SSRIs: Much More Than You Wanted To Know

The claim that “SSRIs don’t work” or “SSRIs are mostly just placebo” is most commonly associated with Irving Kirsch, a man with the awesome job title of “Associate Director Of The Program For Placebo Studies at Harvard”.

(fun fact: there’s actually no such thing as “Placebo Studies”, but Professor Kirsch’s belief that he directs a Harvard department inspires him to create much higher-quality research.)

In 1998, he published a meta-analysis of 19 placebo-controlled drug trials that suggested that almost all of the benefits of antidepressants were due to the placebo effect. Psychiatrists denounced him, saying that you can choose pretty much whatever studies you want for a meta-analysis.

After biding his time for a decade, in 2008 he struck back with another meta-analysis, this being one of the first papers in all of medical science to take the audacious step of demanding all the FDA’s data through the Freedom of Information Act. Since drug companies are required to report all their studies to the FDA, this theoretically provides a rare and wonderful publication-bias-free data set. Using this set, he found that, although antidepressants did seem to outperform placebo, the effect was not “clinically significant” except “at the upper end of very severe depression”.

This launched a minor war between supporters and detractors. Probably the strongest support he received was a big 2010 meta-analysis by Fournier et al, which found that

The magnitude of benefit of antidepressant medication compared with placebo increases with severity of depression symptoms and may be minimal or nonexistent, on average, in patients with mild or moderate symptoms. For patients with very severe depression, the benefit of medications over placebo is substantial.

Of course, a very large number of antidepressants are given to people with mild or moderate depression. So what now?

Let me sort the debate about antidepressants into a series of complaints:

1. Antidepressants were oversold and painted as having more biochemical backing than was really justified
2. Modern SSRI antidepressants are no better than older tricyclic and MAOI antidepressants, but are prescribed much more because of said overselling
3. There is large publication bias in the antidepressant literature
4. The effect size of antidepressants is clinically insignificant
5. And it only becomes significant in the most severe depression
6. And even the effects found are only noticed by doctors, not the patients themselves
7. And even that unsatisfying effect might be a result of “active placebo” rather than successful treatment
8. And antidepressants have much worse side effects than you have been led to believe
9. Therefore, we should give up on antidepressants (except maybe in the sickest patients) and use psychotherapy instead

1. Antidepressants were oversold and painted as having more biochemical backing than was really justifiedTotally true

It is starting to become slightly better known that the standard story – depression is a deficiency of serotonin, antidepressants restore serotonin and therefore make you well again – is kind of made up.

There was never much more evidence for the serotonin hypothesis than that chemicals that increased serotonin tended to treat depression – making the argument that “antidepressants are biochemically justified because they treat the low serotonin that is causing your depression” kind of circular. Saying “Serotonin treats depression, therefore depression is, at root, a serotonin deficiency” is about as scientifically grounded as saying “Playing with puppies makes depressed people feel better, therefore depression is, at root, a puppy deficiency”.

The whole thing became less tenable with the discovery that several chemicals that didn’t increase serotonin were also effective antidepressants – not to mention one chemical, tianeptine, that decreases serotonin. Now the conventional wisdom is that depression is a very complicated disturbance in several networks and systems within the brain, and serotonin is one of the inputs and/or outputs of those systems.

Likewise, a whole bunch of early ’90s claims: that modern antidepressants have no side effects, that they produce miraculous improvements in everyone, that they make you better than well – seem kind of silly now. I don’t think anyone is arguing against the proposition that there was an embarrassing amount of hype that has now been backed away from.

2. Modern SSRI antidepressants are no better than older tricyclic and MAOI antidepressants, but are prescribed much more because of said oversellingFirst part true, second part less so

Most studies find SSRI antidepressants to be no more effective in treating depression than older tricyclic and MAOI antidepressants. Most studies aren’t really powered to do this. It seems clear that there aren’t spectacular differences, and hunting for small differences has proven very hard.

If you’re a geek about these sorts of things, you know that a few studies have found non-significant advantages for Prozac and Paxil over older drugs like clomipramine, and marginally-significant advantages for Effexor over SSRIs. But conventional wisdom is that tricyclics can be even more powerful than SSRIs for certain very severe hospitalized depression cases, and a lot of people think MAOIs worked better than anything out there today.

But none of this is very important because the real reason SSRIs are so popular is the side effect profile. While it is an exaggeration to say they have no side effects (see above) they are an obvious improvement over older classes of medication in this regard.

Tricyclics had a bad habit of causing fatal arrythmias when taken at high doses. This is really really bad in depression, because depressed people tend to attempt suicide and the most popular method of suicide attempt is overdosing on your pills. So if you give depressed people a pill that is highly fatal in overdose, you’re basically enabling suicidality. This alone made the risk-benefit calculation for tricyclics unattractive in a lot of cases. Add in dry mouth, constipation, urinary problems, cognitive impairment, blurry vision, and the occasional tendency to cause heart arrythmias even when taken correctly, and you have a drug you’re not going to give people who just say they’re feeling a little down.

MAOIs have their own problems. If you’re using MAOIs and you eat cheese, beer, chocolate, beans, liver, yogurt, soy, kimchi, avocados, coconuts, et cetera, et cetera, et cetera, you have a chance of precipitating a “hypertensive crisis”, which is exactly as fun as it sounds. As a result, people who are already miserable and already starving themselves are told they can’t eat like half of food. And once again, if you tell people “Eat these foods with this drug and you die” and a week later the person wants to kill themselves and has some cheese in the house, then you’re back to enabling suicide. There are some MAOIs that get around these restrictions in various clever ways, but they tend to be less effective.

SSRIs were the first class of antidepressants that mostly avoided these problems and so were pretty well-placed to launch a prescribing explosion even apart from being pushed by Big Pharma.

3. There is large publication bias in the antidepressant literatureTrue, but not as important as some people think

People became more aware of publication bias a couple of years after serious research into antidepressants started, and it’s not surprising that these were a prime target. When this issue rose to scientific consciousness, several researchers tried to avoid the publication bias problem by using only FDA studies of antidepressants. The FDA mandates that its studies be pre-registered and the results reported no matter what they are. This provides a “control group” by which accusations of publication bias can be investigated. The results haven’t been good. From Gibbons et al:

Recent reports suggest that efficacy of antidepressant medications versus placebo may be overstated, due to publication bias and less efficacy for mildly depressed patients. For example, of 74 FDA-registered randomized controlled trials (RCTs) involving 12 antidepressants in 12,564 patients, 94% of published trials were positive whereas only 51% of all FDA registered studies were positive.

Turner et al express the same data a different way:

. The FDA deemed 38 of the 74 studies (51%) positive, and all but 1 of the 38 were published. The remaining 36 studies (49%) were deemed to be either negative (24 studies) or questionable (12). Of these 36 studies, 3 were published as not positive, whereas the remaining 33 either were not published (22 studies) or were published, in our opinion, as positive (11) and therefore conflicted with the FDA’s conclusion. Overall, the studies that the FDA judged as positive were approximately 12 times as likely to be published in a way that agreed with the FDA analysis as were studies with nonpositive results according to the FDA (risk ratio, 11.7; 95% confidence interval [CI], 6.2 to 22.0; P<0.001). This association of publication status with study outcome remained significant when we excluded questionable studies and when we examined publication status without regard to whether the published conclusions and the FDA conclusions were in agreement

The same source tells us about the effect this bias had on effect size:

For each of the 12 drugs, the effect size derived from the journal articles exceeded the effect size derived from the FDA reviews (sign test, P<0.001). The magnitude of the increases in effect size between the FDA reviews and the published reports ranged from 11 to 69%, with a median increase of 32%. A 32% increase was also observed in the weighted mean effect size for all drugs combined, from 0.31 (95% CI, 0.27 to 0.35) to 0.41 (95% CI, 0.36 to 0.45).

I think a lot of this has since been taken on board, and most of the rest of the research I’ll be talking about uses FDA data rather than published data. But as you can see, the overall change in effect size – from 0.31 to 0.41 – is not that terribly large.

4. The effect size of antidepressants is clinically insignificantDepends what you mean by “clinically insignificant”

As mentioned above, when you try to control for publication bias, the effect size of antidepressant over placebo is 0.31.

This number can actually be broken down further. According to McAllister and Williams, who are working off of slightly different data and so get slightly different numbers, the effect size of placebo is 0.92 and the effect size of antidepressants is 1.24, which means antidepressants have a 0.32 SD benefit over placebo. Several different studies get similar numbers, including the Kirsch meta-analysis that started this whole debate.

Effect size is a hard statistic to work with (albeit extremely fun). The guy who invented effect size suggested that 0.2 be called “small”, 0.5 be called “medium”, and 0.8 be called “large”. NICE, a UK health research group, somewhat randomly declared that effect sizes greater than 0.5 be called “clinically significant” and effect sizes less than 0.5 be called “not clinically significant”, but their reasoning was basically that 0.5 was a nice round number, and a few years later they changed their mind and admitted they had no reason behind their decision.

Despite these somewhat haphazard standards, some people have decided that antidepressants’ effect size of 0.3 means they are “clinically insignificant”.

(please note that “clinically insignificant” is very different from “statistically insignificant” aka “has a p-value less than 0.05.” Nearly everyone agrees antidepressants have a statistically significant effect – they do something. The dispute is over whether they have a clinically significant effect – the something they do is enough to make a real difference to real people)

There have been a couple of attempts to rescue antidepressants by raising the effect size. For example, Horder et al note that Kirsch incorrectly took the difference between the average effect of drugs and the average effect of placebos, rather than the average drug-placebo difference (did you follow that?) When you correct that mistake, the drug-placebo difference rises significantly to about 0.4.

They also note that Kirsch’s study lumps all antidepressants together. This isn’t necessarily wrong. But it isn’t necessarily right, either. For example, his study used both Serzone (believed to be a weak antidepressant, rarely used) and Paxil (believed to be a stronger antidepressant, commonly used). And in fact, by his study, Paxil showed an effect size of 0.47, compared to Serzone’s 0.21. But since the difference was not statistically significant, he averaged them together and said that “antidepressants are ineffective”. In fact, his study showed that Paxil was effective, but when you average it together with a very ineffective drug, the effect disappears. He can get away with this because of the arcana of statistical significance, but by the same arcana I can get away with not doing that.

So right now we have three different effect sizes. 1.2 for placebo + drug, 0.5 for drug alone if we’re being statistically merciful, 0.3 for drug alone if we’re being harsh and letting the harshest critic of antidepressants pull out all his statistical tricks.

The reason effect size is extremely fun is that it allows you to compare effects in totally different domains. I will now attempt to do this in order to see if I can give you an intuitive appreciation for what it means for antidepressants.

Suppose antidepressants were in fact a weight loss pill.

An effect size of 1.2 is equivalent to the pill making you lose 32 lb.

An effect size of 0.5 is equivalent to the pill making you lose 14 lb.

An effect size of 0.3 is equivalent to the pill making you lose 8.5 lb.

Or suppose that antidepressants were a growth hormone pill taken by short people.

An effect size of 1.2 is equivalent to the pill making you grow 3.4 in.

An effect size of 0.5 is equivalent to the pill making you grow 1.4 in.

An effect size of 0.3 is equivalent to the pill making you grow 0.8 in.

Or suppose that antidepressants were a cognitive enhancer to boost IQ. This site gives us some context about occupations.

An effect size of 1.2 is equivalent to the pill making you gain 18 IQ points, ie from the average farm laborer to the average college professor.

An effect size of 0.5 is equivalent to the pill making you gain 7.5 IQ points, ie from the average farm laborer to the average elementary school teacher.

An effect size of 0.3 is equivalent to the pill making you gain 5 IQ points, ie from the average farm laborer to the average police officer.

To me, these kinds of comparisons are a little more revealing than NICE arbitrarily saying that anything below 0.5 doesn’t count. If you could take a pill that helps your depression as much as gaining 1.4 inches would help a self-conscious short person, would you do it? I’d say it sounds pretty good.

5. The effect of antidepressants only becomes significant in the most severe depressionEverything about this statement is terrible and everyone involved should feel bad

So we’ve already found that saying antidepressants have an “insignificant” effect size is kind of arbitrary. But what about the second part of the claim – that they only have measurable effects in “the most severe depression”?

