Abstract
Although there are many effective psychopharmacological and psychotherapeutic treatments for depression, there is ever-increasing evidence that depressive disorders are usually chronic and/or recurrent [1, 2]. This realization has contributed to the McArthur Foundation Research Network reviewing existing usage of terms such as response, recovery and relapse [3] and in proposing consensus definitions for such terms [4]. A response (to treatment) is an improvement of sufficient magnitude so that the depressed patient is no longer fully symptomatic. A remission is a relatively brief period of improvement of sufficient magnitude that the individual is asymptomatic. A recovery is a sustained remission. A relapse occurs when a patient in remission (i.e. before recovery) has a return of depressive symptoms so that they again meet full syndrome criteria. A recurrence is similar to a relapse except that it occurs after recovery rather than during remission. These proposed definitions are to facilitate communication and hopefully lead to a clearer understanding of the long-term outcome of depressive disorders.
Many randomized trials of antidepressant drugs are short term in nature and typically the acute responses are reported after a period of 6–8 weeks. While such short studies can address short-term efficacy and can be analysed to address issues of response and remission, they do not address the more important clinical outcome of recovery. Furthermore maintenance studies usually only address the issues of relapse and recurrence. Another issue in many antidepressant trials is the search for a mythical homogeneous sample and the application of many exclusion criteria [5]. These exclusion criteria may limit the generalizability of many antidepressant trials as in epidemiologic samples and in normal clinical practice [6] depressed patients with comorbid mental disorders are the rule rather than the exception.
The Christchurch Outcome of Depression Study was designed as a long-term treatment study of major depression. Depressed patients have been recruited from a variety of sources and psychiatric exclusion criteria have been minimized so to obtain a broad, heterogeneous and hopefully representative sample of patients with depression. Although in this paper we will just be presenting the results for remission, response and recovery to the first antidepressant to which patients were randomized, all patients are being followed-up for a period of 5 years so that the long-term outcome of these depressed patients can be documented and described. The study was not designed primarily as an efficacy study, but to describe the short- and long-term outcome of depression and to analyse the data for predictors of antidepressant response [7]. In an earlier short-term study we reported that both temperament [8] and neuroendocrine [9] variables predicted antidepressant response. The study was thus designed to provide optimal clinical treatment and examine how demographic, diagnostic, personality, neuroendocrine and genetic markers may predict outcomes.
When the study was commenced we expected that fluoxetine and nortriptyline would show comparable efficacy. Fluoxetine was chosen as the standard selective serotonin reuptake inhibitor (SSRI), and nortriptyline was chosen as it was a favoured tricyclic antidepressant with a different (predominantly noradrenergic) mode of action to fluoxetine and for which blood level monitoring could help optimize treatment outcomes.
In this first paper on treatment outcomes we report on the patterns of remission, response and recovery for these depressed patients in response to the initial antidepressant drug to which they were randomized. We also look at predictors of response and recovery by drug, demographic factors, and DSM-IV depression subtype.
Method
Patients
Depressed patients were recruited from a variety of sources including community mental health centres, the adjacent psychiatric emergency service, self-referral, and by referral from primary care physicians.
Patients were usually screened over the telephone by a research nurse and were then seen for an initial assessment by a psychiatrist or a senior psychiatric registrar. Following this initial assessment, eligible patients were invited to participate in this study after giving written informed consent. Patients were then booked to attend a detailed research assessment. This study has been approved the Canterbury Ethics Committee.
Inclusion and exclusion criteria
For inclusion in this study major, depressive disorder was required to be the principal current diagnosis and the treating psychiatrist needed to consider that treatment with antidepressant medication was appropriate management. Patients were excluded from the study if current moderate to severe alcohol or drug dependence was deemed to be the principal current diagnosis (however, many patients with mild to moderate alcohol or drug dependence were included as the treating clinician considered that major depression was the principal current diagnosis). Patients were also excluded if they had a history of mania (a history of hypomania was allowed), a history of schizophrenia, or a clinical history of severe antisocial personality disorder and considered unlikely to co-operate with the research protocol. Patients were also required to be free of any major physical illness and to be free of prescribed drugs for a minimum of 2 weeks except for the oral contraceptive and an occasional hypnotic for sleep.
