|
| |||||||||||
|
Lilly Research Laboratories, Indianapolis
Massachusetts General Hospital, Boston
University of Pennsylvania, Philadelphia
Princeton Biomedical Research, Princeton
Seattle Clinical Research Center, Seattle
Lilly Research Laboratories, Indianapolis
TOP ABSTRACT INTRODUCTION METHOD RESULTS DISCUSSION Clinical Implications and... APPENDIX REFERENCES |
---|
Correspondence: Dr David Michelson, Lilly Research Laboratories, Lilly Corporate Center, Drop Code 2423, Indianapolis, IN 46285, USA. Tel: (317) 277-6443; Fax: (317) 277-3262; e-mail: dmichelson@lilly.com
ABSTRACT |
---|
TOP ABSTRACT INTRODUCTION METHOD RESULTS DISCUSSION Clinical Implications and... APPENDIX REFERENCES |
---|
Aims Systematically to assess symptoms and effects on daily functioning related to interruption of SSRI therapy.
Method Patients treated with fluoxetine, setraline or paroxetine underwent identical five-day periods of treatment interruption and continued active treatment under double-blind, order-randomised conditions, with regular assessment of new symptoms.
Results Placebo substitution for paroxetine was associated with increases in the number and severity of adverse events following the second missed dose, and increases in functional impairment at five days. Placebo substitution for sertraline resulted in less pronounced changes, while interruption of fluoxetine was not associated with any significant increase in symptomatology.
Conclusions Abrupt interruption of SSRI treatment can result in a syndrome characterised by specific physical and psychological symptoms. Incidence, timing and severity of symptoms vary among SSRIs in a fashion that appears to be related to plasma elimination characteristics.
INTRODUCTION |
---|
TOP ABSTRACT INTRODUCTION METHOD RESULTS DISCUSSION Clinical Implications and... APPENDIX REFERENCES |
---|
METHOD |
---|
TOP ABSTRACT INTRODUCTION METHOD RESULTS DISCUSSION Clinical Implications and... APPENDIX REFERENCES |
---|
Study design
Following an initial assessment, the study
consisted of two five-day periods separated by at least two weeks but
not more than four weeks. Under double-blind, order-randomised
conditions, all subjects underwent placebo substitution during one
five-day period and continued treatment with their usual selective
serotonin reuptake inhibitor (SSRI) during the other five-day
period. Subjects continued treatment with the SSRI at all other
times.
Assessments
Patients completed a 17-item adverse event
scale (see Appendix) daily for five days following study entry and
during the two blinded periods. Items queried were based on reports
from previous studies (Rosenbaum et
al, 1998). Each item was rated from 0 to 3 (absent, mild,
moderate or severe) and scores were reported as the change from the
most symptomatic of the five days immediately following study entry.
At baseline and at the end of each five-day period, the HRSD-21, the
State Anxiety Inventory (SAI; Spielberger,
1983) and a self-rated assessment of social and occupational
functioning during the previous four days (see Appendix) were
administered. Scores for these assessments are reported for
each period as the change from baseline (visit 1). Spontaneous
reports of adverse events were also collected at all visits.
Supine and standing heart rate and blood pressure were measured
at each visit, using an automated monitoring device (Welch-Allyn,
Skaneateles Falls, NY).
Laboratory assessments
Screening
chemistries, urinalysis and complete blood counts were obtained at
the initial visit. In order to determine steady-state and
post-interruption plasma drug concentrations, samples were obtained
at 18.00 h on the fifth day of each blinded period. Drug assays were
performed by commercial laboratories (fluoxetine: Oneida, Whitesboro,
NY; setraline and paroxetine: MedTox Laboratories, St Paul, MN).
Statistical analysis
We conducted analyses comparing the
blinded periods (placebo interruption v. continued active
medication) within each medication group. These analyses were based
on a crossover analysis of variance with sequence, patient within
sequence, period (one or two) and interruption (present or absent)
factors in the model. If the crossover effect was statistically
significant at a level of 0.10, the interruption effect was based on
the analysis of first period results only. Confidence limits
were constructed using the least-squares means and associated
standard errors from the analysis of variance. Analysis of total
adverse events was based on logarithmic transformation of the
data because of non-constant variance (heteroscedasticity).
Confidence limits for total adverse events were constructed using
the non-parametric method of Hodges & Lehmann (1963).
Within-group comparisons of binary measures were performed using
Prescott's (1981)
test.
