Transcript Best Anti-depressant

Depression
G. Michael Allan
Assoc. Prof, Family Med, U of A
Conflict of Interest
• Family Doctor for 12+ yrs
• Academic 8 years
• Pay from U of A and Alberta Health
• Research and Speaking Fees
– Non-Profit Sources (E.g Canadian Expert Drug
Advisory Committee, Alberta College of Family
Physicians, etc)
– No funding from Industry
Objectives
1.
2.
3.
4.
History and Context
Screening and Diagnosis
Challenges with Anti-Depressant Evidence
Do Antidepressants work?
i. What about pediatric (& adolescent) patients?
ii. Safety in Pregnancy
5. How long do they take work?
6. Which is the best
i. Are older worse than newer?
ii. Is one 2nd Generation better than another?
7. Specific Topics
i. What is the evidence for dose or dose adjustment?
ii. What if the first doesn’t work?
iii. When can we stop them?
Sad Truth of Anti-depressants
• Your student just saw Mrs Misery (41 y.o.).
She is feeling down for 4 months, doesn’t
enjoy anything, cries without cause, can’t get
or stay asleep, has no energy, and can’t stay
focused. She has lost 10 lbs (she is glad for
that at least).
• She has no thoughts of self-harm
• As you write a script for an anti-depressant,
the students asks, “But I thought they didn’t
work?”
Sad Truth of Anti-depressants
• Do Anti-depressants Work?
• They work but
– Not as well as we thought
– It varies in some groups
• Psychiatry provides the best examples of
bias in the literature.
• Let’s review some of those bias
Poor Quality in Anti-depressants
• In a series of 46 RCT1
– 91% fair quality – 9% good (poor quality excluded)
– 85% industry funded (11% affiliated, 4% not reported)
• In a series of 50 RCT2
– Intention to treat used only 54%
• In a series of 11 RCT3
– Allocation Concealed (assured randomization) in 9%
1) Ann Intern Med. 2005;143 :415-26.
Lancet 2004;363:1341-5
2) Arch Gen Psychiatry 2006; 63: 1217-23.
3)
Other Specific Biases
• Subjective:
– Clinicians report/score benefit > pts
– E.g.: In one series clinicians found benefit in
33% of scales but patients report benefit in 0%
• Scales:
– ↑ numbers (esp big scales)
– Easier to find statistical significant
– But not clinical significant
– E.g. Improving 2.7 on a 113 scale.
Lancet 2004;363:1341-5. & BMJ 2004;328:879-83.
Hiding the Bad: Part 1
• Selective Publication of Anti-depressant Trials
• FDA records of 12 SSRI/SNRI’s vs Published
• 74 Trials:
– 38 Positive: 37 published, 1 not published.
– 36 Negative: 3 published as negative, 11
published as positive, 22 not published.
• Positive = 94% published vs 51% FDA
• Effect (benefit): Published (0.41) > FDA (0.31)
NEJM 2008; 358: 252.
Hiding the Bad:
Part 2
• Single Trails Pub:
– if trial +ve = 90%
– if trial –ve = 29%
• What Actually happens to
SSRI trials?
– Hiding “Bad” trials
– Multiple pub. “Good”
Melander. et al, BMJ, 2003; 326: 1171-73
Hiding the Bad: Part 2
• What Actually happens to SSRI trials?
– +ve trials published 4.4x each (vs 1.3)
Melander. et al, BMJ, 2003; 326: 1171-73
Getting the most of a small effect
•
•
•
•
•
Initial Severity and Anti-depressant Benefits
35 RCT’s of 4 SSRI/SNRI’s
No diff. between Anti-depressants
Overall, Statistical Sign common BUT,…
Clinical Significance less clear
– Ham D scores range between 17-30.5
– Mean Change was 9.6 for med & 7.8 for placebo
– Mean difference was 1.8 HRSD
– Therefore, 81.5% of the improve seen on antidepressants is from “placebo”
PLOS Med 2008; 5(2): 0260.
Effect size
• These are basically scatter plots
with regression analysis.