A lot of depression research uses a test called the HAM-D, which scores depression from 0 (none) to 52 (max). Kirsch found that the effect size of antidepressants increased as HAM-D scores increased, meaning antidepressants become more powerful as depression gets worse. He was only able to find a “clinically significant” effect size (d > 0.5) for people with HAM-D scores greater than 28. People have come up with various different mappings of HAM-D scores to words. For example, the APA says:

(0-7) No depression
(8-13) Mild depression
(14-18) Moderate depression
(19-22) Severe depression
(>=23) Very severe depression

Needless to say, a score of 28 sounds pretty bad.

We saw that Horder et al corrected some statistical deficiencies in Kirsch’s original paper which made antidepressants improve slightly. With their methodology, antidepressants reach our arbitrary 0.5 threshold around HAM-D score 26. Another similar “antidepressants don’t work” study got the number 25.

Needless to say, when anything over 23 is “very severe”, 25 or 26 still sounds pretty bad.

Luckily, people completely disagree on the meanings of basic words! Very Severely Stupid is a cute article on Neuroskeptic that demonstrates that five different people and organizations suggest five different systems for rating HAM-D scores. Bech 1996 calls our 26 cutoff “major”; Funakawa 2007 calls it “moderate”; NICE 2009 calls it “severe”. APA is unique in calling it very severe. NICE’s scale is actually the exact same as the APA scale with every category renamed to sound one level less threatening. Facepalm.

Ghaemi and Vohringer(2011) go further and say that the real problem is that Kirsch is using the standard for depressive symptoms, but that real clinical practice involves depressive episodes. That is, all this “no depression” to “severe” stuff is about whether someone can be diagnosed with depression; presumably the people on antidepressants are definitely depressed and we need a new model of severity to determine just how depressed they are. As they put it:

the authors of the meta-analysis claimed to use the American Psychiatric Association’s criteria for severity of symptoms…in so doing, they ignore the obvious fact that symptoms differ from episodes: the typical major depressive episode (MDE) produced HDRS scores of at least 18 or above. Thus, by using symptom criteria, all MDEs are by definition severe or very severe. Clinicians know that some patients meet MDE criteria and are still able to work; indeed others frequently may not even recognize that such a person is clinically depressed. Other patients are so severe they function poorly at work so that others recognize something is wrong; some clinically depressed patients cannot work at all; and still others cannot even get out of bed for weeks or months on end. Clearly, there are gradations of severity within MDEs, and the entire debate in the above meta-analysis is about MDEs, not depressive symptoms, since all patients had to meet MDE criteria in all the studiesincluded in the meta-analysis (conducted by pharmaceutical companies for FDA approval for treatment of MDEs).

The question, therefore, is not about severity of depressive symptoms, but severity of depressive episodes, assuming that someone meets DSM-IV criteria for a major depressive episode. On that question, a number of prior studies have examined the matter with the HDRS and with other depression rating scales, and the three groupings shown in table 2 correspond rather closely with validated and replicated definitions of mild (HDRS <24), moderate (HDRS 24–28), and severe (HDRS>28) major depressive episodes.

So, depending on whether we use APA criteria or G&V criteria, an HRDS of 23 is either “mild” (G&V) or “very severe” (APA).

Clear as mud? I agree that in one sense this is terrible. But in another sense it’s actually a very important point. Kirsch’s sample was really only “severe” in the context of everyone, both those who were clinically diagnosable with major depression and those who weren’t. When we get to people really having a major depressive episode, a score of 26 to 28 isn’t so stratospheric. But meanwhile:

The APA seem to have ignored the fact that the HAMD did not statistically significantly distinguish between “Severe” and “Moderate” depression anyway (p=0.1)

Oh. That gives us some perspective, I guess. Also, some other people make the opposite critique and say that the HAM-D can’t distinguish very well at the low end. Suppose HAM-Ds less than ten are meaningless and random. This would look a lot like antidepressants not working in mild depression.

Getting back to Ghaemi and Vohringer, they try a different tack and suggest that there is a statistical floor effect. They quite reasonably say that if someone had a HAM-D score of 30, and antidepressants solved 10% of their problem, they would lose 3 HAM-D points, which looks impressive. But if someone had a HAM-D score of 10, and antidepressants (still) solved 10% of their problem, they would only lose 1 HAM-D point, which sounds disappointing. But either way, the antidepressants are doing the same amount of work. If you adjust everything for baseline severity, it’s easy to see that antidepressants here would have the same efficacy in severe and mild depression, even though it doesn’t look that way at first.

I am confused that this works for effect sizes, because I expect effect sizes to be relative to the standard deviation in a sample. However, several important people tell me that it does, and that when you do this Kirsch’s effect size goes from 0.32 to 0.40.

(I think these people are saying the exact same thing, but so overly mathematically that I’ve been staring at it for an hour and I’m still not certain)

More important, Ghaemi and Vohringer say once you do this, antidepressants reach the magic 0.5 number not only in severe depression, but also in moderate depression. However, when I look at this claim closely, almost all the work is done by G&V’s adjusted scale in which Kirsch’s “very severe” corresponds to their “mild”.

(personal aside: I got an opportunity to talk to Dr. Ghaemi about this paper and clear up some of my confusion. Well, not exactly an opportunity to talk about it, per se. Actually, he was supposed to be giving me a job interview at the time. I guess we both got distracted. This may be one of several reasons I do not currently work at Tufts.)

So. In conclusion, everyone has mapped HAM-D numbers into words like “moderate” in totally contradictory ways, such that one person’s “mild” is another person’s “very severe”. Another person randomly decided that we can only call things “clinically significant” if they go above the nice round number of 0.5, then retracted this. So when people say “the effects of antidepressants are only clinically significant in severe depression”, what they mean is “the effects of antidepressants only reach a totally arbitrary number one guy made up and then retracted, in people whose HAM-D score is above whatever number I make up right now.” Depending on what number you choose and what word you make up to describe it, you can find that antidepressants are useful in moderate depression, or severe depression, or super-duper double-dog-severe depression, or whatever.

Science!

6. The beneficial effects of antidepressants are only noticed by doctors, not the patients themselvesPartly true but okay

So your HAM-D score has gone down and you’re no longer officially in super-duper double-dog severe depression anymore. What does that mean for the patient?

There are consistent gripes that antidepressant studies that use patients rating their own mood show less improvement than studies where doctors rate how they think a patient is doing, or standardized tests like the HAM-D.

Some people try to turn this into a conspiracy, where doctors who have somehow broken the double-blinding of studies try to report that patients have done better because doctors like medications and want them to succeed.

The reality is more prosaic. It has been known for forty years that people’s feelings are the last thing to improve during recovery from depression.

This might sound weird – what is depression except people’s feelings? But the answer is “quite a lot”. Depressed people often eat less, sleep more, have less energy, and of course are more likely to attempt suicide. If a patient gets treated with an antidepressant, and they start smiling more and talking more and getting out of the house and are no longer thinking about suicide, their doctor might notice – but the patient herself might still feel really down-in-the-dumps.

I am going to get angry comments from people saying I am declaring psychiatric patients too stupid to notice their own recovery or something like that, but it is a very commonly observed phenomenon. Patients have access to internal feelings which they tend to weight much more heavily than external factors like how much they are able to get done during a day or how many crying spells they have, sometimes so much so that they completely miss these factors. Doctors (or family members, or other outside observers) who don’t see these internal feelings, are better able to notice outward signs. As a result, it is pretty universally believed that doctors spot signs of recovery in patients long before the patients themselves think they are recovering. This isn’t just imaginary – it’s found it datasets where the doctors are presumably blinded and with good inter-rater reliability.

Because most antidepressant trials are short, a lot of them reach the point where doctors notice improvement but not the point where patients notice quite as much improvement.

7. The apparent benefits of antidepressant over placebo may be an “active placebo” effect rather than a drug effectUnlikely

Active placebo is the uncomfortable idea that no study can really have a blind control group because of side effects. That is, sugar pills have no side effects, real drugs generally do, and we all know side effects are how you know that a drug is working!

(there is a counterargument that placebos very often have placebo side effects, but most likely the real drug will at least have more side effects, saving the argument)

The solution is to use active placebo, a drug that has side effects but, as far as anyone knows, doesn’t treat the experimental condition (in this case, depression). The preliminary results from this sort of study don’t look good for antidepressants:

Thomson reviewed 68 double-blind studies of tricyclics that used an inert placebo and seven that used an active placebo (44). He found drug efficacy was demonstrated in 59% of studies that employed inert placebo, but only 14% of those that used active placebo (?2=5.08, df=1, p=0.02). This appears to demonstrate that in the presence of a side-effect-inducing control condition, placebo cannot be discriminated from drug, thus affirming the null hypothesis.

Luckily, Quitkin et al (2000) solve this problem so we don’t have to:

Does the use of active placebo increase the placebo response rate? This is not the case. After pooling data from those studies in which a judgment could be made about the proportion of responders, it was found that 22% of patients (N=69 of 308) given active placebos were rated as responders. To adopt a conservative stance, one outlier study (50) with a low placebo response rate of 7% (N=6 of 90) was eliminated because its placebo response rate was unusually low (typical placebo response rates in studies of depressed outpatients are 25%–35%). Even after removing this possibly aberrant placebo group, the aggregate response rate was 29% (N=63 of 218), typical of an inactive placebo. The active placebo theory gains no support from these data.

Closer scrutiny suggests that the “failure” of these 10 early studies to find typical drug-placebo differences is attributable to design errors that characterize studies done during psychopharmacology’s infancy. Eight of the 10 studies had at least one of four types of methodological weaknesses: inadequate sample size, inadequate dose, inadequate duration, and diagnostic heterogeneity. The flaws in medication prescription that characterize these studies are outlined in Table 3. In fact, in spite of design measurement and power problems, six of these 10 studies still suggested that antidepressants are more effective than active placebo.

In summary, these reviews failed to note that the active placebo response rate fell easily within the rate observed for inactive placebo, and the reviewers relied on pioneer studies, the historical context of which limits them.

In other words, active placebo research has fallen out of favor in the modern world. Most studies that used active placebo are very old studies that were not very well conducted. Those studies failed to find an active-placebo-vs.-drug difference because they weren’t good enough to do this. But they also failed to find an active-placebo-vs.-inactive-placebo difference. So they provide no support for the idea that active placebos are stronger than inactive placebos in depression and in fact somewhat weigh against it.

8. Antidepressants have much worse side effects than you were led to believeDepends how bad you were led to believe the side effects were

As discussed in Part 2, the biggest advantage of SSRIs and other new antidepressants over the old antidepressants was their decreased side effect profile. This seems to be quite real. For example, Brambilla finds a relative risk of adverse events on SSRIs only 60% of that on TCAs, p = 0.003 (although there are some conflicting numbers in that paper I’m not really clear about). Montgomery et al 1994 finds that fewer patients stop taking SSRIs than tricyclics (usually a good “revealed preference”-style measure of side effects since sufficiently bad side effects make you stop using the drug).

The charmingly named Cascade, Kalali, and Kennedy (2009) investigated side effect frequency in a set of 700 patients on SSRIs and found the following:

56% decreased sexual functioning
53% drowsiness
49% weight gain
19% dry mouth
16% insomnia
14% fatigue
14% nausea
13% light-headedness
12% tremor

However, it is very important to note that this study was not placebo controlled. Placebos can cause terrible side effects. Anybody who experiments with nootropics know that the average totally-useless inactive nootropic causes you to suddenly imagine all sorts of horrible things going on with your body, or attribute some of the things that happen anyway (“I’m tired”) to the effects of the pill. It’s not really clear how much of the stuff in this study is placebo effect versus drug effect.

Nevertheless, it is worth mentioning that 34% of patients declare side effects “not at all” or “a litte” bothersome, 40% “somewhat” bothersome, and 26% “very” or “extremely” bothersome. That’s much worse than I would have expected.

Aside from the sort of side effects that you expect with any drug, there are three side effects of SSRIs that I consider especially worrisome and worthy of further discussion. These are weight gain, sexual side effects, and emotional blunting.

Weight gain is often listed as one of the most common and debilitating effects of SSRIs. But amusingly, when a placebo-controlled double-blinded study was finally run, SSRIs produced less weight gain than placebo. After a year of pill-taking, people on Prozac had gained 3.1 kg; people on placebo had gained 4.3. There is now some talk of SSRIs as a weak but statistically significant agent for weight loss.

What happened? One symptom of depression is not eating. People get put on SSRIs when they’re really depressed. Then they get better, either because the drugs worked, because of placebo, or just out of regression to the mean. When you go from not eating to eating, you gain weight. In the one-year study, almost everyone’s depression remitted (even untreated depressive episodes rarely last a whole year), so everyone went from a disease that makes them eat less, to remission from that disease, so everyone gained weight.