RESULTS |
---|
TOP ABSTRACT INTRODUCTION METHOD RESULTS DISCUSSION Clinical Implications and... APPENDIX REFERENCES |
---|
|
Symptom measures
Placebo substitution, but not continued
active medication, was associated with statistically significant
increases in total numbers of solicited adverse events for patients
treated with paroxetine but not those treated with sertraline or
fluoxetine, by the end of the fourth day (Table 2).
Increases in symptoms for patients treated with paroxetine became
statistically significant as early as the time of the second dose of
placebo (Fig.
1). Mean severity worsened by the end of the fourth day of
placebo substitution for 13 of the 17 items on the solicited
adverse events scale among patients treated with paroxetine, for
three out of 17 among patients treated with sertraline, and for
no items among patients treated with fluoxetine. Among patients
taking paroxetine, mean severity of most items increased by
between 0.5 and 1 on the four-point scale. For both
paroxetine-treated and sertraline-treated patients, dizziness was the
item with the greatest number of patients reporting an increase in
severity (percentage of paroxetine patients worsening: active
treatment 5.7%, placebo 57.1%, P<0.001; percentage of
setraline patients worsening: active treatment 6.1%, placebo 42.4%,
P=0.002). Patients taking paroxetine also experienced
statistically significantly worsened severity in nausea, unusual
dreams, tiredness or fatigue, irritability, unstable or rapidly
changing mood, difficulty concentrating, muscle aches, feeling tense,
chills, trouble sleeping, agitation and diarrhoea during placebo
substitution reactive to active treatment. Patients treated with
sertraline experienced statistically significantly worsened severity
in dizziness, nausea and unusual dreams during placebo
substitution relative to active treatment. Spontaneously reported
adverse events followed a pattern similar to that of solicited
events, with increases for patients treated with paroxetine in
dizziness (placebo substitution 33.3%, active treatment 0.0%;
P<0.001), headache (placebo substitution 27.8%, active
treatment 5.5%; P=0.008), nausea (placebo substitution 16.7%,
active treatment 0.0%; P=0.031) and anxiety (placebo
substitution 16.7%, active treatment 2.8%; P=0.025).
|
|
Among patients treated with sertraline there was an increase in the number spontaneously reporting dizziness during placebo interruption (placebo substitution 35.3%, active treatment 5.9%; P=0.007). Among patients treated with fluoxetine there was no statistically significant increase in spontaneous reports of any symptom during placebo substitution. At the end of the placebo substitution period, patients taking paroxetine, but not those taking fluoxetine or setraline, demonstrated statistically significant increases in HRSD-21 and SAI scores compared with the continued drug period (Table 2). There was no significant relationship between either dose or time on drug and new symptoms.
Functional impairment
Patients treated with paroxetine
reported statistically significant deterioration in functioning at
work, relationships, social activities and overall functioning, while
patients treated with sertraline reported deterioration in overall
functioning, and patients treated with fluoxetine reported no change
in any area of functioning following placebo substitution (Table 2).
Vital signs
Patients treated with paroxetine experienced
a statistically significant increase in standing heart rate and
orthostatic change in heart rate (beats per minute) during placebo
substitution relative to active medication (mean standing heart rate:
active medication 78.7 (s.d.=12.2), placebo substitution 82.3
(s.d.=15.9), P=0.37; mean orthostatic change in heart rate:
active medication 8.5 (s.d.=8.8), placebo substitution 12.5
(s.d.=13.1, P=0.020). There were no statistically significant
changes in either measure among patients treated with sertraline or
fluoxetine, and supine and standing blood pressure were similar for
all groups during both conditions.
Plasma concentrations
Mean plasma drug concentrations
(ng/ml) during active treatment and following placebo substitution,
respectively, were as follows. Fluoxetine/norfluoxetine: active 264.6
(s.d.=160.3), placebo substitution 197.7 (s.d.=132.5), mean
percentage reduction 29.7% (s.d.=15.8%);
sertraline/desmethylsertraline: active 87.7 (s.d.=63.0), placebo
substitution 26.0 (s.d.=33.0), mean percentage reduction 73.5%
(s.d.=11.7%); paroxetine: active 46.7 (s.d.=33.4), placebo
substitution 6.9 (s.d.=11.8), mean percentage reduction 86.7%
(s.d.=12.9%). Percentage reduction in plasma concentrations across
drug groups was statistically significantly correlated with new
adverse events (r=0.56, P<0.01); however, within
individual drug groups, correlations between new events and
percentage reduction in concentration were not significant
(fluoxetine r=0.0, P=0.98; sertraline r=0.19,
P=0.30; paroxetine r=0.27, P=0.13). Neither
absolute drug concentration in the steady state nor absolute change
in concentration after interruption correlated with emergence of new
symptoms following treatment interruption for any group.
