Abstract
Over the last decade, new and more expensive antidepressant compounds have been introduced in Australia. Although most have been from the same class, namely the selective serotonin re-uptake inhibitors (SSRIs; e.g. fluoxetine, sertraline, paroxetine, fluvoxamine, citalopram), a range of other classes have also been introduced including reversible inhibitors of monoamine oxidase inhibitor-A (RIMAs; e.g. moclobemide), serotonin and noradrenalin re-uptake inhibitors (SNRIs; e.g. venlafaxine) and, most recently, serotonin receptor subtype 2 (5HT2) antagonists (e.g. nefazodone). Prior to marketing, each of these compounds had demonstrated antidepressant efficacy in placebo-controlled trials. Comparative studies, however, have usually been limited to short-term studies against the older tricyclic compounds in restricted patient cohorts [1–4]. Such studies have typically had a limited capacity to detect significant differences in efficacy but they do demonstrate that the new agents are better tolerated [5,6]. The lack of relevant comparative data across a wide range of clinical settings, however, has led to an ongoing debate with regard to the place of the less expensive but more toxic tricyclic compounds [7–10].
Consistent with the specified National Health Priorities Areas, a Depression Action Plan is being developed under the second phase of the National Mental Health Strategy (1998–2003) to improve the recognition and treatment of depressive disorders. Within that plan, the delivery of effective pharmacological and non-pharmacological treatments by general practitioners and mental health specialists will be highlighted [11]. Clinical practice guidelines for the treatment of major depression are also to be developed by the Royal Australian and New Zealand College of Psychiatrists (RANZCP) in 1999 for use by mental health clinicians and consumers. Both of these processes will rely strongly on available evidence with regards to efficacy and tolerability of antidepressant compounds. Such evidence, in turn, will be dominated by material available from published clinical trials. Although the tender documents for the RANZCP guidelines note that this process ‘should involve stakeholders representing a range of professional disciplines and interests, and should include consumers’ (p.5), the ways in which the views, practices and experiences of clinical psychiatrists should be incorporated are not specified [12].
Throughout clinical medicine, there is a rush to produce evidence-based guidelines. Such guidelines are designed to influence directly clinical practice by reducing unwarranted variation in treatments provided [13]. This is achieved by removing practices that have been shown to be ineffective and promoting interventions with proven efficacy. However, as a consequence of the lack of extensive evidence about new treatments, and comparisons with older treatments, guidelines for the treatment of depression can only make strong recommendations for a limited number of clinical situations [14–17]. Unfortunately, as it is usually non-systematic, the collective experience of mental health clinicians is in danger of having little impact on the guideline process [18].
The impact of deficits in evidence relevant to decision-making in clinical medicine is exacerbated by the way in which key groups of patients are frequently omitted from treatment trials. In the depression literature groups such as elderly patients, patients with other medical disorders, hospitalised patients, women during the postnatal period, women without adequate contraceptive cover and patients with comorbid anxiety, substance abuse and personality disorders are infrequently evaluated. Further, as most trials are of short duration (6 to 12 weeks of acute treatment), they are at variance with current practice recommendations, which emphasise months to years of maintenance therapy [14–17]. Notably, few studies have been conducted in primary care settings [17], even though the majority of patients with depression and/or anxiety are identified and treated by general practitioners [19].
While translating available evidence and clinical opinions into educational materials for use by general practitioners, we developed summaries of the characteristics of currently available antidepressant agents [20]. However, as these materials were developed by a small working group, they risked being unrepresentative of wider clinical expertise. As the first step in a process of developing wider practice research networks in psychiatry [18,21], we sought to collect more detailed information about the use of antidepressant therapies by specialist psychiatrists.
