David Nutt, one of the co-authors of the citalopram stereochemistry paper (but much better known for being fired from his career as chair of the British Advisory Council on the Misuse of Drugs after he wrote a paper determining that ecstasy was less dangerous than horse-riding), was discussed on SSC a few years ago as the founder of a start-up called GABA Labs, which hoped to create safe alternatives to alcohol (...though as of 2019 they were apparently working with benzodiazepine derivatives, which, uh, yikes).
Anyway, I'm here to report that as of January 2021, GABA Labs actually has an alcohol replacement product on the market called Sentia: https://www.sentiaspirits.com/
But I think I personally would be a bit scared to try it (at least until I know how it compares with MDMA use or horseback riding), since when I went to the FAQ section of the website to determine what exactly Sentia is, the FAQ unhelpfully non-informed me that "Sentia is made through a unique, proprietary series of production methods that preserve the plants’ functional compounds and nutrients, giving you more of what the plants have to offer." https://www.sentiaspirits.com/faq (2nd question)
Those last studies mentioned had no control group, but I sometimes wonder why they can’t compare themselves to some sort of “standard” control group. Should we at some point just collect data on a very large control group (without having an active intervention), and then let studies for the next five years compare their active treatment arm of their antidepressant medication or whatever to this “standard” control group? It seems like it would make trials cheaper, potentially more ethical, and would avoid these sorts of problems. If you have a control group and a treatment group for every trial, then you need to make sure that both of them aren’t abnormal, which seems harder than having one very big control group that you measure every five years which you match for demographics/whatever, and then let every study for five years compare their work to that
First paragraph says the FDA claimed studies don't show any increase in intended effect going above 10 mg, but the studies you cited here all show that. Is the FDA lying or was that statement true at one time and they never updated it when more studies were conducted? Or were they extrapolating from the experience with Celexa without having any studies at all?
What is wrong with Jakubovski et al Dose-Response Relationship Of Selective Serotonin Reuptake Inhibitors?
1. The curves Scott shown in his post (from figure 2) make no sense.
a) In clinical trials, you usually see the improvement rate, largest at 2-4 weeks, peter out and plateau by 8-10 weeks. To the contrary, in Jakubovski et al improvement continues almost linearly beyond 8-10 weeks
b) In clinical trials, the patients improve less and less with each increasing dose of antidepressant. It basically follows the diminishing return law, where, for example, 100 mg helps a lot of patients, 200 mg helps slightly more people, 300 mg makes no difference in comparison to 200 mg. To the contrary, in Jakubovski et al the step between 100 mg and 200 mg improves depression as much as the step between 200 mg and 300 mg and as much as the following step 300-400 mg.
Jakubovski et al dress their model into a lot of fine words: “Previous research has demonstrated that a logarithmic model provided the best fit for the time course of SSRI response” and “The effects of SSRI were modeled using an autoregressive variance function, and the model with the lowest values on the Akaike Information Criterion was selected (14)”, etc. But, in my view, the matter is simple: if your model does not correspond to real life – discard the model.
2. Scott puts a lot of faith in Jakubovski et al, but this is based on a misapprehension. Scott believes that Jakubovski et al. “tries to figure out which doses of which antidepressants are equivalent to each other, and comes up with the following suggestion … 100 mg of imipramine=120 mg of sertraline=100 mg of fluvoxamine=20 mg of paroxetine or fluoxetine=33.3 mg of citalopram=16.7 mg of escitalopram” That is not so!
Jakubovski et al postulate (in other words, conjure from thin air) these ratios as the condition of their model. They try to disguise this choice as something objective: “Dose of SSRI was converted into imipramine equivalents based on previously defined methodology based on the therapeutic dose range of each medication (10, 15)… For SSRI analysis, the dose equivalents were as follows: 100 mg of imipramine=120 mg of sertraline=100 mg of fluvoxamine=20 mg of paroxetine or fluoxetine=33.3 mg of citalopram=16.7 mg of escitalopram.” But description of the dose equivalence choice is too short and perfunctory to be reproducible, thus revealing it as mostly discretional. By the way, this also renders all the subsequent analyses in Jakubovski et al worthless, because the dose equivalency choice, to a significant degree, determines the outcome of their analyses of imipramine dose equivalence v. efficacy, imipramine dose equivalence v. dropouts, etc.
High-Dose Citalopram and Escitalopram and the Risk of Out-of-Hospital Death
The incidence of sudden unexpected death ranged from 27.2 per 10,000 person-years for sertraline to 58.8 per 10,000 person-years for escitalopram. The incidence of all deaths ranged from 50.3 per 10,000 person-years for sertraline to 84.9 per 10,000 person-years for paroxetine.
Lexapro and Celexa are selective serotonin reuptake inhibitors (SSRIs). Both drugs are used to treat depression. Lexapro is also used for anxiety. In 2018, the drugs’ labels included a black box warning for an increased risk of suicide. Other side effects include birth defects.