DISCUSSION |
---|
TOP ABSTRACT INTRODUCTION METHOD RESULTS DISCUSSION Clinical Implications and... APPENDIX REFERENCES |
---|
These data are consistent with reports from controlled clinical studies (Rosenbaum et al, 1998) and epidemiological data (Price et al, 1996) suggesting that abrupt interruption of SSRI treatment is associated with the emergence of physical and psychological symptoms in a manner that suggests a relationship to drug plasma half-life. Our study differs from previous interruption studies in assessing the time course of symptom onset, symptom severity and the association of symptoms with functional impairment as well as changes in plasma drug concentrations.
Limitations
Several factors limit the interpretation of
these data. Although the groups were well matched for baseline
characteristics, patients were not randomised to treatment groups,
and selection bias or expectations about the treatment groups could
have influenced the results. This is unlikely to have been a
significant factor, however, since the comparison periods were
double-blind and order-randomised, and any effects due to patient or
clinician expectations should have been observed during both
periods. Also, a study in which patients were prospectively
randomised to different treatments demonstrated results consistent
with those observed here (Fava et al,
1998).
Another limitation relates to the comparability of doses of the individual agents and to how these doses affected outcomes. The manufacturers' recommended doses were empirically derived from efficacy studies conducted separately for each drug, and could theoretically have differing biological activity at relevant loci. However, the drugs have similar preclinical serotonin (5-HT) profiles, the mean dose for each drug was modestly above the initial recommended starting dose and within its reported effective antidepressant range, and the doses reflect usual clinical practice, suggesting that they are comparable. Furthermore, the paroxetine mean dose was somewhat closer to the initial recommended starting dose than either the fluoxetine or sertraline dose, and any bias would be expected to be in the direction of symptom reduction for patients treated with paroxetine relative to the other treatment groups. Hence, it seems unlikely that the observed results are an artefact of dosing differences among the drugs.
Finally, crossover designs can be vulnerable to carryover effects. However, tests for carryover suggest that it was not a problem for most variables analysed. The few variables that reached statistical significance (P<0.1) were re-analysed using only patients from the first period, and the results were unchanged.
Perhaps the most important limitation of this study is the restriction of treatment interruptions to five days. It is possible that longer periods could be associated with the onset of symptoms among patients treated with fluoxetine or worsened symptoms among any of the treatment groups. We note, however, that interruptions of up to eight days demonstrate patterns of effects consistent with those reported here (Rosenbaum et al, 1998), and it seems unlikely that clinically relevant interruptions (as opposed to discontinuation) would last significantly longer than this. With respect to discontinuation of medication, longer periods may be more relevant to fluoxetine. In one trial (Zajecka et al, 1998) patients with major depression successfully treated with fluoxetine for 12-14 weeks and then abruptly switched to placebo reported a modest but statistically significant increase in dizziness (which was reported by approximately 10% of discontinuing patients over the following six weeks compared with approximately 4% of those continued on fluoxetine), without evidence of other signs and symptoms. These figures probably represent low estimates of the incidence of this symptom, since assessments were obtained by spontaneous rather than solicited reports.
Clinical relevance
A previous report (Rosenbaum et
al, 1998) suggests that SSRI-related discontinuation
syndromes, although uncomfortable, are self-limited and generally
resolve within 1-2 weeks; our results are consistent with these
findings. In this context, the most important clinical risks seem
more likely to be related to appropriate recognition and management
than to the morbidity of the symptoms as such. Discontinuation
symptoms can include prominent psychological manifestations, and
patients who experience discontinuation symptoms after stopping
medication could be misdiagnosed as having relapsed, and as a result
have therapy reinstituted prematurely. Similarly, a body of data
suggests that many patients have gaps in medication compliance or
stop medication spontaneously, and our results suggest that some
patients, particularly those taking paroxetine, will develop
interruption-related symptoms that may be viewed as breakthrough
depressive symptoms or some other condition (e.g. influenza). The
degree to which such problems actually result in inappropriate or
unnecessary treatment, however, is not known.
Also of clinical interest is time of exposure. We did not find an increased risk related to longer exposures. This finding is consistent with the hypothesis that a minimum period is required to establish new physiological conditions related to drug administration, but that drug-related changes are stable once in place.