Method
‘SPHERE: A National Depression Project’ is a collaborative program of academics, clinical psychiatrists and general practitioners. It was launched in February 1998 to provide comprehensive training and disease management programs for the treatment of depression and anxiety by general practitioners [21]. For the initial practice survey, we decided to target psychiatrists involved with the SPHERE Project. They are drawn from each major region across the country. As these specialists were in active clinical practice, likely to be using antidepressants frequently and were currently involved in an ongoing professional education activity, we believed that they were also likely to have characteristics relevant to expressing opinions in this area: that is, they were likely to be aware of recent advances in pharmacotherapy and to have also gained experience with the relevant compounds in clinical practice situations.
In the first phase, psychiatrists whose names had been forwarded to the SPHERE Project as possible facilitators for educational sessions were sent a detailed questionnaire asking them to assess the benefits, risks and clinical utility of each of the antidepressant compounds currently available on the Australian market. While 321 surveys were sent (representing approximately one in six psychiatrists nationally), only 41% were current facilitators or had already participated in some aspect of the SPHERE Project.
Details of the clinical practice of each practitioner were collected as well as an estimate of the number of scripts written for each compound over the last 6 months. Psychiatrists were asked to rate a series of attributes of each compound (antidepressant efficacy, anti-anxiety efficacy, side-effect burden, usefulness in general practice settings, usefulness in specialist psychiatrist settings, speed of onset and toxicity in overdose) on a 0 (very low) to 5 (very high) scale. No more detailed description of each of these constructs was provided. While the proportions of psychiatrists who gave each product a ‘4’ or ‘5’ rating for each attribute are shown in Table 1, statistical evaluation consisted of comparisons of the mean ratings for each compound.
In the second phase, a modified (brief) questionnaire was included with the specialist affairs journal Australasian Psychiatry, which is distributed to 1950 Australian Fellows of the RANZCP. This was intended principally to access opinions among a wider group of psychiatrists with regard to antidepressant efficacy and tolerability not included in the first survey. The questionnaire contained two sections: the first asked psychiatrists to rate each of the major classes (rather than individual products) on three of the same factors as in the first survey (antidepressant efficacy, anti-anxiety effect, and side-effect burden) on a 1 (low) to 5 (high) scale; the second part asked respondents to rank (1, first choice, to 7, last choice) their antidepressant class preference for 13 specific clinical situations (Table 2).
Statistical analyses
For each parameter (e.g. antidepressant efficacy, anti-anxiety efficacy, side-effect burden, etc.) an overall repeated measures ANOVA was initially conducted. For the ratings of efficacy and tolerability, we used the three most popular SSRIs, SNRIs, 5HT2 antagonists, RIMAs, mianserin, high and low-dose TCAs and phenelzine to represent older monoamine oxidase inhibitors (MAOIs). Where significant differences were detected between the means across the whole group then further paired t-tests were conducted to determine subcategories of efficacy. As this involved multiple comparisons, only differences in means which were significant at the p < 0.01 level are reported. For the clinical situation, ratings each of the seven classes were treated as an individual group (i.e. SSRIs, SNRIs, 5HT2 antagonists, RIMAs, tetracyclics, TCAs and MAOIs). For each situation, the percentage of psychiatrist first and first/second preferences for antidepressant classes were reported.
Summary of relative rankings (0, very low, to 5, very high) of attributes of available antidepressants by 103 Australian psychiatrists
Percentage of first and first/second preferences (1, first choice, to 7, last choice) for antidepressant classes across a variety of clinical situations by 86 Australian psychiatrists
Results
In the first survey, 32% (103/321) of the psychiatrists returned questionnaires. Of those who did return the surveys, 37% (n = 38) were also SPHERE facilitators. Respondents were predominantly male (76%), with a mean of 11 years (SD = 8.1; range = 1–35) in specialist practice. A third of participants were in full-time private practice, 15% in full-time hospital-based practice, 19% had some degree of academic practice, and 44% were in a combination of public and private practice. Practitioners also represented a wide variety of subspecialty interests with 56% reporting an interest in mood disorders, 40% in anxiety disorders, 30% in psychotic disorders, 22% in consultation-liaison psychiatry, 16% in psychogeriatrics, and 18% in child and adolescent psychiatry. Consistent with the national coverage of the SPHERE Project, 25% of practitioners were from New South Wales, 34% from Victoria, 16% from Queensland and 25% from the other states and territories. The psychiatrists reported treating a mean of 27 patients (SD = 21.7; range = 2–150) each week for mood and/or anxiety disorders and that they had written a mean of 11 scripts (SD = 7.9; range = 0–40) over the last week for antidepressant medications.