Nausea, dry mouth, trouble sleeping, constipation, tiredness, drowsiness, dizziness, and increased sweating may occur. If any of these effects persist or worsen, tell your doctor promptly.
Your doctor has prescribed this medication because he or she has judged that the benefit to you is greater than the risk of side effects. Many people using this medication do not have serious side effects.
Rarely, males may have a painful or prolonged erection lasting 4 or more hours. If this occurs, stop using this drug and get medical help right away, or permanent problems could occur.
Great read, Scott. Can you comment on whether the dual action of Lexapro could explain the most difficult side effects related to lower sexual desire and sexual (dys)function? This is the most difficult part of taking a higher dose (30 mg) for me.
Scott is mistaken here: “Sometime around 2011, the FDA freaked out that high doses of citalopram might cause a deadly heart condition called torsade de pointes, and lowered the maximum dose to prevent this… When escitalopram was invented, it inherited its parent chemical's unusually-low maximum dose, and remains at that level today.”
Escitalopram was approved in 2002, which was before 2011 when the FDA freaked out about torsade de pointes. In the original, 2002, label of escitalopram the maximum dose is 20 mg and is based on the lack of improvement at the higher dose, not on torsade de pointes (see the 2002 label here https://www.accessdata.fda.gov/drugsatfda_docs/nda/2002/21-440.pdf_Lexapro_Prntlbl.pdf p.20): “The recommended dose of Lexapro™ is 10 mg once daily. A fixed dose trial of Lexapro™ demonstrated the effectiveness of both 10 mg and 20 mg of Lexapro™, but failed to demonstrate a greater benefit of 20 mg over 10 mg (see Clinical Efficacy Trials under Clinical Pharmacology). If the dose is increased to 20 mg, this should occur after a minimum of one week.”
However, the good news is that based on the data from this review https://sci-hub.se/10.1007/s40263-014-0196-9 40 mg escitalopram would have QTc prolongation slightly higher than 40 mg of citalopram, and thus should be safe: Change in the mean QTc interval from baseline was 8.5 ms with citalopram 20 mg/day, 18.5 ms with citalopram 60 mg/day … 4.5 ms with escitalopram 10 mg/day, and 10.7 ms with escitalopram 30 mg/day.
Have antidepressants or any other psychiatric medications been shown to increase sex drive or romantic/social interest? Or is it mainly a mood fix. What about energy levels?
This is only tangentially related to the article, but I assume as a psychiatrist you have experience with fluvoxamine. What's your take on the studies showing fluvoxamine prevents cytokine storms in COVID-19 patients drastically reducing the risk of death or hospitalization? My read is the FDA and CDC have been predictably stupid in not authorizing and recommending it's use, but I'd be interested in your thoughts on the safety efficacy tradeoff
I was prescribed escitalopram at 20mg for several years. I did score better on the phq-9 compared to untreated, but was still unable to participate normally in life. After five years, it just stopped working - I crashed like a truck going off a cliff. With the help of a new psychiatrist, I switched to sertraline at a much higher dose (titrated, with serum checks) and finally got to moderate functionality. The persistent caution about higher doses of escitalopram, even in patients with stubborn depression, has caused years of my life to be lost in the weeds. This is combined with the fact that most MHPs I've met have been reluctant to change medications, and seem to choose Lex as their first choice. Given the imposed ceiling on dosage, I'm at a loss as to why it's so popular.
"I will add my own anecdotal evidence here, which is that I had a patient who spent a year pretty depressed, got very modestly better on Lexapro 20 mg, and agreed to let me experiment with higher doses of Lexapro on him. He got better still at 30 mg, and even better when we increased to 40 mg. I am relatively confident these were real effects."
How can you distinguish that from "remission needed a long time, and you kept raising their Lexapro dose while they waited"?
In my experience, research has this annoying feature where you have some observation of an effect, and it's pretty clear. But then you do an experiment and the data still shows an effect. And then you do the stats and it looks marginal.
And then you do a fuck-ton of experiments and more stats and it still looks pretty shit. And your paper eventually gets published but looks worse than those of people who do much less.
I'm reasonably convinced that it's the inherent nature of a supposedly 'fair test' experiment to
give poor-looking effect levels on results which are individually convincing.
At least two other factors may participate in the phenomenon analyzed here: 1) GI absorption differences and 2) changes in intestinal microbial ecosystem.
1) Different SSRIs have different gastrointestinal absorption. In this paper, “Clinical Pharmacokinetics of Selective Serotonin Reuptake Inhibitors” (https://tinyurl.com/w6zc5mep), Table III shows GI absorption differences between, for example, Fluoxetine, Paroxetine and Citalopram of 80%, >= 64% and ~100%, respectively. The primary datasets analyzed and referenced above use the by-mouth dose instead of plasma levels of the different SSRIs. OTOH, the graph above labeled “(a) Drug (SERT occupancy)” uses a dosing method of micro dialysis. This means a microprobe was inserted into the interstitial area of a rodent brain to administer and/or measure directly the SSRI available to the main targeted site of the serotonin transporter. There are serotonin receptors in the epithelia of the GI tract where by-mouth dosing of SSRIs also interact. The micro dialysis method validates the targeted SSRI effect but doesn’t measure the GI tract serotonin receptor effects.
