|
Clinical research assistants with extensive experience
and established interrater reliability administered the Structured
Clinical Interview for DSM-IV Axis I Disorders (SCID-I) (7) and
the HRSD-17 to all subjects prior to treatment (Time 1). The treating
physician completed the TSES.
ADs were selected at the discretion of the treating psychiatrist
from among various medications used in the Depression Clinic; however,
during this study only 5 medications were prescribed at a rate that
provided suitable numbers for data analysis. During the study, clinicians
provided standard clinical management every 2 weeks, while repeat
measures of HRSD-17 and TSES (completed by the same research staff,
who were blind to which of the 5 medications had been prescribed)
were collected after 8 weeks of treatment (Time 2). Remission was
defined as a final HRSD-17 score < 7 after 8 weeks of treatment.
Statistical Analyses
We calculated mean differences for continuous variables across
AD medication groups using a 1-way analysis of variance (ANOVA).
We used chi-square analyses to examine differences in the incidence
rates of side effects between pairs of AD medications (for example,
venlafaxine vs moclobemide) for the noncontinuous variables. We
used point biserial correlations to examine the relation between
response to treatment and side effects, and we used a repeated-measures
ANOVA to examine treatment response from Time 1 to Time 2.
Results
Sociodemographic and Clinical Variables
There were no significant differences
among AD medication conditions with respect to age, socioeconomic
status (8), duration of current episode, age at first onset, education,
or baseline HRSD-17 scores. During the time interval for this study
(from January 1, 1995, to March 31, 2000), 217 patients (86 men
and 131 women) completed baseline assessments. The sample was predominantly
white. Of the sample, 193 (73 men and 120 women) completed the protocol
after at least 8 weeks of treatment; 24 patients dropped out of
treatment before the week 8 assessment. There were no significant
differences between completers and noncompleters. There were no
significant differences in dropout rates among the 5 AD medications
(bupropion 12%, venlafaxine 13%, moclobemide 16%, paroxetine 23%,
and sertraline 24%), nor were there significant differences in rates
of remission or in reduction in HRSD scores (Table
1). A post hoc power analysis using a medium effect size (that
is, Cohen’s d > 0.50 < 0.79) produced a power of 0.84
for the 193 patients who completed the study. Thus, it is unlikely
that these nonsignificant results were attributable to insufficient
power.
Intensity of Side Effects
When the intensity of potential side effects across the AD medications
was examined after Bonferroni correction, only 1 side effect (delayed
ejaculation) produced a statistically significant difference among
drugs (F = 6.89, P < 0.001). Unplanned, post hoc testing using
the Scheffé test indicated that for this side effect patients
on paroxetine scored significantly higher on the intensity rating
than did patients on other ADs.
Incidence Rates of Side Effects
We then examined mean score differences for the frequency
or absolute incidence rate of side effects. Paired comparisons (with
Bonferroni correction) revealed significant differences between
at least 1 drug pair for 16 of 32 items: 7 CNS side effects, 7 GI
side effects, and 2 sexual functioning (SF) side effects (see Tables
2, 3, and 4
respectively).
CNS Side Effects
No CNS side effects were reported significantly more frequently
with venlafaxine or paroxetine than with any other drug pairing.
Sertraline produced a significantly higher incidence of tremor and
sweating, compared with moclobemide, and a greater incidence of
tremor, compared with paroxetine. Moclobemide yielded a significantly
higher incidence of nervousness and edema, compared with venlafaxine
and paroxetine, and a significantly higher rate of agitation, compared
with paroxetine. Bupropion was associated with significantly higher
rates of nervousness, agitation, and postural hypotension, compared
with paroxetine. It was also associated with a higher incidence
of nervousness and edema, compared with paroxetine and venlafaxine.
GI Side Effects
Dry mouth was reported by 35% or more patients across
all 5 groups but was not significantly different across drugs. The
incidence of nausea was significantly greater with venlafaxine,
compared with moclobemide and bupropion, while the incidence of
constipation was significantly greater with paroxetine, compared
with moclobemide.
|
|
Bupropion treatment was associated with a significantly
increased frequency of reported weight loss, compared with paroxetine.
