Association between antidepressant prescribing and suicide in Australia, 1991-2000: trend analysis
BMJ 2003; 326 doi: https://doi.org/10.1136/bmj.326.7397.1008 (Published 10 May 2003) Cite this as: BMJ 2003;326:1008All 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.
Responses to Our Critics
We welcome the debate that our paper has prompted.
The following are our responses to the various
criticisms made of our paper in the Rapid Responses.
1. Misconstruals of our views
A number of critics (De Leo, Draper, Verberne)
attributed to us views that are not expressed in our
paper, namely, that the decline in suicide in Australia
between 1990 and 2001 was wholly due to the
pharmacological effects of increased antidepressant
prescribing, and that suicide rates are not affected by
the availability of methods of suicide, unemployment,
alcohol consumption, social factors etc. It should be
clear from the factors that we considered as alternative
explanations of the trend that we accept neither of these
views. We explicitly argued that: "the association that
we observed between antidepressant prescribing and
suicide may reflect increased recognition, diagnosis
and treatment of depression by general practitioners as
much as any pharmacological effects of antidepressant
medication"(p1012).
The observation that suicide rates declined in older
women before the antidepressant era would only
undermine our paper if we had claimed that all
declines in suicide were attributable to antidepressant
prescribing. We accept that there are multiple causes
of suicide; that integrated efforts are required to reduce
suicide deaths; and that the prescribing of
antidepressants to people who are depressed is not
sufficient although it should be part of an integrated
response. Nor do we claim that reducing suicide is the
only goal of antidepressants prescribing (Allen).
Improved quality of life is an important goal of treatment
in depression.
2. Methods of Data Analysis
Ankarberg and De Leo argue that antidepressants did
not decrease suicide rates between 1990 and 2001
because there was no decrease in young males
among whom antidepressant prescribing increased
over the same period (although from a very low base
rate). It is always possible to select data points that run
counter to a statistical trend; that’s why we use
statistics. The critical finding was the relationship
between exposure to antidepressants across age
groups in both sexes. Antidepressants are clearly not
all that influences in suicides among young males. We
noted in our paper that others have argued that the
increased suicide rate in young males has been driven
by rising unemployment.
De Leo has argued that correlation coefficients
calculated using the base rate as covariate produce a
different result, namely, no correlation between change
in antidepressant prescribing and suicide trend. This
analysis would answer a different question: whether
after taking account of baseline antidepressant use, the
increased rate of prescribing predicted suicide trend.
We do not think that this is an appropriate analysis
because controlling for the base rate of prescribing
removes the effects of antidepressant exposure which
we found is related to change in suicide rate.
De Leo also argues that larger correlation between
antidepressant exposure and decline in suicide than
between change in antidepressant prescribing and
suicide decline signals a spurious result. We think the
more likely explanation is statistical: a score based on
an average of DDD over three time points will be more
reliable than a score based on a difference in DDD over
the same period.
Senn has argued that the relationship we reported was
not robust to method of statistical analysis. He briefly
reports two Poisson regression analyses, only one of
which found a similar relationship to the one that we
reported. We disagree that our results are not robust to
method of analysis. We originally performed a Poisson
regression analysis of suicide rates and calculated a
correlation between the logarithm of the incident rate
ratio (IRR) for the two periods and the two measures of
DDD across the eight age groups for each and sex.
This analysis produced the same patterns of result as
the analysis that we reported in our paper. The
correlations between log IRR and average DDD were
-0.81 (p < 0.05) in females and –0.79 in males (p <
0.05) while the correlations with change in DDDs was
–0.81 in females (p < 0.05) and –0.43 in males (not
statistically significant). We reported the correlations
between the difference in suicide rate and the two DDD
measures because we believed that this analysis
would be more easily understood than one based on
log IRRs.