2) There is increasing evidence that the gut microbiome influences/participates in/?causes? psychiatric effects both “good’” and “bad”. Different antidepressants have demonstrated different antimicrobial effects. See “Unravelling the antimicrobial action of antidepressants on gut commensal microbes” (https://tinyurl.com/yfu3zeeh). Those antidepressants with less GI absorption travel throughout the length of the GI tract influencing the microbial ecosystem and its outputs along the way. The outputs of the microbial ecosystem include neurotransmitters, SCFA and many other small molecules that get absorbed and influence the host health. These phenomenon may cause sufficient GI distress that the patient stops taking the SSRI. The effect on the microbiome is still present even if there is no overt GI distress.
"I hope the three of you who understand stereochemistry appreciate that title as much as it deserves." I understand stereochemistry and I fucking loved that title.
I'd be surprised if the original title of this post wasn't something to the effect of "Oh, the places you'll go, when you're dosing Lexapro". I wonder why it got the more dry title
1/4 of the "minimal dose" of Sertraline worked for me.
I was given the tablets with the lowest dose available and told to split them to ramp up to the minimal therapeutic dose with was 1 tablet. Being the very careful person that I am, I didn't see why I shouldn't ramp up even more gently than that and split the tablet into fourths. I stayed at that dose, because it worked. My psychiatrist mocked me, calling it homeopathy. But I had the antidepressant effect with side effects and everything. Anyways, that "therapeutic relationship" didn't last for some reason.
Basically the same occurred years later with Mirtazapine. I feel lucky, because Mirtazapine is a very effective antidepressant and I managed to dodge the side effects. Considering that during the first week I almost fell asleep at work (on 1/4 of the "minimal dose") I understand why it gets lousy patient reviews. I definitively feel low dose Mirtazapine should be prescribed more.
Also, pet peeve: the lowest therapeutic dose of psychiatric medication is way lower than conventionally believed. Because psychiatric diseases are often several things lumped together you will always have responders and non responders (or less responsive responders). So you need to give a high enough dose that the effect on the responders gives you a statistical signal even though the other half (or something) is not responding. But in clinical practice you try an SSRI and if that doesn't work, Wellbutrin, so you are matching people with "their" antidepressant.
In my case I respond tremendously well to 10mg of fluoxetine with no side effects, but if I take 20mg I get constant headaches. It has a much longer half-life than all the others which probably makes it safer in terms of accidentally missing or accidentally doubling doses and easier to get off of. If the other SSRIs are roughly equivalent in terms of efficacy, has the world benefitted at all from other SSRIs existing? Is it just a bunch of rent-seeking pharma companies making me-too drugs that are no better than prozac except for having a patent?
"But the details here are really confusing and seem to mostly involve differences in the placebo group that don't make sense."
I wonder if this could be explained by a difference in the mechanics of a fixed-dose placebo vs. a variable dose placebo. A fixed-dose placebo is simple, but a variable-dose placebo involves potentially repeated conversations with the care provider that result in changing the "dose" of the placebo. It seems at least possible that this interaction could result in an increased chance of a placebo being suspected, which could result in a decreased placebo effect, which could result in an increased difference between treatment and placebo groups. Although if this story is correct, I think that would mean that variable-dose is no more effective than fixed-dose.
* * *
"I hope the three of you who understand stereochemistry appreciate that title as much as it deserves."
While I appreciate the flattery, I'm betting a lot more than three of us got the joke.
Forgive me for being confused, but if the FDA says that 20mg is the maximum safe dose of Lexapro, how are Scott and others able to prescribe higher doses?
Why do most studies use intention-to-treat despite the problems Scott details above?
It's to avoid issues with selection bias. If you only count patients that actually take the drug instead of people who quit after day 1, you bias the results because there might be some fundamental underlying difference between quitters and non-quitters. For instance, people who successfully follow instructions tend to have better health results across the board, so if you only measure outcomes for those who don't quit, it will tend to make your treatment look more effective than it actually is.
Using ITT makes it harder to detect a real result, but in exchange when we do have a result we are more confident that there's actually a causal relationship.
Come. The obvious explanation is the stereochemistry. With racemic citalopram, for example, half the drug you're getting gives you all the side effects (and burden on drug clearance pathways) with no benefit at all. For a real hair-raising illustration of how *important* stereochemistry is to biochemistry, just compare the effects of R- and S-thalidomide.
I started with 20mg Lexapro and found that it cured much of my Autism: my social reflexes moved so fast I could actually keep up with a conversation involving more than 2 people! I could turn into a total motormouth at times, thinking of interesting things to say faster than I could say them. I had the sort of verbal intelligence and wit that I otherwise have only had while being very angry.
This effects went away, so I increased the dose to 20mg. The pattern repeated. I didn't increase the dose beyond that, I just kept taking this pointless pill for a few years because quitting made me irritable. I eventually quit cold turkey, which turned out to be a lot easier than I expected.