Assessment
Following the initial clinical assessment and after giving consent all patients underwent a detailed clinical and neurobiological assessment. All patients were assessed using the Structured Clinical Interview for DSM-III-R (SCID-P) [10]. The SCID had been extended to include all DSM-III-R and DSM-IV melancholic and atypical criteria of depression. Patients were also rated on the Montgomery Asberg Depression Rating Scale (MADRS) [11], on the expanded 27-item Hamilton Depression Rating Scale (HDRS) [12], on the mental state examination for melancholia [13]. During this assessment phase patients also completed a series of self report questionnaires including the Hopkins Symptom Checklist (SCL-90) [14], the Social Adjustment Scale (SAS) [15], the SCID-PQ [10], the Temperament and Character Inventory (TCI) [16], the Parental Bonding Instrument (PBI) [17], the Dysfunctional Attitude Scale (DAS) [18], and the Interpersonal Sensitivity Scale (IPSS) [19].
Patients were also independently interviewed by a research nurse who collected detailed information on recent life events and of events in childhood including abuse. The research nurse also interviewed a nominated significant other of the patient for collaborative history and the significant other completed a complementary series of self-report questionnaires about the nominated patient.
In the weeks following the initial baseline assessment when the results from the SCID-PQ were available all patients were assessed for axis II psychopathology using the SCID-II [10].
During the assessment the patient attended for a full afternoon for a neurobiological assessment. This included serial blood samples to obtain a mean afternoon cortisol level, a thyrotropin releasing challenge test, plasma cholesterol, large neutral amino acids, plasma proteins; and DNA was obtained for genetic markers. Routine biochemical (e.g. glucose, electrolytes, renal and liver function tests) and haemotological tests were also completed.
Treatment
After completion of the assessment, depressed patients were randomly allocated to receive either fluoxetine or nortriptyline for an initial period of 6 weeks. If patients were randomized to fluoxetine they initially received 20 mg per day for a period of 3 weeks. After 3 weeks of treatment with fluoxetine, the clinician was able to adjust the dose to 40, 60 or 80 mg but had to ensure the patient was on a stable dose for the 10 days prior to the 6-week assessment. In exceptional circumstances (usually side-effects), the clinician could reduce the dosage to 10 mg. At 6 weeks the mean fluoxetine dose was 28.1 mg with the dose ranging from 10–80 mg and the most common dose was 20 mg.
If patients were randomized to nortriptyline they initially received 25 mg at night for one night, then 50 mg at night for one night and then 75 mg nocte. After 1 week on 75 mg of nortriptyline a blood level was obtained for nortriptyline levels. Clinicians were then free to adjust the dose based upon blood levels, plus clinical response and side-effects to try and ensure optimum treatment with nortriptyline over the period of 6 weeks. At 6 weeks the mean nortriptyline dosage was 93.5 mg with the dose ranging from 50 to 175 mg. After 6 weeks an adequate trial was judged to have been completed if the patient took the antidepressant, as prescribed for the majority of the 6-week period. In the case of nortriptyline, this also required evidence of adequate blood levels. In most instances, patients who did not complete an adequate treatment trial, stopped the drug within the first 2 weeks (usually due to sideeffects). One patient on nortriptyline continued with the drug for 6 weeks, but never exceeded 50 mg daily and never obtained blood levels approaching the recommended therapeutic levels due to postural hypotension. She was deemed not to have completed an adequate 6-week trial of the antidepressant.