Sexual side effects are a less sanguine story. Here the direction was opposite: the medical community went from thinking this was a minor problem to finding it near-universal. The problem was that doctors usually just ask “any side effects?”, and off Tumblr people generally don’t volunteer information about their penis or vagina to a stranger. When they switched to the closed-ended question “Are you having any sexual side effects?”, a lot of people who denied side effects in general suddenly started talking.

Numbers I have heard for the percent of people on SSRIs with sexual side effects include 14, 24, 37, 58, 59, and 70 (several of those come from here. After having read quite a bit of this research, I suspect you’ve got at least a 50-50 chance (they say men are more likely to get them, but they’re worse in women). Of people who develop sexual side effects, 40% say they caused serious distress, 35% some distress, and 25% no distress.

So I think it is fair to say that if you are sexually active, your chances with SSRIs are not great. Researchers investigating the topic suggest people worried about sexual side effects should switch to alternative sexual-side-effect-free antidepressant Serzone. You may remember that as the antidepressant that worked worst in the efficacy studies and brought the efficacy of all the other ones down with it. Also, it causes liver damage. In my opinion, a better choice would be bupropion, another antidepressant which has been found many times not to cause sexual side effects and which may even improve your sex life.

(“Bupropion lacks this side effect” is going to be a common theme throughout this section. Bupropion causes insomnia, decreased appetite, and in certain rare cases of populations at risk, seizures. It is generally a good choice for people who are worried about SSRI side effects and would prefer a totally different set of side effects.)

There is a certain feeling that, okay, these drugs may have very very common, possibly-majority-of-user sexual side effects, but depressed people probably aren’t screwing like rabbits anyway. So after you recover, you can wait the appropriate amount of time, come off the drugs (or switch to a different drug or dose for maintenance) and no harm done.

The situation no longer seems so innocuous. Despite a lack of systematic investigation, there are multiple reports from researchers and clinicians – not to mention random people on the Internet – of permanent SSRI-induced sexual dysfunction that does not remit once the drug is stopped. This is definitely not the norm and as far as we know it is so rare as to be unstudyable beyond the occasional case report.

On the other hand, I have this. I took SSRIs for about five to ten years as a kid, and now I have approximately the pattern of sexual dysfunction associated with SSRIs and consider myself asexual. Because I started the SSRIs too early to observe my sexuality without them, I can’t officially blame the drugs. But I am very suspicious. I feel like this provides moderate anthropic evidence that it is not as rare as everyone thinks.

The last side effect worth looking at is emotional blunting. A lot of people say they have trouble feeling intense emotions (sometimes: any emotions at all) when on SSRIs. Sansone and Sansone (2010) report:

As for prevalence rates, according to a study by Bolling and Kohlenberg, approximately 20 percent of 161 patients who were prescribed an SSRI reported apathy and 16.1 percent described a loss of ambition. In a study by Fava et al, which consisted of participants in both the United States and Italy, nearly one-third on any antidepressant reported apathy, with 7.7 percent describing moderate-to-severe impairment, and nearly 40 percent acknowledged the loss of motivation, with 12.0 percent describing moderate-to-severe impairment.

A practicing clinician working off observation finds about the same numbers:

The sort of emotional “flattening” I have described with SSRIs may occur, in my experience, in perhaps 10-20% of patients who take these medications…I do want to emphasize that most patients who take antidepressant medication under careful medical supervision do not wind up feeling “flat” or unable to experience life’s normal ups and downs. Rather, they find that–in contrast to their periods of severe depression–they are able to enjoy life again, with all its joys and sorrows.

Many patients who experience this side effect note that when you’re depressed, “experiencing all your emotions fully and intensely” is not very high on your list of priorities, since your emotions tend to be terrible. There is a subgroup of depressed patients whose depression takes the form of not being able to feel anything at all, and I worry this effect would exacerbate their problem, but I have never heard this from anyone and SSRIs do not seem less effective in that subgroup, so these might be two different things that only sound alike. A couple of people discussing this issue have talked about how decreased emotions help them navigate interpersonal relationships that otherwise might involve angry fights or horrible loss – which sounds plausible but also really sad.

According to Barnhart et al (2004), “this adverse effect has been noted to be dose-dependent and reversible” – in other words, it will get better if you cut your dose, and go away completely when you stop taking the medication. I have not been able to find any case studies or testimonials by people who say this effect has been permanent.

My own experience was that I did notice this (even before I knew it was an official side effect) that it did go away after a while when I stopped the medications, and that since my period of antidepressant use corresponded with an important period of childhood socialization I ended out completely unprepared for having normal emotions and having to do a delicate social balancing act while I figured out how to cope with them. Your results may vary.

There is also a large research on suicidality as a potential side effect of SSRIs, but this looks like it would require another ten thousand words just on its own, so let’s agree it’s a risk and leave it for another day.

9. Therefore, we should give up on medication and use psychotherapy insteadMakes sense right up until you run placebo-controlled trials of psychotherapy

The conclusion of these studies that claim antidepressants don’t outperform placebo is usually that we should repudiate Big Pharma, toss the pills, and go back to using psychotherapy.

The implication is that doctors use pills because they think they’re much more effective than therapy. But that’s not really true. The conventional wisdom in psychiatry is that antidepressants and psychotherapy are about equally effective.

SSRIs get used more than psychotherapy for the same reason they get used more than tricyclics and MAOIs – not because they’re better but because they have fewer problems. The problem with psychotherapy is you’ve got to get severely mentally ill people to go to a place and talk to a person several times a week. Depressed people are not generally known for their boundless enthusiasm for performing difficult tasks consistently. Also, Prozac costs like 50 cents a pill. Guess how much an hour of a highly educated professional’s time costs? More than 50c, that’s for sure. If they are about equal in effectiveness, you probably don’t want to pay extra and your insurance definitely doesn’t want to pay extra.

Contrary to popular wisdom, it is almost never the doctor pushing pills on a patient who would prefer therapy. If anything it’s more likely to be the opposite.

However, given that we’re acknowledging antidepressants have an effect size of only about 0.3 to 0.5, is it time to give psychotherapy a second look?

No. Using very similar methodology, a team involving Mind The Brain blogger James Coyne found that psychotherapy decreases HAM-D scores by about 2.66, very similar to the 2.7 number obtained by re-analysis of Kirsch’s data on antidepressants. It concludes:

Although there are differences between the role of placebo in psychotherapy and pharmacotherapy research, psychotherapy has an effect size that is comparable to that of antidepressant medications. Whether these effects should be deemed clinically relevant remains open to debate.

Another study by the same team finds psychotherapy has an effect size of 0.22 compared to antidepressants’ 0.3 – 0.5, though no one has tried to check if that difference is statistically significant and this does not give you the right to say antidepressants have “outperformed” psychotherapy.

If a patient has the time, money, and motivation for psychotherapy, it may be a good option – though I would only be comfortable using it as a monotherapy if the depression was relatively mild.

10. Further complications

What if the small but positive effect size of antidepressants wasn’t because they had small positive effects on everyone, but because they had very large positive effects on some people, and negative effects on others, such that it averaged out to small positive effects? This could explain the clinical observations of psychiatrists (that patients seem to do much better on antidepressants) without throwing away the findings of researchers (that antidepressants have only small benefits over placebo) by bringing in the corollary that some psychiatrists notice some patients doing poorly on antidepressants and stop them in those patients (which researchers of course would not do).

This is the claim of Gueorguieva and Krystal 2011, who used “growth modeling” to analyze seven studies of new-generation-antidepressant Cymbalta and found statistically significant differences between two “trajectories” for the drug, but not for placebo. 66% of people were in the “responder” trajectory and outperformed placebo by 6 HAM-D points (remember, previous studies estimated HAM-D benefits over placebo at about 2.7). 33% of people were nonresponders and did about 6 HAM-D points worse than placebo. Average it out, and people did about 3 HAM-D points better on drug and placebo, pretty close to the previous 2.7 point estimate.

I don’t know enough about growth modeling to be sure that the researchers didn’t just divide the subjects into two groups based on treatment efficacy and say “Look! The subsection of the population whom we selected for doing well did well!” but they use many complicated statistics words throughout the study that I think are supposed to indicate they’re not doing this.

If true, this is very promising. It means psychiatrists who are smart enough to notice people getting worse on antidepressants can take them off (or switch to another class of medication) and expect the remainder to get much, much better. I await further research with this methodology.

What if there were actually no such thing as the placebo effect? I know dropping this in around the end of an essay that assumes 75% of gains related to antidepressants are due to the placebo effect is a bit jarring, but it is the very-hard-to-escape conclusion of Hróbjartsson and Gøtzsche’s meta-analysis on placebo. They find that three-armed studies – ie those that have a no-treatment group, a placebo-treatment group, and a real-drug-treatment group – rarely find much of a difference between no-treatment and placebo. This was challenged by Wampold et al here and here, but defended against those challenges by the long-name-Scandinavian-people here. Kirsch, who between all his antidepressant work is still Associate Director of Placebo Studies, finds here that 75% of the apparent placebo effect in antidepressant studies is probably a real placebo effect, but his methodology is a valiant attempt to make the most out of a total lack of data rather than a properly-directed study per se.

If placebo pills don’t do much, what explains the vast improvements seen in both placebo and treatment groups in antidepressant trials? It could be the feeling of cared-for-ness and special-ness of getting to see a psychiatrist and talk with her about your problems, and the feeling of getting-to-contribute-something you get from participating in a scientific study. Or it could just be regression to the mean – most people start taking drugs when they feel very depressed, and at some point you have nowhere to go but up. Most depression gets better after six months or so – which is a much longer period than the six week length of the average drug trial, but maybe some people only volunteered for the study four months and two weeks after their depression started.

If Hróbjartsson and Gøtzsche were right, and Kirsch and the psychiatric establishment wrong, what would be the implications? Well, the good implication is that we no longer have to worry about problem 7 – that antidepressants are merely an active placebo – since active placebos shouldn’t do anything. That means we can be more confident they really work. The more complicated implication is that psychiatrists lose one excuse for asking people to take the drugs – “Sure, the drug effect may be small, but the placebo effect is so strong that it’s still worth it.” I don’t know how many psychiatrists actually think this way, but I sometimes think this way.

What if the reason people have so much trouble finding good effects from antidepressants is that they’re giving the medications wrong? Psychiatric Times points out that:

The Kirsch meta-analysis looked only at studies carried out before 1999. The much-publicized Fournier study examined a total of 6 antidepressant trials (n=718) using just 2 antidepressants, paroxetine and imipramine. Two of the imipramine studies used doses that were either subtherapeutic (100 mg/day) or less than optimal (100 to 200 mg/day)

What if we’ve forgotten the most important part? Antidepressants are used not only to treat acute episodes of depression, but to prevent them from coming back (maintenance therapy). This they apparently do very well, and I have seen very few studies that attempt to call this effect into question. Although it is always possible that someone will find the same kind of ambiguity around maintenance antidepressant treatment as now clouds acute antidepressant treatment, so far as far as I know this has not happened.

What if we don’t understand what’s going on with the placebo effect in our studies? Placebo effect has consistently gotten stronger over the past few decades, such that the difference between certain early tricyclic studies (which often found strong advantages for the medication) and modern SSRI studies (which often find only weak advantages for the medication) is not weaker medication effect, but stronger placebo effect (that is, if medication always has an effect of 10, but placebo goes from 0 to 9, apparent drug-placebo difference gets much lower). Wired has a good article on this. Theories range from the good – drug company advertising and increasing prestige and awareness of psychiatry have raised people’s expectations of psychiatric drugs – to the bad – increasing scientific competence and awareness have improved blinding and other facets of trial design – to the ugly – modern studies recruit paid participants with advertisements, so some unscrupulous people may be entering studies and then claiming to get better, hoping that this sounds sufficiently like the outcome the researchers want that everyone will be happy and they’ll get their money on schedule.

If placebos are genuinely getting better because of raised expectations, that’s good news for doctors and patients but bad news for researchers and drug companies. The patient will be happy because they get better no matter how terrible a prescribing decision the doctor makes; the doctor will be happy because they get credit. But for researchers and drug companies, it means it’s harder to prove a difference between drug and placebo in a study. You can invent an excellent new drug and still have it fail to outperform placebo by very much if everyone in the placebo group improves dramatically.