Finally, an important clinical question is the degree to which new symptoms following treatment interruption represent a specific, drug-related phenomenon rather than depressive relapse. The characteristic presence and predominance following interruption of specific physical symptoms including dizziness and nausea are not typical of depression, and suggest instead an acute disruption of a drug-induced homoeostasis. Our findings are also consistent with effects observed with other drugs, such as rebound hypertension following discontinuation of antihypertensives, again suggesting a specific interruption effect. None the less, because the diagnoses of both discontinuation syndromes and depression are based on descriptive findings rather than markers of underlying pathophysiological processes, we cannot definitively rule out the possibility that some or all of the observed increases in symptoms are related to relapse of underlying illness.
Potential mechanisms underlying symptom production
The
mechanisms which underlie discontinuation phenomena are incompletely
understood, but symptom production appears to be most closely related
to the rate at which internal disruptions occur. Although fluoxetine,
sertraline and paroxetine have some differences in their in
vitro receptor profiles (Richelson, 1996, 1998), the
most apparent difference is in their pharmacokinetic half-lives, and
the resulting rate of clearance of parent drug and active metabolites
from relevant pharmacodynamic targets. We observed a pattern of
symptom emergence and increased severity which parallels the plasma
half-lives of the drugs, strongly suggesting that half-life is indeed
the most important factor. This finding is consistent with data from
other drug classes, such as antihypertensives, implicating shorter
plasma half-life in producing these phenomena (Rickels et
al, 1988; Schweizer et
al, 1990; Noyes et
al, 1991).
The nature of the symptoms observed could potentially be related to a primary effect on serotonin production, release or receptors, to secondary effects on systems modulated by serotonergic pathways, or some combination of these. Although secondary effects may be important, the pattern of individual symptom frequency observed here supports a primary aetiological role for alterations in serotonin homoeostasis. Consistent with previous studies, dizziness was the most common symptom for both paroxetine- and sertraline-treated patients (reported spontaneously in approximately a third of both groups). Dizziness is commonly observed in the context of 5-HT1A receptor stimulation, and its high incidence during placebo substitution is consistent with a primary effect on serotonergic neurotransmission (Grof et al, 1993). Another common symptom was nausea, thought to be mediated by the 5-HT3 receptor, and serotonin is believed to have important roles in modulating psychological symptoms observed in this study, such as nervousness and agitation (Kilpatrick et al, 1990; Richelson, 1998). It is likely, however, that other factors also influence symptoms. The changes in heart rate observed during treatment interruption in the paroxetine group, for example, could represent alterations in noradrenergic-sympathetic nervous system function.
Relationships to plasma drug concentrations
Neither doses
nor absolute plasma drug levels correlated with symptoms associated
with treatment interruption for any group. Plasma concentrations
achieved at a given dose of an SSRI vary widely between individuals
and do not correlate with efficacy (Nielsen et
al, 1991; Amsterdam et
al, 1997), and plasma concentrations may not accurately
reflect brain exposure. In this regard, a recent report on brain
paroxetine and fluoxetine concentrations measured by magnetic
resonance spectroscopy before and after treatment interruption
suggests a relationship between higher steady-state brain
concentrations of paroxetine and new symptoms experienced when
treatment is interrupted (details available from the first author
upon request). However, it is also likely that the lack of
correlation between symptoms and plasma concentrations reflects
individual differences in concentration-effect relationships at the
receptor level, and that the lack of a dose-symptom relationship
parallels the lack of a therapeutic dose-response relationship.
By contrast, there was a statistically significant relationship across all drug groups between percentage reduction in plasma concentration and the appearance of new symptoms. Within each individual drug group this relationship was not statistically significant (although correlations appeared to increase from fluoxetine to sertraline to paroxetine in the predicted direction). This is probably because in the group with the most symptoms, namely those treated with paroxetine, virtually all drug had been eliminated in most patients at the time of measurement, because differences in half-life across drugs are much greater and more important than those among individuals taking any single drug, and because inter-individual differences in plasma concentration reflect not only half-life but also absorption, protein binding and distribution. These findings, while not providing definitive proof of a role for half-life in the development of new symptoms after treatment interruption, are consistent with the hypothesis that it is the major risk factor.
Clinical Implications and Limitations |
---|
TOP ABSTRACT INTRODUCTION METHOD RESULTS DISCUSSION Clinical Implications and... APPENDIX REFERENCES |
---|
LIMITATIONS
APPENDIX |
---|
TOP ABSTRACT INTRODUCTION METHOD RESULTS DISCUSSION Clinical Implications and... APPENDIX REFERENCES |
---|
Self-assessment of occupational and social functioning
REFERENCES |
---|
TOP ABSTRACT INTRODUCTION METHOD RESULTS DISCUSSION Clinical Implications and... APPENDIX REFERENCES |
---|
Demyttenaere, K. (1997) Compliance during treatment with antidepressants. Journal of Affective Disorders, 43, 27-39.[CrossRef][Medline]
Fava, M., Rosenbaum, J., Hoog, S. L., et al (1998) A comparison of symptoms following treatment interruption: evidence from a randomized, double-blind trial with fluoxetine, sertraline, and paroxetine. European Psychiatry, 13 (suppl. 4), 204.
