and mirtazapine should be the drugs of choice for use in primary care. Phenelzine, amitriptyline (and adjunctive olanzapine) should be recommended as additional drug treatments for use under supervision of mental health specialists.

6.10. Clinical practice recommendations

6.11. Research recommendations