The course of depression

T H E C O U R S E O F D E P R E S S I O N
R O B E R T J . B O L A N D
M A R T I N B . K E L L E R
Long-term naturalistic studies have changed the way we remission, an individual still has more than minimal view depression. Whereas it was often previously viewed as symptoms. Full remission is defined as the point at an episodic disease, the past two decades of research have which an individual no longer meets criteria for the dis- underscored the importance of understanding depression as order and has no more than minimal symptoms.
a lifelong disease, with a number of possible courses.
3. Recovery, defined as a full remission that lasts for a de- An appreciation of this longitudinal data is crucial to fined period of time. Conceptually, it implies the end understanding all aspects of depression. Cross-sectional of an episode of an illness, not of the illness per se.
judgments of symptomatic severity provide limited prog- 4. Relapse, defined as a return of symptoms sufficient to nostic information. A full understanding of a patient’s prog- satisfy full criteria for an episode. It occurs in an interval nosis or likely treatment response also requires a longitudi- of time before what is defined as ‘‘recovery.’’ Concep- nal perspective. Which patient is likely to recover fully, and tually, this refers to the return of an episode, not a new who will suffer from a chronic mood disorder? What length of treatment will be sufficient for such patients? 5. Recurrence, defined as a return of full symptomatology Studies within the last decade have helped to shed light occurring after the beginning of the recovery period.
on these questions. This chapter examines some of these Conceptually, this represents the beginning of a new studies, and discusses their implications for our approach to depression. Limitations of the data will be discussed aswell.
REPRESENTATIVE STUDIES
THE CHANGE POINTS OF DEPRESSION
A relatively small number of studies have been particularly Considerable confusion has resulted from the use of various influential in shedding light on the course of depression.
terms to denote the different change points in the courseof depression. Similar terms, such as ‘‘relapse’’ and ‘‘recur-rence’’ have been used interchangeably and inconsistently The Collaborative Depression Study
in different studies. As a result, the MacArthur Foundations Research Network on the Psychobiology of Depression (1) The CDS (2) is a prospective long-term naturalistic study of the natural course of depression. Subjects were recruited 1. Episode, defined as a certain number of symptoms for a from patients with depression seeking psychiatric treatment at one of several sites (university or teaching hospitals in 2. Remission, defined as a period of time in which an indi- Boston, Chicago, Iowa City, New York, and St. Louis).
vidual no longer meets criteria for the disorder. In partial This study included programs in biological and clinicalstudies. The data presented here are from the clinical studiesprogram; 555 subjects in the clinical studies program had anindex episode of unipolar major depression. Subjects were Robert J. Boland: Department of Psychiatry and Human Behavior,
Brown University; Department of Psychiatry, Miriam Hospital, Providence, examined at 6-month intervals for 5 years and then annually for a minimum of 18 years. Recent National Institute of Martin B. Keller: Department of Psychiatry and Human Behavior, Brown
University; Department of Psychiatry, Butler Hospital and Brown Affiliated Mental Health (NIMH) funding will extend the follow-up Hospitals, Providence, Rhode Island.
to at least 23 years on all subjects.
Neuropsychopharmacology: The Fifth Generation of Progress The Zurich Study
(5). However, for those patients who did not recover in thefirst year, most still had not recovered within 5 years. Thus Angst (3), in Zurich, has conducted the only other long- by 2 years, about 20% of the original sample were still term prospective study of mood disorders. In that study, depressed—two-thirds of those still depressed at 1 year were 173 hospitalized patients with unipolar depression were still in their index episode of depression at 2 years. At 5 identified between 1959 and 1963. This group was then years, 12% of patients had still not recovered (6), by 10 evaluated every 5 years for up to 21 years of follow-up.
years 7% had not recovered (7), and by 15 years, the num- The Medical Outcomes Study(MOS)
bers seem to have leveled off at 6%. These data are presentedin Fig. 69.1.