Conclusion

An important point I want to start the conclusion section with: no matter what else you believe, antidepressants are not literally ineffective. Even the most critical study – Kirsch 2008 – finds antidepressants to outperform placebo with p < .0001 significance. An equally important point: everyone except those two Scandinavian guys with the long names agree that, if you count the placebo effect, antidepressants are extremely impressive. The difference between a person who gets an antidepressant and a person who gets no treatment at all is like night and day. The debate takes place within the bounds set by those two statements. Antidepressants give a very modest benefit over placebo. Whether this benefit is so modest as to not be worth talking about depends on what level of benefits you consider so modest as to not be worth talking about. If you are as depressed as the average person who participates in studies of antidepressants, you can expect an antidepressant to have an over-placebo-benefit with an effect size of 0.3 to 0.5. That's the equivalent of a diet pill that gives you an average weight loss of 9 to 14 pounds, or a growth hormone that makes you grow on average 0.8 to 1.4 inches. You may be able to get more than that if you focus on the antidepressants, like paroxetine and venlafaxine, that perform best in studies, but we don't have the statistical power to say that officially. It may be the case that most people who get antidepressants do much better than that but a few people who have paradoxical negative responses bring down the average, but right now this result has not been replicated. This sounds moderately helpful and probably well worth it if the pills are cheap (which generic versions almost always are) and you are not worried about side effects. Unfortunately, SSRIs do have some serious side effects. Some of the supposed side effects, like weight gain, seem to be mostly mythical. Others, like sexual dysfunction, seem to be very common and legitimately very worrying. You can avoid most of these side effects by taking other antidepressants like bupropion, but even these are not totally side-effect free. Overall I think antidepressants come out of this definitely not looking like perfectly safe miracle drugs, but as a reasonable option for many people with moderate (aka "mild", aka "extremely super severe") depression, especially if they understand the side effects and prepare for them.

This entry was posted in Uncategorized and tagged , , , . Bookmark the permalink.

129 Responses to SSRIs: Much More Than You Wanted To Know

  1. Greyfox says:

    You wander through the Internets to quill your thirst for data and knowledge and come across much “meh” and talk aimed at the intellectually challenged. Than there is this nugget of gold like your blog here of mediocre fame. Seriously, why does stuff like this have the quality of a dusty book you find hidden in the commode of an old lady living in the backstreet of a small remote village? The world is not fair, but I sincerely value your work and thoughts, so here is my “good job” comment of appreciation.

  2. AntiSocialite says:

    Dear author (and others)

    I just wanted to ask what you can make of my situation and whether or not I should seek professional help. Also I’d be interested in knowing whether antidepressants may or may not alleviate my “symptoms” (or whatever the heck’s going on with me)

    I was definitely depressed from the age of 14 to 17, had zero self-esteem, hated myself, saw no light at the end of the tunnel and was basically an emotional wreck. Much of this originated from my need for female validation which I’ve freed myself from (I opted out of the “dating market” and live in voluntary celibacy). Other issues such as acne and borderline obesity I’ve freed myself from as well due to roaccutane, diet and exercising.

    From 17 to 25 I’ve had occasional relapses and especially annoying symptoms such as insomnia and lethargy. But what has as of late become a more shocking problem for me is a deep seated anger at each and everyone around me. Normally I can control this (except for the random angry/hateful though popping up) but when I’m drunk, well it all goes to the shitter … I literally verbally and physically lash out against whoever comes into my field of vision. I’m lucky to have avoided assault charges so far and of course I’ve made the decision to avoid drinking as much as possible, especially places with an abundance of alcoholic drinks (such as night clubs and bars).
    But even when sober when someone pisses me off I can easily imagine doing violent things to that person and it feels very satisfying to me. Now don’t get me wrong, I realize that the consequences of such an action would be severe and I’d never actually do it (well unless I was blackout drunk) but it’s still a little scary at times. The thing is of course that anger also feels powerful and I enjoy the feeling and have somewhat grown attached to it – might be a reason why I can’t let it go.

    I don’t know I guess that instead of getting better I feel like I’m getting worse the older I get. Physically not really, I’d say I’m in (on of the) best shape in my life, but mentally … well, I’m just getting more and more antisocial. Interestingly that’s exactly what personality tests spit out: Anti-social personality disorder. Meh, I don’t know if that’s true though – if it’s a spectrum, then yes, if not, then definitely no – I still feel guilt, remorse and all that emotional mambo jambo.

  3. Anonymous says:

    I wasn’t going to mention this because your article is about SSRIs and I thought this meant that discussing other types of anti-depressants would be off-topic. Then I saw that you mentioned Cymbalta. Since you have apparently chosen to include other anti-depressants as well, you I thought you might be interested in my response to this:

    “There is a subgroup of depressed patients whose depression takes the form of not being able to feel anything at all, and I worry this effect would exacerbate their problem, but I have never heard this from anyone and SSRIs do not seem less effective in that subgroup, so these might be two different things that only sound alike.”

    The depression did not take the form of not being able to feel anything at all, but it did (before medication) result in me feeling less intensely. I tried Cymbalta for about a month. At first, it was great and it made me feel very happy. Then, I began having side-effects and could barely feel. I noticed that it was impossible to feel excited, and that my inability to even slightly excite myself made it impossible to feel motivated. I stopped caring about things completely.

    Examples: One of the most difficult things for me was getting out of bed to go to work in the morning. This became much harder after the Cymbalta side-effects began. Before Cymbalta, I at least had a little bit of anxiety that created an aversion to getting fired. This was just enough to make me get out of bed in time to go to work. After one month on Cymbalta, that anxiety was gone. I just didn’t care at all. I don’t remember how I managed to make myself go to work despite that when I first noticed this, but I realized that every day that I let this continue, there was a significant risk of me simply not getting out of bed. I also realized that because I had lost my ability to motivate myself with anxiety and excitement, I had become incapable of getting myself to work on any of the things that would improve my life situation.

    So, in addition to Cymbalta causing severe emotional blunting that I believe posed a risk to my livelihood, Cymbalta caused severe apathy that made me incapable of doing the extra work it takes to build a meaningful and desirable life. I also experienced intolerable gastrointenstinal symptoms. For those reasons, I quit Cymbalta. Even without the intolerable gastrointestinal symptoms, I would have concluded that Cymbalta was too detrimental for me to continue using it. Even if getting out of bed had been easy enough to be sure that I’d do it consistently, I saw the irony involved in taking a pill that perpetuated the problem it was supposed to help me solve. An anti-depressant that makes me too apathetic to bother with building a more worthwhile life is not serving it’s purpose and is therefore unacceptable to me. Now you have heard of at least one person who experienced severe, potentially debilitating apathy on anti-depressants, and found that specifically Cymbalta caused severe apathy as a side-effect by exacerbating it.

  4. Pingback: The (depression) drugs don’t work | Philosophies of a Disenchanted Scholar

  5. Anthony David Adams says:

    Curious if you’ve analyzed the research suggesting that gratitude journaling / writing gratitude letters is just as effective as prozac. Would be curious to see your ideas on that.

  6. Pingback: Increase Longevity And Intelligence With Boosted Klotho Levels | Rockstar Research Magazine

  7. Pingback: Friday links | Meta Rabbit

  8. Eva H. says:

    Just want to say thank you for a fantastic post.

  9. Pingback: Monday Miscellany: Cotard’s CBT, Clients | Gruntled & Hinged

  10. Oligopsony says:

    Is using Art or whatever to process unwanted emotions healthy, or is it just a high-status excuse to engage in rumination?

  11. Pingback: Cavalcade of Mammals » Blog Archive » Links for July 2014

  12. Pingback: Depression, symptoms and definition | Aceso Under Glass

  13. Pingback: Links 7/12/14 | Mike the Mad Biologist

  14. druc says:

    Do you have any hope that one day you will be able to reverse your SSRI-induced sexual dysfunction?

  15. Fronken says:

    On the other hand, I have this. I took SSRIs for about five to ten years as a kid, and now I have approximately the pattern of sexual dysfunction associated with SSRIs and consider myself asexual. Because I started the SSRIs too early to observe my sexuality without them, I can’t officially blame the drugs. But I am very suspicious. I feel like this provides moderate anthropic evidence that it is not as rare as everyone thinks.

    So … this has been bugging me. You’re poly … and ace? At the same time.

    Is that … I suppose I can conceive of various possibilities, but I’m just going to ask.

    How?

  16. Pingback: Is Depression Really on The Rise?

  17. Pingback: Antidepressant vs placebo | mathew

  18. Lance Bush says:

    Hi Scott, great post, thanks for sharing this. I wanted to report on my own experience with taking an SNRI which you mention here, as I have always suspected long-term/permanent side effects of SSRIs/SNRIs are more common that is recognized in by the medical community. I am about 28 now, and was prescribed effexor when I was 14, though I no longer recall the dosage. After being on Effexor for about two weeks, I appeared to have developed the following side effects:

    1. Complete loss of sexual function and major reduction in libido
    2. Extreme emotional blunting – my emotional states rarely moved far from completely flat affect
    3. Tinnitus in my right ear
    4. Recurrent bouts of depersonalization
    5. I began sleeping 10+ hours a day, rather than my normal amount (~7-8 hours).

    Incidentally, it was completely unhelpful with my depression, and if anything I felt significantly more depressed, in addition to having a foggy, confused state of mind both while on the medication and for several weeks afterwards.

    I eventually discontinued the medication, but as I began to wean myself off of it I began having extremely frequent “brain-zaps” occurring several times a day along with extreme nausea. It was one of, if not the most agonizing few weeks of my life, as I was constantly miserable all day every day for days and days without end.

    Here’s where I’m at today:

    The “brain zaps” lasted for about 3 and a half years, slowly declining in frequency until they went away completely. However, 15 years later, and I still have tinnitus. I still have almost total loss of sexual function, lower libido, aorgasmia, and reduced sexual sensitivity. I had an even stranger possible reaction though, and not one I’ve read about having occurred in others. After I stopped taking effexor, I never developed facial hair, muscle tone, or significant body hair.

    At 19, I discovered that I had hyponadism; a pituitary tumor was ruled out as unlikely after two MRIs, and I do not have Kallmann syndrome, nor have I ever had any significant head injuries. I currently take hCG, and while this has increased my libido and secondary male sexual characteristics (facial hair, etc.) it had at best little influence on sexual function and capacity for orgasms, and appears to have left my libido largely undirected – I seem to find a much smaller proportion of the population attractive than most people.

    I think it is at least *possible* that the Effexor caused some sort of permanent disruption in my endocrine system, although I’ve never heard of this happening even in case studies (I have never looked). I have suggested the possible role of Effexor in causing my hypogonadism to two urologists and three endocrinologists, along with the other side effects – all of them dismissed this possibility outright, sagely informing me that “SSRIs do not cause permanent side effects.”

    The lack of clinical curiosity in the cause of my conditions is a bit puzzling, and I wonder if the number of people who have permanent side effects is under reported in part because those clinicians tell patients who suggest this possibility that this doesn’t occur, and those clinicians fail to catalog these subjects as possible candidates.

  19. Troy says:

    OT request for Scott: in the wake of the Hobby Lobby case many people on the Internets have been arguing over whether the contraceptives Hobby Lobby won’t cover are in fact abortifacients or not. As far as I can tell (which isn’t very far), the debate is over whether the contraceptives work only by preventing implantation, or whether in some cases (for all we know?) they work by preventing fertilization as well. Basically I don’t trust any sources I can find online on this question, and so I’m wondering if you’re able to weigh in (either in response to this comment or in a post), as I consider you a relatively impartial judge.

    • Anonymous says:

      A prerequisite for the medical question is identifying the contraceptives at issue. The ones that Hobby Lobby objected to are: Plan B, Ella, and copper IUDs.

      Why do you care whether they qualify as abortifacients under some particular definition that other people care about? Shouldn’t you first figure out what you care about and only later apply it to particular drugs? You have the debate backwards: people think preventing implantation is worse than preventing fertilization. Note that implantation follows fertilization. But if you don’t know this, you don’t seem to care about the difference.

      A few comments to bound the situation:

      The first two are emergency contraceptives. Often they are taken after fertilization, so if they continue to work in those circumstances, it must be by preventing implantation. But they are sufficiently unreliable that it is possible that they only stop fertilization.