No GI side effects were more frequent with moclobemide.
SF Side Effects
Delayed ejaculation in men was reported significantly more frequently
with paroxetine, compared with bupropion or sertraline, and more
frequently with venlafaxine, compared with sertraline. Increased
libido in both men and women was reported significantly more frequently
with bupropion, compared with paroxetine or sertraline. No more
sexual side effects were reported by patients on moclobemide, compared
with any of the other ADs.
Discussion
Equivalent effectiveness and remission rates among antidepressants
in a natural practice setting were associated with heterogeneity
of side effects across drug classes and, in the case of sertraline
and paroxetine, within the same class. Contrary to what was proposed,
no single AD produced a significantly greater side-effect burden,
yet among the 32 side effects examined, 16 showed significant differences
in incidence across a total of 10 drug pairings. This validates
clinician AD selection based on individual tolerability, while assuming
uniform effectiveness.
These results are surprisingly consistent with those
found in RCTs comparing 2 or more ADs. In a metaanalysis designed
to compare adverse effects associated with SSRIs and tricyclic antidepressants
(TCAs), nausea, anorexia, diarrhea, insomnia, nervousness, anxiety,
and agitation occurred significantly more often with SSRIs, compared
with TCAs (4). However, these authors did not include sexual dysfunction
in their analysis. Several other investigations of sexual dysfunction
produced findings similar to ours. Montejo-González and others
reported significantly more anorgasmia associated with paroxetine
than with sertraline (48% and 37%, respectively), as well as a higher
rate of delayed ejaculation (9). Modell and others reported significantly
lower rates of sexual side effects with bupropion, compared with
SSRIs (10). Kennedy and others reported lower levels of sexual dysfunction
in women during treatment with moclobemide and venlafaxine, compared
with paroxetine and sertraline (11). Our results were also consistent
with those found for agitation, nausea, and diarrhea in a cohort
study carried out in a primary care setting (12).
Metaanalyses of PDR and CPS data were performed respectively by
Dewan and Anand (1) and Vida and Looper (13). Using categories similar
to those in our study, Dewan and Anand (1) assigned “penalty”
scores for CNS, GI, and sexual side effects across SSRIs and novel
antidepressants, based on calculated drug placebo differences. Bupropion,
citalopram, nefazodone, and mirtazapine received the lowest overall
scores, while fluvoxamine, paroxetine, sertraline, venlafaxine,
and fluoxetine were all ranked higher.
What this study demonstrates is an important degree of variability
in side-effect reporting across studies and a need for clinical
studies to account for the effect of past therapeutic interventions
on patient response. Most patients in naturalistic settings are
not pharmacologically naive and do not remain on the same antidepressant
dosage for the duration of treatment, which may result in cross-tolerance
or change side-effect reporting. Despite a need for consistency
in side-effect definitions, it may also be argued that using a tool
such as the TSES introduces suggestibility and increases side-effect
reporting. Trindade and others concluded that their results did
not depend on the method of eliciting adverse effects (4). A search
of Medline, however, revealed that most such studies to date do
not use specific side-effect questioning, and incidence rates were
higher in our study than in these, despite relatively low mean dosing
across ADs. In addition, including a greater number of side effects
decreases the likelihood that differences will achieve statistical
significance, although between-drug differences may be extremely
clinically significant. The additional dimension of intensity, calculated
as the product of frequency and severity, did not add sensitivity
to our measures and only distinguished one drug from the others
on the basis of sexual dysfunction.
In summary, we have developed and evaluated a clinical
interview side-effect questionnaire (the TSES) that assesses both
the frequency and the severity of common AD side effects. The absence
of a placebo treatment group limits any conclusions about the prevalence
of these effects, compared with placebo. However, using a natural
practice population to compare side effects across frequently prescribed
ADs provides important complementary data to iniformation derived
from RCTs. Until newer agents with significantly superior rates
of remission, compared with existing agents, are available, selecting
ADs based on side-effect profiles appears to be justified.
Acknowledgements
The authors thank Susan Dickens, MA, Jason Bacchiochi,
MSc, and Ms Natasha Owen.
|