I often feel like a character on a sitcom where the you can have wacky plotlines for one episode - In this episode, my life doesn't suck! - but everything has to go back to normal in time for the end credits. I get just a glimpse into the world of normal people.
I have also discovered that increasing the dose of lexapro up to 30 will often work in patients for whom 20mg has stopped working. I have never tried going higher than that, but maybe I will now that I see it's reasonably well tolerated. However, I have also noticed in my practice that lexapro causes more sexual dysfunction than other SSRIs, which could be related to it's slightly different mechanism of action. Lexapro has been my go-to starting SSRI for depression/anxiety because it seems to have pretty high pt satisfaction, and I will often just switch pts if they experience sexual side effects. However between the sexual side effects and the withdrawal I do wonder if I should just be starting most people on zoloft or prozac and switching if they don't work. *shrug*
I think TSC is right about the timing and approval of Lexapro dosing; it had nothing to do with Citalopram FDA debacle which came much later. Scott, I think you can add support to you argument (in addition to the collective clinical experience and some questionable effectiveness studies you cited) from the receptor binding site literature:
In short, the paper seems to support the notion that increasing to a high dose of an SSRI is a valid strategy in patients who haven't responded to a lower dose; that is, in some patients, high-dose SSRIs are better than low-dose SSRIs.
The title of the paper is "Double-Blind Study of High-Dose Fluoxetine Versus Lithium or Desipramine Augmentation of Fluoxetine in Partial Responders and Nonresponders to Fluoxetine", which pretty much says it all. In a sample comprising an equal mix of nonresponders and partial responders to fluoxetine 20mg/day, patients were randomized to high-dose fluoxetine (increasing to 40-60mg/day), or adding lithium or desipramine (two common strategies for patients who haven't responded to an SSRI). There were no significant differences in response rates between the three groups, though there was a nonsignificant trend toward fluoxetine being superior.
I do notice that their dose of desipramine was, in my (unqualified) opinion, too low, and their lithium blood levels were definitely on the low end. This is a limitation of this study, but probably not enough to entirely impeach the result.
I was prescribed anti depressants - after a week I felt awful so I quit taking them. When my brain de-fogged, I changed jobs, left my husband and joined a sports club. Sometimes drugs are not the answer.
Hi Scott, thanks for this. Furukawa is really interesting!
Just to be clear, there are zero randomised trials comparing head-to-head different doses of Escitalopram? If that's right, then I think Furukawa is the best evidence we have.
Notable that in Figure 1 of the appendix, Citalopram plateaus at ~30mg not ~20mg. This fact, plus the idea that the only _downside_ of a higher dose seems to be the higher risk of side effects, would imply that at least trying Escitalopram 30mg or 40mg is a great idea.
I wish Furukawa gave us scatterplots in addition to the splines, the absence of scatterplots feels a tad suspicious (am I misundertanding something? I only skimmed). I might try to make the scatter plots based on the table in the appendix myself.
Another thing that would be really cool is a Bayesian hierarchical model of SSRIs dose-response schedules, to generalise this idea of learning from one SSRI and, if approrpriate, transfering to another. I'm pretty sure this specific idea hasn't been done. [Added a few minuted later] Holy Shit it _has_ been done, first page of Google: "A Bayesian dose–response meta-analysis model: A simulations study and application" (Hamza, Cipriani, Furukawa, Egger, Orsini, and Salanti, 2021) [1]. My reading of Figure 3 is they don't extend the x-axis beyond 30mg fluoxtetine-equivalents for Escitalopram, which makes this pointless for answering our question? 🙄
Thank you very much for this article, it's very interesting. I've been on 20mg Lexapro for a long time, along with Welbutrin to counter some of the side effects. I have had a manic episode before so I'm definitely cautious about using too high a dosage. I've tried 40mg for just a few days and don't recall notice anything much, but it may not have been long enough to tell.
How long was your patient on 30mg and subsequently 40mg before they noticed differences?
Ignorant related question: Have there been any laboratory studies of supplementing SSRIs with serotonin precursors, such as 5-HTP?
IANAC but I love a good chirality joke!
Anyone know our best, easy to interpret evidence that SSRIs are not just placebo?
I would be interested in a similar explanation but for ADHD medication (ie how long until it stops working, and how to maximize that time)
David Nutt, one of the co-authors of the citalopram stereochemistry paper (but much better known for being fired from his career as chair of the British Advisory Council on the Misuse of Drugs after he wrote a paper determining that ecstasy was less dangerous than horse-riding), was discussed on SSC a few years ago as the founder of a start-up called GABA Labs, which hoped to create safe alternatives to alcohol (...though as of 2019 they were apparently working with benzodiazepine derivatives, which, uh, yikes).
