Meta-analysis of effects and side effects of low dosage tricyclic antidepressants in depression: systematic review
BMJ 2002; 325 doi: https://doi.org/10.1136/bmj.325.7371.991 (Published 02 November 2002) Cite this as: BMJ 2002;325:991All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
The systematic review of Furukawa et al (1) suggests
that low doses/ subtherapeutic doses of tricyclic
antidepressants can be efficacious in the treatment of
depression. This result goes against the grain of
clinical guidelines for the continued prescribing of
standard doses of greater than 100mg/day. However,
this result has a parallel with the changing guidelines
for the treatment of psychosis. In early-onset
psychosis, low doses of antipsychotics (Haloperidol 2-
5 mg/day) are clinically effective (2), with the justification
that this stage of illness needs a different
pharmacotherapy strategy.
The use of low dose antidepressants may thus be
seen within an early intervention framework. A
subgroup analysis of early onset depression would be
difficult from past studies, but planned prospective
studies with this remit would inform the validity of this
systematic review.
1 Furukawa TA, McGuire H, Barbui C. Meta-analysis of
effects and side effects of low dosage tricyclic
antidepressants in depression: systematic review. BMJ
2002;325:991-5.
2 Remington G, Kapur S, Zipursky RB.
Pharmacotherapy of first-episode schizophrenia.
BJPsych 1998; 172 (suppl 33): 66-70.
Competing interests:
None declared
Competing interests: No competing interests
Sir,
We agree with Furukawa and colleagues on the enduring importance of
the dosage of tricyclic antidepressants. Unfortunately their analysis is
misleading and poorly describes the available data. We concur that there
is some evidence from small trials that doses of tricyclic
antidepressants between 75-100mg are superior to placebo. The limited
available evidence presented by the authors actually indicates that
standard doses may, on average, be more efficacious than lower doses,
although they may cause more side-effects. This is clinically important,
because it would mean that some patients who can tolerate higher doses
would be sub-optimally treated on doses <100mg.
Furukawa and colleagues presented the responder analysis in the key
comparison between low and standard dose drug as random effects relative
risks and highlighted the fact that there was no statistically significant
difference between doses. In fact, the results at 4 weeks are highly
dependent on the choice of statistical method -. Using the responder data
from this 4 week comparison, we found that the evidence suggests that
standard dose is statistically superior to low dose using practically any
other sort of analysis than the one reported. The full random effects odds
ratio meta-analysis provides an odds ratio of 0.59 (0.35 to 0.95), which
reflects the lack of heterogeneity on the theoretically more efficient
odds ratio scale (the exact fixed effects and DerSimonian & Laird
methods produce very similar results on the odds ratio scale). The
DerSimonian and Laird random effects risk difference analysis suggests
that higher dose tricyclics bring about 12% more responses, and the
plausible range of response difference described by the 95% confidence
interval is from 2% to 21%. This is certainly not equivalence and, indeed,
it means that there is limited evidence that standard dose is more
effective than low dose although there clearly remains substantial
uncertainty.
The authors should have treated this as an equivalence question.
Focusing on the finding that standard drugs were not statistically
significantly superior to low dose drugs is not the relevant question.
Estimation is more appropriate than hypothesis testing, and the authors
should have considered the width of confidence intervals as providing a
plausible range for the estimate of treatment effect. Although the
broader confidence interval of the random effects model may be
conservative in most situations, this is not the case in equivalence
analyses. Even using their own results, the best estimate is that low dose
treated patient were 11% less likely to respond than patients treated with
standard dose although this could be much as 26% less likely or as low as
7% more likely; the estimates, of course, become less plausible as we move
to the edges of the limits. This does not establish convincingly that the
higher dose is not superior to the lower dose, although probably does rule
out the possibility that low dose is superior to standard dose to a
clinically significant degree.
Finally, given the paucity of comparative data and the importance of
the clinical question, we are surprised that the authors did not consider
a meta-regression analysis including all trials of a specific drug v.
placebo adding a factor for dose, to add weight to the head to head
comparisons. This would have made better use of the available data and
avoided the somewhat arbitrary dichotomisation of dose used in their
analysis.
Yours sincerely,
John Geddes, Professor of Epidemiological Psychiatry, University of
Oxford
Nick Freemantle, Professor of Clinical Epidemiology and
Biostatistics, University of Birmingham.
