Mojtabai, 2011 The public health impact of antidepressants

[brighteningup]

Just found this interesting study

 

http://www.sciencedirect.com/science/article/pii/S0165032711002837

 

The public health impact of antidepressants: An instrumental variable analysis

 

Ramin Mojtabai

 

Journal of Affective Disorders

Volume 134, Issues 1-3, November 2011, Pages 188-197

Abstract

Background

 

There has been a marked increase in antidepressant medication prescription and use over the past three decades with unclear effects on the mental health status of the population. This study examined the impact of expansion of antidepressant use on prevalence and characteristics of depression and suicidal ideations in the community.

 

Method

 

Instrumental variable models were used to assess the impact of antidepressant treatments on the prevalence of depressive episodes, mixed anxiety and depression states and suicidal ideations in 22,845 participants of the 1993, 2000 and 2007 National surveys of psychiatric morbidity of Great Britain who were between 16 and 64 years of age.

 

Results

 

Increased prevalence of antidepressant treatment did not impact the prevalence of depressive episodes or mixed anxiety and depression states. However, antidepressant treatment was associated with decreased prevalence of severe and, to a lesser extent, mild depressive episodes and suicidal ideations and a corresponding increase in prevalence of moderate depressive episodes.

Limitations

 

The data were cross-sectional and based on self-report of symptoms in the past month and current medication use with no information on dose and duration of medication treatment.

 

Conclusions

 

Expansion of antidepressant treatments in recent years has not changed the community prevalence of depression overall, but it has reduced the prevalence of more severe depression and suicidal ideations. The findings call for better targeting and more judicious use of antidepressants in cases of more severe depressive episodes which are more likely to respond to such treatments.

I've been able to access the full text main article and thought I'd include a few quotes from the conclusions as they are interesting:

 

The prevalence of a condition is the product of its incidence

and duration. Antidepressants are typically prescribed and used

after the onset of depression and thus do not change the

incidence of depressive episodes. Furthermore, as the present

data show, greater use of antidepressants in the community was

not associated with a shorter self-reported duration of depres-

sive episodes. Thus, the prevalence of depressive episodes could

not have changed appreciably as a result of antidepressant

medication treatment.

However, the study found that antidepressants were

associated with a decrease in the prevalence of severe depression

and a corresponding increase in the prevalence of moderate

depression suggesting that many depressive episodes that would

have been rated as severe without antidepressant treatment,

were now categorized as moderate after antidepressant treat-

ment. A large proportion of treated patients in randomized

clinical trials of antidepressants respond only partially to

treatment (Arroll et al., 2009; Trivedi et al., 2006). Thus, the

increased prevalence of moderate depressive episodes as a result

of increased antidepressant treatments may reflect the increase

in the number of antidepressant-treated patients with initially

severe illness who have responded partially to treatment and

now meet the criteria for moderate depression.

(Mojtabai,2011,p195)

 

 

Next paragraph follows on immediately and refers to specific mathematical methods and is a little obscure but I include it for completeness.

 

The association of antidepressant treatment with a de-

creased prevalence of severe depression is consistent with

some past research indicating a greater efficacy of antidepres-

sants in the treatment of more severe forms of depression

(Fournier et al., 2010; Khan et al., 2002). The results of the

bivariate probit analysis for severe depressive episodes suggest

that the non-significant result of the naïve probit analysis

(Table 3) likely reflected the countervailing effects of increased

use of antidepressants among individuals with severe depres-

sive episodes and the decreased prevalence of this type of

episodes among antidepressant-treated individuals. The strong

effect of antidepressants only emerged when the effect of

selection into treatment was removed in the instrumental

variable model (Table 3).

(Mojtabai,2011,p195)

 

Then perhaps the most interesting bit is the next paragraphs here:-

A puzzling finding of the study was the association of

increased antidepressant use with reduced prevalence of minor

depressive episodes. As noted, previous research indicates that

antidepressants may be less effective in treatment of these

milder forms of depression (Fournier et al., 2010; Khan et al.,

2002), although this finding is debated and other studies have

found a similar effect of antidepressant across the range of

initial severity (Fountoulakis and Moller, 2011; Kirsch et al.,

2008). The contrast between the effects of antidepressants on

mild and moderate depressive episodes is intriguing, especially

in the context of some past arguments regarding the possible

negative effects of antidepressants (Fava, 2003). Future

research needs to further assess the impact of antidepressants

on milder forms of depression, especially in view of the growing

use of antidepressants in individuals with these milder forms.