Moncrieff made a number of criticisms of the DDD
measure. She argued that we should have used RR
rather than absolute changes in DDD. We disagree: the
questions of interest were about the relationship
between suicide and overall exposure to
antidepressants and change in exposure to
antidepressants over time. A ratio of DDDs would
provide an alternative measure of change in exposure
but it would not measure overall antidepressant
exposure. She also argued that DDD is a potentially
misleading measure because it will underestimate
prescribing levels of tricyclic antidepressants in the
elderly. We accept that this may be the case but believe
that, if anything, this would underestimate any
correlation between antidepressant prescribing and
suicide trend by underestimating elderly exposure to
antidepressants.
Alternative Explanations of Our Findings
Fabrication of Data
Bagnall implies that our findings have been fabricated
to serve the interests of the pharmaceutical industry
because several of the authors (AM, PM and IB) have
accepted funding from the industry. No evidence is
offered for this ad hominem attack or for the assertion
that a history of pharmaceutical industry funding is
incompatible with conducting valid scientific research.
Changes in Methods of Suicide
Blisker suggested a more plausible alternative
explanation of our findings, namely, that it : "could
reflect a proportionately greater reliance upon
antidepressant overdose as a suicide method in older
age groups. If the proportion of individuals using
antidepressant overdose as a suicide method
increases with age, and if individuals in these older
age groups are unlikely to substitute other methods,
one would expect that improved safety of
antidepressant medication would have its greatest
impact upon suicide rates in older age groups". De
Leo made the same point, citing in support of this
hypothesis Australian data that there had been a larger
proportionate decline in suicides by drug overdoses
among older than younger adults.
This is a plausible hypothesis that cannot be easily
excluded because detailed data are not available on
changes in method of suicide over time by age and sex.
The Australian Bureau of Statistics (2003) reports that
the proportion of suicides in which poisoning was the
method among males declined from 11.3% in 1991 to
7.8% in 2001. The decline was even larger among
females, from 35.1% to 25.8%. The small proportion of
older male suicide deaths by poisoning using drugs
makes a decrease in self-poisoning an unlikely
explanation of age-related changes in suicides among
men. The fact that the decline in suicide rate was
greater among men than women is also inconsistent
with Blisker and De Leo’s hypothesis because there
was a greater decline in self-poisoning among women
than men. De Leo and colleagues (2001) report that in
1997 all forms of poisoning by solids or liquids
accounted for only 6% of suicides among males aged
over 65 years compared with 33% of suicide deaths in
females. The most common methods of suicide
among males in this age group were: hanging (30%),
carbon monoxide poisoning (27%) and firearms (15%).
Omission of other factors
Sheikal argued that our results may have been different
if we had included data on the use of other treatments
e.g. ECT and psychological interventions. We think this
unlikely for ECT because this is a specialist
intervention that has never been widely used by general
practitioners, it is not often used in specialist settings
and its frequency of use has probably declined over the
study period. We agree that the increased use of
psychological interventions by GPs has probably played
a role. We argued as much in our paper.
Ineffectiveness of Antidepressants in Reducing Suicide
Ankarberg and De Leo argue that antidepressants do
not reduce suicide because meta-analyses of placebo
controlled RCTs have not shown a lower rate of suicide
among patients receiving antidepressants than in
those receiving placebos. We were aware of these
data but believe that they are of limited value in
assessing the impact of antidepressant prescribing on
suicide in the population because most of these
studies involved follows up of less than six months and
patients at risk of suicide are often excluded from
placebo-controlled trials.
Allen and Bagnall claim that SSRIs may increase
suicide risk in the short-term in individual patients. This
is not supported by the meta-analyses of controlled
trials and is not supported by our data. This possibility
is not inconsistent with a correlation between
antidepressant prescribing and suicide deaths
because rare adverse events that affect individuals do
not preclude an aggregate benefit from
antidepressants at the population level.
Ineffectiveness of GP Management of Depression
Allen argues from his personal experience that
antidepressants work but he is sceptical that
psychological treatments are effective for depression.
We agree that patients with severe depression require
more than talk but the evidence is that for mild to
moderately severe depression psychological
interventions can be as effective as medication.
Psychological interventions inevitably accompany
medication as well.
Allen also argues that GPs do not have skills to
dispense antidepressants. We agree that more
serious cases of depression require specialist
treatment but GPs provide an important gateway to
specialist psychiatric care and, when appropriately
trained and supported, they can diagnose and manage
a range of the more common depression of mild to
moderate severity.