Andrea Cipriani, M.D., Research Fellow, University of Oxford
Competing interests:
None declared
Competing interests: No competing interests
I was pleased to see the review of low dose tricyclic antidepressants
providing support for what many GPs see as normal practice. The great
variation in metabolism and blood levels has always suggested that some
patients will require higher doses than others. But it is not a simple
high versus low dose debate about whether low doses do or don’t work.
There is now reasonable evidence that they will be effective for some
patients. The evidence base is there to guide our shared decision making
with the individual patient. The authors could have gone further in
describing the implications of their findings for clinical decision
making.
While for patients in urgent need of improvement aiming for 125mg or
more in seven days is necessary, in other situations starting with a low
dose can be a logical approach for clinicians and patients to take. It can
be useful for those who are wary of medication – ‘low dose’ can be a
selling point and there is less chance of medication cessation or
unwillingness to make a return appointment following side effects. It is
probably sensible to inform patients of the option to increase to a
‘standard dose’ if there is no improvement. Even for those keen to achieve
a ‘standard dose’, aiming for this at one month rather than one week may
reduce the impact of side effects. If some patients return well after 2
weeks they then be content to continue on maintenance at a lower dose. If
they remain well this is a good outcome. Others may achieve remission at
standard doses but have intolerable side effects and agree to continuation
treatment at a reduced dose. Thoughtful clinicians and patients follow
these and other pragmatic courses to ensure that the dose is indeed
therapeutic.
The important caveat to these remarks is that symptoms must be
reviewed and patients should not be left languishing still depressed on
doses of less than 100mg. It is the responsibility of clinicians to ensure
all means are used to engage patients in a review process which will
ensure that as many as possible receive a therapeutic dose – one that
makes them better.
Competing interests:
None declared
Competing interests: No competing interests
To the Editor –
Furukawa et al should be congratulated for this comprehensive meta-
analysis of tricyclic antidepressants addressing the important question of
dose response.(1) This meta-analysis, along with the author’s editorial
response to subsequent comment on the publication, provides a further
learning opportunity that should not be foregone – that is the concept of
‘equivalence’ in meta-analysis.
A key question practitioners will ask of this meta-analysis is
whether low dose tricyclics are equivalent to standard doses. However, in
the absence of a clearly devoted research question, it remains unclear
whether this review was intended as a superiority or equivalence trial.
This should be clarified for the practitioner, as these concepts are not
immediately intuitive.
The bottomline message that low dose tricyclics “may or may not be as
effective as standard dosage tricyclics” was correctly asserted in the
original paper. However, a misleading statement was made in subsequent
correspondence when the author stated that an earlier Canadian meta-
analysis had shown “SSRIs in standard dosage are as effective as tertiary
amine antidepressants in low dosage” and “that SSRIs in standard dosage
are again simply as effective as tertiary amine antidepressants in
standard dosage.”(2) In contrast to the authors’ earlier statement
suggesting non-superiority between standard and low dose tricyclics in
their meta-analysis, the latter statement claims the Canadian meta-
analysis showed SSRIs are “as effective” as tricyclics. These concepts,
non-superiority and equivalence, are keenly different.(3) Equivalence
cannot be safely defined simply as overlapping confidence intervals, as is
the inference given above regarding the Canadian study, which did not
explicitly set out to prove equivalence. Rather, if proving equivalence is
the intention, then objective criteria should be defined a priori, as for
example, 95% CI within the bounds of 0.90 to 1.10 for an aggregate
relative effect size of 1.00.
Theory aside, how shall the practitioner interpret this meta-
analysis? As the authors have aptly stated, standard dose is not
significantly better than low dose. But, this does not infer low or
standard dose is better or worse than the other. Equally important,
neither does it mean they are the same. The bottom-line is that we remain
uncertain whether standard and low doses are the same or different. The
outer limits of the 95% confidence intervals define just how different
they might be, and the possibility of clinically important differences
cannot be ruled out. Much larger studies are needed to bring these bounds
within the threshold of an acceptable definition of equivalence.(4)
1. Furukawa TA, McGuire H, Barbui C. Meta-analysis of effects and
side effects of low dosage tricyclic antidepressants in depression:
systematic review. BMJ 2002;325:991-1000.
2. Trindade, E, Menon, D. Selective serotonin reuptake inhibitors (SSRIs)
for major depression. Part I: Evaluation of the clinical literature
(Report 3E). Ottawa: Canadian Coordinating Office for Health Technology
Assessment, 1997.