– Putting on mean outcomes for
placebo and drug from each trial
and drawing a best fit line for the
effect of placebo and drug.
– You then figure out when the
two separate to predefined level.
– This uses effect sizes, which is
common in summary statistics
for scales (like depression or
pain)
• In this case, they used 0.5 which
they got from NICE.
• It separates at green, about HSRD
27
• Note: If they used an easier clinical
sign (let’s say 0.2, like in OA),
separation would be earlier (around
24 in the OA case).
Getting the most of a small effect
• Differences Only reach Clinical Significance when
Depression Severe
– The actual number varies by graph but seems like HSRD
26 (from figure 4), 27 (from figure 3) and 28 (from figure
2)
– Maybe lower (perhaps Moderate) if “clinical significance”
defined more liberally
• Example: 0.2 is defined as the minimum effect size in OA
• Bottom-line: drug effect stable but placebo less
effective as depression worsens
– Meaning, benefit of drug over placebo greater with
worsening depression.
PLOS Med 2008; 5(2): 0260.
Getting the most: NEW
• Meta-analysis of 6 trials: with patient level
data!
– Searched 3 databases (pubmed, Psychinfo &
cochrane).
– RCT, Adult, ≥ 6 wks, Ham D, pt level data.
• 6 studies (3 imipramine, 3 Paroxetine), 6-11
weeks (median 8), baseline Ham D= 14-23.
• Results: Same as before, severe,  benefit
– Clin Sign Diff (NICE definition (diff 3 on Ham-D)
at Ham-D 25
Getting the most: NEW
• Continued results: used 3 categories
– Mild/mod (Ham-D ≤18), severe (19-22), very
severe (≥23)
– Using effect size (measuring diff from
comparator): both mild/mod & severe never got
better 0.20. At very severe 0.47.
• NNT (for better outcome): 16, 11 and 4
(mild/mod, severe, very severe).
JAMA. 2010;303(1):47-53
Getting to the Actual Numbers
• Paroxetine is well studied and a meta-analysis of
the drug helps see the individual outcomes
• Standard Mean difference was -0.31, -0.40 to 0.22)
• Actual numbers that get ≥50% of improvement; If
you treat 100 pts with paroxetine
– 53% taking paroxetine
– 42% with placebo
– Difference is 11% (or NNT 9)
• If you look at other outcomes
– Those who discontinue for any reason: no difference
– Adverse events: reporting (NNH 9), discontinuing due to
(NNH 17)
– Experience Suicidal tendencies (NNH 142).*
CMAJ 2008;178(3):296-305
Bottom-line
• Antidepressants help reduce symptoms of
(moderate to severe) depression in up to
70% of patients
– (but likely lower & maybe a lot lower)
• Remission Less.
• In mild Depression, 90% of that effect is from
“Placebo”
• As the Depression is more severe, the drug
gives more of that effect (compared to
placebo).
Best Anti-depressant
• A 27 year old woman has been slipping into
depression for 4 months and non-medicinal
measures have been ineffective. She is now
clearly depressed (not suicidal). You have
heard from psychiatrists and different drug
reps certain meds are better.
Best Anti-depressant
•
1)
2)
3)
4)
5)
6)
Which was better in a well-designed metaanalysis?
Venlafaxine
Buproprion
Esocitalopram
Citalopram
They’re all the same
None Ever Work (Dr Thomas Cruise)
Best Anti-Depressant = Any
• They are all about the same
• 46 RCT’s (11.5 K pts), ≥3 months
• No Diff Quality of Life (& for Ham D)
– Venlafaxine > Fluoxentine: RB 1.12 (1.02-1.23) & NNT16
– Sertraline > Fluoxetine: RB 1.1 (1.01-1.2) & NNT 17
– SE similar (? Venlafaxine N&V), rapidity of response No diff.
• Benefit = always 5% in favour of sponsored drug
(NNT 20)
• SE similar (? Venlafaxine N&V), rapidity of
response No diff.
Ann Intern Med. 2005;143 :415-26.