In this regard, it is also notable that prevalence of depressive

episodes with longer duration increased across the survey

years. However, the increase was not associated with antide-

pressant treatment in the bivariate probit regression analyses.

(Mojtabai,2011,p195-6)

 

{Seems to me there could be an alternate explanation here - could the increase in moderate depression also be partly be due to people moving from mild to moderate symptoms (I'm not clear if the study has controlled for this specifically).}

 

A little further down the page are these concluding remarks

 

In conclusion, the findings of the study suggest that the

increased use of antidepressants might have been associated

with a decline in the prevalence of severe depression and

suicidal ideations. Concerns about the unmet need for care of

depression and other common mental disorders have mainly

focused on disorders with more severe and disabling

symptoms. It is therefore reassuring that the severe forms

of depression have become less common with the expansion

of antidepressant medication use. The decline in the preva-

lence of suicidal ideations associated with the increased use

of these medications is also reassuring in view of the ongoing

debate about the possible risks of suicidal behavior in

individuals treated with these medications (Akiskal and

Benazzi, 2006; Rihmer and Akiskal, 2006).

 

However, it is sobering that antidepressant use was not

associated with a decline in the overall prevalence of

depressive episodes or the mixed anxiety and depression

cases or a reduction in the duration of depressive episodes.

The lack of effect on duration of episodes is especially

puzzling given the efficacy of antidepressant medications in

clinical trials. However, real world effectiveness of antide-

pressant medication treatment often falls short of efficacy in

randomized clinical trials (Brugha et al., 1992)

(Mojtabai,2011,p196)

 

[Skyler]

It would be nice to know more about the data supporting this observation:

 

>It is therefore reassuring that the severe forms

of depression have become less common with the expansion

of antidepressant medication use. The decline in the preva-

lence of suicidal ideations associated with the increased use

of these medications is also reassuring in view of the ongoing

debate about the possible risks of suicidal behavior in

individuals treated with these medications (Akiskal and

Benazzi, 2006; Rihmer and Akiskal, 2006).

[Altostrata]

The conclusion in these types of surveys always depend on what is defined as depression. Like the rate of inflation, you may find it is defined differently over the years, which makes conclusions like these, favorable to pharmapsychiatry, questionable.

 

It's all a big political game, lots of propaganda being generated, you have to look into the interests and associations of the authors, too.

[brighteningup]

Georgie and Alto (and everyone else whose interested),

 

Following up on your comments, as I've got access to the full text, I've done a bit of checking up.

 

Before I get into this, I'd also add that I completely agree there is a lot of politics, and self-interests in this, and this goes right down to how difficult it can be to get articles published, you usually have to get past peer reviewers (i.e. other academic experts in the field). This will have happened in this case. Authors then either then get the article rejected or accepted, but sometimes they have to make changes based on reviewers’ comments, or the article won't get published. This is both a strength and weakness of academic research - no-one - including the author, know who these reviewers are (the reviewers aren't supposed to know who the author is either when they review the article) - no-one gets to know what the reviewers affiliations and funding is or why they reject or accept papers except journal editors. On the plus side it almost certainly leads to better quality in research, on the negative side things can happen like a reviewer can be really critical of an article because it contradicts his or her pet theory or other self interest. [And no I haven't experienced this personally but I have talked to people who have.]

 

If you look carefully at the top of the abstract on the link you will see:

Received 27 September 2010; revised 2 May 2011; accepted 17 May 2011. Available online 17 June 2011

This probably means the author was asked to make some revisions following the reviewers’ comments (this is very common).

 

Another issue with this study is that although the overall study sample population is quite large, some of the sub-sample numbers looked at in the actual statistical analysis are quite small. This can be very problematic in statistical terms, i.e. small numbers can give you misleading results. This is based on looking at Table 1 in the full text article (Mojtabai, 2011, page 191).

 

{However, I don't understand in detail the statistical methods being used here, it might well be ok, but as a general rule a bigger sample than this is often better, especially when you're controlling for other factors as this study was}.

 

I do quantitative analysis as part of the graduate research I do, but I'm not a statistician and don't consider myself an expert - anyone know any tame statisticians who could look at this study?? What I can say is that whilst the relevant analyses show significant results (statistically) and the conclusions drawn accurately reflect this, it's seems for some models to be a very small number of cases to be using for the analysis. This is especially true for the models which are used for the conclusions on suicidal ideation, mild, moderate and severe depression. The numbers look ok for depression and anxiety disorders and probably ok for all depression (all types lumped together).