In this survey, psychiatrists perceived critical differences in antidepressant efficacy (F = 46.4, df = 9, p < 0.001), anti-anxiety efficacy (F=10.4, df=9, p <0.001), side-effect burden (F = 84.5, df=9, p < 0.001), usefulness in general practice (F = 125.7, df = 9, p < 0.001) usefulness in specialist psychiatrist settings (F = 44.6, df = 9, p < 0.001), speed of onset (F=8.6, df=9, p <0.001), toxicity in overdose (F = 258.3, df = 9, p < 0.001; Table 1), and number of scripts written for each compound in the last 6 months (F = 14.8, df = 9, p < 0.001). When ranked for each of the key parameters, statistically significant differences (p < 0.01) emerged in the following ways: (i) antidepressant efficacy of venlafaxine > high-dose TCAs, paroxetine, fluoxetine and sertraline > older MAOIs and nefazodone > mianserin and low-dose TCAs > moclobemide; (ii) anti-anxiety efficacy of nefazodone, paroxetine, older MAOIs and high-dose TCAs >sertraline, low-dose TCAs, venlafaxine and fluoxetine > mianserin and moclobemide; (iii) side-effect burden of high-dose TCAs > older MAOIs and low-dose TCAs > mianserin, venlafaxine and nefazodone > paroxetine and fluoxetine > sertraline > moclobemide; (iv) usefulness in general practice settings of sertraline and paroxetine > fluoxetine > nefazodone, moclobemide and venlafaxine > mianserin and low-dose TCAs > high-dose TCAs >older MAOIs; (v) usefulness in specialist psychiatrist settings of venlafaxine and sertraline> paroxetine > nefazodone and fluoxetine > high-dose TCAs and moclobemide >older MAOIs, mianserin and lowdose TCAs; (vi) speed of onset of venlafaxine, paroxetine and sertraline > nefazodone, fluoxetine, older MAOIs and high-dose TCAs > mianserin, moclobemide and low-dose TCAs; and (vii) toxicity in overdose with high-dose TCAs, older MAOIs and low-dose TCAs > mianserin > venlafaxine, nefazodone and moclobemide >fluoxetine and paroxetine> sertraline.
The relative volumes of scripts written indicated psychiatrists' strong preference for the newer SSRIs (notably sertraline, mean = 41 scripts, and then paroxetine, mean = 24 scripts) and the SNRI venlafaxine (mean = 35 scripts) for the treatment of depression. Individual rankings for the most-recently released SSRIs (citalopram and fluvoxamine) are not shown in Table 1 due to the much lower volume of scripts written. These compounds received very similar rankings to the more widely prescribed SSRIs. The rankings of another relatively new agent, the 5HT2 antagonist (nefazodone), should also be viewed cautiously due to the low volume of scripts written (mean = 17). A number of agents, which have been on the market for longer periods and have enjoyed a substantial market share, are now not favoured by psychiatrists. These include the TCAs (mean = 14 scripts), mianserin (mean = five scripts), fluoxetine (mean = 15 scripts), the older MAOIs (phenelzine = three, and tranylcypromine = three) and moclobemide (mean = 14 scripts).