Best Anti-Depressant = Any
•
•
Added since 2005
Meta-analysis1: 203 studies (171 RCT’s) – using
some indirect comparisons
–
Effectiveness similar. Few stat sign relative benefit, but
• None clinically significant E.g. MADRS 60 pt scale:
esocitalopram 1.13 > citalopram (min clinically
important diff=2)
– Sponsorship may play a role in these subtle differences
– Adverse Events similar in amount (61% of patients had
≥1) but types varied
• E.g. Venlafaxine 11% more nausea & vomiting,
Sertraline 3% more diarrhea,
1. Ann Intern Med. 2008;149:734-750. 2. Lancet 2009; 373: 746–58
Best Anti-Depressant = Any
• Meta-analysis2 (117 RCTs), examined response to
treatment and withdrawal, used indirect methods
– Identified some small differences
– Efficacy Top 4: mirtazapine, escitalopram, venlafaxine,
sertraline
– Acceptability Top 4: escitalopram, sertraline, bupropion,
citalopram
– Used Odds Ratios, treated scales as =, Funding bias
• Interpret in context 1) Antidepressant research
suffers significant bias, 2) ≤10% of studies are high
quality
• Bottom-line: Similar Effectiveness, different AE
1. Ann Intern Med. 2008;149:734-750. 2. Lancet 2009; 373: 746–58
Best Antidepressant = Any: Quality Concerns
• Other reviews have attempted to determine if one
antidepressant is superior but the results are inconsistent.1
• The 2009 review2 has important concerns regarding validity
including,
– They treated all depression scales as the same (and they are not),
– Using odds ratios exaggerated the differences they found,
– Importantly, when they tried to account for sponsorship bias, the
differences between the drugs were reduced.
• Both reviews2,3 did some indirect comparison of drugs from
different studies, which is less reliable than direct
comparison in the same study.
• The 2008 review3 did not have quite as many concerns with
validity.
1) Cochrane 2005;(4): CD004185. Curr Med Res Opin. 2009; 25 :161-75. Cochrane
2009;(2): CD006117. Ann Intern Med. 2005;143 :415-26. 2) Lancet 2009; 373: 746–
58. 3) Ann Intern Med. 2008;149: 734-750.
Miscellaneous
Management
Issues
Is Bigger Better?
• Mrs Low (31 year old) is in. 2 weeks ago
she was diagnosed with moderate
depression. She is now ready to trial a
medication.
• Your resident says he heard (from an
expert) that “primary care doctors always
under-treat depression”
• As you write the script, he wants to know
what your target dose is.
Is Bigger Better
•
1)
2)
3)
4)
Based on two literatures reviews, you
can say;
Moderate dose SSRI (e.g. 30-40) is the
general target
Starting low (10mg) is reasonable
Increases into high doses (>20mg) may
not be much help.
Forget the prescription, send her to this
know it all expert.
Bigger Is NOT Better
• Low doses of anti-depressants seem as
effective as high doses.
– Flouxetine (5 vs 20 vs 40mg)1
– Tricyclics (50-100 vs >100mg)2
• Increasing doses in non- responsive patients
doesn’t seem to help much.3
– At least not until 8 weeks have past.
1) Psychopharm Bull 1988; 24: 183-8. 2) BMJ. 2002;325:991-5. 3) Br J Psychiatry
Switching Quick: New
• What constitutes a reasonable trial?
When should we say, “this isn’t working,
lets’ try another instead.”
• Only three studies have looked at this and
can’t see a difference in switching after 6-7
weeks and not.
• Note STAR*D waited a mean of 12 weeks
Acta Psychiatr Scand 2010;121:174–9.
New: Combination
• Combined regular anti-depressants at the
start may be helpful,
– RCT of 105 pts x 6 weeks, Fluoxetine (20) vs Mirtazapine
(30) plus Fuoxetine (20) or Venlafaxine (225) or
Bupropion (150)
– Remission = Ham D ≤7 (mean start =22)
– Fluox 25%, Fluox+M 52%, Ven+M 58%, Bup+M 46% (all
comb stat sign (p=0.01) vs Fluox alone); NNT 4
– AE common but No diff in AE quiting
• Few other studies also find (nothing big yet
though).