The MOS (4) examined the course of several diseases (myo- The long duration of the CDS allowed the investigators cardial infarction, congestive heart failure, hypertension, di- to observe subsequent episodes of major depression begin- abetes, and depression) in a variety of health care settings, ning during the study. This was particularly useful, as the including large medical group practices, small group prac- onset of symptoms could be identified more accurately than tices, and solo practices, in three cities (Los Angeles, Boston, for the retrospective determination done for an index epi- and Chicago). A representative sample of different medical sode. It was found that, for each new episode of depression, specialties—including psychiatry—was chosen, and all pa- the rates of recovery were similar to that seen during the tients seen from February through October 1986 were asked index episode. Thus, for the second episode (first prospec- to participate in the study. In all, over 20,000 patients par- tively observed episode) approximately 8% of subjects did ticipated, and were evaluated yearly for 3 years.
not recover after 5 years. An analysis of subsequent episodes(second, third, and fourth prospectively observed episodes) THE COURSE OF DEPRESSION: CHANGE
showed similar findings. By the fifth episode, the rate de- creases, but not significantly so (8). It appears that for eachepisode of depression, some individuals—about 10%—re- Traditionally, depression was pictured as an acute illness, self-limited, and lasting approximately 6 to 9 months from The seemingly high rate of chronicity was surprising. A time of onset to full recovery. A number of studies, includ- reasonable concern about this result was that the patient ing those mentioned above, however, show the potential population studied may have been unusually treatment re- for great variation from this traditional model. Recovery sistant. The study used a convenient sample of patients seek- may take much longer, or not occur at all (i.e., chronic ing inpatient or outpatient treatment at any one of five depression). Furthermore, the risk of relapse and recurrence major medical centers. However, most patients studied re- ceived either no treatment or subtherapeutic doses given forvery brief durations (9). Thus, the CDS cohort does not Recovery
seem to be biased in the direction of treatment resistance.
In the CDS, approximately 70% of patients recovered from Furthermore, other studies show comparable data. In the the index episode of major depression within the first year Zurich study, Angst et al. (10) reported that during the FIGURE 69.1. Outcome of maintenance therapy for de-
pressed patients initially stabilized on imipramine plus
lithium.
Chapter 69: The Course of Depression follow-up evaluations, about 13% of patients did not re- MITIGATING FACTORS
cover from their episode of major depression. In the MOS, Comorbidity
patients were divided by severity: of those with milderdepression, about 65% recovered within 2 years, whereas 54% of the more severely depressed group recovered in the There are few longitudinal studies looking at the outcome of depression in medically ill patients, partly because of the Shorter studies also give similar results. Rounsaville and difficulties inherent in recruiting such an unstable popula- colleagues (12), in a prospective follow-up of 96 patients tion. Studies that exist suggest that comorbid medical illness with major depression, found that 12% of subjects had not predispose individuals to a worse course of depression. The recovered after 16 months. Kerr et al. (13), following ini- MOS, for example, found an additive effect on patient func- tially hospitalized patients, found that 6% remained ill for tioning when depression and other chronic medical illnesses For the 141 patients in the CDS who recovered from their Double depression refers to the presence of concurrent dys- index episode of major depression, 22% relapsed within 1 thymia and major depression. In this disorder, the episodes year of follow-up (14). Factors predicting relapse included of major depression are superimposed on a more chronic multiple episodes of major depression, older age, and a his- depressive disorder. It appears to be common—studies sug- tory of nonaffective psychiatric illness. The characteristics gest that between one-fourth and two-thirds of patients with of this relapsed group were also examined, and it was found major depression will also have a comorbid dysthymia.
