Well, I can think of two possible take homes from this data.
First the flippant one.
Don't prescribe Xanax.
Xanax isn't substantially different from other benzos.
Don't prescribe benzos.
Now the more serious one.
Anyone who thinks they can pin down the inevitably subjective effects of psychiatric drugs by citing the various objective biochemical performance indicators is pissing into the wind. The measure is how they effect the patient.
Most uncontroversially it's because pain and mood are the two symptom classes most susceptible to the placebo effect. Psychiatric drugs largely live or die on how effective they are as placebos. So you need to ask the person taking them, not the instruments measuring them in vivo or in vitro.
But perhaps more importantly, if only because it's a huge epistemological black hole at the centre of psychiatry into which few practitioners are prepared to gaze, the mind can neither be meaningfully reduced to chemical reactions nor to whatever subset of the attributes of those reactions we're able to quantify.
This isn't to say the mind isn't entirely emergent from physics and chemistry. I'm not flogging Cartesian dualism here. But the proximate and remote causes giving rise to it are so multitudinous and interact so complexly that imagining you can read them off the instruments of chemical analysis (or the symptom checklists in the DSM and ICD bibles) is not science. It's pseudoscience built on overextended, oversimplified scientism.
I've long been really puzzled that diazepam doesn't have a much longer duration of action. The fact that taking another dose the next seems to have very similar effects as the first dose makes me skeptical of a purely tolerance based effect.
I noticed the remark about diazepam disassociating from the receptors more quickly but I'm not sure I followed why/if that would explain why it's effects wear off so much more quickly than the half-life would predict. Is it somehow being stored in an unavailable form?
There may be more than one thing going on with diazepam, including redistribution given its lipophilicity. I'm not sophisticated enough in my knowledge about tolerance mechanisms to comment about how rapidly you can see tolerance develop and decay. I don't see why this couldn't happen, though. I also came across a paper that showed that diazepam activates different GABA-A receptor isoforms depending on concentration, so there might be some other receptor oddness going on.
That remark you're referring to isn't meant to comment on why the offset happens so quickly, but rather why the onset might lead serum levels.
Yes, I didn't mean to suggest that tolerance couldn't be part of the mechanism but the fact that taking another dose doesn't feel substantially less efficacious at that time even though based on the half life it would seem like there must be a substantial decrease in the dose-response makes it seem like there must be something more complex going on.
Thanks for clarifying about the disassociation, I'd misunderstood that.
The first time I thought at all past the received wisdom about benzos (Xanax is the devil, Librium is the one for outpatient alcohol withdrawal - also, aside, is Librium ever prescribed for anyhing else and what's up with its non "-pam" suffix?) was when I was trying to taper patients off of chronic BZDs and saw how incredibly imprecise conversion between different BZDs was.
Using the MDCalc conversion calculator, for example, 1mg of clonazepam = as little as 0.5mg or maybe as much as 4mg lorazepam.
Because it will probably make no functional difference, I may just choose to believe that alprazolam does have a very short half-life after all so I don't have to take benzos for my own disabling cognitive dissonance
Just wanted to say thanks for another myth-checking post. As an addiction psychiatrist, I’ve always heard (and repeated) the idea that alprazolam is the worst, most addictive benzo – while diazepam is the long-acting, “safer” one we use to treat sedative use disorder. Your breakdown made me question that and actually go looking for some hard data.
Turns out… I couldn’t find much! I was surprised to realize that I couldn’t track down any meaningful papers comparing benzodiazepines (or Z-drugs) head-to-head in terms of addiction potential or misuse risk. It feels like a lot of the "common knowledge" on this topic really might be more lore than evidence.
Anyway, I really appreciate the work you put into this – it’s not often I read something that genuinely shifts how I think about clinical practice. Kudos!
Thanks very much for the kind words, it means a lot to hear that my work is having a real impact on how others practice (hopefully for the better!)