Conclusion
We agree with our critics that one has to be cautious in
interpreting time series data. As we acknowledged in
our article. We do not claim to have proved that
increased antidepressant prescribing was responsible
for the decline observed in suicide deaths in Australia
between 1991 and 2000. Any such inference will
require a convergence of evidence from a variety of
different studies. . We hope that our paper will
encourage other investigators to test relationships
between antidepressant prescribing and suicide in
other countries. It would be of particular interest to see
if the pattern of decline in suicide deaths that we
observed in Australia has also occurred in countries
that have not seen an increase in antidepressant
prescribing over the same period.
Australian Bureau of Statistics. Information Paper
Suicides 2001. Catalogue No. 3309.0.55.001.
Australian Bureau of Statistics, 2003.
De Leo D, Hickey PA, Neulinger K and Cantor CH.
Ageing and Suicide. Commonwealth Department of
Health and Aged Care, Canberra, 2001.
Competing interests:
Disclosures re
Competing Interests:
Funding for the project
was provided by
beyondblue: the
National Depression
Initiative; Wayne Hall's
time was funded by the
Strategic Fund of the
ViceChancellor,
University of
Queensland.
Wayne Hall has no
competing interests to
declare.
Andrea Mant was a
consultant on Quality
Use of Medicines to
Merck, Sharp and
Dohme Australia
(1997), has been a
member of advisory
boards for Pfizer and
Sanofi-Synthelabo
(1999-2000) and was
sponsored to attend
Global Health Care
2000 Conference
(Eli-Lilly).
Philip Mitchell has
received research
funding and honoraria
in the last five years
from several
pharmaceutical
companies which
manufacture
antidepressant
medications.
Valerie Rendle has no
competing interests to
declare.
Ian Hickie has received
research funding and
honoraria in the last
five years from several
pharmaceutical
companies for conduct
of General Practice
training programs,
notably SPHERE: A
National Depression
Project.
Pharmaceutical
industry support
(Wyeth) has been
received for
participation in
international meetings
detailing the economic
and social costs of
depression.
Peter McManus has no
competing interests to
declare.
Competing interests: No competing interests
Hall et al's analysis of the beneficial effects on suicide of higher
levels of antidepressant prescribing(1) in Australia may be misleading.
Their suggestion of such profound potential effects of increased levels of
antidepressant prescribing on population suicide rates contrasts with
those of a recent review of antidepressant trials.(2) In Khan's review,
perhaps surprisingly, suicide rates were no different in those treated
with placebo than those receiving antidepressants.
The authors' approach of correlating age- and sex-specific
prescribing, and changes in prescribing, with changes in each age group's
suicide rate is unusual. Indeed in five of the sixteen age- and sex-
groups (younger males and females) Hall studied, increases in prescribing
were actually accompanied by rises in suicide. A more conventional, and
direct, approach would have been to contrast trends in age-specific
prescribing with trends in suicide. Inspection of UK data on
antidepressant prescribing (1976-1998)(3) and Office for National
Statistics data on suicide and undetermined deaths(4) for England and
Wales reveals that a 3-4 fold rise in antidepressant prescribing in 20-29
and 30-39 old males has been accompanied by 50% rises in their suicide
rates (see attached figure). Even in 40-54 year olds, an age group in
which an almost four fold increase in prescribing occurred, suicide rates
have remained relatively stable. It is only amongst those aged 55+ that
any effects are seen - these groups do indeed show an accelerated rate of
decline in the late 1980s when antidepressant prescribing increased
dramatically. In women, declines in suicide in all but 15-19 year olds
have been accompanied by rises in antidepressant prescribing, however, the
rise in prescribing in the late 1980s was not accompanied by any
noticeable change in trends in their rates of suicide.