3. Greene WL, Concato J, Feinstein AR. Claims of equivalence in medical
research: are they supported by the evidence? Ann Intern Med 2000 May
2;132(9):715-22.
4. Edmonds ML, Camargo CA Jr, Brenner BE, Rowe BH. Replacement of oral
corticosteroids with inhaled corticosteroids in the treatment of acute
asthma following emergency department discharge: a meta-analysis. Chest
2002;121:1798-805.
Competing interests:
None declared
Competing interests: No competing interests
It should be emphasised that Furukawa et al’s interesting study is in
fact two separate analyses which look at quite different things(1). It is
difficult, therefore, to draw any firm conclusions.
The first meta-analysis of 35 studies suggests that low dose
tricyclics may improve certain symptoms, some of which may be secondary to
depression. This has already been established and as previous
correspondents have pointed out, low dose tricyclics are commonly used to
treat chronic pain, insomnia and anxiety, whether these symptoms occur
secondary to a depressive illness, or as part of self-limiting adjustment
and anxiety disorders or exist independently. The question which remains
unanswered is whether low dose tricyclics can adequately treat an episode
of clinical depression. The authors are careful not to suggest this, but
quite correctly only go so far as to state that their meta-analysis only
shows that low dose tricyclics are more effective than placebo.
The second meta-analysis, which looks at the more interesting
question of whether low dose tricyclics are as effective as recommended
doses in the treatment of depressive illness, is unfortunately much
smaller than the first - six studies comprising of 551 patients. The
authors acknowledge some of the weaknesses in their analysis
(heterogeneity of the studies, questions over the quality of some studies,
over the adequacy of blinding, and the lack of operational criteria and
interview schedules to diagnose depression) and other correspondents have
pointed out, crucially, that the cut-off point for a therapeutic dose is
incorrect, as most authorities recognize that 125mg and not 100mg of a
tricyclic is the minimum recommended dose. The only useful result to come
out of this small analysis is, therefore, that patients do not tolerate
higher doses as well as low doses of tricyclics; again this is an already
established finding and adds nothing new to the debate.
My concern is that this meta-analysis will be misinterpreted and will
have the effect of deterring clinicians from prescribing recommended doses
of tricyclics and to thereby fail to adequately treat depressive illness.
A report appearing this week in Hospital Doctor “Use of low-dose
tricyclics in depression ‘vindicated’” confirms my fear(2).
I conducted a study in one teaching general practice in Merthyr
Tydfil, South Wales and identified 222 patients receiving long-term
antidepressant prescription (mostly tricyclics) for a depressive illness
as diagnosed by their General Practitioner (3,4).
62% reported moderate-severe depressive symptoms on the 13-item Beck
Depression Inventory (77% response rate). Only a half of these were in
receipt of a recommended dose (125mg) of their antidepressant) These
patients consulted their GPs more frequently than those reporting few or
no depressive symptoms and were likely to consult for psychological
complaints, suggesting that GPs were aware of these patients’ continuing
distress.
In 1998, five years later, the same cohort of 171 patients was traced
and there was disappointingly little change; 62% of the group reported
moderate-severe depressive symptoms (response rate 80%)(5). Closer
analysis of the individual patients revealed that 56% of the total cohort
from 1993 reported moderate –severe depressive symptoms both in 1993 and
again at follow–up in 1998. The finding of most concern was however that
in 1998 only 59% of this group was in receipt of 125mg or more of their
tricyclic antidepressant.
This study identified a large group of chronic (moderate-severely)
depressed patients who were receiving low dose tricyclics. If this is
representative of other general practices, then it highlights the need for
a depression review clinic in primary care. These patients need regular
review and appropriate intervention. If low dose tricyclics work then
nothing needs to be done other than the occasional review to consider
whether the depression has remitted and whether the medication can be
discontinued or alternatives substituted.
If, however, patients receiving low dose tricyclics continue to
report moderate or severe symptoms, then a review is required followed by
appropriate intervention, which may include a trial at a recommended dose
of the tricyclic, a change of antidepressant to another class,
consideration of psychological and social interventions, or referral to
secondary care for consideration of a second opinion or further treatments
for resistant depression. The thing not to do is to leave patients to
suffer on low dose tricyclics in the belief that these doses are no more
effective than recommended doses. I am afraid that this meta-analysis may
have the undesirable effect of inducing such therapeutic inertia.