As an aside, I've grown increasingly more skeptical about the whole narrative around benzodiazepines being particularly addictive (as opposed to causing physical dependence) even with long-term use. The latest blow I think comes from the Rosenqvist paper in AJP from 2024: https://ajp.psychiatryonline.org/doi/10.1176/appi.ajp.20230075
A mentor of mine who is a bit of an authority on the benzodiazepines has made a compelling argument that a big part of their reputation for abuse comes from:
1. Being used as accessory drugs in the context of another primary SUD (i.e. the individual is not actually addicted to benzodiazepines, but we assume that they are); his reading of the literature indicates that primary benzodiazepine use disorders are quite rare in a way that OUD and AUD are not.
2. They are often prescribed to individuals with Cluster B traits (e.g. for panic attacks or "mood swings"), and these patients are quite likely to overuse just about any prn medication given to them because they so rarely provide any sort of meaningful relief.
That’s a really interesting perspective. Reflecting on my experience running an addiction psychiatry clinic over the past few years, I’d have to agree—cases of "pure" sedative use disorder are indeed rare. The ones that come to mind were in patients with significant personality disorders and physical comorbidities.
We do have good data showing that the percentage of patients on chronic benzodiazepines increases with age, even though the risks associated with benzodiazepine use also rise as patients get older. There’s strong evidence linking chronic use to increased morbidity and mortality, and interestingly, discontinuation itself appears to raise mortality rates as well (https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2813161).
It feels like a lot of the "common knowledge" surrounding benzodiazepines could benefit from a fresh injection of data-driven analysis.
Those rising numbers with age might reflect the fact that there are some segment of patients that just do really well on benzodiazepines and nothing else, and so their numbers in a population naturally increase as those individuals eventually figure that out.
I also can't help but wondering if chronic use is really a proxy for a lot of other things that predict poor outcomes (e.g. like having a really really bad anxiety disorder that you need to take benzos for every day), but that's so obvious that I'm fairly sure efforts were made to control for that.
I've been relatively surprised to come across conflicting evidence regarding the problems with chronic benzo use. There's that 2007 paper that showed no difference in rates of falls in elderly populations despite a 50% drop in prescribing rates (https://www.acpjournals.org/doi/10.7326/0003-4819-146-2-200701160-00004). There's also that 2020 AJP article that showed no association with benzos or Z-drugs in terms of dementia risk; in fact, the population with the highest use look like they had a *reduced* risk relative to the others. (http://ajp.psychiatryonline.org/doi/10.1176/appi.ajp.2019.19030315)
Do you think that these are just a couple of null results that you would expect in any large body of research, or do you think that the association between chronic benzo use and poor outcomes is maybe more murky than we're taught?
Any thoughts on dosing? People are often surprised when they see that the original studies went up to 10 mg a day. I suspect that this medication partially has a bad reputation because people are persistently moderately underdosing it
Seems unlikely to me that dosing is the issue here if you look at the relationship between serum level and psychomotor effects in the Ellinwood paper. Return to near-baseline performance happens while within a few ng/mL of peak serum concentrations, it's not like there just isn't enough in the system to sustain an anxiolytic response.
I think the often-cited long-lasting action of Diazepam has much to do with its active metabolites that are especially noticable in the first two weeks, where I and many others report waking up with a sort of "afterglow" in the morning that persists into midday. However, this effect disappears (or better, becomes much less pronounced) as soon as the slightest bit of tolerance sets in. It reminds me of the Wellbutrin-honeymoon, which also has people report fantastical effects in the first two weeks, only to then disappear quickly. Combine that with modern benzo-regimens, where they are rarey prescribed for periods above 6 weeks (although Diazepam is one of those few benzos that are still considered feasible for long-term by some), I think it at least makes perfect memetic sense how they would gain that reputation.
The UpToDate article on benzodiazepine withdrawal has an interesting bit of information that I didn't know previously:
"[...] аlprаzolam and triazolam are triazolobenzodiazepines with a triazole moiety not found in other bеոzοԁiаzерines, which may confer a binding affinity to a unique subset of bеոzοԁiazерine receptors and thus limit cross-reactivity"
That is to say, withdrawal from chronic alprazolam use might not respond (as well) to other benzodiazepines - a good reason not to prescribe it in the first place I'd say.
Well, I can think of two possible take homes from this data.
First the flippant one.