There are well-recognised problems with interpreting such ecological
data. Analyses such as Hall's and others,(5) where the effects of only
single factors are investigated, may produce misleading results. Many
factors may influence population trends in suicide.(4) It is of note,
however, that in a recent time series analysis taking account of such
factors, we found some evidence that rises in antidepressant prescribing
in the elderly may indeed have had some beneficial effects on their
suicide rates.(4)
David Gunnell1
Nicos Middleton1
Jonathan Sterne1
1Department of Social Medicine, Canynge Hall, Whiteladies Road,
Bristol BS8 2PR
Email: d.j.gunnell@bristol.ac.uk
Fax: 0117 9287325
Phone: 0117 9287253
References
1. Hall WD, Mant A, Mitchell PB, Rendle VA, Hickie IB, McManus P.
Association between antidepressant prescribing and suicide in Australia,
1991-2000: trend analysis. BMJ 2003;326:1008-11
2. Khan A, Khan S, Kolts R, Brown WA Suicide rates in clinical trials
of SSRIs, other antidepressants, and placebo: analysis of FDA reports. Am
J Psychiatry 2003;160:790-2.
3. Middleton N, Gunnell D, Whitley E, Dorling D, Frankel S. Secular
trends in antidepressant prescribing in the UK, 1975-1997. J Pub Health
Med 2001;23:262-267
4. Gunnell D, Middleton N, Whitley E, Frankel S, Dorling D. Why are
suicide rates rising in young men but falling in the elderly? - a time
series analysis of trends in England and Wales 1950-1998. Soc Sci Med (in
press 2003)
5. Isaacson G Suicide prevention - a medical breakthrough? Acta
Psychiatr Scand 2000;102:113-7.
Competing interests:
None declared
Competing interests: No competing interests
Hall et al suggest that reductions in suicide rates in older people
in Australia could be attributed to their increased exposure to
antidepressants, in particular selective serotonin reuptake inhibitors
(SSRIs) [1]. Of course, reduction of suicide rates due to improved
treatment of depression is an outcome we all hope could occur. But the
associations appear to be attribute possible causality inappropriately.
According to the Australian Bureau of Statistics, there has been a
gradual reduction in suicide rates in persons over the age of 40 in
Australia since 1961-65, the peak period of suicide deaths for these age
groups since the Great Depression. This is true for each 5-year age group
from age 40 years over, and for both men and women. Indeed, in most age
groups, the greatest reduction occurred in the period before 1990. For
example in 1961-65, suicide rates in men aged 45-49 years were 34.9 per
100 000 people and in men aged 50-54 years 39.4 per 100 000 people. In
1986-90 these had dropped to 23.5 and 24.7 per 100 000 respectively (or
24.1 for the ten year group 45-54 quoted by Hall et al). Yet by 1996-2000
this had only dropped to 23.8 per 100 000. Similarly if we examine males
aged 55-64 years, from 1961-65 to 1986-90 suicide rates dropped from 37.8
per 100 000 to 25.1 per 100 000, and to 20.7 in 1996-2000. This scale of
pre-1990 suicide rate reductions is also found in females. For example in
1961-65, the suicide rates of females aged 65-69 was 18.2 per 100 000 and
in those aged 70-74, 15.1 per 100 000, which dropped to 8.5 and 8.2 per
100 000 respectively in 1986-90 (or 8.4 for the ten year group 65-74
quoted by Hall et al). By 1996-2000 this had only reduced to 5.8 per 100
000 [2].
Clearly there were more substantial historical trends of suicide
reduction long before the introduction of the SSRIs. The reasons for this
reduction are unclear but may include improved physical health in middle
age and old age, better community services for the disabled to enhance
independence, and greater financial security in later life, all of which
are risk factors for depression in older people. Improved detection and
treatment of depression is unlikely to be as effective in preventing
suicide as the prevention of depression itself.
While it is important to examine these types of associations, a
broader appreciation of the issues involved would provide a more balanced
report.
1. Hall DW, Mant A, Mitchell PB, Rendle VA, Hickie IB, McManus P.
Association between antidepressant prescribing and suicide in Australia,
1991-2000: trend analysis. Br. Med. J. 2003;326:1008-1012.