References
1 Furukawa AT, McGuire H & Barbui C. Meta-analysis of effects
and side-effects of low dosage tricyclics antidepressants in
depression:systematic review. BMJ 2002;325:991-5. (2 November)
2 Hitche L & Roberts R. Use of low-dose tricyclics in depression
‘vindicated’. Hospital Doctor 14 Nov 2002 p10.
3 Ali IM Depresion in primary care: a study of long-term
antidepressant users.[MSc Thesis]. Cardiff: University of Wales,1994.
4 Ali IM . Long-term treatment with antidepressants in primary care.
Are sub-therapeutic doses still being used? Psychiatric Bull 1998;22:15-
19.
5 Ali IM The long-term outcome of depressed patients in primary care.
Presented & published in the Proceedings of the Royal College of
Psychiatrist’s Annual Meeting, Birmingham, 30 June 1999.
Competing interests:
I have received a small donation from Lilly for my study and a few small honorariums for talking to professionals on the management of depression. I have attended conferences and symposia, some of which have been sponsored by various pharmaceutical companies.
Competing interests: No competing interests
As I read the above article, I realized that though most of the
information was quite accurate, I disagree on the point that lowdose
tricyclics are NOT better than the usual dose for depression. I have seen
patients who have been put on the usual dose of TCA's and though the
initial response was encouraging, most of them did not show satisfactory
compliance. The side effect seen were quite alarming [postural
hypotension] and distressing [drymouth, constipation, sedation] to the
patient as well.
Though the usual level of TCA used is 50mg, in the PHc's), this is below
the level, experts recommend as "THERAPEUTIC"[Donoghue and Tylee, 1996;
Donoghue et al 1996]. A recent Metaanalysis found that, although low doses
anti depressants were less effective than higher dose, they were more
effective than placebo, They even caused fewer adverse events, such as
noncompliance and sideeffects, than HIGHER DOSES[Bollini et al, 1999]
Finally even if we compare the efficacy of TCA's and SSRI's at their
therapeutic dosage, the two classes of drugs have equal efficacy when
prescribed in therapeutic doses {Song et al 1993; Hotopf et al 1997].
So I feel Low dose TCA's definately are better if we consider it from the
patients point of view and compliance, as depression would any way need
treatment for more 4-6 months after the first episode.
Competing interests:
None declared
Competing interests: No competing interests
Dr Jones questions our motivation for undertaking this meta-analysis.
It was simple. I made a survey of psychiatric practices in Japan and found
that (a) benzodiazepines were often co-administered with antidepressants
and that (b) tricyclic antidepressants were often given in "inadequate
dosage." I naturally condemned these practices. (1)
Then, however, I wondered if I had enough evidence to say this, and I
wasn't sure. So I undertook two meta-analyses, one dealing with
antidepressant-benzodiazepine combined therapy and another on low dosage
tricyclic. The first one found that my reproof might be wrong; the
combined treatment brought about more response than antidepressant alone
up to 4 weeks. The question is now how to judge the balance between this
benefit against the known harms of benzodiazepines. (2)
And my second meta-analysis again showed that my reprimand (along with
similar claims by many others) were ill-founded. And you know what we
found.
However, we must admit that some of the concerns expressed in the
rapid responses are quite understandable. For lack of space, we could not
report all sensitivity analyses examining possible confounders. Firstly,
we limited our analyses to those employing opertionalized criteria; the
results were essentially identical. Secondly, we conducted several
subgroup analyses (planned a priori) on older people, primary care
patients and psychiatric patients. Although some of the results were
underpowered and allowed wide confidence intervals, they were all
compatible with the overall results. The details of these sensitivity and
subgroup analyses will soon be available in the Cochrane Library.
So, we should not turn the argument upside down. If a meta-analysis
does not support a certain practice recommended by a guideline, is it this
part of the guideline that may be wrong or is it the meta-analysis?
There is much corroborative (and sometimes circumstantial) evidence
that low dosage antideressants do work and may be better than the
recommended range.