Anyway, I'm here to report that as of January 2021, GABA Labs actually has an alcohol replacement product on the market called Sentia: https://www.sentiaspirits.com/
But I think I personally would be a bit scared to try it (at least until I know how it compares with MDMA use or horseback riding), since when I went to the FAQ section of the website to determine what exactly Sentia is, the FAQ unhelpfully non-informed me that "Sentia is made through a unique, proprietary series of production methods that preserve the plants’ functional compounds and nutrients, giving you more of what the plants have to offer." https://www.sentiaspirits.com/faq (2nd question)
I've been on 40mg of Lexapro for the past two years. I think it has been the optimal dose for me.
I think you underrate your readership to think only 3 of us got the chirality joke. Maybe even more than 3% of us did!
Those last studies mentioned had no control group, but I sometimes wonder why they can’t compare themselves to some sort of “standard” control group. Should we at some point just collect data on a very large control group (without having an active intervention), and then let studies for the next five years compare their active treatment arm of their antidepressant medication or whatever to this “standard” control group? It seems like it would make trials cheaper, potentially more ethical, and would avoid these sorts of problems. If you have a control group and a treatment group for every trial, then you need to make sure that both of them aren’t abnormal, which seems harder than having one very big control group that you measure every five years which you match for demographics/whatever, and then let every study for five years compare their work to that
Okay, maybe one day it will be useful for me to know that two mirrored stereoisomers are called "left-handed" and "right-handed" by scientists.
First paragraph says the FDA claimed studies don't show any increase in intended effect going above 10 mg, but the studies you cited here all show that. Is the FDA lying or was that statement true at one time and they never updated it when more studies were conducted? Or were they extrapolating from the experience with Celexa without having any studies at all?
What is wrong with Jakubovski et al Dose-Response Relationship Of Selective Serotonin Reuptake Inhibitors?
1. The curves Scott shown in his post (from figure 2) make no sense.
a) In clinical trials, you usually see the improvement rate, largest at 2-4 weeks, peter out and plateau by 8-10 weeks. To the contrary, in Jakubovski et al improvement continues almost linearly beyond 8-10 weeks
b) In clinical trials, the patients improve less and less with each increasing dose of antidepressant. It basically follows the diminishing return law, where, for example, 100 mg helps a lot of patients, 200 mg helps slightly more people, 300 mg makes no difference in comparison to 200 mg. To the contrary, in Jakubovski et al the step between 100 mg and 200 mg improves depression as much as the step between 200 mg and 300 mg and as much as the following step 300-400 mg.
Jakubovski et al dress their model into a lot of fine words: “Previous research has demonstrated that a logarithmic model provided the best fit for the time course of SSRI response” and “The effects of SSRI were modeled using an autoregressive variance function, and the model with the lowest values on the Akaike Information Criterion was selected (14)”, etc. But, in my view, the matter is simple: if your model does not correspond to real life – discard the model.
2. Scott puts a lot of faith in Jakubovski et al, but this is based on a misapprehension. Scott believes that Jakubovski et al. “tries to figure out which doses of which antidepressants are equivalent to each other, and comes up with the following suggestion … 100 mg of imipramine=120 mg of sertraline=100 mg of fluvoxamine=20 mg of paroxetine or fluoxetine=33.3 mg of citalopram=16.7 mg of escitalopram” That is not so!
Jakubovski et al postulate (in other words, conjure from thin air) these ratios as the condition of their model. They try to disguise this choice as something objective: “Dose of SSRI was converted into imipramine equivalents based on previously defined methodology based on the therapeutic dose range of each medication (10, 15)… For SSRI analysis, the dose equivalents were as follows: 100 mg of imipramine=120 mg of sertraline=100 mg of fluvoxamine=20 mg of paroxetine or fluoxetine=33.3 mg of citalopram=16.7 mg of escitalopram.” But description of the dose equivalence choice is too short and perfunctory to be reproducible, thus revealing it as mostly discretional. By the way, this also renders all the subsequent analyses in Jakubovski et al worthless, because the dose equivalency choice, to a significant degree, determines the outcome of their analyses of imipramine dose equivalence v. efficacy, imipramine dose equivalence v. dropouts, etc.
High-Dose Citalopram and Escitalopram and the Risk of Out-of-Hospital Death
The incidence of sudden unexpected death ranged from 27.2 per 10,000 person-years for sertraline to 58.8 per 10,000 person-years for escitalopram. The incidence of all deaths ranged from 50.3 per 10,000 person-years for sertraline to 84.9 per 10,000 person-years for paroxetine.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5916769/
Lexapro and Celexa are selective serotonin reuptake inhibitors (SSRIs). Both drugs are used to treat depression. Lexapro is also used for anxiety. In 2018, the drugs’ labels included a black box warning for an increased risk of suicide. Other side effects include birth defects.
https://www.drugwatch.com/ssri/lexapro-celexa/
https://www.webmd.com/drugs/2/drug-63990/lexapro-oral/details
Nausea, dry mouth, trouble sleeping, constipation, tiredness, drowsiness, dizziness, and increased sweating may occur. If any of these effects persist or worsen, tell your doctor promptly.