that the likelihood of remaining depressed for at least a The comorbid presence of dysthymia can have an impor- year after relapse was 22%. Predictors of prolonged time to tant effect on the course of depression. In the collaborative recovery included a longer length of the index episode, older study, it was found that patients with double depression recovered more rapidly from episodes of major depression Most studies look at relapse in terms of how it is affected than those with major depression alone. However, the au- thors also found that the recovery tended to be not to oneof ‘‘normalcy,’’ but to one of dysthymia. Relapse is alsomore frequent in patients with double depression than those Recurrence
with major depression alone—almost twice as likely in onestudy of 32 double-depressed subjects followed for 2 years Angst (15), reporting on a 10-follow of patients in the Zu- (20). The MOS also found that full recovery was less likely rich study, found that only 25% of patients had only a single for patients with double depression—these patients had a episode of depression. Thus, three-fourths of the sample threefold risk of continued disease when compared with had a recurrent depression, with one or more recurrences.
those with major depression alone (21).
Though Angst examined a number of sociodemographicvariables, none significantly predicted the likelihood of re-currence.
Similarly high rates of recurrence have been found in Substance Abuse
other long-term studies. Weissman and Kasl (16) found that Clearly, comorbid substance abuse has a detrimental effect two-thirds of woman seen over 1 year had a recurrence of on the course of depression. The CDS found that subjects depression. Rao and Nammalvar (17), examining over 100 who were currently alcoholic were half as likely as nonalco- cases of depression in India for a follow-up of between 3 holic depressed subjects to recover from their episode of and 13 years, found that only about a fourth of the original major depression (22). Patients with a previous, but not group reported no recurrence of symptoms.
current, history of alcoholism had a recovery rate compara- The rate and timing of recurrence seems most dependent on the type of recovery. Patients in the CDS who fullyrecovered (i.e., were asymptomatic on follow-up evaluation) Anxiety Disorders
had a much lower rate of recurrence (66%) than those with Anxiety disorders are commonly comorbid with depression.
some residual symptoms (87%). The time to recurrence was The presence of such comorbid disorders appears to exert also much longer in the asymptomatic group: mean of 180 a negative effect on the course of depression. Coryell and weeks in the asymptomatic group compared with 33 weeks colleagues (23) found that depressed patients with panic in the group with residual symptoms (18).
disorder had a slower time to recovery than those without Neuropsychopharmacology: The Fifth Generation of Progress comorbid panic. The CDS similarly found that patients TABLE 69.1. RELAPSE RATES VS. PLACEBO:
with higher symptom ratings of anxiety had longer times CONTINUATION STUDIES
to recovery from major depression (24).
(Reference)
Treatment
(Placebo) % P Value
FamilyHistory
A family history of depression appears to predispose an indi- vidual to depression. Two studies have looked at the rela- tionship between parental history of depression and course of depression in the offspring. Though these studies had relatively small sample sizes, they both suggested that pa-tients with a parental history of depression had a longertime to recovery than other patients (25,26).
take inhibitors for continuation therapy, including fluoxe-tine (27), paroxetine (28), sertraline (29), citalopram (30), Treatment Variables
and mirtazapine (31). Nefazodone has also been shown in Clearly, one of the questions of most practical interest is continuation treatment (32). These studies are summarized whether pharmacologic treatment is capable of significantly in Table 69.1. However, when looking beyond continua- altering the course of depression for a patient. Antidepres- tion therapy to the maintenance treatment of recurrent sants are generally used at all stages of depression—to hasten recovery, prevent relapse, and prevent recurrence of depres- Prien and colleagues (33) reported on a 2-year mainte- sion. However, as will be discussed, the further one looks nance trial for depression. Patients who were successfully down the course of depression, the less is really known about treated for acute depression were randomized to receive lith- the ability of antidepressant and other pharmacologic treat- ium carbonate, imipramine, both, or placebo. Treatments ments to alter the course of depression.
were continued for 2 years with doses maintained at acute A wealth of data supports the efficacy of all available treatment levels. Of 150 patients beginning maintenance antidepressants in shortening the time to recovery from treatment, 36% were successfully treated. The lowers rate major depression. However, when one goes beyond the of recurrence was found it the group treated with imipra- acute phase and examines pharmacotherapy during later mine (Fig. 69.1). However, even this group had a 47% points in the course of depression, the data become more A second study, the Pittsburgh Study of Maintenance Data support the efficacy of most of the serotonin reup- Treatment for Recurrent Depression (34), reports on up to FIGURE 69.2. Pittsburgh Study of Maintenance Therapies in Recurrent Depression.