Patients would normally be seen at least weekly for a period of 20–40 min depending upon the clinical need. During these sessions the objectives were to optimize antidepressant therapy, encourage compliance, assess progress and provide education about depressive disorders. Patients did not receive formal psychotherapy, but the clinical sessions were supportive, educational and delivered so as to optimize the results from antidepressant therapy. After the initial 6-week treatment trial, a decision was made as to whether the patient had had an adequate treatment trial and whether to continue with the initially prescribed antidepressant or to change treatment strategies. For patients who did not improve or who had only improved minimally over the initial 6 weeks with either fluoxetine or nortriptyline the protocol was for the patient to be swapped to the alternative antidepressant drug. Patients who had some improvement or a marked improvement with the initial drug to which they had been randomized were encouraged to continue with the initial drug for a minimum period of 6 months. During the subsequent phase of treatment patients would be seen as often as indicated by clinical need.
Outcome measures
Prior to treatment and at weeks 3, 6, 9, 13, 20 and 26, patients completed the SCL-90 and the SAS. At these times patients were also rated by the clinician using the HDRS, the MADRS and a global clinical judgement of improvement. During the course of treatment the research nurse regularly saw all patients and at 6 months the research nurse completed a structured interview that assessed depressive symptoms over the prior 6-month period.
Response was defined as a 60% or greater improvement on the MADRS over the first 6 weeks of treatment. Recovery was defined as a sustained global clinical improvement of much improved or very much improved for a minimum period of 2 months, and was a consensus rating based upon the assessments of both the treating psychiatrist and the research nurse.
Analyses
All data from the study were entered into the relational database Paradox and transferred to SYSTAT for statistical analyses. Statistical analyses included descriptive statistics, t-tests, analysis of variance, chi-squared tests, Fisher exact probability tests and logistic regression. Results presented in this paper relate to the baseline characteristics and outcome with the first antidepressant to which the patients were randomized. The key outcome criteria were the dichotomised measures of response (greater than or equal to 60% improvement on the MADRS) and recovery (2 months sustained improvement). Results were also analysed using the dimensional measure of mean percentage improvement on the MADRS but with any deterioration over the initial 6-week period coded as a 0% improvement. For the continuous measure of percentage improvement we utilized analysis of variance with subsequent Tukey post hoc tests. For the dichotomised measures of response and recovery logistic regression was utilized.
Results
Patient characteristics
Of the 195 depressed patients who were entered into this study, 111 (57%) were female and the mean age was 31.6 years. Using DSM-IV, 18 (10%) met criteria for bipolar II disorder, 86 (44%) met criteria for melancholia, 16 (8%) met criteria for atypical depression, 121 (62%) had recurrent depression, and 125 (64%) had chronic depression. Chronic depression was defined as being depressed for more than 2 years of the past 5. The baseline MADRS score was 31.0 (± 6.6), and the 17-item HDRS score was 19.9 (± 4.4). It is of note that just over 60% of the depressed patients in the sample had never been prescribed an antidepressant in their life.
Dropouts
Of the 195 patients who were randomized for treatment, 100 received fluoxetine and 95 nortriptyline. As can be seen from Table 1, 14 of the patients randomized to fluoxetine did not complete an adequate 6-week trial whereas 27 of those randomized to nortriptyline did not complete an adequate 6-week trial (p = 0.02). Table 1 also shows a further analysis of these dropouts by drug and gender, from which it can be seen that women were especially likely to dropout if prescribed nortriptyline whereas men were more likely to dropout if they were randomized to fluoxetine (p = 0.002).
Patients completing an adequate 6-week randomized trial of either fluoxetine or nortriptyline
Six-week outcome including response
After 6 weeks, for the 86 patients who completed an adequate trial of fluoxetine, their MADRS score had decreased 64% to 12.1 (± 10.3). For the 68 patients with nortriptyline, their MADRS score had decreased 58% to 13.4 (± 11.3). These changes or 6 week scores are not significantly different.
Using a definition of response as a 60% or greater improvement on the MADRS, and using an intention to treat analysis, then 54% responded to fluoxetine and 41% responded to nortriptyline (OR = 1.92; p = 0.025; 95% CI = 1.09–3.41). This difference in response rate largely reflects the greater dropout rate with nortriptyline.