      Copper IUDs can also be used for emergency contraception and are so reliable that they must be stopping implantation. But this is quite rare. The normal use is ahead of time. If, as is commonly believed, they stop fertilization in normal use, then there is no opportunity to stop implantation. Also Hobby Lobby objected to another product, hormonal IUDs. Those are never used as emergency contraception, which is evidence that they doctors really believe it when they say that they do not prevent implantation.

      • anon1 says:

        Nitpick: it’s evidence that doctors don’t believe hormonal IUDs are a reliable option for preventing implantation. It would still be plausible (based only on their lack of usage, since I’m too lazy to look for actual studies) that they prevent implantation but do so less effectively than other options.

      • Troy says:

        I mixed up my words in my initial post, which significantly altered its meaning. I meant to say that the debate was over whether they always worked by preventing fertilization or whether they sometimes prevented implantation as well.

  20. ifonly says:

    In most psychiatric research one group objectifies or ‘thing-izes’ another, then manipulates exquisitely perceptive, responsive, creative minds without reference to meaning, revelation, dialogue, or love. Do we want to be I & Thou to one another or I & It?
    Cf. youtube, Daniel Mackler critique of psychotherapy (well-informed advice on finding effective therapy)
    madnessradio.net (esp. 9-1-11 broadcast with David Webb on the meaning of suicidal ideas as an impetus to transform one’s life)
    madinamerica.com
    Anatomy of an Epidemic by Robert Whitaker

  21. Paul Torek says:

    How about recurrence? How does the recurrence of depression among the SSRI-treated compare to that among the spontaneously recovering, or the talk therapy-treated? (Self-selection is an issue, especially for the spontaneously recovering, so randomized assignment would be nice – which sounds unethical for spontaneous recovery since that’s equivalent to non-treatment. But maybe you could compare before/after drug treatments are available. Still, I’d like to see that data.)

  22. Pingback: Linkblogging For 09/07/14 | Sci-Ence! Justice Leak!

  23. Kaj Sotala says:

    So it makes me feel like I’m committing a mild sin for posting anecdotes on what’s a well-researched piece on rigorous statistical trials… but it just happens that I’ve been on SSRIs for one week now, and seeing this post pop up just now means I have to share my experience so far.

    There have been two major improvements: an increased reserve of mental energy, and an elevation of my general mood. As an example of the former: I work part-time, two days of the week. It used to be that work would leave me feeling drained, so that I wouldn’t have the energy for pretty much anything on the evenings after work, nor for two or so days afterwards. But now my experience is that I don’t even feel particularly drained right after leaving work. When I do take my lunch break at work, I feel recharged and ready to do more, rather than having difficulties refocusing afterwards. It actually feels like I’d be on vacation, because work has just stopped being stressful.

    I also used to have days when I wasn’t at all motivated to do anything: everything just felt meh. I haven’t really had those since starting the meds.

    And I just generally feel better. Previously, my happiness setpoint hovered somewhere between mildly negative and mildly positive; now I tend to generally feel good, or at least neutral.

    Some of the improvement is probably a direct effect of the meds, some of it comes indirectly. Having more mental energy means that I’m capable of doing more exercise and meditation, both of which also help improve my mood. Because my mood is pretty good in general, I’m not as desperate for any immediate short-term reward: instead of compulsively checking social media, I’m more capable of doing things like reading novels that requires more of a long-term focus but is ultimately more gratifying.

    And also because I feel better, I get less anxiety in social situations, making them more enjoyable. I also notice that I’m now more capable of feeling empathy towards other people and focusing on them, instead of worrying about what they think of me. If I just put in a little effort towards consciously fostering my sense of empathy, then just seeing strangers on the street makes me feel more happy.

    As for side effects from the drugs, I haven’t noticed a single one.

    I’m not sure of what’s going to happen once the time comes to quit the meds, and I’m not particularly keen to find out. But I’ll do my best to use this time to build up new habits that should help keep my mind in good shape.

    • a person says:

      This might be a dumb question, but were you prescribed SSRIs for depression, or are you taking them for some other reason? The tone of your post makes it sound like you went from normal to above average, whereas I would expect the typical person who just got put on SSRIs to go from horrible to “maybe things will get better after all”.

      Of course if you don’t want to answer, don’t feel obligated.

      • Kaj Sotala says:

        The psychiatrist actually didn’t tell me any specific diagnosis, but the original reason why I went to see him was that I was having major problems staying focused on anything. I was fine as long as the thing in question was easy to do or I was excited about it, but once I got to the part where I actually needed to consciously force myself to focus on it, I started having major problems. And I don’t mean just the ordinary “oh, I’m having issues focusing on boring stuff” level of difficulty, I mean a “I’ve been at work for seven hours today and I only managed to work for maybe fifteen minutes in total over the whole day” level of difficulty (this actually happened).

        This was likely related to the following description that I wrote some time before realizing that I should maybe talk to a psychiatrist:

        It feels that, large parts of the time, my mind is constantly looking for an escape, though I’m not entirely sure what exactly it is trying to escape _from_. But it wants to get away from the current situation, whatever the current situation happens to be. To become so engrossed in something that it forgets about everything else.

        Unfortunately, this often leads to the opposite result. My mind wants that engrossment right now, and if it can’t get it, it will flinch away from whatever I’m doing and into whatever provides an immediate reward. Facebook, forums, IRC, whatever gives that quick dopamine burst. That means that I have difficulty getting into books, TV shows, computer games: if they don’t grab me right away, I’ll start growing restless and be unable to focus on them. Even more so with studies or work, which usually require an even longer “warm-up” period before one gets into flow.

        Worse, I’m often sufficiently aware of that discomfort that my awareness of it prevents the engrossment. I go loopy: I get uncomfortable about the fact that I’m uncomfortable, and then if I have to work or study, my focus is on “how do I get rid of this feeling” rather than on “what should I do next in this project”. And then my mind keeps flinching away from the project, to anything that would provide a distraction, on to Facebook, to IRC, to whatever. And I start feeling worse and worse.

        (Some of you are probably thinking that this sounds like ADD: but all the sources that I’ve looked at say that ADD cannot develop in adulthood, and I’m completely certain that I didn’t yet have this as a child.)

        I also had more general problems with maintaining motivation or having the energy to do anything, would typically spend a day or two each week just feeling miserable and being unable to do pretty much anything other than laying in bed or being obsessively on Facebook, etc.

        I guess the general problem was that my level of mental resilience was really low: any stress or setback would easily just end up paralyzing me. This was making it really difficult for me to hold a job or make progress in my studies, and I was starting to fear that I’d never be capable of feeling not-miserable in any job for more than a few months.

        • a person says:

          Wow, everything you said also describes me. Maybe I should get a prescription for SSRIs. :\

          Thanks for the reply, I didn’t know SSRIs were used for things like that, very interesting. Good luck with everything.

        • Kaj Sotala says:

          Thanks, and good luck. 🙂

        • CAE_Jones says:

          Anecdotally: I tried Prozac for what sounds like the same situation, and didn’t really notice a difference (this was over the course of maybe 6-8 months; I forget exactly when it ended). There may have been other changes I didn’t really notice (as mentioned in the post), but I didn’t really feel like it helped much on the Akrasia-from-Hell front. Your mileage may vary, of course.

        • Anonymous says:

          Have you tried burnout as an explanation for this? One of the main symptoms of burnout is difficulty concentrating. If you seem to be excessively sensitive to burnout, you could also consider fatigue to be a symptom. There are over 100 potential causes of fatigue, but some of them are as simple as a vitamin deficiency.

  24. Tim says:

    It’s worth mentioning another possible side effect of anti-depressants: Drastically higher risk of developing type 2 diabetes. I’ve experienced this firsthand. I have severe, severe depression. I am near my ideal weight, eat healthily, and exercise (though I could stand to do a bit more.) After a decade on SSRIs I became borderline diabetic. I was often lightheaded with even less energy than normal. When I quit the medication, my blood glucose returned to normal and symptoms disappeared. No one in my family has ever had diabetes. Maybe it’s my depression talking, but whatever marginal relief I experienced on the meds was certainly not worth getting another chronic disease.

  25. M. Mackey says:

    The Mrs was on an MAOI many years ago, and it kinda worked, but the side effects weren’t great. The SSRI she’s on now makes her feel “normal.” Having been with her for about 15 years, I have to say that, given the severe depression she’s been fighting, they work. And given how much trouble she had finding anything that did work before SSRIs came along, I’m dubious that a placebo effect is the main cause.

  26. Glenn says:

    I don’t have time to read all the comments to see if anyone else has mentioned this, but…with regard to the “active placebo” possibility: There is a long history of SSRI testing on animals, both depressed and non-depressed, and most found anti-depressant effects of various kinds. Active placebo effects should be entirely absents in these cases. The usual caveats about publication bias apply, of course.

    Also, I’m guessing that when they do studies on depressed rats, the degree of depression they induce in them would count as very, very severe.

  27. Kaminiwa says:

    Huh. I was on SSRIs when I was 15-16 for about a year. I was celibate until 24, and I’d consider myself asexual until I was 28 and suddenly realized what all the fuss was about (there’s still only two people I’m really “in to” in that way)

    Conversely, I suspect I had a natural inclination to asexuality, and it’s probably notably *improved* my life.

    I can also speak to the apathy: soon before I refused to keep taking them, I was laying in bed thinking that it was very *peculiar* that I was being actively abused and didn’t seem to particularly mind this. Even flunking a few classes for a complete failure to attend them didn’t really bother me. I’ve never had apathy that strong in my life.

  28. Katie says:

    “I am going to get angry comments from people saying I am declaring psychiatric patients too stupid to notice their own recovery or something like that…”

    Not from me. When I was prescribed my current regimen (SNRI + Abilify), the friend I was living with was the first to notice the positive effects. Depression was putting a huge strain on our relationship and every other human relationship in my life. The changes in my behavior that were too subtle for me to notice made a HUGE difference to him and significantly improved our relationship, which made a HUGE difference to me. They gave him hope for our friendship, which somewhat relieved my fear that I was going to lose everyone important to me.

  29. File 13 says:

    Around the time I became very depressed, I also suffered symptoms of drowsiness, fatigue, weight-gain and light-headedness. I began taking fluoxetine (prozac), but suffered from nausea so switched to citalopram.

    I think it’s helped me a lot.
    Much more importantly, my wife thinks it’s helped me a LOT.

    I still suffer from drowsiness, fatigue, weight-gain and light-headedness, but those symptoms seem no worse now than when I was depressed, so I feel that it would be facetious of me to cry “side-effects”. My 2c.

  30. NeshSelvar says:

    I’d still bet I could get ‘clinically significant’ results that exceed the placebo effect(ie less then 50% of gains related to placebo effect) in decreasing depression by injecting combinations of short acting direct mono-amine agonists, with different receptor activity profiles, into the cerebral spinal fluid, until a good combination was found and then switch to a long term combination with the same profiles.
    (side effects may include basically any psychological state
    that can co-exist with the definition of not ‘depressed’)

    But more feasibly, it would be nice if psychiatrists at least laid out the range options with pros and cons, like Scott does here, to their patients.

    • anon1 says:

      What’s your plan for preventing tolerance here? I don’t know much about monoamine agonists in general, but 5HT-2A agonists produce a tolerance very quickly.

      • NeshSelvar says:

        A substantial part of tolerance for most monoamine drugs is down-regulation in production of endogenous dopamine, serotonin or nor-epinephrine . Most drugs used to alter these systems are re-uptake inhibitors or releasing agents, but for direct agonists this doesn’t apply since endogenous chemical are not central to there action. While receptor based tolerance is a issue, 5HT-2A is a exception. 5HT-2A builds and loses tolerance at a much faster rate then most. I don’t remember the exact source with me but some studies on the tolerance to the perceptual effects of psychedelics (mediated in part by 5HT-2A) can more then double over a few days and go back down in almost as short as time. Most other receptors/effects seem to be able to maintain increased total activity.

    • Phil Goetz says:

      Getting long-term, repeated injections into your cerebral spinal fluid might be worse than depression.

  31. Anonymous says:

    Sometimes they work and sometimes they don’t. Sometimes they are miraculous and sometimes they are useless. Then there is me. I have ridiculously horrible reactions to them. They make things worse with me and destroyed my life. I have been on several and now am on none and am depressed as hell but at least I am stable and predictable. Friends and coworkers were on the same ones that I was to different effect. For example, Welbutrin was my ex’s chill pill and a coworker stopped taking hers because while it worked, it made her so somnolent that driving became dangerous. Me, on the other hand, Welbutrin made me psycho and physically violent.