2. Australian Bureau of Statistics. Suicides 1921-1998. ABS Catalogue No.
3309.0, Canberra, 2000
Competing interests:
None declared
Competing interests: No competing interests
Dear Editor,
As already noted Hall et al’s (1) data on suicide rates and antidepressant
prescribing contradict their own conclusions. The conventional and
intuitive way of analysing trends in levels of antidepressant use by age
group would be to look at ratios of levels of use in 1998-2001 compared
with 1990-1991. Using the data on bmj.com, this shows that levels of use
increased 7.5 times for men aged 25-34 and 2.1 times for men aged 75-84,
for example. Spearman correlations were conducted using these ratios for
all age groups and the difference in suicide rates as defined by the
authors. This revealed a strong positive correlation. Increases in suicide
rates were associated with higher rates of increase of antidepressant use.
For men, Spearman’s Rho was 0.86, p=0.007. For women, it was 0.76, p=0.03.
The authors appear to have reached their figures by analysing
absolute differences in levels of prescribing. The use of Daily Dependent
Dose (DDD) is also problematic and not clearly explained. DDDs represent
units of World Health Organisation defined therapeutic doses. Thus the
increase in antidepressant use in older age groups is probably accounted
for partly by a change from lower dose tricyclic antidepressant
prescribing to prescribing selective serotonin reuptake inhibitors at
standard doses. However, evidence that lower doses of tricyclics are less
efficacious than standard doses is not strong (2). In addition, DDDs
relate to the general adult population. They will therefore underestimate
prescribing levels of tricyclic antidepressants in the elderly where it is
accepted that therapeutic and tolerable doses will be lower.
It remains very difficult to prove that the massive increase in
antidepressant prescribing has had any objective and positive impact on
the health of populations. Other commentators have suggested that there is
no impact on suicide trends, and pointed out that rates of self harm are
rising, not falling (3). Long-term incapacity due to depression in the UK
continued to rise throughout the 1990s (4).
Yours,
Joanna Moncrieff.
References:
1) Hall WD, Mant A, Mitchell PB, Rendle VA, Hickie IB, McManus P.
Association between antidepressant prescribing and suicide in Australia,
1991-2000: trend analysis. BMJ 2003; 326:1008-11
2) Furukawa TA, McGuire H, Barbui C. Meta-analysis of effects and
side effects of low dosage tricyclic antidepressants in depression:
systematic review. BMJ 2002;325:991-995.
3) Van Praag HM Why has the antidepressant era not shown a
significant drop in suicide rates? Crisis 2002; 23:77-82
4) Moncrieff J, Pommerleau J. Trends in sickness benefits in Great
Britain and the contribution of mental disorders. Journal of Public Health
Medicine 2000;22:59-67.
Competing interests:
None declared
Competing interests: No competing interests
One should be cautious about using correlated time series to judge
cause and effect. At the very least it seems prudent to adjust figures for
any possible overall secular trend. There are also very many alternative
models that could have been envisaged by Hall et al1. I tried a number of
different approaches, two of which I report here. For both of these I used
data on numbers of suicides for 1991-1992, 1994-1996 and 1998-2001 for
males and females and by ten-year age group starting at age 15 and
finishing at age 75, as my response variable. (These values were obtained
from the Australian Bureau of Statistics2.) Main effects of age and sex as
categorical variables and their interactions as well as year as a main
effect, also treated as a categorical variable, were fitted as
standardising predictors. Poisson regression (with re-scaling by
residual deviance) and negative binomial regression, with total (log)
years of exposure as an offset in both cases, were used as fitting
approaches and use of antidepressant (defined daily dose/100 people/day)
as given in Table 2 of Hall et al was used as a predictor.
A negative
association with antidepressant use and suicide was observed, as was in
the paper by Hall et al. However, for the first approach this was not
significant, estimate (est) =-0.0010, standard error (SE) =0.0009,
p=0.27, and for the second it was just significant, est=-0.0014,
SE=0.0006, p=0.02. No doubt many other approaches could have been
envisaged. It seems to me, however, that the relationship is not so strong
as to be robust to changes in modelling approaches. As such, it is less
than compelling and, in any case, very many difficulties of interpretation
remain, to some of which the Hall et al referred1.