Recovery
While acute treatment studies of depression typically focus on the 6-week outcome clinicians and patients are most interested in recovery, which for this study has been defined as a minimum period of 2 months during which time they are globally rated as much or very much improved. Of the 100 patients randomized to fluoxetine, 57% met the criteria for recovery which is significantly higher than the 37% recovery rate achieved with nortriptyline (OR = 2.27; p = 0.005; 95% CI = 1.28–4.04). A further breakdown of this data shows that with fluoxetine 33 patients neither responded nor recovered and 44 patients both responded and recovered. However, 10 patients met criteria for response but this was not sustained and so they were not deemed to have recovered but conversely 13 patients had not met criteria for response at 6 weeks but were deemed to have recovered with fluoxetine by sustaining a long-term improvement. With nortriptyline 53 patients neither responded nor recovered and 29 patients both responded and recovered. However, seven patients responded but did not sustain this to meet criteria for recovery, and six patients at 6 weeks had not met criteria for response but were ultimately deemed to have recovered with nortriptyline. What this means is that of the 195 patients there were 23 patients on fluoxetine and 13 on nortriptyline for whom response and recovery are not the same. This suggests that for 20% of patients there is a discrepancy between being classified as having achieved a response or recovery. For studies such as this one which are interested in predictors of response this misclassification rate could have significant potential to obscure significant findings.
Predictors of response and recovery
Table 2 shows the mean percentage improvement in depressed patients by age, sex, depression subtype and drug. The data were analysed using analysis of variance and whenever there was a significant finding, subsequent post hoc tests were performed. From this table which only presents the results of those completing an adequate 6-week trial it can be seen that those aged less than 25 have a poorer response to nortriptyline, that patients meeting criteria for atypical depression have a poorer response to nortriptyline, and that patients with chronic depression have a poorer outcome than those with nonchronic depression regardless of drug.
Mean percent improvement (MADRS) by age, sex, depression subtype and drug
Table 3 shows response and recovery rates by age, sex, depression subtype and drug. In this table analyses were undertaken using intention to treat. From this table it can be seen that females have a lower response rate and lower recovery rate with nortriptyline compared with fluoxetine. In the subjects under the age of 25 nortriptyline is inferior to fluoxetine on both response and recovery measures. Similarly in those meeting criteria for atypical depression nortriptyline is inferior to fluoxetine for both measures of response and recovery. Of note in these results is that there is no trend for nortriptyline to be superior to fluoxetine in those with melancholia. The only instance where the odds ratio is less than 1 (i.e. in favour of nortriptyline) is that the response rate in men slightly favours nortriptyline although this is lost when one looks for recovery.
Response (≥60% improvement ΔMADRS3) and recovery (2 months much improved) rates by age, sex, depression subtype and drug. Analyses undertaken using intention to treat on all 195 depressed patients
Discussion
Six-week outcome
The Christchurch Outcome of Depression Study is a long-term prospective naturalistic study designed to describe the short- and long-term outcome of depression and to examine for predictors of outcome. We attempted to optimize the psychopharmacological treatment by allowing the use of flexible dosages and the use of blood level monitoring in the case of nortriptyline. Although the study can be described as a long-term naturalistic study it has the additional strength that the depressed patients were randomized to the initial treatment; however, treatment was not blind. In this first major paper describing the outcome of the depressed patients in the study we report a number of important findings which have implications for the treatment of depression and for the design of future studies of antidepressant efficacy.
The first important finding is that more patients randomized to nortriptyline were less likely to complete an adequate trial than those randomized to fluoxetine. Furthermore there was an important interaction between gender and type of medication. Thus women were more likely to drop out of the study if prescribed nortriptyline and men more likely if they were prescribed fluoxetine. These results are essentially compatible with the recent findings of Kornstein et al. [20] who, in a randomized treatment trial of the SSRI sertraline versus the TCA imipramine, found that women did not continue taking imipramine and men did not take sertraline. Thus, both studies suggest that men discontinue SSRIs, and women tricyclic antidepressants at higher rates.