    So it all depends on the unique individual I think. For some strange reason what actually helps me is phenergan.

  32. Chamomile Geode says:

    Aren’t SSRI’s sometimes used specifically for the sexual side effects, especially in institutional settings? Like, patients with dementia are getting handsy with the nurses, so they’e prescribed SSRI’s? Th

    • Scott Alexander says:

      I’ve never seen that happen and the sexual side effects aren’t nearly certain nor extreme enough to solve the problem. In any ethical hospital person who recommended that would be kicked out immediately; in any unethical hospital, they’d move up to something more likely to work, like haloperidol.

      • Ialdabaoth says:

        I’ve never seen that happen and the sexual side effects aren’t nearly certain nor extreme enough to solve the problem. In any ethical hospital person who recommended that would be kicked out immediately; in any unethical hospital, they’d move up to something more likely to work, like haloperidol.

        You’re thinking hospitals, I bet Geode is thinking hospice care, mental wards, etc. – places where the patient is unlikely to recover, leave, and sue.

        • Chamomile Geode says:

          Definitely! I actually found some info on that: this (http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1470648/) is a case study about a dude treated with citalopram for sexual impropriety, this (http://www.ncbi.nlm.nih.gov/pubmed/9334516) seems to indicate that the cognitively impaired elderly should be given SSRI’s as the first pharmacological line of treatment for sexual impropriety, and this review (http://www.ncbi.nlm.nih.gov/pubmed/9988296) also seems to consider SSRI’s an appropriate treatment for sexual impropriety.

          Also–and this is the main reason I commented here–I wonder how often SSRI’s might be used to moderate the sexual urges of adolescents. (I’m pretty sure that this happened to me; the first place I saw the linked reviews and case reports was on my very religious parents’ computer, around the time I was started on SSRI’s, along with google searches like “medication to control teen sexuality” and “pills to reduce sexual urges”).

          Is using SSRI’s for this kind of thing…considered unethical? I’ve been wondering about this for years.

  33. Anon says:

    “It could be the feeling of cared-for-ness and special-ness of getting to see a psychiatrist and talk with her about your problems, and the feeling of getting-to-contribute-something you get from participating in a scientific study.”

    When you’re at the point going to the store to buy milk feels like an accomplishment enough to set you on the upward path, you bet your ass doing something actually interesting and useful would make you feel better.

  34. gwern says:

    The reason effect size is extremely fun is that it allows you to compare effects in totally different domains. I will now attempt to do this in order to see if I can give you an intuitive appreciation for what it means for antidepressants…To me, these kinds of comparisons are a little more revealing than NICE arbitrarily saying that anything below 0.5 doesn’t count. If you could take a pill that helps your depression as much as gaining 1.4 inches would help a self-conscious short person, would you do it? I’d say it sounds pretty good.

    I think it would be more helpful if the effect size were instead simply characterized in terms of depressive symptoms. Those other domains are fine because we already have intuition about what the population looks like and how variable it is and what sort of loss function we might have for it – but depression? Who knows?

    I’d rather see something like ‘an effect size of d=1 for depression means tha:t of 100 ultra-depressed patients who have a 5% annual risk of suicide, about 40 of them will be merely medium-depressed with a 1% annual risk of suicide within 6 months; as compared to 30 such patients given only placebo’

  35. Phil Goetz says:

    I continue to be dumbfounded by Scott’s ability to write so much dense material so regularly.

    How much of the placebo effect can be explained by regression to the mean, in all studies? Have placebo studies included a control (no treatment, no placebo) group?

    I’m not a placebo doubter, but “placebo effect” takes the position in modern medicine that “God” had in medieval philosophy. Why is the sky blue? Because God made it blue. Why does the sun rise and fall? God made it that way. The God hypothesis halts all inquiry. Similarly, any symptoms that deviate from the program—chronic Lyme patients who claim antibiotics make them feel better, DVT patients who claim warfarin makes them feel better—are dismissed out-of-hand as psychological, so current beliefs need not be questioned.

    The effect of taking anti-depressants, for me, was comparable in magnitude to, say, having a broken leg versus not having a broken leg. Some drugs are even more obviously effective, like sumatriptin for migraines, which relieves in half an hour migraines that have never responded to any medication before, ever. There’s something deeply wrong with medicine when even these drastic improvement of symptoms are not considered evidence unless they’re part of a double-blind placebo-controlled study.

    • Mary says:

      ” Why is the sky blue? Because God made it blue. Why does the sun rise and fall? God made it that way.”

      This is a historical ignorant view of medieval philosophy. Indeed, this was the era in which the philosophical views of causes was worked out that enabled modern science.

    • Phil Goetz says:

      My doc, BTW, says that bupropion must have a placebo effect on me, because the dose I take is about 1/10th of the usual dose. To which I always reply, “So gimme my placebo.”

    • Pablo Stafforini says:

      There’s something deeply wrong with medicine when even these drastic improvement of symptoms are not considered evidence unless they’re part of a double-blind placebo-controlled study.

      I recall reading similar complaints a while ago from a guy who claimed his life had been transformed by crystal healing.

      • Phil Goetz says:

        Let’s compare our cases.

        The kind of thing I’m talking about: A patient has DVT and is prescribed warfarin. The hematologist wants to keep his INR below 2.5, because of a double-blind placebo-controlled study that said there were no benefits to higher INRs. The guy claims that he still has severe pain unless he takes enough warfarin to raise his INR to 3.5. The hematologist dismisses the patient’s claims, tells him that warfarin dose has no influence on pain levels (though the hematologist has never had DVT himself), and that it is impossible that more warfarin will help, because a double-blinded placebo-controlled study said it wouldn’t.

        This happens to the hematologist once a week for twenty years, and the hematologist remains secure in his belief that all of his patients are deluded. Because he’s got a double-blinded placebo-controlled study that says so.

        The double-blinded placebo-controlled study did not actually say that INR above 2.5 provided no benefits. It said that the number of contrary incidents caused by raising INR above 3.5 were greater than the number of contrary incidents prevented. What it actually meant was that the excess of bleeding incidents in patients with an INR over 3.5 that required treatment were equal to the excess number of deaths in patients with INR under 3.5. The number of bleeding incidents in patients with INR over 3.5 was 6. The hematologist was told to keep INR under 2.5 rather than 3.5 to “be safe”.

        The kind of thing you’re talking about: Meeting a guy at Burning Man who wears sandals and talks about crystals.

        These are not comparable.

        The fraction of double-blinded, placebo-controlled studies which are obviously fatally flawed is, in my estimation, perhaps 1/3. The rate of fatal flaws in studies is so high that each individual study can be no more than weak evidence. When there are a half-dozen such studies to choose from, the person writing the review article or committee report will choose whichever ones support his favorite theory, and not mention the others. I therefore expect that the conclusions of reviews and committees on controversial subjects are mostly useless.

        Every step higher you go in the abstraction hierarchy of anecdote / study / review / meta-analysis introduces new opportunities for error, bias, and statistical misinterpretation. It is not clear which level contains the most signal. I lean towards the view that the mass of anecdotes accumulated by a practicing physician over 20 years can provide more information than review articles.

        • If you want to be contrarian about modern medicine while still dismissing crystal healing as pseudoscience, you need to find a heuristic that would vindicate your form of contrarianism when you apply it, but won’t vindicate crystal healing when applied by a proponent of that practice. You give many concrete details about the medical case you have in mind, but without further elaboration it’s hard to identify the underlying criterion you are relying upon. In particular, I would like to know what evidential value you assign in your process of belief formation to your own personal experience relative both to the personal experiences of other people and to the double-blind, placebo-controlled studies that you criticize modern medicine for relying so heavily upon.

        • Anonymous says:

          Pablo, much of the point is that Phil read the study and the doctor didn’t understand it, but just went with a headline. I see a parallel: I read Phil’s comment…

        • Anonymous,

          That seems like an uncharitable interpretation of Phil’s comment. If you are right and his point is merely that the particular doctor he consulted failed to understand a study, that hardly provides support for the claim that is the subject of this exchange (“There’s something deeply wrong with medicine…”).

  36. Chris T. says:

    Can you speak to SSRI discontinuation syndrome as a side effect? More commonly called “the zaps”. For me, this was a terrible side effect. I ended up having to taper VERY gradually from what was billed to me as a “non-habit forming” drug. Discontinuation syndrome to me seems like a really nice way to say withdrawals.

  37. Mary says:

    ” Patients have access to internal feelings which they tend to weight much more heavily than external factors like how much they are able to get done during a day or how many crying spells they have, sometimes so much so that they completely miss these factors. ”

    One wonders if it would help if doctors urged patients to keep track of such incidents in a daily journal.

    • Danielle W. says:

      Background: I spent 2.5 years dealing with a rather nasty depression. I started out with therapy but, after about a year, approached my doctor for a medical solution, who prescribed Lexapro.

      My therapist suggested that method and it really helped in a lot of ways – it helped him gain insight into what I was feeling and how often, and it helped me gain a better perspective on the actual balance (it was easy for me to weigh the down times far more than the better ones.) Thanks to this exercise I was able to see that I was gaining a healthy relationship with “negative” emotions.

      I continued therapy while on Lexapro because I saw the two addressing different domains. Put another way, the Lexapro made the emotional load such that I could at least function while therapy helped me gain the ability to move forward.

  38. Steve says:

    What would the negative side effects be of allowing medicine trials to look for bimodal distributions of results?

    • Phil Goetz says:

      See Stevenson et al. 2010, The Role of Histamine Degradation Gene Polymorphisms in Moderating the Effects of Food Additives on Children’s ADHD Symptoms for a semi-good example.

      Back in the 1970s, some doctors claimed that food dyes made hyperactivity worse in some children. Studies followed which tested the hypothesis that, for every child, food dyes made them more hyperactive. The results failed to reach statistical significance, so the medical establishment declared–very forcefully; I think doctors lost their licenses over this–that artificial food dyes do NOT cause hyperactivity in anybody. !∀xP(x) implies ∀x!P(x), after all.

      Except that it doesn’t.

      This is an example of what I called “The universal medical journal article error”. I called it “universal” because most medical journal articles that fail to find an effect in a population either commit this error, or are interpreted by others in a way that commits this error.

      The t-test and the F-test test the hypothesis that an intervention creates a change in outcome, assuming that it causes the same change in outcome in all subjects. This never happens in medicine, particularly if (as may happen in controversial conditions such as hyperactivity, Lyme, and depression) half your subjects don’t actually have the condition being studied. So failing an F-test for significance is not significant evidence that the effect does not exist. (It would be weak evidence for that hypothesis, if you didn’t actually have the data in hand to check for the direction of the effect. That data trumps merely knowing that the intervention does not create a large effect in all subjects; hence that knowledge should be discarded and the data checked, if one feels compelled to interpret the data beyond “failure to prove P(X) for every X”.)

      What the Stevenson 2010 study did was to look at the theory (food dye aggravating ADHD), come up with a list of 6 common gene polymorphisms that might cause such an effect, divide subjects into groups using those polymorphisms, and test the hypotheses that each polymorphism resulted in food dye aggravating ADHD. They did not, I think, adequately adjust for multiple-hypothesis testing, but since all the strong results were for the same histamine-degrading gene, HNMT, I expect an excruciatingly-correct analysis would still find that result significant. (The results in the paper are a little confusing, because Table 2 seems to indicate that the SNPs in question made hyperactivity decrease, but the data in Figures 1 and 2 give a clearer picture: Food dye caused a great increase in hyperactivity compared to placebo for children with the HNMT SNPs.)

      This is a pretty-good model of how to do studies without assuming that everything affects all people the same way. (Checking for a bimodal distribution is a special case of this.)

    • Phil Goetz says:

      Er, I guess I should’ve highlighed the answer to your question: The penalty for looking for bimodal response distributions is multiple hypothesis testing.

  39. Nestor says:

    Would be interested in your take on meditation, as it has been compared to medication for issues like depression.

    In my experience it does seem to have an effect in blunting involuntary reactions to annoying stimulus, i.e. flatmate makes an annoying noise, I still intellectually think it’s annoying but I no longer have the visceral annoyance reaction in my gut, same goes for cringe-making memories of past mistakes – the memory is disconnected from the embarrassment.

    • Error says:

      cringe-making memories of past mistakes – the memory is disconnected from the embarrassment

      Owowowowow….ow.