Am J Psychiatry 2010; 167:281–288
Waiting for the World to Change
• Mrs low is back 7 days later. She is on
Fluoxetine 10mg. She reports she is
feeling a little better (more energy
concentration).
• She read that it should take 2-4 weeks to
work so wants to know how this can be
helping?
• How fast do anti-depressants work?
Waiting for the World to Change
• They work as fast as 7 days:
• From a meta-analysis1 of 50 trials (10,121 depressed
patients) looking at response to SSRI medications,
compared to placebo, at a series of time intervals.
– On average, 1/3 of the total benefit occurs in the first 7 days (whole
effect by 6 weeks).
– For every 25 people treated, one more will be 50% improved over
placebo at 7 days.
• Hard to find because the “n” needed to show a benefit is 9x
bigger that most trials!
• The results of the meta-analysis above confirm those of
another meta-analysis2
– That study also found antidepressants result in statistically significant
improvement in clinically important outcomes within the first week.
1) Arch Gen Psychiatry 2006; 63: 1217-23. 2) J Clin Psychopharmacol 2006;26:56–60.
A Trial of Separation?
• Mrs Low is now 6 months on Antidepressants and has done well. This was
her first case of depression. She wants to
know if she can stop?
• What if this was her 2nd or 3rd case?
A Trial of Separation?
• In Meta-analysis of 31 RCT (of all types)1
– Meds stopped after 4-28 weeks (most 6-16)
– Relapse at 12 months: 41% Placebo vs 18%
– NNT 5.
• Dose reduction similar (5 RCT)2
– 25.3% low dose vs 15.1% in previous dose
• Recurrence (hard to separate out one)3
–
–
–
–
–
From a cohort of 318 depressed pts, 60% had previous depression
After 1 yr, 25% of the cohort had a recurrence
If second, 41% in 1 year.
Add 16% for each subsequent episode, but some repeaters
36% did not have a recurrence in 5 years.
1) Lancet 2003; 361: 653–51. 2) Psychother Psychosom. 2007;76(5):266-70 3) Am J
Psychiatry 2000; 157:229–233
Missing the Diagnosis
• Ms V Blue is a 25 year old female coming
for her complete medical. She is mentions
she is a little more fatigued of late.
• You want to screen for depression.
• Is there a quick screen?
2 Question Screen
•
During the past month have you often been
bothered by,
1) Feeling down, depressed, or hopeless?
2) Little interest or pleasure in doing things?
• No to both: virtually rules out depression
– LR 0.05 (Sensitivity 97%)
•
Yes to either: go through full depression questions
– Not specific (can’t make Dx, many false positives)
•
Another study of these found same (LR 0.07)
BMJ 2003; 327:1144-46. J Gen Intern Med. 1997;12:439-45.
Patients Like Ours?
Shooting Stars: Fall to Earth
• So what really happens to the average
patient seen for depression?
Shooting STAR*D: Findings
• 2876 people were put on Citalopram
• More like real patients
– (mix of general and specialty)
• 80% had chronic or recurrent depression
• Many complicating Psychiatric conditions.
– 18% had attempted suicide.
• Mean HRSD = 21.8
• Mean exit dose of citalopram = 42 mg/day
Shooting STAR*D: Findings
• Citalopram remission rates
– HRSD = 27.5% and QIDS-SRTreatment 33%
• Response rate = 47%
• Similar rates between GP’s and specialists
• If a patient hadn’t attained remission, they
were offered to continue.
– 2086 patients were offered to continue.
Am J Psychiatry 2006; 163:28–40
Shooting STAR*D: 2 Findings
• Substituting 727 patients
• Drop Citalopram and immediate start:
– Bupropion-SR, Sertraline, or Venlafaxine-XR
• Remission rates about 25% across the
board (for both tools & all 3 meds).
• Tolerability and adverse events similar.
N Engl J Med 2006;354:1231-42.
Shooting STAR*D: 2 Findings
• Augmenting 565 patients
• Drop Citalopram continue and add:
– Bupropion or Buspirone
• HSRD Remission rates around 30% for both.