If one examines the results at 6 weeks of the 154 depressed patients who completed an adequate antidepressant trial there are no significant differences in outcome after 6 weeks. This result is consistent with other results of SSRIs versus tricyclics, which suggest they are of comparable efficacy [21]. However, fluoxetine was superior to nortriptyline using an intention to treat analysis, although most of this difference can be attributed to the higher dropout rate with nortriptyline.
Recovery
While at 6 weeks it is difficult to conclude that there is any real difference between fluoxetine and nortriptyline, when we examined recovery rates (i.e. 2 months much improved), an important new finding clearly emerges. It is almost certainly of interest to clinicians to note that, of patients randomized to fluoxetine 57% achieved a recovery, whereas of those randomized to nortriptyline only 37% achieved recovery. We are unaware of any other treatment studies which have reported recovery rates (as opposed to symptom change, remission or response) and consider this an important but unanticipated finding which will require further replication. It is perhaps of note that some clinicians have believed that the SSRIs have been more effective than the tricyclic antidepressants but the data from trials never supported this view. However, as trials trend to report only short-term and not long-term results this could potentially explain the discrepancy between the experience of some clinicians and the findings of randomized controlled trials.
There are perhaps indications in the literature which support this finding. Given that depressed patients seldom continue to take an antidepressant drug which is not working there are a number of studies which have reported on the probability of staying with the initial antidepressant to which a depressed patient was allocated. Thus, Simon et al. [22] in a prospective naturalistic study in a health maintenance organization reported that of 173 initially treated with fluoxetine, 61% completed 90 days of adequate treatment. In contrast, 48% of 182 patients treated with imipramine, and 49% of 181 patients treated with desipramine, completed 90 days of adequate treatment. Croghan et al. [23] reported that 40% of patients with fluoxetine received continuous treatment for 6 months, but for the second generation of tricyclic antidepressants the comparable figure was 31%, and for the first generation tricyclic antidepressants the figure was only 24%. Katon et al. [24] reported that on fluoxetine about 50% of patients initially commenced continued to receive it for 5 months, but with tricyclics less than 30% of patients who commenced received the drug for 5 months. One interpretation of these studies is that staying on the initially prescribed antidepressant is a proxy measure of satisfaction with drug treatment and perhaps therefore of recovery. In these three large studies patients commenced on fluoxetine are much more likely to stay with their initial drug than those who were prescribed a tricyclic antidepressant. Together these results suggest that the minimum duration of an acute antidepressant treatment should be 3 months with recovery as the key outcome measure.
Predictors of response and recovery
In this study we have found four important predictors of outcome, namely chronicity, age, gender and atypical depression. Of these four only the chronic/non-chronic distinction was independent of drug and will be discussed first. It is not surprising that those with chronic depressions have a lower percentage improvement than those with non-chronic (acute) depression. This is consistent with other findings, but it should be borne in mind that this was not a placebo-controlled study. Studies which utilize placebo and examine duration of depression report considerably higher placebo-response rates in acute depression than in chronic depression. This probably reflects a higher rate of natural recovery in patients with short duration depressions. Indeed the drug placebo difference may be greater in more chronic depressions than in depressions of shorter duration [25]. We thus interpret these findings as meaning that those with acute depression have a greater improvement because of a markedly greater natural recovery rate.
The finding that patients with atypical depression do poorly with nortriptyline is not original. Indeed a key part of the rationale for introducing atypical depression as a diagnostic specifier in DSM IV was because of the well documented evidence of poor response to tricyclic antidepressants. However, the majority of these studies on atypical depression had compared phenelzine with imipramine [26–28] and have shown the clear superiority of phenelzine. There have been relatively few studies which have looked at SSRIs in atypical depression but the limited data which exists suggests that SSRIs such as fluoxetine are effective in atypical depression [29].