Grof, P., Joffe, R., Kennedy, S., et al (1993) An open study of oral flesinoxan, a 5-HT1A receptor agonist, in treatment-resistant depression. International Clinical Psychopharmacology, 8, 167-172.[Medline]
Hamilton, M. A. (1967) Development of a rating scale for primary depressive illness. British Journal of Social and Clinical Psychology, 6, 278-296.
Hodges, J. L. & Lehmann, E. I. (1963) Estimates of location based on rank tests. The Annals of Mathematical Statistics, 34, 598-611.
Kilpatrick, G. J., Bunce, K. T. & Tyers, M. B. (1990) 5-HT3 receptors. Medicinal Research Reviews, 10, 441-475.[Medline]
Nielsen, O. A., Morsing, I., Petersen, J. S., et al (1991) Paroxetine and imipramine treatment of depressive patients in a controlled multicentre study with plasma amino acid measurements. Acta Psychiatrica Scandinavica, 84, 233-241.[Medline]
Noyes, R. Jr, Garvey, M. J., Cook, B., et al (1991) Controlled discontinuation of benzodiazepine treatment for patients with panic disorder. American Journal of Psychiatry, 148, 517-523.[Abstract]
Prescott, R. J. (1981) The comparison of success rates in cross-over trials in the presence of an order effect. Applied Statistics, 30, 9-15.
Price, J. S., Waller, P. C., Wood, S. M., et al (1996) A comparison of the post-marketing safety of four selective serotonin re-uptake inhibitors including the investigation of symptoms occurring on withdrawal. British Journal of Clinical Pharmacology, 42, 757-763.[Medline]
Richelson, E. (1996) Synaptic effects of antidepressants. Journal of Clinical Psychopharmacology, 16 (suppl. 2), 1-7.[CrossRef][Medline]
Richelson, E. (1998) Serotonin: and what about its side effects? Depression and Anxiety, 7 (suppl. 1), 18-20.[CrossRef][Medline]
Rickels, K., Fox, I. L., Greenblatt, D. J., et al (1988) Clorazepate and Iorazepam: clinical improvement and rebound anxiety. American Journal of Psychiatry, 145, 312-317.[Abstract]
Rosenbaum, J. F., Fava, M., Hoog, S. L., et al (1998) Selective serotonin reuptake inhibitor discontinuation syndrome: a randomized clinical trial. Biological Psychiatry, 44, 77-87.[CrossRef][Medline]
Schweizer, E., Rickels, K., Case, W. G., et al (1990) Long-term therapeutic use of benzodiazepines. II. Effects of gradual taper. Archives of General Psychiatry, 47, 908-915.[Abstract]
Spielberger, C. D. (1983) Manual for the State Trait Anxiety Inventory. Palo Alto, CA: Consulting Psychologists.
Zajecka, J., Fawcett, J., Amsterdam, J., et al (1998) Safety of abrupt discontinuation of fluoxetine: a randomized, placebo-controlled study. Journal of Clinical Psychopharmacology, 18, 193-197.[CrossRef][Medline]
Received for publication February 26, 1999. Revision received June 25, 1999. Accepted for publication August 17, 1999.
This article has been cited by other articles:
Consequences of Interrupting SSRI Treatment Journal Watch Women's Health, June 1, 2000; 2000(601): 13 - 13. [Full Text] |
|||||
|
W. E. E. MEIJER, M. L. BOUVY, E. R. HEERDINK, J. URQUHART, and H. G. M. LEUFKENS Spontaneous lapses in dosing during chronic treatment with selective serotonin reuptake inhibitors Br. J. Psychiatry, December 1, 2001; 179(6): 519 - 522. [Abstract] [Full Text] [PDF] |
|||||
|
S. J. Kingsbury and G. M. Simpson Psychopharmacology: Considerations in Choosing an Antidepressant Psychiatr. Serv., November 1, 2001; 52(11): 1435 - 1436. [Full Text] [PDF] |
|||||
|
I. M. Anderson and J. G. Edwards Guidelines for choice of selective serotonin reuptake inhibitor in depressive illness Advan. Psychiatr. Treat., May 1, 2001; 7(3): 170 - 180. [Full Text] |
|||||
|
|
HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Psychiatric Bulletin | Advances in Psychiatric Treatment | All RCPsych Journals |