• QIDS-SR-16: Buproprion > than buspirone
– Reduction 25% vs 17%
– Lower end score 8 vs 9
– Lower intolerance drop-out rate 12.5% vs 20.6
N Engl J Med 2006;354:1243-52.
Shooting STAR*D: 3 Findings
• Augmenting 142 patients
• Add: Lithium or triiodothyronine (T3)
• Remission rates around 20% for both.
• Substituting 234 patients
• Switch: nortriptyline or mirtazapine
• Remission rates between 10-20% for both
EBMH November 2008; 11(4): 97-9.
Shooting STAR*D: Summary
• Perhaps trials reporting the ideal response 70% (or
more “realistic” 53% from the paroxetine study) are
actually an exaggeration
• Efficacy population = 51.6% response versus 39.1% in effectiveness or
pragmatic STAR*D population.
• Maybe choosing the type of alternative antidep
doesn’t matter (they seem the same).
• Maybe specialist care is not a lot different from GP
• Choice of augmentation uncertain (guidelines2 put
lithium and olanzapine ahead of the first choices
here).
Am J Psychiatry 2009; 166:599–607. 2. J Psychopharmacol 2008;22:343–96.
Related to STAR*D: New
• Some related citations
• Treating Depression in Primary Care does work1
– Treatment response NNT 7-9,
– Withdrawal due to AE (4-30 TCA or 20-90 SSRI)
• If you need to switch,2
– it doesn’t matter which one
– The more previous treatments, the higher the chance of
a poor outcomes.
1) Cochrane Database Syst Rev. 2009 Jul 8;(3):CD007954.
2) J Clin Psychiatry. 2006 Dec;67(12):1836-55.
What About the Children?
Pediatric Depression: Meds or not
• You are seeing a 12 year old boy whom
a has many symptoms of depression.
This is new for him (last 6 weeks). His
mother worried and wonders if
antidepressants might help (she has
been successfully treated in the past).
• In addition to possible referral, looking
into contributing factors, etc: What can
we say about antidepressants in
pediatrics.
Pediatric Depression: Meds or not
•
What can we say:
o
o
o
There are concerns that the drugs are ineffective.
There are concerns that the drugs are not safe.
There is some evidence that perhaps fluoxetine may
be a reasonable choice.
•
SR/Met: 5 Databases + refs & authors, 6
drugs (5 pub, 6 unpub from 5000+), f/u 42-96
days F/U poor (54-83%). (BMJ= 4 drugs, 6
published trials)
Lancet 2004;363:1341-5 (BMJ 2004;328:879-83)
Pediatric Depression: Meds or not
• New anti-depressants vs placebo
• Results: see table
• Good use of unpublished data, excellent
combo, but poor f/u, short duration
• Industry bias (under-reporting SE & Hiding
trials).
• Fluoxetine only (for now) & Some question it’s
benefit.
Lancet 2004;363:1341-5 (BMJ 2004;328:879-83)
Study 1 & 2: SSRIs in childhood dep,… &
Efficacy & safety of anti-depressants,...
# of
Trials
(n)
NNT to
Benefit
[CI]
NNH to
Serious
SE [CI]
NNH to
Suicide Risk
[CI]
< placebo
< placebo
Fluoxetine
2 (315) 6 [4-15]
Paroxetine
3 (658) 20 [7- harm] 15 [8-50]
Sertraline
1 (376) 34 [5- harm] ∞ [34-ben] 50 [50-benefit]
100 [25-benefit]
Citralopram 2 (422) insignificant 12 [6-100] 25 [13- benefit]
Venlafaxine 3 (374) insignificant 15 [9-50]
12 [10-34]
In 6 trials, 42 measures used but only 14 showed any improve
(0/10 patient/parent measures. Scales clinically questionable
(Improving 2.7 on a 113 scale ?)
Suicide
• Adolescents: Odds ratio 1.92 (1.51-2.44)
• Adults: 0.57 (0.47-0.70)
• Elderly: 0.46 (0.27-0.79)
• Absolute numbers not given (but likely
low).
CMAJ. 2009;180:291-7.
Questions