(a) A Canadian meta-analysis of 15 RCTs involving over 1,000 patients with
major depression has shown that selective serotonin-reuptake inhibitors
(SSRIs) in standard dosage is as effective as tertiary amine
antidepressants in low dosage (100 mg or less) and a meta-analysis of 37
RCTs involving over 3,000 patients has shown that SSRIs in standard dosage
is again simply as effective as tertiary amine antidepressants in standard
dosage (above 100 mg); in both instances the 95% confidence intervals of
the pooled odds ratio were quite narrow. (3)
(b) Another very well-conducted meta-analysis became recently available to
shed light on this issue from a different perspective (4). It pooled all
RCTs comparing two or more doses of the same antidepressant and concluded
that doses below 100 mg/day of imipramine equivalents are marginally less
effective than the therapeutic range (100-200 mg/day) but produces
significantly fewer adverse events. Closer inspection of the included
studies reveals, however, that this meta-analysis is based largely on
trials involving SSRIs and other newer antidepressants. Thus, although
their conclusion is phrased in terms of imipramine equivalent, the results
should be interpreted as pertaining mostly to SSRIs and newer
antidepressants.
(c) Something similar is happening to antipsychotics. It now seems that we
used to prescribe an overdose of classical antipsychotics and that the
seeming advantage of atypical antipcyhotics in terms of side effects is
only because these were compared with an overdose of classical
antipsychotics. (5)
(1) Furukawa TA, Kitamura T, Takahashi K. Treatment received by
depressed patients in Japan and its determinants: naturalistic observation
from a multi-center collaborative follow-up study. J Affect Disord
2000;60:173-9.
(2) Furukawa T, Streiner DL, Young LT. Antidepressant plus benzodiazepine
for major depression. In: The Cochrane Library. Oxford: Update Software,
2002; Issue 3.
(3) Trindade, E, Menon, D. Selective serotonin reuptake inhibitors (SSRIs)
for major depression. Part I: Evaluation of the clinical literature
(Report 3E). Ottawa: Canadian Coordinating Office for Health Technology
Assessment, 1997.
(4) Bollini P, Pampallona S, Tibaldi G, Kupelnick B, Munizza C.
Effectiveness of antidepressants: meta-analysis of dose-effect
relationships in randomised clinical trials. Br J Psychiatry 1999;174:297-
303.
(5) Geddes J, Freemantle N, Harrison P, Bebbington P. Atypical
antipsychotics in the treatment of schizophrenia: systematic overview and
meta-regression analysis. BMJ 2000;321:1371-6.
Competing interests:
The author has received fees for speaking from several pharmaceutical companies, some of which manufacture various types of antidepressants such as paroxetine, fluvoxamine and milnacipran.
Competing interests: No competing interests
This metanalysis must be considered downright naughty. While
masquerading as a contribution to an academic debate about appropriate
antidepressant treatment it actually does little more than endorse the
widely held prejudice in favour of using low dose tricyclics in
depression. Published in a widely read UK general medical journal it will
inevitably encourage a practice that is not encouraged by either the Royal
College of Psychiatrists or American Psychiatric Association.
The subject of optimal tricyclic dose remains controversial but the
value of standard dose tricyclic antidepressant treatment compared to
placebo is abundantly clear. As the authors here concede but perhaps do
not emphasise sufficiently the evidence for low dose tricyclic
antidepressant is of generally poor quality. Many of the trials used in
their analysis took place before standardised diagnostic or outcome
criteria were commonplace.
What is less clear is the motivation for their undertaking.
Fluoxetine is now available in generic preparations for about £7 per month
significantly reducing the financial advantage of older antidepressants.
There is some evidence to support the use of tricyclics in severe
depression, but presumably the authors would not recommend low dose
tricyclics in such patients. The chronic severe nature of depression would
make it unethical to recommend a therapy without a secure evidence base.
Being charitable one can only hope the therapeutic advice was added to add
colour to an otherwise unexceptional meta-analysis.
Competing interests:
I have received a small grant for a research study on olanzapine by Eli Lilly
Competing interests: No competing interests
I seriously doubt that it is possible to rely upon a clinical
impression of a response to a trial of treatment, derived from a history
alone, to avoid the potentially harmful effects of antidepressants? What
of an increased risk of developing organ dysfunctions and failure during
an acute illness or of developing chronic diseases from the prolonged use
of medications especially in elderly patients with co-morbidities?
Objective measurements of oxidative phosphorylation should provide the
means to assess the risk in a meaningful way.
Competing interests:
None declared
Competing interests: No competing interests
Re: Low dose tricyclics: making the best use of the available evidence
We would first like to thank Professor Geddes and his colleagues for
their thorough and meticulous reading of our systematic review. We feel
that such is indeed the best any authors could hope for for their work.