With regard to the recommendations for use of these compounds in general practice as compared with the specialist sector, the newer SSRIs were perceived as most useful in both sectors. Although venlafaxine was thought highly efficacious for depression it was perceived as a drug more useful to specialists, presumably reflecting its higher side-effect burden and more complex prescribing pattern. Nefazodone was somewhat preferred for use in the specialist sector, while moclobemide was thought to be useful only in the general practice sector. Although psychiatrists noted some differences in speed of onset, which largely paralleled overall efficacy ratings, no agent(s) stood out as having a particularly fast onset of action.
In the second survey, data sets were returned by 86 psychiatrists, of which 8% were current facilitators of the SPHERE Project. The psychiatrists included were again predominantly male (69%), aged 45 years (range = 29–70), with a mean of 12 years of specialist practice. The results of the efficacy and tolerability ratings were largely comparable with those of the first survey. For antidepressant efficacy the percentage of respondents that rated each class ‘4’ or ‘5’ were, in order of highest to lowest: TCAs (95%), SNRIs (89%), SSRIs (88%), MAOIs (74%), 5HT2 antagonists (46%), tetracyclics (41%) and RIMAs (27%). For anti-anxiety effects the order was: SSRIs (36%), 5HT2 antagonists (35%), TCAs (31%), MAOIs (27%), RIMAs (25%), SNRIs (19%) and tetracyclics (15%). For side-effect burden the order from lowest to highest was: RIMAs (2%), SSRIs (7%), 5HT2 antagonists (18%), tetracyclics (22%), SNRIs (30%), MAOIs (73%) and TCAs (82%).
With regard to the order of preference in specific clinical situations, the details are outlined in Table 2. Importantly, there were significant differences in ordering preferences across the variety of clinical situations described, with SSRIs emerging as the preferred class in most cases. Other classes, however, were rated as the first or second choice in a number of clinical situations. For instance, SNRIs were the preferred choice for treatment-resistant depression and patients who had failed to respond to one SSRI. Serotonin receptor subtype 2 antagonists were the second choice to SSRIs for mixed anxiety and depression, and major depression with sleep disturbance. Reversible inhibitors of monoamine oxidase inhibitor-Awere the second choice to SSRIs for adolescents with major depression, patients over age 65, patients with serious medical illnesses and patients with chronic fatigue. Tricyclic antidepressants were the preferred choice for patients with chronic pain, and second choice to SSRIs for patients with major depression with panic disorder, postnatal disorders and patients with psychotic depression.
Discussion
The surveys reported have a number of important limitations. Taken together, data regarding clinical opinions were only obtained from 10 to 15% of practising psychiatrists. Further, a high proportion of these psychiatrists was in contact with a specific national education program that contains advice with regard to choice of antidepressant compounds. That program emphasises that clinicians should actively choose different antidepressant compounds for different clinical situations. Therefore, clinicians in contact with the SPHERE Project may be more likely to report differences between existing antidepressant compounds.
Those psychiatrists who did participate, however, had characteristics that suggested that their clinical opinions should be valued: that is, it was important to obtain opinions from psychiatrists who were likely to be informed about recent drug developments and also to have had experience prescribing the newer compounds. The subspecialty interests of the psychiatrists and the volume of scripts written confirmed that these clinicians were active in the relevant areas of clinical psychiatry and had considerable clinical experience with the newer compounds. We would suggest, therefore, that while the overall response rate was less than ideal data were returned largely by general psychiatrists with relevant clinical expertise. Similarly, when we next move to surveying the use of various types of non-pharmacological strategies, it will be important to ensure that those surveyed have actual experiences with the various non-pharmacological types of treatments described.