Chapter 69: The Course of Depression 5 years of maintenance treatment. In this study, subjects To date, there are only two published studies on the long- first underwent open treatment for acute depression, using term treatment of chronic depression. Kocsis and colleagues imipramine with interpersonal therapy (IPT). Patients who (36) report on a placebo-controlled trial of desipramine for achieved recovery for at least 4 months were then random- the treatment of chronic depression. The study included ized into one of five treatment conditions: (a) IPT alone, patients with chronic major depression (n ס 14).
given monthly; (b) imipramine treatment alone; (c) IPT After successful acute-phase and continuation treatment, plus placebo; (d) placebo plus medication clinic; and (e) patients were continued on treatment for a total of 2 years.
imipramine plus IPT. This portion of the study was contin- During this maintenance period, patients on the placebo ued for 3 years. Results from this study are summarized in had four times the recurrence rate of those receiving desipra- mine. This rate was consistent for all diagnostic groups, The authors intentionally chose patients who had highly including those with chronic depression.
recurrent depression (to maximize the likelihood of seeing a Keller and colleagues (37) investigated the treatment of statistical difference in the groups). They found that patients chronic depression in a larger sample of patients. Here, 161 who received imipramine (with or without IPT) had an patients who were successfully treated during an acute-phase approximately 20% risk of recurrence after 3 years. This and continuation phase were randomized to receive with compares significantly with the other conditions: IPT alone sertraline or placebo for a 76-week maintenance period. The had a 60% risk of recurrence, and the placebo condition study included both chronic major depression and double depression patients in roughly equal numbers; as the results This study was extended, with a smaller sample, for an- did not differ between the groups, the data was pooled.
other 2 years. The subjects who had remained well during Patients who received the placebo during the mainte- the first 3 years of the study were randomized to receive nance phase of treatment were four times more likely to either imipramine or placebo. At that point only 20 patients have a recurrence of depression than those receiving sertra- remained in the study. After the two additional years, less line. The time to recurrence was delayed for patients treated than 10% of the patients (one patient out of 11) receiving with sertraline compared to those treated with the placebo.
imipramine had a recurrence, where approximately two- Using a less stringent criterion of reemergence of depressive thirds of the placebo group had a recurrence.
symptoms (though no necessarily meeting full criteria for Thus, the world literature of placebo-controlled studies depression), it was found that only 26% of patients taking of the treatment of recurrent depression beyond 3 years sertraline experienced a reemergence of depressive symp- consists of only 200 subjects, who had a history of highly toms, compared with 50% of those on the placebo. Simi- recurrent depression and received a medication that is rarely larly, only 34% of patients on sertraline maintenance ther- apy shored first symptoms of depression, compared with60% of patients taking the placebo (Table 69.2).
OTHER SUBTYPES OF DEPRESSION
The data are also limited when one considers the effect oftreatment on the course of specific subtypes of depression.
TABLE 69.2. RECURRENCE RATES OF MAJOR
Two such subtypes will be considered here: chronic major DEPRESSION DURING MAINTENANCE STUDY
TREATMENT
Sertraline
Treatment Of Chronic Depression
(n = 77)
(n = 84)
P Value
Chronic depression is thought to respond more poorly to antidepressant treatment. Thus, studies of the acute and long-term treatment for this subtype are of great impor- tance. Particularly lacking are studies of psychotherapy in Keller and colleagues (35) recently compared antidepres- sant treatment and cognitive-behavioral therapy, both alone and combined. For the 519 subjects completing the study, 55% of the antidepressant (nefazodone) group and 52% of the psychotherapy group responded to treatment. However, when treatment was combined, the response rate jumped to 85%. Thus, this study gives strong support to the clinical From Keller MB, Kocsis JH, Thase ME, et al. Maintenance phase wisdom that combined treatment is preferable to either efficacy of sertraline for chronic depression: a randomized controlled trial. JAMA 1998;280:1665–1672, with permission.