 

The author of the study is from John Hopkins University - his web page is here

http://www.jhsph.edu/faculty/directory/profile/4936/Mojtabai/Ramin

There's no obvious info about who funded this study or where his funding usually comes from not sure where you'd look for, for this.

 

Warning the rest of this is quite long - only read on if you want to know some more about the data used for the study above.

 

The data used for this study is taken from 3 separate social surveys at seven year periods within Great Britain. The surveys were of the general population - that means the survey was set up to give a reasonable sample of all people living in the areas surveyed (not just those who were sick or had visited a GP). They were carried out independently either by the Office for National Statistics or the National Centre for Social Research in collaboration with the University of Leicester for the NHS Information Centre for health and social care.

 

The main report of the 2007 survey has quite a lot of methodological information about the 2007 survey and some of the previous surveys it's full citation is:-

McManus et al., 2009b S. McManus, H. Meltzer, T. Brugha, P. Bebbington and R. Jenkins, Adult psychiatric morbidity in England, 2007: results of a household survey, appendices and glossary, The NHS Information Centre for Health and Social Care (2009).

and you can download it here (only for the really keen, these are long and hard going unless your passionate about social surveys ...)

http://www.ic.nhs.uk/statistics-and-data-collections/mental-health/mental-health-surveys/adult-psychiatric-morbidity-in-england-2007-results-of-a-household-survey

 

A primary purpose of the 2007 survey was to assess change in the population prevalence of

disorders over time. For this reason maintaining comparability with the 2000 survey was a

priority, so both the questionnaire and the way it was administered were largely the same.

Differences between the adult psychiatric morbidity surveys include: the 2007 survey

covered England only (the populations of Scotland and Wales were also sampled in 1993

and 2000 also sampled Scotland and Wales) and there was no upper age limit to the 2007

survey (capped at 64 in 1993 and 74 in 2000). Several new disorders and topics were

covered in 2007, including eating disorder, PTSD, ADHD and gambling behaviour.

[McManus et al, 2009, - the main report p18)

 

Yes that does mean that it's not quite comparing the same population like with like for each survey, but the 2007 survey was aiming to try and be comparable across time. I'm not sure how clear this is in either the abstract of the article above, or the methods section of the full text. You can come to your own conclusions on the suitability of using this for the purposes the article does.

 

Psychiatric disorders were assessed by a structured interview—the Clinical Interview Schedule—Revised (CIS—R) (Lewis et al., 1992). Symptom data obtained in CIS—R were subjected to a computer algorithm providing diagnostic categories according to the 10th edition of the International Classification of Diseases (ICD-10) (World Health Organization., 1992)

 

[Mojtabai, 2011, page 189]

 

The Lewis et al.., 1992 refers to a standardised assessment interview tool designed for lay interviewers more info here:

http://journals.cambridge.org.ezproxy.webfeat.lib.ed.ac.uk/action/displayAbstract?fromPage=online&aid=5036972

 

Cutting through the academic jargon - this means a trained interviewer who wasn't a medical expert, took out a specific set (a very very long carefully defined set) of questions which they went through systematically with each person. It was almost certainly done in the person’s home. The interview questions were designed to be asked by someone who wasn't a medical expert. Responses were then fed into a computer which did some clever maths programming logic stuff based on definitions from the WHO ICD10 definitions of depression to assess if someone was depressed or not. Some of the questions the interviewee would have been asked they would have answered directly, some they would have been able to complete themselves if they felt uncomfortable - I'm not sure which this would have applied to. The answers to the questions were generally based how the person reported they were feeling in the last week.

 

{{Only for the really really keen - For more info see the McManus et all 2009 report on the 2007 survey - the appendices pdf, especially from page 10, and the questionnaire at Appendix D which shows the questions actually asked in 2007 in order that they were able to assess the symptoms in the lists above, there are all kinds of questions about how your sleeping, eating, concentrating / forgetting etc etc. The questionnaire must have taken a long while to administer. If it was anything like other social surveys I know about, the interviewer probably filled in the answers on a laptop in 2007 which were then automatically fed in a big computer system to analysis (in earlier years pen paper may have been used and then some persons would have had to a massive data entry task).}}

 

So for example in the 2007 study, the following list generated from results of the questionnaire was used to assess the most mild type of depression.