      So, tangent completely unrelated to the original post: “Cringe Memories” is a great term for a horrible experience. I have an array of bad memories (mostly public social errors) that are practically incapacitating if my train of thought wanders across them; the train gets derailed by sudden, seething self-contempt and rage. I also have a mild case of Tourette’s, and stumbling on a cringe memory is almost guaranteed to trigger an episode. Which in the wrong circumstances can form a new cringe memory. Yay positive feedback loops.

      I’m glad you mentioned it just so I can internalize that I’m not the only one to experience them, even if you don’t mean it quite as badly as the above.

      • Matthew says:

        Scott had a post on dreaming; I’d like to see a post on memory and remembering.

        I don’t think I have much in the way of cringe memories, but to the extent that I’m depressed and angry, it’s because I have near perfect recall for emotionally salient moments (particularly negative emotions). When I say it feels like I got my heart broken yesterday (it was in January), I mean it literally, because I have an obsessive personality and the memory of it is just as intense as if it just happened.

      • Creutzer says:

        No, you’re not alone with that. I have those kinds of memories, too; when my train of thought comes across them, I experience a sudden outburst of stinging emotional pain, which sometimes even results in involuntary outward signs of pain (e.g. facial expression).

        • CaptainBooshi says:

          Ha! This very thing happened to me today. I was just reading something online which reminded me of a time when I said something particularly stupid while hanging out with people in college. I physically winced and said something out loud without even meaning to.

  40. Zorgon says:

    Thanks hugely, SA. Probably going to spread this around a bit, I have a fair few pharma/depression type friends who would find this intriguing.

  41. I’m a fairly self-conscious short person. I’m 4′ 11″ and was teased a fair amount about my height when I was a kid. I have no idea about the typical self-conscious short person, if there is such a thing.

    I don’t think 1.4″ would mean very much to me, let alone .8″. 1.4″ would probably mean I could reach all the shelves in supermarkets, I think, but I’d still be pretty short. I’d never thought about it before, but I think it would take 3″ for me to feel as though I’d moved into the low end of the normal range.

  42. Justis Mills says:

    Not sure if this comment passes the “necessary” gate, so, sorry, but: as someone who was on SSRIs for a little while a couple years ago, this was a really interesting read. Good to know more about the little pill I ate every day while I went to free therapy in college.

    • Earnest_Peer says:

      The “necessary” thing is mostly a defense against assholes, so don’t worry.

  43. Andrew Hunter says:

    Scott, I’d like to hear anything more you have to say about bupropion in particular. You seem to toss it off as “and of course it doesn’t have any of these side effects, but eh, whatever” which seems a bit, well, underselling it! Insomnia isn’t great, naturally, but (according to at least one doctor) that’s mostly a function of it being a stimulant (and isn’t a problem if you time your doses properly): for the fraction of depressed people who have excessive sleepiness it’s practically a benefit. Similarly, a large number of people see decreased appetite as a _good_ thing: it’s pretty easy to convince myself to overeat if I’m worried about losing weight. Seizure risk, obvi, is nothing to laugh at, but I don’t have any (personal) reason to be particularly worried (no family history or other conditions that would contraindicate), and I figure most people are in that bucket.

    I mean, unless you think it’s substantially less effective than your other choices, why is this post not saying “and basically unless you are anorexic or a seizure risk, you should obviously try bupropion (…and possibly something else after that if bupropion doesn’t work)”? In short, you seem less positive on it than your text would imply you should be.

    • Creutzer says:

      Seconded. I’ve always been wondering why Bupropion isn’t basically the very first choice for an anti-depressant.

      • badgerchild says:

        I had a trial with an SSRI (Cymbalta?) that produced severe emotional blunting in me to the point where I dropped a full jar of spaghetti sauce on the floor and cut my feet because I didn’t care enough to clean it up (or to bandage my feet, come to think of it), and had bizarre, awful dreams, and this within a week of starting the pill. People noticed my changed demeanor at work and asked, “Are you OK? You sound different and you move differently too”. Obviously I was a strong negative responder. My psychiatrist said, “That cannot happen”. Oh, really?

        I had tried Zoloft years before and felt some significant emotional blunting on that within a few days of taking it, which was a blessed relief since I was actually going through one of “those” relationships where I was being triggered again and again. I thought that was actually the therapeutic purpose of the drug. I realize now that the three weeks I was taking it weren’t enough to really affect the actual depression. I guess that was a mild negative response.

        After the second SSRI gave me such a bad time, I refused to take anything but buproprion, and fired my psychiatrist when he put up a fight on the grounds that I also had anxiety and buproprion sometimes worsened anxiety. In subsequent therapy I came to understand that my anxiety was related to my depression and it is not unlikely that successfully treating the depression would have treated the anxiety as well.

        Anyway, in much the same way patients often have to try many medications before finding a proper one, I had to try four or five talk therapists before I found one with whom I clicked and whose sessions I found helpful. That’s not to say the others weren’t competent! Undoubtedly the choice of therapist and therapy has everything to do with the measured efficacy of talk therapy. In my individual case, giving me coping tools and thinking techniques that I had never learned before turned my bad thinking around a lot. I still get frequent episodes of fatigue and anxiety, but I manage them and they don’t deepen into severity anymore like they used to.

        Therapies that mildly affect dopamine, or that lower my borderline-high blood sugar (which has a lot to do with cortisol), seem to work better on me than any other chemical therapies. I still don’t know what the mechanism of action of loratadine is, however, only that it has a beneficial effect on whatever’s going on in my head.

      • Sarah says:

        I was once told that it can worsen anxiety. Haven’t seen confirmation, but, y’know, it makes intuitive sense for dopamine to do that.

    • anon1 says:

      > and isn’t a problem if you time your doses properly

      Anecdote warning: my roommate took bupropion for a while (every morning immediately on waking, like you do) and couldn’t sleep more than six hours in a night. Typically less. He’d spontaneously wake up at seven in the morning or even earlier. It did improve somewhat over several weeks, but not that much, and it continued to be enough of a lack of sleep to make it hard to function.

      • Desertopa says:

        Was waking up at seven in the morning particularly undesirable for him?

        Personally, I consider six hours to be a full night’s sleep, and haven’t found any noticeable benefit to getting more (insomnia, for me, was when I was getting about three and a half on average.) Obviously, not all people function well on that amount, but when I hear that someone is getting six hours a night, I don’t assume that it’s undesirable by default.

        • Matthew says:

          I’m having trouble locating the study at the moment, but there definitely was one that found people who get 6-7 hours of sleep a night often feel fine subjectively, but severely underestimate the objective decline in cognitive performance.

          There is about 7% of the population with a gene variant that makes 4 hours of sleep sufficient. Pretty much everyone else seems to need around 8, even if they don’t think they do.

        • anon1 says:

          Six hours was the maximum, if he spent substantially more time lying in bed. It was pretty strongly undesirable. Anyway, the main point is that extended-release bupropion can interfere with sleep for at least 24 hours so timing is not necessarily a solution.

  44. Doug S. says:

    Oddly, sexual side effects are not always unwanted. For example, SSRIs are known to be effective for reducing premature ejaculation…

  45. Jai says:

    Slightly apropos: A few years ago, when I started treating my depression, my psychiatrist enthusiastically recited the depression-is-simple-neurochemistry-something-something-seratonin spiel, and I was too tired and too apathetic about everything to question it.

    She put me on Bupropion and talk therapy, and six months later I was functional again.

    Which is when she brought up the healing power of quantum thought waves.

    That’s how I came to realize that (1) having accurate beliefs about the world is not a prerequisite for a medical license (2) I’d let someone who believes in basically-pixie-dust to decide what chemicals I should put in my brain and (3) it all worked out anyway.

    I’m not quite sure what the takeaway is here.

    • suntzuanime says:

      Nobody has accurate beliefs – a system that relied on having sane people for it to work wouldn’t work, and it would therefore be replaced.

      • Jai says:

        Right, but I kind of expected that a credential reporting almost a decade of study to acquire would result in approximately-correct beliefs about the main thing it’s supposed to be focused on, in a broad sense.

        Like, run of the mill sanity failures, like that cancer screening Bayes problem doctors famously flunk – I expect that. Parroting popular ideas without checking them? I’m just like everyone else in expecting that. Confirmation bias? I expected to see that all along. But getting the core mechanics of the thing you’re supposed to be an expert in completely wrong in a pretty obvious way that has been explicitly taken down by tons of people in your field?

        The level of wrongness you can get while still being able to practice is apparently higher than I intuited.

        • Anonymous says:

          core mechanics of what? did she attribute talk therapy or drugs to quantum thought waves?

        • suntzuanime says:

          The thing is, she can’t prescribe healing quantum thought waves until they’re FDA-approved, and the FDA-approval process will screen them out. The system works!

          Her area of expertise is knowing which FDA-approved medications to give you. Unapproved medications are the job of an expert at the FDA, who knows enough to look at studies conducted by people who know how to test whether quantum healing waves work. The only person who actually has to have a correct opinion about quantum healing waves is the person who decides whether or not to fund the study – everything else will be taken care of by the system.

          And even if this one person totally blows it on quantum healing waves, the consequence is you waste a few million dollars on a failed study – not “no big deal” exactly, but it’s not going to bring civilization to a halt. This is properly a cause for celebration, not horror: look how robust our society is!

  46. Douglas Knight says:

    While it is an exaggeration to say they have no side effects (see above)

    What does “see above” refer to? Perhaps it should be “see point 8”?

  47. gattsuru says:

    There is a subgroup of depressed patients whose depression takes the form of not being able to feel anything at all, and I worry this effect would exacerbate their problem, but I have never heard this from anyone and SSRIs do not seem less effective in that subgroup, so these might be two different things that only sound alike.

    I’m thinking the latter. English has very few good terms to describe different types of lack of emotional responsiveness, and many of them. Alexythimia is a drastically different thing from depressive akrasia, yet many surveys will happily lump the two together. I’ve heard a lot of folk who have problems with SSRI-family drugs describe the emotional blunting as feeling ‘like a psychopath’, which doesn’t really match up with their pre-medication mentality (esp since many report social anxiety behaviors, if not emotional responses, before use of drug).

    • I’ve heard a lot of folk who have problems with SSRI-family drugs describe the emotional blunting as feeling ‘like a psychopath’

      Not addressed in this post: Mass shootings were much less common in the US before the 1980s, despite just as easy, or easier, access to firearms than today. Simple question: How many mass shooters have been on anti-depressants? Quick perusal of the internet says, maybe all of them. Or those that weren’t. were on anti-psychotics, or were on something. Have there been any mass shooters not on drugs? Yes, before the 1980s, when such shootings were much less common.

      How many people taking CBT have shot up a school? Far as I can tell, zero.

      And here I’m tempted to add: QED.

      • gattsuru says:

        Survey by internet reporting is fraught with peril. While school shooters are probably (even tautologically) mentally ill, most don’t have a psychiatric history of the sort that points to SSRI use.

        This isn’t much of a surprise. SSRIs were first marketed in America starting in 1987 (zimelidine, followed by fluoxetine as Zimelidine was withdrawn from the marketplace), while the rise in mass shootings is supposed to have started about years before that, including the Edmond post office shooting in 1986.

        Drug /abuse/ is common, but that tends to be about marijuana or harder illegal drugs, and even then isn’t very useful in a predictive manner.

        • Thanks for the reply, but it doesn’t exactly address the broader point–how many mass shooters were not on prescription psychotropics? And why back in, say, 1929, when there weren’t any such thing, didn’t at least a few bullied kids or stressed college students take dad’s Winchester 30-30 to school and kill eight or ten people? Pre-1987 there were many non-SSRIs being doled out. Correlation is not causation, but…cable news came along then, too. What a combo.

        • Desertopa says:

          There are a lot of possible reasons, and I suspect the most plausible ones are cultural. In general, people are a lot more likely to take actions that they have an existing social script for, so I think one of the major causes of mass shootings is probably mass shootings. Running amok is an example of a comparable behavior which is frequently exhibited in some societies where it’s part of an existing social script, but generally is not generally exhibited in the same manner in other societies.

          The fact that people engaged in mass shootings today are very likely to be using or have a history of use of prescription psychotropic drugs doesn’t necessarily suggest a direct causation. Rather, most if not all people engaged in mass shootings are probably mentally ill, and mentally ill people are much more likely now than in the past to have a history of use of prescription psychotropic drugs.

          There might be other causative factors involved, but SSRIs would be way, way down my list of candidates. Mass shootings are still rare enough that you don’t need a very strong causative factor to account for the increase.