Let us admit from the outset that it came as a big surprise to us
ourselves that fixed effects model odds ratio, relative risk and risk
difference were indeed statistically significant for the per protocol
comparison between low dosage and standard dosage tricyclics at 4 weeks.
We looked around and experimented with our data and found that, for the
same data set, random effects model odds ratio and relative risk were not
significant whereas random effects model risk difference was. We further
found that, at 6-8 weeks, odds ratio or relative risk or risk difference
were not significant by any analytical method; nor were they at 4 or at 6-
8 weeks if we examined the data on the worst case scenario intention-to-
treat analysis.
We agree with Professor Geddes et al that estimation is more
appropriate here than hypothesis testing. The following line of argument
then, we believe, still leads us to the same conclusion as in our original
review that �gthey [low dosage tricyclics] may or may not be as effective
as standard dosage tricyclics but result in fewer dropouts due to side
effects.�h
1) A separate empirical analysis of ours on the best summary effect
measure of meta-analyses showed that random effects model relative risk
was the most generalisable and interpretable among fixed or random effects
model odds ratios, relative risks and risk differences (1). We therefore
planned a priori to stick to this method (Please see our protocol for this
meta-analysis in the Cochrane Library). We felt any further
experimentation with analytical methods would be a fishing expedition.
2) The very fact that the results vary according to meta-analytic
methods weakens our confidence in the statistically significant findings
of some of these analyses. Moreover, these significant findings do not
hold when we incorporate the dropouts and attempt the worst case scenario
intention-to-treat analysis whereby dropouts were considered non-
responders in the standard dose arm but as responders in the low dose arm
(a fair assumption if we are to recommend standard dose which we know
causes more side effects).
Furthermore, the findings at 4 weeks are at odds with those at 6-8 weeks,
which further decreases our confidence in the superiority of standard
dosage regimen.
3) The appropriate estimates for the effectiveness of standard dose
over low dose would therefore be, based on the random effects model
relative risk and the plausible range of the patient expected event rate
(we assumed response rate ranging between 50 and 70% among those receiving
standard dosage), absolute benefit increase of 6 to 8% (95% confidence
intervals: -5% to 18%) at 4 weeks and of -6 to -8% (95% confidence
intervals: -43% to 17%) at 6-8 weeks. At the same time the available data
show that the absolute risk increase of dropouts due to side effects might
well be 9% (95% confidence interval: 4% to 12%).
4) Thus the currently available best evidence does not rule out
clinically important benefit increase on standard dosage over low dosage,
while it clearly points to an approximately equal number of patients
dropping out for side effects.
We therefore remain uncertain whether standard and low doses are the
same or different but it seems clear to us that we should not try to
increase the dosage above 100 or 125 mg for patients with depression
upfront. When patients do not respond to the low dosage regimen, there
seem to be several options: increase the same antidepressant, switch to a
different antidepressant or augment the antidepressant with lithium or
thyroid hormone. It is painfully regrettable to realize that we do not
have evidence base to choose among these alternatives.
We did not essay meta-regression on our current dataset because we
limited placebo-controlled RCTs to those administering 100 mg or less of
tricyclics, and there was indeed small variation in dosage. If we were to
examine the dose response relationship for tricyclics by meta-regression,
we would need to assemble all placebo-controlled trials with tricyclics
irregardless of their dosage. That might be an interesting analysis but
probably very hard to interpret clinically for two reasons. First we admit
that our choice of grouping trials according to dosages was somewhat
arbitrary but we thought, and planned a priori accordingly, that such
dichotomization is clinically sound and easily understandable; on the
other hand, findings from meta-regression would have been translated into
practice with more difficulties. Second, we are afraid that meta-
regression analyses would still be fraught with (and indeed might even
inadvertently mask) the nasty problem of dropouts which would make
impossible any firm conclusion to be drawn.
(1) Furukawa TA, Guyatt GH, Grifitth LE. Can we individualize the
Number Needed to Treat (NNT)? An empirical study of summary effect
measures in meta-analyses. Int J Epidemiol 2002;31:72-6.
Competing interests:
TAF has received fees for speaking from several pharmaceutical companies, some of which manufacture various types of antidepressants such as paroxetine, fluvoxamine and milnacipran.
Competing interests: No competing interests