Your doctor has prescribed this medication because he or she has judged that the benefit to you is greater than the risk of side effects. Many people using this medication do not have serious side effects.
Rarely, males may have a painful or prolonged erection lasting 4 or more hours. If this occurs, stop using this drug and get medical help right away, or permanent problems could occur.
Great read, Scott. Can you comment on whether the dual action of Lexapro could explain the most difficult side effects related to lower sexual desire and sexual (dys)function? This is the most difficult part of taking a higher dose (30 mg) for me.
Scott is mistaken here: “Sometime around 2011, the FDA freaked out that high doses of citalopram might cause a deadly heart condition called torsade de pointes, and lowered the maximum dose to prevent this… When escitalopram was invented, it inherited its parent chemical's unusually-low maximum dose, and remains at that level today.”
Escitalopram was approved in 2002, which was before 2011 when the FDA freaked out about torsade de pointes. In the original, 2002, label of escitalopram the maximum dose is 20 mg and is based on the lack of improvement at the higher dose, not on torsade de pointes (see the 2002 label here https://www.accessdata.fda.gov/drugsatfda_docs/nda/2002/21-440.pdf_Lexapro_Prntlbl.pdf p.20): “The recommended dose of Lexapro™ is 10 mg once daily. A fixed dose trial of Lexapro™ demonstrated the effectiveness of both 10 mg and 20 mg of Lexapro™, but failed to demonstrate a greater benefit of 20 mg over 10 mg (see Clinical Efficacy Trials under Clinical Pharmacology). If the dose is increased to 20 mg, this should occur after a minimum of one week.”
However, the good news is that based on the data from this review https://sci-hub.se/10.1007/s40263-014-0196-9 40 mg escitalopram would have QTc prolongation slightly higher than 40 mg of citalopram, and thus should be safe: Change in the mean QTc interval from baseline was 8.5 ms with citalopram 20 mg/day, 18.5 ms with citalopram 60 mg/day … 4.5 ms with escitalopram 10 mg/day, and 10.7 ms with escitalopram 30 mg/day.
Have antidepressants or any other psychiatric medications been shown to increase sex drive or romantic/social interest? Or is it mainly a mood fix. What about energy levels?
Hi Scott,
This is only tangentially related to the article, but I assume as a psychiatrist you have experience with fluvoxamine. What's your take on the studies showing fluvoxamine prevents cytokine storms in COVID-19 patients drastically reducing the risk of death or hospitalization? My read is the FDA and CDC have been predictably stupid in not authorizing and recommending it's use, but I'd be interested in your thoughts on the safety efficacy tradeoff
I was prescribed escitalopram at 20mg for several years. I did score better on the phq-9 compared to untreated, but was still unable to participate normally in life. After five years, it just stopped working - I crashed like a truck going off a cliff. With the help of a new psychiatrist, I switched to sertraline at a much higher dose (titrated, with serum checks) and finally got to moderate functionality. The persistent caution about higher doses of escitalopram, even in patients with stubborn depression, has caused years of my life to be lost in the weeds. This is combined with the fact that most MHPs I've met have been reluctant to change medications, and seem to choose Lex as their first choice. Given the imposed ceiling on dosage, I'm at a loss as to why it's so popular.
Does body weight affect dosage? If I'm twice as big, do I need twice as much?
"I will add my own anecdotal evidence here, which is that I had a patient who spent a year pretty depressed, got very modestly better on Lexapro 20 mg, and agreed to let me experiment with higher doses of Lexapro on him. He got better still at 30 mg, and even better when we increased to 40 mg. I am relatively confident these were real effects."
How can you distinguish that from "remission needed a long time, and you kept raising their Lexapro dose while they waited"?
In my experience, research has this annoying feature where you have some observation of an effect, and it's pretty clear. But then you do an experiment and the data still shows an effect. And then you do the stats and it looks marginal.
And then you do a fuck-ton of experiments and more stats and it still looks pretty shit. And your paper eventually gets published but looks worse than those of people who do much less.
I'm reasonably convinced that it's the inherent nature of a supposedly 'fair test' experiment to
give poor-looking effect levels on results which are individually convincing.
At least two other factors may participate in the phenomenon analyzed here: 1) GI absorption differences and 2) changes in intestinal microbial ecosystem.
1) Different SSRIs have different gastrointestinal absorption. In this paper, “Clinical Pharmacokinetics of Selective Serotonin Reuptake Inhibitors” (https://tinyurl.com/w6zc5mep), Table III shows GI absorption differences between, for example, Fluoxetine, Paroxetine and Citalopram of 80%, >= 64% and ~100%, respectively. The primary datasets analyzed and referenced above use the by-mouth dose instead of plasma levels of the different SSRIs. OTOH, the graph above labeled “(a) Drug (SERT occupancy)” uses a dosing method of micro dialysis. This means a microprobe was inserted into the interstitial area of a rodent brain to administer and/or measure directly the SSRI available to the main targeted site of the serotonin transporter. There are serotonin receptors in the epithelia of the GI tract where by-mouth dosing of SSRIs also interact. The micro dialysis method validates the targeted SSRI effect but doesn’t measure the GI tract serotonin receptor effects.