Neuropsychopharmacology: The Fifth Generation of Progress TABLE 69.3. PHARMACOTHERAPY OF DYSTHYMIA: SELECTED AGENTS
DEMONSTRATING A POSITIVE EFFECT IN RANDOMIZED-CONTROLLED
TRIALS
Duration of Study
Compared to:
Treatment Of Dysthymia
depression research. Berlin/Heidelberg: Springer-Verlag, 1986:26–35.
Few studies have examined the pharmacotherapy of dysthy- 4. Wells KB, Burnam A, Rogers B, et al. The course of depression mia, possible because of long-held beliefs that nonmajor in adult outpatients: results from the Medical Outcomes Study.
depressions were less responsive to pharmacotherapy. What Arch Gen Psychiatry 1992;49:788–794.
5. Keller MB, Shapiro R, Lavori PW, et al. Recovery in major de- data do exist, however, do not support this belief. Most pressive disorder: analysis with the life table and regression studies are of a relatively short duration of treatment, rang- models. Arch Gen Psychiatry 1982;39:905–910.
ing from 4 to 12 weeks. For this time period there are data 6. Keller MB, Lavori PW, Mueller TI, et al. Time to recovery, to support the use of most classes of antidepressants. These chronicity, and levels of psychopathology in major depression. A studies are summarized in Table 69.3. Thus, the weight of 5-year prospective follow-up of 431 subjects. Arch Gen Psychiatry evidence suggests that most agents that are effective for 7. Mueller TI, Keller MB, Leon A, et al. Recovery after five years major depression are also effective for dysthymia, at least of unremitting major depressive disorder. Arch Gen Psychiatry 8. Solomon DA, Keller MB, Leon AC. Recovery from major depres- sion. A 10-year prospective follow-up across multiple episodes.
ACKNOWLEDGMENTS
Arch Gen Psychiatry 1997;54:1001–1006.
9. Keller MB, Klerman GL, Lavori PW, et al. Treatment received by depressed patients. JAMA 1982;248:1848–1855 Dr. Keller has received research support and/or served as a 10. Angst J, Baastrup P, Grof, et al. The course of monopolar depres- consultant or on an advisory board for a number of different sion and bipolar psychoses. Psychiatr Neurol Neurochir 1973;76: pharmaceutical companies including Pfizer, Bristol-Myers Squibb, Forrest Laboratories, Wyeth-Ayerst Laboratories, 11. Wells KB, Burnam A, Rogers B, et al. The course of depression Merck, Janssen, Eli Lilly, Organon, Pharmacia/Upjohn, in adult outpatients: results from the Medical Outcomes Study.
Arch Gen Psychiatry 1992;49:788–794.
SmithKline Beecham, Zeneca, Mitsubishi Pharmaceuticals, 12. Rounsaville BJ, Prusoff BA, Padian N. Chronic mood disorders Scirex, Janus Pharmaceuticals, Sepracor Pharmaceuticals, in depression outpatients: a prospective 16-month study of ambu- Somerset Pharmaceuticals, and Sanofi-Synthelabo.
latory patients. J Nerv Ment Dis 1980;168:406–411.
13. Kerr TA, Roth M, Schapira K, et al. The assessment and predic- tion of outcome in affective disorders. Br J Psychiatry 1972;121: REFERENCES
14. Keller MB, Lavori PW, Lewis CE, et al. Predictors of relapse in 1. Frank E, Prien RF, Jarrett RB, et al. Conceptualization and ra- major depressive disorder. JAMA 1983;250:3299–3304.
tionale for consensus definitions of terms in major depressive 15. Angst J. How recurrent and predictable is depressive illness? In: disorder: remission, recovery, relapse, and recurrence. Arch Gen Montgomery S, Rouillon F, eds. Long-term treatment of depres- Psychiatry 1991;48:851–855.