In short, I think one
should be sceptical about claiming a benefit of increased anti-depressant
use in terms of reduced risk of suicide on the basis of such data. It may
be mere coincidence but is certainly worth studying in other societies.
References
1. Hall DW, Mant A, Mitchell PB, Rendle VA, Hickie IB, McManus P.
Association between antidepressant prescribing and suicide in Australia,
1991-2000: trend analysis. Br. Med. J. 2003;326:1008-1012.
2. Australian Bureau of Statistics. Information Paper: Suicides 2001,
2002.
Competing interests:
The author is a consultant to the pharmaceutical. This note was prepared completely independently of any other party.
Competing interests: No competing interests
As a person with severe Bipolar Mood disorder, who has been, but is
no longer, on several different antidepressants, I feel that I can add
something - personal experience.
1. Antidepressants work. I was transformed from a wreck who was
physically shattered, suicidal and terrified of everything, especially
other people. After taking anti-depressants I began to improve. It is
simply not credible that such an improvement, which I have seen replicated
in many other patients, is caused by a friendly chat with a doctor, or the
placebo effect. I was terrified of the doctor, and had no hope that the
pills would help me; no hope of any kind.
2. Reducing suicide is not the only goal of antidepressants. If
suicides remained constant, but patient's quality of life improved whilst
they were alive then the drugs would be worthwhile. Being depressed is
hell; there is nothing worse. People who have not been suicidally
depressed cannot imagine what it is like. Unfortunately this fact colours
much discussion and study in the medical community.
3. I believe that one reason SSRIs increase suicide risk may be
that, when you are in the depths you wants life to end, but don't have the
energy or will to act on the desire. As one starts to recover, the
physical and mental torpor lift faster that the despair. There is then a
dangerous crossover period when you still have feelings of darkness,
hopelessness and fear, but you are full of energy. This is when there is a
real danger of self destructive behaviour. It is very important that
patients are monitored over this period, but the depressive thoughts may
also breakthrough later. Being suicidal once makes it easier to
contemplate later - the taboo has been nroken and it now seems a viable
option, always there.
4. In my personal experience, GPs do not have the skills necessary
to be dispensing antidepressants. They prescribe them to patients who
present with depression, without finding out if the depression is a
manifestation of something else. In my instance, I was twice pescribed
SSRIs with no other mood stabilisers as GPs failed to diagnose my Bipolar
Disorder. The second time I had a severe reaction. I am sure that order
conditions are missed by busy GPs who see a depressed individual and
automatically prescribe an SSRI from whichever company last gave him or
her a freebie.
Competing interests:
None declared
Competing interests: No competing interests
The conclusions presented in Hall et al.’s (2003) paper on the
association between antidepressants prescribing and suicide are
questionable. The authors claim that their findings support the contention
that there is a clear association and, perhaps, causal relationship
between antidepressant prescribing and suicide, especially in older males
and females. Unfortunately, the authors fail to duly acknowledge that
their data also indicate a lack of impact of antidepressants on suicide in
younger individuals, which, indeed, might have been used to corroborate an
opposite standpoint to that presented in the paper. For instance, tables 1
and 2 show that, despite an increase in antidepressant use, both male and
female subjects aged 15 to 44 reported an increase in suicide rates.
Additionally, the relative increase in antidepressant use in this age
segment of the population was much higher (e.g., OR = 11.9 in 15-24 year
old males) than that observed in the elderly (e.g., OR = 1.97 in 65-74
year old males). These results question the existence of an unequivocal
association between antidepressant prescriptions and suicide rates. Of
course, it could be argued that without antidepressants the rates of
suicide in the younger segment of the population could have been even
higher. However, the research design adopted in the present study cannot
provide a reliable answer to these questions.
It should be also noted that if we calculate the correlation between
absolute changes in antidepressant use and suicide rates in the age
groups, controlling for the differences in initial levels (1986-1990) of
antidepressant use and suicide rates, the strength of association between
the two variables appears to be much lower and in the opposite direction
than that reported by Hall et al. (eg., r = 0.04 for males and r = 0.33
for females; p<05). Furthermore, the fact that the authors found a
higher correlation between age groups and changes in suicide rates than
between estimated changes in antidepressant prescriptions and suicide
rates implies that some age-related factors may be responsible for the
observed association. We believe that these findings should have raised at
least some concerns about the spuriousness of the observed antidepressant-
suicide relationship.