Don't prescribe Xanax.
Xanax isn't substantially different from other benzos.
Don't prescribe benzos.
Now the more serious one.
Anyone who thinks they can pin down the inevitably subjective effects of psychiatric drugs by citing the various objective biochemical performance indicators is pissing into the wind. The measure is how they effect the patient.
Most uncontroversially it's because pain and mood are the two symptom classes most susceptible to the placebo effect. Psychiatric drugs largely live or die on how effective they are as placebos. So you need to ask the person taking them, not the instruments measuring them in vivo or in vitro.
But perhaps more importantly, if only because it's a huge epistemological black hole at the centre of psychiatry into which few practitioners are prepared to gaze, the mind can neither be meaningfully reduced to chemical reactions nor to whatever subset of the attributes of those reactions we're able to quantify.
This isn't to say the mind isn't entirely emergent from physics and chemistry. I'm not flogging Cartesian dualism here. But the proximate and remote causes giving rise to it are so multitudinous and interact so complexly that imagining you can read them off the instruments of chemical analysis (or the symptom checklists in the DSM and ICD bibles) is not science. It's pseudoscience built on overextended, oversimplified scientism.
I've long been really puzzled that diazepam doesn't have a much longer duration of action. The fact that taking another dose the next seems to have very similar effects as the first dose makes me skeptical of a purely tolerance based effect.
I noticed the remark about diazepam disassociating from the receptors more quickly but I'm not sure I followed why/if that would explain why it's effects wear off so much more quickly than the half-life would predict. Is it somehow being stored in an unavailable form?
There may be more than one thing going on with diazepam, including redistribution given its lipophilicity. I'm not sophisticated enough in my knowledge about tolerance mechanisms to comment about how rapidly you can see tolerance develop and decay. I don't see why this couldn't happen, though. I also came across a paper that showed that diazepam activates different GABA-A receptor isoforms depending on concentration, so there might be some other receptor oddness going on.
That remark you're referring to isn't meant to comment on why the offset happens so quickly, but rather why the onset might lead serum levels.
Yes, I didn't mean to suggest that tolerance couldn't be part of the mechanism but the fact that taking another dose doesn't feel substantially less efficacious at that time even though based on the half life it would seem like there must be a substantial decrease in the dose-response makes it seem like there must be something more complex going on.
Thanks for clarifying about the disassociation, I'd misunderstood that.
OMG. This was a fabulous and humbling post.
The first time I thought at all past the received wisdom about benzos (Xanax is the devil, Librium is the one for outpatient alcohol withdrawal - also, aside, is Librium ever prescribed for anyhing else and what's up with its non "-pam" suffix?) was when I was trying to taper patients off of chronic BZDs and saw how incredibly imprecise conversion between different BZDs was.
Using the MDCalc conversion calculator, for example, 1mg of clonazepam = as little as 0.5mg or maybe as much as 4mg lorazepam.
Because it will probably make no functional difference, I may just choose to believe that alprazolam does have a very short half-life after all so I don't have to take benzos for my own disabling cognitive dissonance
Just wanted to say thanks for another myth-checking post. As an addiction psychiatrist, I’ve always heard (and repeated) the idea that alprazolam is the worst, most addictive benzo – while diazepam is the long-acting, “safer” one we use to treat sedative use disorder. Your breakdown made me question that and actually go looking for some hard data.
Turns out… I couldn’t find much! I was surprised to realize that I couldn’t track down any meaningful papers comparing benzodiazepines (or Z-drugs) head-to-head in terms of addiction potential or misuse risk. It feels like a lot of the "common knowledge" on this topic really might be more lore than evidence.
Anyway, I really appreciate the work you put into this – it’s not often I read something that genuinely shifts how I think about clinical practice. Kudos!
Thanks very much for the kind words, it means a lot to hear that my work is having a real impact on how others practice (hopefully for the better!)