2. Katz M, Klerman GL. Introduction: overview of the clinical stud- 16. Weissman MM, Kasl SV. Help-seeking in depressed out-patients ies program. Am J Psychiatry 1979;136:49–51.
following maintenance therapy. Br J Psychiatry 1976;129: 3. Angst J. The course of major depression, atypical bipolar disorder, and bipolar disorder. In: Hippius H, et al., eds. New results in 17. Rao AV, Nammalvar N. The course and outcome in depressive Chapter 69: The Course of Depression illness: a follow-up study of 122 cases in Madurai, India. Br J tenance therapies in recurrent depression. Arch Gen Psychiatry Psychiatry 1977;130:392–396.
18. Keller MB, Boland RJ. Implications of failing to achieve success- 35. Keller MB, McCullough JP, Klein DN, et al. A comparison of ful long-term maintenance treatment of recurrent unipolar major nefazodone, the cognitive behavioral-analysis system of psycho- depression. Biol Psychiatry 1998;44:348–360.
therapy, and their combination for the treatment of chronic 19. Wells KB, Steward A, Hays RD, et al. The functioning and well- depression. N Engl J Med 2000;342:1462–7140.
being of depressed patients—results of the Medical Outcomes 36. Kocsis JH, Friedman RA, Markowitz JC, et al. Maintenance Study. JAMA 1989;262:914–919.
therapy for chronic depression. A controlled clinical trial of desi- 20. Keller MB, Lavori PW, Rice J, et al. The persistent risk of chron- pramine. Arch Gen Psychiatry 1996;53:769–774.
icity in recurrent episodes of non-bipolar major depressive disor- 37. Keller MB, Kocsis JH, Thase ME, et al. Maintenance phase effi- der: a prospective follow-up. Am J Psychiatry 1984;143:24–28.
cacy of sertraline for chronic depression: a randomized controlled 21. Wells KB, Burnam A, Rogers W, et al. The course of depression trial. JAMA 1998;280:1665–1672.
in adult outpatients: results from the medical outcomes study.
38. Kocsis JH, Frances AJ, Voss C, et al. Imipramine treatment for Arch Gen Psychiatry 1992;49:788–794.
chronic depression. Arch Gen Psychiatry 1985;45:253–257.
22. Mueller TI, Lavori PW, Keller MB, et al. Prognostic effect of 39. Stewart JW, McGrath PJ, Quitkin FM, et al. Chronic depression: the variable course of alcoholism on the 10-year course of depres- response to placebo, imipramine and phenelzine. J Clin Psycho- sion. Am J Psychiatry 1994;151:701–706.
23. Coryell W, Endicott J, Andreasen NC, et al. Depression and 40. Stewart JW, Quitkin FM, Liebowitz MR, et al. Efficacy of desi- panic attacks: the significance of overlap as reflected in follow- pramine in depressed outpatients. Response according to research up and family study data. Am J Psychiatry 1998;145:293–300.
diagnosis criteria diagnoses and severity of illness. Arch Gen Psy- 24. Clayton PF, Grove WM, Coryell W, et al. Follow-up and family study of anxious depression. Am J Psychiatry 1991;148: 41. Hellerstein DJ, Yanowitch P, Rosenthal J, et al. A randomized double-blind study of fluoxetine versus placebo in the treatment 25. Warner V, Weissman MM, Fendrich M, et al. The course of of dysthymia. Am J Psychiatry 1993;150:1169–1175.
major depression in the offspring of depressed parents: incidence, 42. Vanelle JM. Controlled efficacy study of fluoxetine in dysthymia.
recurrence and recovery. Arch Gen Psychiatry 1992;49:795–801.