  48. Sniffnoy says:

    For example, Horder et al note that Kirsch incorrectly took the difference between the average effect of drugs and the average effect of placebos, rather than the average drug-placebo difference (did you follow that?)

    My immediate reaction to that is “but expected value respects differences!” — but I’m guessing this is some weird statistical thing where because you’re looking at samples rather than the whole population it doesn’t work that way?

    • Douglas Knight says:

      Horder et al do not explain the problem or provide evidence that they know about linearity of expectation, but a potential problem is that you are assuming a constant ratio of treatment and placebo arm sizes. In that case your argument applies, but if the treatment arm is twice the placebo arm, one effectively counts each placebo patient twice. If the ratio varies, then there is the choice between weighting a patient by the ratio in his study or the ratio in the total population.

      For example, if I have two studies, one with small treatment arm that does well and one with a small placebo arm that does badly, both studies produce large differences, but if I mix together the populations, I am mainly comparing the large mediocre treatment arm to the large mediocre placebo arm and the difference is small.

      ━━━━━━━━━

      I was very surprised to learn that one of the authors of Horder et al is economist Robert Waldmann, one of the authors of the Noise Trader paper.

  49. Sarah says:

    I’d been meaning to do a SSRI infodump one of these days but was daunted by the sheer volume of literature. So thanks.

    But I still have all the questions I started with.
    *why would serotonin fix depression?
    *what does taking an SSRI actually do to mood/psychology?
    *is there a sub-population it works on much better than “depressives” in general?
    *in other words, is there a cleaner explanation somewhere than “seems to work, sorta, somewhat, sometimes”?

    • gattsuru says:

      *why would serotonin fix depression?

      We don’t know. We don’t know how depression causes things to break in the first place.

      There are a lot of theories, though. SSRIs are more about moderating serotonin (and other neurochemicals) by keeping them active in the neurons longer at individual times. That may all that it needs to do and it just takes some time to feel the effects, or it may be some downstream effect — a lot of neurochemicals are tied up into the same control circuitry, and the long time before effectiveness correlates more with downstream regulatory chemicals than with the blood life of the SSRI itself.

      Scott’s brought up BDNF, the acetylcholine system is a fan favorite, and my personal vote for so-stupid-it-might-work revolves around the sleep cycle and not getting enough of it. There’s also an interesting theory that it’s all actually getting too much cortisol and related effects on neurogenesis.

      *is there a sub-population it works on much better than “depressives” in general?

      There are some DNA tests that show contraindicatives for individual drugs, and some often-weird correlations, but in most cases it looks to be near-random at best, and Texas Sharpshooter at worst.

      *in other words, is there a cleaner explanation somewhere than “seems to work, sorta, somewhat, sometimes”?

      Nope.

      Keeps things interesting, I suppose.

      • Sarah says:

        What about atypical vs. melancholic depression?

        I was under the impression that the atypical type (which goes with sleeping and eating a lot and sensitivity to social humiliation) is less likely to respond to SSRIs than the classic or melancholic type (which goes with not eating, not sleeping, and not responding to positive events.)

        If I completely butcher the neuroscience here, melancholic depression sounds like lots of cortisol, and atypical depression…doesn’t.
        http://neurotransmitter.net/Gold.pdf

  50. I love all of these articles about depression and antidepressants. Having had a major depressive episode myself, and having had several family members as well, I take all of this info as invaluable, even if I know nothing about psychiatry beyond what you put in these posts.

  51. Anonymous53 says:

    Measuring the effect in terms of standard deviations on the score of a questionare seems questionable for many reasons. With a large meta-analysis, they should be powered to at least attempt an analysis with a hard end point, such as suicide rate. Did they do this?

  52. Ialdabaoth says:

    Doctors (or family members, or other outside observers) who don’t see these internal feelings, are better able to notice outward signs. As a result, it is pretty universally believed that doctors spot signs of recovery in patients long before the patients themselves think they are recovering.

    So, there’s a good chance that the last few times I took $5 knock-off Prozac and noticed no change, that I was actually improving but didn’t have anyone to tell me that I was?

    This… changes several assumptions.

    Is there any way to find a *good* psychiatrist, who will actually pay attention to how I’m actually responding to treatment and not attempt to enforce their narrative?

    I’m currently in Twin Falls, Idaho, so I’m assuming that I’d basically have to move to gain access to a large enough community; but where would I have the most likely chances of finding a not-fully-Institutionalized psychiatrist AND a company willing to employ/pay medical insurance for a programmer with obvious and severe mental health issues?

    • gattsuru says:

      I am not a doctor, this is not medical advice, yadah yadah.

      So, there’s a good chance that the last few times I took $5 knock-off Prozac and noticed no change, that I was actually improving but didn’t have anyone to tell me that I was?

      Prozac is an SSRI. Like /all/ SSRIs (and most SDRIS and SNDRIs), full effectiveness of treatment doesn’t begin until at a week or even two weeks of reliable dosage. This is especially true for drugs with a long blood half-life like Prozac: individual pills have small doses because you’re supposed to have several days worth floating around in your bloodstream. Some folk do get benefits before that, but it’s a pretty consistent rule that it takes two weeks to reliably test results.

      And /then/, on top of that, yes, it’s very common that you’ll be the last person to note that you’re feeling better, and it’ll probably not even come as you’re expecting. It’s often the revelation that someone Did Something and wasn’t tired afterward, not that they were suddenly happy, that tends to get the attention.

      That said, some folk don’t respond to individual SSRIs (or any psychiatric med), so after three to four weeks, doctors seeing little response may move you to a different drug that’s more likely to have response.

      Is there any way to find a *good* psychiatrist, who will actually pay attention to how I’m actually responding to treatment and not attempt to enforce their narrative?

      Search online, or on yellow pages, or check with your general practitioner for advice, if you have one. (Hell, in many cases, your general practitioner can get you started.)

      … but where would I have the most likely chances of finding a not-fully-Institutionalized psychiatrist AND a company willing to employ/pay medical insurance for a programmer with obvious and severe mental health issues?

      Most towns around the i84 corridor in Idaho have psychiatric outpatient services, ranging from small practices to clinicians operating as part of a hospital network to actual hospital folk.

      While Idaho hasn’t expanded Medicaid at this time, you may find it worthwhile checking for coverage anyway. Even if you don’t qualify, many if not most mental health centers provide a “sliding scale” payment based on income, which generally have half- or quarter-price options for unemployed or underemployed individuals. Look for the phrase “all incomes levels”. Some larger hospitals also offer to set you up with an MSW or MS operating under a PhD that double-checks their work and signs prescriptions, in exchange for reduced price care.

      • Moshe Zadka says:

        I am also not a doctor, in fact I am even *less* a doctor than the other people who have replied who are not doctors.

        I have, however, been working as a software engineer in the Silicon Valley for five years. If you’re a good programmer, mental issues are not such a big hurdle. A previous manager of mine was eager to hire a guy with Asperger’s (Apple stole him from us, though). At FB, one of my colleagues publicly posted about being depressed, to the point of being suicidal.

        Please feel free to reach out to me (my work e-mail is my first name, first letter of my last name @fb.com) if you want to apply to Facebook — we’re always looking for people, especially experienced people.

    • Anonymous says:

      I think he’s just talking about HAMD scores, so you can self-administer.

    • CaptainBooshi says:

      So, there’s a good chance that the last few times I took $5 knock-off Prozac and noticed no change, that I was actually improving but didn’t have anyone to tell me that I was?

      From my own personal experience, this is absolutely true. The few times when I was extremely severely depressed (unable to leave my room, even to eat, horribly afraid of even just talking to people I knew, etc., etc.), I was able to feel the effects of anti-depressants, often within a few days of taking them. At any level of depression less than that, I wouldn’t feel any different at first, but family and friends close to me would all tell me they could see the difference. At one point, my mother was able to tell I had gone onto anti-depressants again simply though conversations over the phone, without my informing her of it at any point.

    • Zorgon says:

      Just adding to the chorus, here – the last time I was on antidepressants it was a relatively minor episode triggered by work stress, rather unlike my much longer and more severe early 20s episodes. I noticed no mood difference whatsoever, but within about 3 weeks I had colleagues who had no obvious way of knowing I was medicated commenting on how much better I seemed to be doing.

      It is, of course, also quite possible that the episode itself was already beginning to clear up. But it has for a long time seemed like SSRIs help with “cope levels” long before they do anything about “happy levels”.

      • I’m reminded of someone who had a wild talent for saying the wrong thing, and then her social skills improved rather suddenly. I asked her about it, and she said she was on an anti-depressant. (Sorry, I’m out of touch with her and don’t know which anti-depressant it was.)

        She hadn’t noticed that her behavior was different. From her point of view, other people had become much nicer.

        • fubarobfusco says:

          I’ve recently been on an SSRI and experienced something perhaps related: I’ve noticed that my expected value of social interactions is ① higher and ② more accurate.

          I’d previously noticed, from time to time, that I would feel dubious about going to a social event, but go anyway and end up having a good time — in other words, that I was poorly calibrated at predicting my own enjoyment. Now I more accurately predict when I’ll have a good time.

      • Doug S. says:

        It is, of course, also quite possible that the episode itself was already beginning to clear up. But it has for a long time seemed like SSRIs help with “cope levels” long before they do anything about “happy levels”.

        My personal experience matches this. The primary effect of antidepressant medication on me has been to take me from “I hate doing this stupid shit – fuck the consequences, I’m just not going to do it” to “I hate doing this stupid shit – okay, Mom, I’ll get to work on it.” I don’t necessarily feel any better, but they make me much more compliant and willing to tolerate things that I wouldn’t otherwise.

        I’m going to quote a New York Times Magazine article from a few years ago:

        “I remember one patient who came in and said she needed to reduce her dosage,” he says. “I asked her if the antidepressants were working, and she said something I’ll never forget. ‘Yes, they’re working great,’ she told me. ‘I feel so much better. But I’m still married to the same alcoholic son of a bitch. It’s just now he’s tolerable.’ ”

        • Ialdabaoth says:

          My personal experience matches this. The primary effect of antidepressant medication on me has been to take me from “I hate doing this stupid shit – fuck the consequences, I’m just not going to do it” to “I hate doing this stupid shit – okay, Mom, I’ll get to work on it.” I don’t necessarily feel any better, but they make me much more compliant and willing to tolerate things that I wouldn’t otherwise.

          This SO MUCH THIS OMFG this. (My first run of medication was to permit me to conform to my mother’s narcissistic requirements for me. It worked exceptionally well; she’s been dead 18 years, and I’m STILL conforming to those roles to this day.)

        • An Anon says:

          I think multiple people here could use this: Dysfunctional Families Support Group

          Also, today I cried for all of you subjected to chemically induced apathy and/or compliance. (Enough so to get a friendly visit from concerned neighbours.) If you find it helpful, please let it be a token of the feelings and noncompliance you weren’t allowed to have.

        • There are a lot of books and websites about recovery from living with narcissists, including having a narcissistic mother, but I don’t know which are best.

          Does anyone here have recommendations?

        • Ialdabaoth says:

          Well, I have kind of a mental block; “blaming your parents for your own shittiness” is something only whiny losers who can’t face their own utter uselessness do.

          Unless you manage to write an inspirational book on it and go on the 700 club and talk about how Jesus’s love saved you, you aren’t allowed to blame your parents. You must gaze into the abyss of your own failure and not flinch.

          EDIT: Actually, let me frame that another way.

          If your life is a mess, it is *suspiciously easy* to decide that you had terrible parents or bad luck or any other number of systematic failures / oppressions. Thus, we have to assume that anyone who blames anything but their own utter failure as a human being for their setbacks is actually an utter failure as a human being.

          SUCCESSFUL people are allowed to blame their parents or environment, because they aren’t motivated to find excuses for their failure-as-people; in fact, doing so makes them all the more inspirational.

          (Yes, I know this is clearly BS. I’m describing how my internal value system works, not asserting its validity.)

  53. The Anonymouse says:

    Scott, how in the world do you manage to keep up your posting regimen? I’m knackered just getting up, reading on the toilet, spending the day at work, then coming home and reading half of the comments on your blog. 🙂

    (Which is my way of saying, keep it up!)

    • Qetchlijn says:

      I’m becoming increasingly convinced that Scott is the Nicolas Bourbaki of the rationalist blogging community.

      • GG says:

        The secret of Scott’s productivity is that he was bitten by a radioactive lukeprog.