 

F32.00 Mild depressive episode without somatic symptoms

1. Symptom duration =2 weeks

2. Two or more from:

• Depressed mood

• Loss of interest

• Fatigue

3. Two or three from:

• Reduced concentration

• Reduced self-esteem

• Ideas of guilt

• Pessimism about future

• Suicidal ideas or acts

• Disturbed sleep

• Diminished appetite

4. Social impairment

5. Fewer than four from:

• Lack of normal pleasure /interest

• Loss of normal emotional reactivity

• A.M. waking =2 hours early

• Loss of libido

• Diurnal variation in mood

• Diminished appetite

• Loss of =5% body weight

• Psychomotor agitation

• Psychomotor retardation

 

[McManus et al, 2009, - the appendices, p11]

 

 

 

and the following list for a severe depressive episode

 

F32.2 Severe depressive episode

1. All three from:

• Depressed mood

• Loss of interest

• Fatigue

2. Four or more from:

• Reduced concentration

• Reduced self-esteem

• Ideas of guilt

• Pessimism about future

• Suicidal ideas or acts

• Disturbed sleep

• Diminished appetite

3. Social impairment

4. Four or more from:

• Lack of normal pleasure /interest

• Loss of normal emotional reactivity

• A.M. waking =2 hours early

• Loss of libido

• Diurnal variation in mood

• Diminished appetite

• Loss of =5% body weight

• Psychomotor agitation

• Psychomotor retardation

 

[McManus et al, 2009, - the appendices p13]

 

The study quoted by Mojtabai then used bits of the survey for the purposes of his research. These types of social survey are generally made available for academics to use in data archives, along with comprehensive information

 

This study focused on three ICD-10 categories of depressive episodes based on severity (mild, moderate and severe), and the CIS—R mixed anxiety and depression category. Depressive episodes in ICD-10 are defined by episodes lasting at least two weeks during which the individual experienced a number of more common or “typical” symptoms and a number of other depressive symptoms (World Health Organization., 1992). [Mojtabai, 2011, page 189]

....

Duration of depressive episode was assessed based on one question about the length of time the person has been “feeling sad, miserable or depressed or unable to enjoy or take an interest in things”. Responses to this question were categorized for this study into < 6 months, 6 months or more but less than 1 year, 1 year or more but less than 2 years and 2 years or more.[Mojtabai, 2011, page 189-190]

 

For suicidal ideation the study specifically says which bit of the questionnaire it used

 

Suicidal ideations were assessed in CIS—R by asking the participant if in the past week, they had thought of killing themselves. This question was included in the CIS—R versions used in 1993 and 2007 but not 2000. The 2000 survey included a more detailed module for assessment of suicidal ideations and behaviors which did not include this specific question. Therefore the analyses for suicidal ideations were limited to survey years 1993 and 2007.

[Mojtabai, 2011, page 190]

 

So yes this means the suicide ideation conclusions are based on results from two out of the 3 surveys, one in 1993 and one in 2007, not all 3 years (this is because the 2000 question was different to the one used in 1993 and 2007). Also, if you plough through the full text of the article to table 1, you find in 1993 0.8% (97 present 10,011 absent) of the respondents had suicidal ideation present (in the last week), as defined for the study, present and in 2007 0.6% (40 present 5385 absent) had suicidal ideation present (see Mojtabai, 2011, 191); this is a small sample to draw conclusions on when comparing the two years...

 

Whilst the relevant statistics from the subsequent analysis are significant (statistically) and the conclusions drawn accurately reflect this, it's a very small number of cases to be using which can be very problematic in statistical terms. {as discussed above).

 

Finally, to give you an idea of the type of questions asked on suicidal thinking. As an example, I've looked at the 2007 survey questions and the actual questions on suicidal ideations used then were as follows (this also give you an idea of what the interviewers script looks like).

 

DSHIntro

There may be times in everyone's life when they become very miserable and depressed and may

feel like taking drastic action because of these feelings.

1 Continue

DSH1

Have you ever felt that life was not worth living?

1 Yes

2 No

IF DSH1 = Yes THEN

DSH1a

Was this....

READ OUT AND CODE FIRST THAT APPLIES

1 ...in the last week,

2 …in the last year,

3 or at some other time?

DSH2

Have you ever wished that you were dead?