Br J Psychiatry 1997;170:345–350.
26. Hammen C, Burge D, Burney E, et al. Longitudinal study of 43. Ravindran AV, Bialik RJ, Lapierre YD. Therapeutic efficacy of diagnoses in children of women with unipolar and bipolar affec- specific serotonin reuptake inhibitors (SSRIs) in dysthymia. Can tive disorder. Arch Gen Psychiatry 1990;47:1112–1117.
J Psychiatry 1994;39:21–26.
27. Montgomery SA, Dunfour H, Brion S, et al. The prophylactic 44. Thase ME, Fava M, Halbreich U, et al. A placebo-controlled, efficacy of fluoxetine in unipolar depression. Br J Psychiatry 1988; randomized clinical trial comparing sertraline and imipramine for the treatment of dysthymia. Arch Gen Psychiatry 1996;53: 28. Montgomery SA, Dunbar G. Paroxetine is better than placebo in relapse prevention and the prophylaxis of recurrent depression.
45. Keller MB, Harrison W, Fawcett JA, et al. Treatment of chronic Int Clin Psychopharmacol 1993;8:189–195.
depression with sertraline or imipramine: preliminary blinded 29. Doogan DP, Caillard V. Sertraline in the prevention of depres- response rates and high rates of undertreatment in the commu- sion. Br J Psychiatry 1992;160:271–222.
nity. Psychopharmacol Bull 1995;31:205–212.
30. Montgomery SA, Rasmussen JG, Tanghol P. A 24-week study 46. Bakish D, Ravindran A, Hooper C, et al. Psychopharmacological of 20 mg citalopram, 40 mg citalopram and placebo in the pre- treatment response of patients with a DSM-III diagnosis of dys- vention of relapse of major depression. Int Clin Psychopharmacol thymia disorder. Psychopharmacol Bull 1994;30:53–59.
47. Botte J, Evrard JL, Gilles C, et al. Controlled comparison of 31. Montgomery SA, Reimitz PE, Zivkov M. Mirtazapine versus RO-11-1163 (moclobemide) and placebo in the treatment of amitriptyline in the long-term treatment of depression: a double- depression. Acta Psychiatry Belg 1992;92:355–369.
blind placebo-controlled study. Int Clin Psychopharmacol 1998; 48. Versiani M, Amrein R, Stabl M. Moclobemide and imipramine in chronic depression (dysthymia): an international double-blind, 32. Feiger AD, Bielski RJ, Bremner J, et al. Double-blind placebo- placebo-controlled trial. International Collaborative Study substitution study of nefazodone in the prevention of relapse Group. Int Clin Psychopharmacol 1997;12:183–193.
during continuation treatment of outpatients with major depres- 49. Boyer P, Lecrubier Y. Atypical antipsychotic drugs in dysthymia: sion. Int Clin Psychopharmacol 1999;14:19–28.
placebo controlled studies of amisulpride versus imipramine, ver- 33. Prien RF, Kupfer DJ, Mansky PA, et al. Drug therapy in the sus amineptine. Eur Psychiatry 1996;11(suppl 3):135S–140S.
prevention of recurrences in unipolar and bipolar affective disor- 50. Smeraldi E. Amisulpride versus fluoxetine in patients with dys- ders. Arch Gen Psychiatry 1984;41:1096–1104.
thymia or major depression in partial remission: a double-blind, 34. Kupfer DJ, Frank E, Perel JM, et al. Five-year outcome for main- comparative study. J Affect Disord 1998;48:47–56.
Neuropsychopharmacology: The Fifth Generation of Progress. Edited by Kenneth L. Davis, Dennis Charney, Joseph T. Coyle, and Charles Nemeroff. American College of Neuropsychopharmacology ᭧ 2002.

Source: http://www.themostfromlife.co.uk/assets/downloads/KellerRecurringDepression.pdf