As an aside, I've grown increasingly more skeptical about the whole narrative around benzodiazepines being particularly addictive (as opposed to causing physical dependence) even with long-term use. The latest blow I think comes from the Rosenqvist paper in AJP from 2024: https://ajp.psychiatryonline.org/doi/10.1176/appi.ajp.20230075
A mentor of mine who is a bit of an authority on the benzodiazepines has made a compelling argument that a big part of their reputation for abuse comes from:
1. Being used as accessory drugs in the context of another primary SUD (i.e. the individual is not actually addicted to benzodiazepines, but we assume that they are); his reading of the literature indicates that primary benzodiazepine use disorders are quite rare in a way that OUD and AUD are not.
2. They are often prescribed to individuals with Cluster B traits (e.g. for panic attacks or "mood swings"), and these patients are quite likely to overuse just about any prn medication given to them because they so rarely provide any sort of meaningful relief.
Curious if you have any thoughts on this!
That’s a really interesting perspective. Reflecting on my experience running an addiction psychiatry clinic over the past few years, I’d have to agree—cases of "pure" sedative use disorder are indeed rare. The ones that come to mind were in patients with significant personality disorders and physical comorbidities.
We do have good data showing that the percentage of patients on chronic benzodiazepines increases with age, even though the risks associated with benzodiazepine use also rise as patients get older. There’s strong evidence linking chronic use to increased morbidity and mortality, and interestingly, discontinuation itself appears to raise mortality rates as well (https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2813161).
It feels like a lot of the "common knowledge" surrounding benzodiazepines could benefit from a fresh injection of data-driven analysis.
Those rising numbers with age might reflect the fact that there are some segment of patients that just do really well on benzodiazepines and nothing else, and so their numbers in a population naturally increase as those individuals eventually figure that out.
I also can't help but wondering if chronic use is really a proxy for a lot of other things that predict poor outcomes (e.g. like having a really really bad anxiety disorder that you need to take benzos for every day), but that's so obvious that I'm fairly sure efforts were made to control for that.
I've been relatively surprised to come across conflicting evidence regarding the problems with chronic benzo use. There's that 2007 paper that showed no difference in rates of falls in elderly populations despite a 50% drop in prescribing rates (https://www.acpjournals.org/doi/10.7326/0003-4819-146-2-200701160-00004). There's also that 2020 AJP article that showed no association with benzos or Z-drugs in terms of dementia risk; in fact, the population with the highest use look like they had a *reduced* risk relative to the others. (http://ajp.psychiatryonline.org/doi/10.1176/appi.ajp.2019.19030315)
Do you think that these are just a couple of null results that you would expect in any large body of research, or do you think that the association between chronic benzo use and poor outcomes is maybe more murky than we're taught?
Got news for you. One of those cats is actually a skunk. Don't get him angry.
Any thoughts on dosing? People are often surprised when they see that the original studies went up to 10 mg a day. I suspect that this medication partially has a bad reputation because people are persistently moderately underdosing it
Seems unlikely to me that dosing is the issue here if you look at the relationship between serum level and psychomotor effects in the Ellinwood paper. Return to near-baseline performance happens while within a few ng/mL of peak serum concentrations, it's not like there just isn't enough in the system to sustain an anxiolytic response.
I think the often-cited long-lasting action of Diazepam has much to do with its active metabolites that are especially noticable in the first two weeks, where I and many others report waking up with a sort of "afterglow" in the morning that persists into midday. However, this effect disappears (or better, becomes much less pronounced) as soon as the slightest bit of tolerance sets in. It reminds me of the Wellbutrin-honeymoon, which also has people report fantastical effects in the first two weeks, only to then disappear quickly. Combine that with modern benzo-regimens, where they are rarey prescribed for periods above 6 weeks (although Diazepam is one of those few benzos that are still considered feasible for long-term by some), I think it at least makes perfect memetic sense how they would gain that reputation.
The UpToDate article on benzodiazepine withdrawal has an interesting bit of information that I didn't know previously:
"[...] аlprаzolam and triazolam are triazolobenzodiazepines with a triazole moiety not found in other bеոzοԁiаzерines, which may confer a binding affinity to a unique subset of bеոzοԁiazерine receptors and thus limit cross-reactivity"
That is to say, withdrawal from chronic alprazolam use might not respond (as well) to other benzodiazepines - a good reason not to prescribe it in the first place I'd say.