As the manufacturers of antidepressants claim equal efficacy for available agents, current marketing of the newer and more expensive agents largely emphasises differences in side-effect burden and toxicity. Further, as government licensing for these types of compounds depends on demonstrated efficacy for major depression (DSM or ICD defined) only, the characteristics of the compounds across the broader range of depressive disorders are typically not explored extensively in clinical trials. Contrary to the current marketing of antidepressants, the results of these surveys suggest that clinical psychiatrists do perceive important differences in key aspects of efficacy. For example, although moclobemide is rated as having the least side-effect burden it is also rated lowly for antidepressant and anti-anxiety effects. Consistent with these opinions, it now has a low script volume in the specialist sector. Additionally, despite their proven efficacy (and largely as a consequence of their side-effect burden and toxicity), psychiatrists report that they have moved away from using the older tricyclics (except for chronic pain) and the older MAOIs. Although venlafaxine is perceived as a very efficacious treatment for depression, there is the qualification that it has significant side-effects and is viewed as a drug better suited to the specialist sector. Nefazodone and paroxetine are viewed as the best products for the treatment of comorbid anxiety, which is a major clinical issue in both specialist psychiatrist and general practice settings [22,23]. The SSRIs (particularly sertraline and paroxetine) are viewed as the most appropriate products for use in general practice, while the SSRIs, SNRIs (venlafaxine) and 5HT2 antagonists (nefazodone) are seen to have key roles in the specialist sector.
Psychiatrists perceive several meaningful levels of difference in antidepressant and anti-anxiety efficacy across antidepressant classes but perceive few significant differences between the multiple SSRIs available. This provides some validation for the education system produced by our earlier reference group [20,21]. Psychiatrists' views, however, appear to be at variance with current market share data, which indicate the substantial ongoing usage of TCAs, fluoxetine, mianserin and moclobemide. As the specialist sector is likely to change its prescribing patterns most rapidly, the ongoing higher market share of these compounds is likely to reflect their continued widespread use by general practitioners and other physicians.
The opinions expressed by psychiatrists in our survey are not necessarily consistent with the limited data available from controlled trials. The factors that determined clinical opinions were not tested in this study. The more sceptical views are that they simply represent the rather idiosyncratic views of individual practitioners, the undue influence of key opinion leaders, the effects of non-clinical pharmaceutical company promotion strategies and/or the results of successful marketing strategies for individual compounds (e.g. paroxetine and nefazodone being better than other SSRIs for anxiety). An alternative view is that while successful marketing and novelty factors contribute to the initial usage of specific compounds, long-term usage among specialists is increasingly influenced by their own actual experiences. The latter view is supported to some degree by the opinions expressed with regard to differential use and clinical indications for the older drugs in this survey. That is, some of the newer drugs which have been actively promoted but have also been available for significant periods (e.g. mianserin and moclobemide) are now not used often by clinical psychiatrists and/or were not often recommended in the specific clinical situations surveyed. By contrast, the TCAs which are not only not promoted but are often actively discouraged [8], are still held to be important in a range of common clinical situations (e.g. chronic pain, psychotic depression, treatment-resistant depression). Thus, the more optimistic view of clinical surveys is that actual clinical experience does have a significant impact on the opinions expressed.
Such practice surveys should be viewed simply as one part of a larger practice-based system designed to increase the knowledge base for improving mental healthcare [18]. The SPHERE Project has as one of its aims the collection of patient outcome data sets from routine clinical practice in both general practice and specialist psychiatry [21]. Such practice-based approaches do not dispute the importance of properly controlled trials. However, they can provide additional knowledge with regard to long-term efficacy and utility in specific clinical situations where few or no controlled trials will ever be conducted. If we encourage such repeated practice-based approaches then as new classes of drugs move into the market we may be able to provide practitioners, consumers and governments with a more balanced view of the ongoing status of each class. Such balanced views are difficult to obtain from either the pharmaceutical industry or limited academic reviews. While the current emphasis on evidence-based medicine appears to have reinforced the notion that only data from randomised-controlled trials counts, it should also highlight the deficits in knowledge that need to be filled by alternative professional and practice-based approaches [18,21].
Potential conflict of interest
This project was conducted as part of ‘SPHERE: A National Depression Project’. One of the major sponsors of the SPHERE Project is Bristol-Myers Squibb, manufacturer of Serzone (nefazodone). The survey was designed, evaluated and reported without reference to Bristol-Myers Squibb.