One of the important new findings from this study is that depressed patients under the age of 25 years have poorer response and recovery rates with nortriptyline compared with fluoxetine. We have reanalysed the data from an earlier antidepressant study [8], which compared clomipramine with desipramine in depressed patients. In that study we have also found that the depressed patients under the age of 25 had an inferior response to desipramine clomipramine [30]. This finding is also consistent with emerging literature on the pharmacotherapy of childhood and adolescent depression. In adolescence it appears that tricyclic antidepressants are not effective but that fluoxetine and presumably other SSRIs are effective [31]. Our data therefore suggest that the superiority of SSRIs over tricyclic antidepressants which is observed in adolescents continues up to the age of about 25 years. The key issue for future study is the underlying developmental neurobiological basis for this differential response to serotonergic antidepressants in children, adolescents and young adults. With the development of selective nordrenaline re-uptake inhibitors, it will be important to determine whether these are effective in the under 25-year-old-age group as this may help determine whether this is a serotonergic versus noradrenergic finding or whether the tricyclic antidepressants are not effective at younger ages. However, the results from our earlier study which found that clomipramine was more effective than desipramine suggests this lack of efficacy is more likely to be related to a predominantly noradrenergic mechanism rather than to the tricyclic antidepressant class.
Another key finding from this study is the influence of gender on response and recovery rates. Our findings again are consistent with the large study in dysthymia and chronic depression completed by Kornstein et al. [20]. We would reiterate the conclusion from that study that gender needs to be considered when choosing an appropriate antidepressant for a depressed patient. In the Kornstein study, it was reported that this differential response was particularly notable in the premenopausal woman. In postmenopausal women there were comparable responses to sertraline and imipramine. As the mean age of our sample was only 31.6 years and we only had about 10 postmenopausal women in the sample we are unable to comment as to the generalizability of these findings to the postmenopausal phase of life.
One of the notable negative findings in this study was that those meeting DSM IV criteria for melancholia did not have a better response to nortriptyline than fluoxetine. A number of authors (e.g.[32, 33]), suggest that tricyclic antidepressants are superior to the SSRIs in melancholia. It could be that such a difference in favour of tricyclic antidepressants is most likely to be seen in severely depressed hospitalized melancholic patients and may not generalize to younger outpatients with melancholia. In a later paper we will examine different definitions of melancholia in relationship to the differential response to fluoxetine and nortriptyline.
Limitations
In the interpretation of the results from this study there are at least two important limitations. The first is that the treatment was not double blind although patients were randomized. The decision not to undertake treatment in a double blind manner was made after a number of issues were considered. In the first instance the primary area of interest was to describe the long-term patterns of outcome and to look at predictors of response rather than to compare the efficacy of fluoxetine and nortriptyline. When the study began we accepted the data that fluoxetine and nortriptyline were of comparable efficacy and we chose nortriptyline because of our own favourable experience with it. Another issue in the decision not to make the study double blind was that we saw this as a long-term treatment study with few psychiatric exclusions. Thus we included depressed patients who were highly suicidal (if treatable as outpatients); who had comorbid disorders, including alcohol and drug dependence. We therefore consider that these depressed patients are more representative of normal clinical practice than is the case in many antidepressant trials. For similar reasons we also chose not to include a placebo arm in this study. However, as this study was neither double blind nor placebo controlled we accept that our findings of recovery should be taken as suggestive rather than definitive, but would urge further investigators to look at recovery as the key outcome variable in depression studies rather than shorter term measures such as depression symptom changes, remission or response.
Another feature of this study is that the mean age of our sample was relatively young, and that the majority had chronic depression of early onset. This makes our findings very comparable with the Kornstein et al. study [20] on dysthymia and chronic depression.
Conclusions
As the paradigm for the treatment of depression shifts from short-term to long-term treatment, it is important that studies which bridge the interface between trials and real life effectiveness [34] are undertaken. That recovery rates differ between drugs which appear comparable in the short-term deserves further evaluation. Indeed for clinicians and patients recovery is more relevant than many variables chosen as endpoints in current studies of depression.
Footnotes
Acknowledgements
The clinical research unit is supported by the University of Otago, the Mental Health Division of Canterbury Health and a programme grant from the Health Research Council of New Zealand. This project also received a grant from Lottery Health and an unrestricted grant from Eli Lilly (New Zealand).