It should be also noted that Hall at al.’s paper presents an overly
biased overview of the current findings on the relationship between
antidepressants and suicidal behaviour. Pharmacoepidemiological
investigations, pooled analyses, and randomised control trials have
yielded conflicting evidence in this respect and it is not clear whether
antidepressants may prevent or even provoke suicidal behaviours (Baldwin,
2000).
Another problem associated with the conclusions reached in Hall et
al.’s study regards the authors’ claim that in the examined timeframe
there was “no evidence of marked changes in method of suicide”. This is
incorrect. There was indeed a substantial decline in suicide by drugs
overdosing. This phenomenon was observed in all age groups, but was more
pronounced in older adults, reaching a 40% decrease in the over 75 and
affecting proportionately more women than men, who are notoriously more
prone to attempting suicide with prescribed medications (ABS, 2001). As
suggested by another commentator of Hall et al.’s paper (D. Blisker), the
possible impact of newer, safer antidepressants on suicide rates in
elderly could be explained by a reduction in access to lethal means of
suicide. Not to mention that suicide rates, especially in older women in
Australia, have demonstrated a nearly constant decline since the pre-
antidepressants era. This happened in most, if not all, Anglo-Saxon
countries, suggesting the (powerful) influence of socio-cultural factors
(De Leo, 2001).
Finally, it is important to acknowledge that suicide is not simply a
function of depression but requires integrated efforts at all levels of
the community, all round. In fact, a recent study (Bertolote et al., in
press) showed that by treating with a 70% efficacy index ALL depressed
people, the obtainable reduction in suicide rates would not be higher than
12%. Consequently, although depression is one of the factors associated
with higher suicide risk, concentrating most efforts only in this
direction would be dangerously reductionistic and unjustifiable (De Leo,
2002).
References:
1. Australian Bureau of Statistics (2001). Suicide, Australia.
Information paper, Canberra, Australian Bureau of Statistics
2. Baldwin, D.S. Pharmacological provocation and prevention of suicidal
behaviour. Intern Rev Psychiatry 12: 54-61.
3. Bertolote, J.M., Fleishman, A., De Leo, D., et al. Mental disorders and
suicide (editorial). Br J Psychiatry (in press).
4. Blisker, D.E. Another plausible alternative explanation. Rapid
responses to: Association between antidepressant prescribing and suicide
in Australia, 1991-2000: trend analysis. BJM 2003, May 10.
5. De Leo, D. (2001). Suicide and Euthanasia in Older Adults. A Trans-
cultural Journey. Hogrefe/Huber, Goettingen.
6. De Leo, D. (2002). Why are we not getting any closer to preventing
suicide? (editorial). Br J Psychiatry 181: 372-374.
Competing interests:
None declared
Competing interests: No competing interests
Hall et al. (1) state: ". . . the association we observed between
antidepressant prescribing and suicide may reflect increased recognition,
diagnosis, and treatment of depression by general practitioners as much as
any pharmacological effects of antidepressant medication." The time scale
they took into account was the decade 1991-2000.
If they are right in their conjecture, the processes involved must have
started some sixty or seventy years ago, if not in Australia then in the
United States.
In the U.S., a steady decline in the suicide rate of people aged 65
and over has been well-documented: from 45.3, in 1933, to 21.5, in 1986.
(1) Murphy & Wetzel (2) showed the same decline for the period 1940-
1975. The following table, presenting age-specific white male suicide
rates in five-year intervals, has been extracted from their Table 6.
1940 1975 65-69 57.5 35.0 70-74 59.3 37.6 75-79 65.8 44.9 80-84 66.4 44.6
The decline of the suicide rates in older age groups started well
before the introduction of the tricyclics, that is, before there was any
effective pharmaceutical treatment of depression.