1 Yes

2 No

IF DSH2 = Yes THEN

DSH2a

Was this…

READ OUT AND CODE FIRST THAT APPLIES

1 ...in the last week?

2 …in the last year?

3 or at some other time?

DSH3

Have you ever thought of taking your life, even if you would not really do it?

1 Yes

2 No

IF DSH3 = Yes THEN

DSH3a

Was this…

READ OUT AND CODE FIRST THAT APPLIES

1 ...in the last week,

2 …in the last year,

3 or at some other time?

 

[see McManus et al, 2009, - the appendices

 

I don't have the references to hand, but research suggests that people are often actually surprisingly willing to talk about and self report quite personal things and interviewers will be trained in how to ask sensitive questions, some people will have refused to answer some of the questions though.

 

Ok looking into this has been interesting, and a positive way to procrastinate, but I really really need to do some actual work for a bit now.

 

Best,

 

Bright

[Altostrata]

Here's how Johns Hopkins spun this paper (in bold) http://www.jhsph.edu/publichealthnews/press_releases/2011/mojtabai_antidepressant_prescriptions.html

 

Mojbatai published a second paper, The Proportion of Antidepressants Prescribed Without a Psychiatric Diagnosis is Growing, in August 2011. There's a topic on it here.

 

August 4, 2011

 

Prescriptions for Antidepressants Increasing among Individuals with no Psychiatric Diagnosis

 

Americans are no strangers to antidepressants. During the last 20 years the use of antidepressants has grown significantly making them one of the most costly and the third most commonly prescribed class of medications in the U. S. According to the Centers for Disease Control and Prevention, from 2005-2008 nearly 8.9 percent of the U.S. population had at least one prescription in this drug class during any given month. A new study led by researchers at the Johns Hopkins Bloomberg School of Public Health examines national trends in antidepressant prescribing and finds much of this growth was driven by a substantial increase in antidepressant prescriptions by non-psychiatrist providers without any accompanying psychiatric diagnosis. The results are featured in the August 2011 issue of Health Affairs.

 

“We’ve seen a marked increase in antidepressant use among individuals with no psychiatric diagnosis. Nearly four out of every five antidepressant prescriptions are written by non-psychiatrist providers,” said Ramin Mojtabai, MD, PhD, MPH, lead author of the study and an associate professor with the Bloomberg School’s Department of Mental Health. “Between 1996 and 2007, the number of visits where individuals were prescribed antidepressants with no psychiatric diagnoses increased from 59.5 percent to 72.7 percent and the share of providers who prescribed antidepressants without a concurrent psychiatric diagnosis increased from 30 percent of all non-psychiatrist physicians in 1996 to 55.4 percent in 2007.”

 

Using data from the 1996-2007 National Ambulatory Medical Care Surveys, researchers reviewed a national sample of office-based physician visits by patients ages 18 years and older during a one-week period. They conducted two sets of logistic regression analyses, comparing antidepressant visits lacking psychiatric diagnoses with antidepressant visits including psychiatric diagnoses and visits lacking both prescriptions for antidepressants and psychiatric diagnoses. In addition, Mojtabai and colleagues assessed physician practice-level trends in antidepressant visits without psychiatric diagnosis and found that in the general medicine practice, antidepressant use was concentrated among people with less severe and poorly defined mental health conditions.

 

An earlier study led by Mojtabai and published in the Journal of Affective Disorders examined the impact of expansion of antidepressant use on the prevalence and characteristics of depression and suicidal ideations. That study found that antidepressant use significantly reduced the prevalence of more severe depression and suicidal ideations among individuals with sever depressive episodes. The findings lead researchers to recommend that antidepressants be prescribed primarily to individuals with severe depression or a confirmed psychiatric diagnosis.

 

“With non-specialists playing a growing role in the pharmacological treatment of common mental disorders, practice patterns of these providers are becoming increasingly relevant for mental health policy,” adds Mojtabai. “To the extent that antidepressants are being prescribed for uses not supported by clinical evidence, there may be a need to improve providers’ prescribing practices, revamp drug formularies or undertake broad reforms of the health care system that will increase communication between primary care providers and mental health specialists.”

 

“The Proportion of Antidepressants Prescribed Without a Psychiatric Diagnosis is Growing” was written by Ramin Mojtabai and Mark Olfson and was supported in part by the Agency for Healthcare Research and Quality.

 

“The public health impact of antidepressants: An instrumental variable analysis” was written by Ramin Mojtabai.