2) There is increasing evidence that the gut microbiome influences/participates in/?causes? psychiatric effects both “good’” and “bad”. Different antidepressants have demonstrated different antimicrobial effects. See “Unravelling the antimicrobial action of antidepressants on gut commensal microbes” (https://tinyurl.com/yfu3zeeh). Those antidepressants with less GI absorption travel throughout the length of the GI tract influencing the microbial ecosystem and its outputs along the way. The outputs of the microbial ecosystem include neurotransmitters, SCFA and many other small molecules that get absorbed and influence the host health. These phenomenon may cause sufficient GI distress that the patient stops taking the SSRI. The effect on the microbiome is still present even if there is no overt GI distress.
How do you decide whether a post like this will be posted here or on the pysch blog (Lorien?)?
"I hope the three of you who understand stereochemistry appreciate that title as much as it deserves." I understand stereochemistry and I fucking loved that title.
I'd be surprised if the original title of this post wasn't something to the effect of "Oh, the places you'll go, when you're dosing Lexapro". I wonder why it got the more dry title
1/4 of the "minimal dose" of Sertraline worked for me.
I was given the tablets with the lowest dose available and told to split them to ramp up to the minimal therapeutic dose with was 1 tablet. Being the very careful person that I am, I didn't see why I shouldn't ramp up even more gently than that and split the tablet into fourths. I stayed at that dose, because it worked. My psychiatrist mocked me, calling it homeopathy. But I had the antidepressant effect with side effects and everything. Anyways, that "therapeutic relationship" didn't last for some reason.
Basically the same occurred years later with Mirtazapine. I feel lucky, because Mirtazapine is a very effective antidepressant and I managed to dodge the side effects. Considering that during the first week I almost fell asleep at work (on 1/4 of the "minimal dose") I understand why it gets lousy patient reviews. I definitively feel low dose Mirtazapine should be prescribed more.
Also, pet peeve: the lowest therapeutic dose of psychiatric medication is way lower than conventionally believed. Because psychiatric diseases are often several things lumped together you will always have responders and non responders (or less responsive responders). So you need to give a high enough dose that the effect on the responders gives you a statistical signal even though the other half (or something) is not responding. But in clinical practice you try an SSRI and if that doesn't work, Wellbutrin, so you are matching people with "their" antidepressant.
In my case I respond tremendously well to 10mg of fluoxetine with no side effects, but if I take 20mg I get constant headaches. It has a much longer half-life than all the others which probably makes it safer in terms of accidentally missing or accidentally doubling doses and easier to get off of. If the other SSRIs are roughly equivalent in terms of efficacy, has the world benefitted at all from other SSRIs existing? Is it just a bunch of rent-seeking pharma companies making me-too drugs that are no better than prozac except for having a patent?
"But the details here are really confusing and seem to mostly involve differences in the placebo group that don't make sense."
I wonder if this could be explained by a difference in the mechanics of a fixed-dose placebo vs. a variable dose placebo. A fixed-dose placebo is simple, but a variable-dose placebo involves potentially repeated conversations with the care provider that result in changing the "dose" of the placebo. It seems at least possible that this interaction could result in an increased chance of a placebo being suspected, which could result in a decreased placebo effect, which could result in an increased difference between treatment and placebo groups. Although if this story is correct, I think that would mean that variable-dose is no more effective than fixed-dose.
* * *
"I hope the three of you who understand stereochemistry appreciate that title as much as it deserves."
While I appreciate the flattery, I'm betting a lot more than three of us got the joke.
Forgive me for being confused, but if the FDA says that 20mg is the maximum safe dose of Lexapro, how are Scott and others able to prescribe higher doses?
Why do most studies use intention-to-treat despite the problems Scott details above?
It's to avoid issues with selection bias. If you only count patients that actually take the drug instead of people who quit after day 1, you bias the results because there might be some fundamental underlying difference between quitters and non-quitters. For instance, people who successfully follow instructions tend to have better health results across the board, so if you only measure outcomes for those who don't quit, it will tend to make your treatment look more effective than it actually is.
Using ITT makes it harder to detect a real result, but in exchange when we do have a result we are more confident that there's actually a causal relationship.
That's such a beautiful title for a paper...
Come. The obvious explanation is the stereochemistry. With racemic citalopram, for example, half the drug you're getting gives you all the side effects (and burden on drug clearance pathways) with no benefit at all. For a real hair-raising illustration of how *important* stereochemistry is to biochemistry, just compare the effects of R- and S-thalidomide.
I started with 20mg Lexapro and found that it cured much of my Autism: my social reflexes moved so fast I could actually keep up with a conversation involving more than 2 people! I could turn into a total motormouth at times, thinking of interesting things to say faster than I could say them. I had the sort of verbal intelligence and wit that I otherwise have only had while being very angry.