1. Hall WD, Mant A, Mitchell PB, Rendle VA, Hickie IB, McManus P.
Association between anti-depressant prescribing and suicide in Australia,
1991-2000: trend analysis BMJ 2003; 326: 1008-1011
2. McIntosh JL. In: Life Span Perspectives of Suicide, A. A. Leenaars,
editor, Plenum Press 1991:60-61.
3. Murphy GE, Wetzel RD. Suicide risk by birth cohort in the US, 1949-
1974. Arch Gen Psychiatry 1980; 37: 519-523.
Thomas J P Verberne
Clinical Neuropsychologist
Competing interests:
None declared
Competing interests: 1940 197565-69 57.5 35.070-74 59.3 37.675-79 65.8 44.980-84 66.4 44.6
Rapid response to Hall et als article in BMJ 10 May 2003:
“Association between antidepressant prescribing and suicide in Australia,
1991-2000: trend analysis”
Peter Ankarberg, clinical psychologist, licensed psychotherapist
In their article in BMJ 10 May 2003: “Association between
antidepressant prescribing and suicide in Australia, 1991-2000: trend
analysis” Hall et al came to the conclusion that the increase in
antidepressant use in Australia have contributed to the declining suicide
rates. This conclusion is not supported by their own data or by recent
meta-analyses of controlled clinical trials for antidepressants.
In Hall et als article it is shown that the biggest increase in
antidepressant use 1990-2000 has been in the ages 15-44. At the same time
the rate of suicide for this age span has increased in Australia. For men
aged 25-34 the use of antidepressants has increased more than six times
during the period. At the same time the suicide rate has increased with
almost 17 percent. Using this correlation in time in the same way that
Hall et al use the total correlation for all ages one could argue that
antidepressants increase the risk of suicide for people younger than 45
and particularly for males 25-34 years. Neither this conclusion nor Hall
et als are of course scientifically valid. To make valid conclusions we
need controlled studies.
Hall et al state that there is little direct evidence that
antidepressants reduce the suicide rate because even large clinical trial
have limited power to detect a reduction. They fail to mention that since
the year 2000 three large meta-analyses on antidepressants and suicide
have been published. Together they include over 60 000 depressed patients
in randomised clinical trials comparing antidepressants to placebo. The
majority of trials were short term but several long term trials were
included with a total of 1949 patients. The results of all meta-analyses,
both for short and long term treatment, were that suicide was slightly
more common among patients taking antidepressants compared to patients
taking placebo.
To summarise: Hall et als material does not show that antidepressants
reduce the rate of suicide. At the same time placebo controlled randomised
clinical trials show that the rate of suicide is higher with
antidepressants than it is with placebo. From this it seems clear that the
only reasonable conclusion is that antidepressants do not reduce the rate
of suicide. There might however be a slight increase in suicide rates for
patients treated with antidepressants although further studies are needed
to make any firm conclusions.
1 Hall WD, Mant A, Mitchell PB, Rendle VA, Hickie IB, McManus P.
Association between antidepressant prescribing and suicide in Australia,
1991-2000: trend analysis. British Medical Journal 2003; 326
2 Khan A, Khan S, Kolts R, Brown WA. Suicide rates in clinical trials of
SSRIs, other antidepressants, and placebo: Analysis of FDA reports. Am J
Psychiatry 2003; 160: 790-92.
3 Khan A, Warner HA, Brown WA: Symptom reduction and suicide risk in
patients treated with placebo in antidepressant clinical trials: An
analysis of the Food and Drug Administration database. Arch Gen Psychiatry
2000; 57: 311-317.
4 Storosum JG, van Zwieten BJ, van den Brink W, Gersons B, Broekmans AW.
Suicide risk in placebo-controlled studies of major depression. Am J
Psychiatry 2001; 158: 1271-75.
Competing interests:
None declared
Competing interests: No competing interests
Re: Responses to our critics
The bracketing of my name with the names of professors De Leo and
Draper as a critic who attributed to the authors views not expressed in
their paper must be due to a word processing error. I am sure that when
they reread my letter, they will agree.
Thomas J P Verberne
Competing interests:
None declared
Competing interests: No competing interests