This effects went away, so I increased the dose to 20mg. The pattern repeated. I didn't increase the dose beyond that, I just kept taking this pointless pill for a few years because quitting made me irritable. I eventually quit cold turkey, which turned out to be a lot easier than I expected.
I often feel like a character on a sitcom where the you can have wacky plotlines for one episode - In this episode, my life doesn't suck! - but everything has to go back to normal in time for the end credits. I get just a glimpse into the world of normal people.
I have also discovered that increasing the dose of lexapro up to 30 will often work in patients for whom 20mg has stopped working. I have never tried going higher than that, but maybe I will now that I see it's reasonably well tolerated. However, I have also noticed in my practice that lexapro causes more sexual dysfunction than other SSRIs, which could be related to it's slightly different mechanism of action. Lexapro has been my go-to starting SSRI for depression/anxiety because it seems to have pretty high pt satisfaction, and I will often just switch pts if they experience sexual side effects. However between the sexual side effects and the withdrawal I do wonder if I should just be starting most people on zoloft or prozac and switching if they don't work. *shrug*
I think TSC is right about the timing and approval of Lexapro dosing; it had nothing to do with Citalopram FDA debacle which came much later. Scott, I think you can add support to you argument (in addition to the collective clinical experience and some questionable effectiveness studies you cited) from the receptor binding site literature:
https://pubmed.ncbi.nlm.nih.gov/17497139/
https://academic.oup.com/ijnp/article/10/6/777/804783
https://www.sciencedirect.com/science/article/pii/S0925492715301335?casa_token=Zp9WEOeEQc0AAAAA:sCIUMuc2CZpeqsAvRSVsnXMcuM3RDn3x-j5XWAYbjv1z0duZOxFKy1tgihPhRyoJAG-62JUbjQ5w
Intro to use of PET for drug development if needed:
https://onlinelibrary.wiley.com/doi/full/10.1002/npr2.12084
This post made me think of this paper, by Fava et al (paywalled, but can be found elsewhere for free): https://journals.lww.com/psychopharmacology/Abstract/2002/08000/Double_Blind_Study_of_High_Dose_Fluoxetine_Versus.8.aspx
In short, the paper seems to support the notion that increasing to a high dose of an SSRI is a valid strategy in patients who haven't responded to a lower dose; that is, in some patients, high-dose SSRIs are better than low-dose SSRIs.
The title of the paper is "Double-Blind Study of High-Dose Fluoxetine Versus Lithium or Desipramine Augmentation of Fluoxetine in Partial Responders and Nonresponders to Fluoxetine", which pretty much says it all. In a sample comprising an equal mix of nonresponders and partial responders to fluoxetine 20mg/day, patients were randomized to high-dose fluoxetine (increasing to 40-60mg/day), or adding lithium or desipramine (two common strategies for patients who haven't responded to an SSRI). There were no significant differences in response rates between the three groups, though there was a nonsignificant trend toward fluoxetine being superior.
I do notice that their dose of desipramine was, in my (unqualified) opinion, too low, and their lithium blood levels were definitely on the low end. This is a limitation of this study, but probably not enough to entirely impeach the result.
I was prescribed anti depressants - after a week I felt awful so I quit taking them. When my brain de-fogged, I changed jobs, left my husband and joined a sports club. Sometimes drugs are not the answer.
Hi Scott, thanks for this. Furukawa is really interesting!
Just to be clear, there are zero randomised trials comparing head-to-head different doses of Escitalopram? If that's right, then I think Furukawa is the best evidence we have.
Notable that in Figure 1 of the appendix, Citalopram plateaus at ~30mg not ~20mg. This fact, plus the idea that the only _downside_ of a higher dose seems to be the higher risk of side effects, would imply that at least trying Escitalopram 30mg or 40mg is a great idea.
I wish Furukawa gave us scatterplots in addition to the splines, the absence of scatterplots feels a tad suspicious (am I misundertanding something? I only skimmed). I might try to make the scatter plots based on the table in the appendix myself.
Another thing that would be really cool is a Bayesian hierarchical model of SSRIs dose-response schedules, to generalise this idea of learning from one SSRI and, if approrpriate, transfering to another. I'm pretty sure this specific idea hasn't been done. [Added a few minuted later] Holy Shit it _has_ been done, first page of Google: "A Bayesian dose–response meta-analysis model: A simulations study and application" (Hamza, Cipriani, Furukawa, Egger, Orsini, and Salanti, 2021) [1]. My reading of Figure 3 is they don't extend the x-axis beyond 30mg fluoxtetine-equivalents for Escitalopram, which makes this pointless for answering our question? 🙄
[1] https://journals.sagepub.com/doi/10.1177/0962280220982643
Thank you very much for this article, it's very interesting. I've been on 20mg Lexapro for a long time, along with Welbutrin to counter some of the side effects. I have had a manic episode before so I'm definitely cautious about using too high a dosage. I've tried 40mg for just a few days and don't recall notice anything much, but it may not have been long enough to tell.
How long was your patient on 30mg and subsequently 40mg before they noticed differences?