Thread by @psybalazs

Here is my breakdown of the new #psychedelic #microdosing paper published yesterday (tinyurl.com/mrbhk3xd) that was 🔥🔥🔥 since all these flyers were going around at the wonderland meeting👇🍄💊🚨

Healthy males were randomised into LSD (n=40) and placebo (n=40) groups. They received 14 doses of either 10μg LSD or placebo every 3 days for 6 weeks.

Participants were naïve to MDing, but 70% in both groups had prior psychedelic experience (last use over a year ago), so by in large this was NOT a psychedelic naïve sample.

The long term outcomes (that is difference from baseline to post-treatment) are easy: "No credible evidence of longitudinal changes to traits, mood, or cognition was found", even if blinding integrity is not considered.

For context: doses similar to the 14*10=140μg LSD used to MD here have shown to produce persistent positive effects on anxiety in macrodose therapy setting tinyurl.com/wh3cx8bn

The acute effects are more interesting, here a number of outcomes were positive when ignoring blinding integrity (-angry, +creative, +energy, +happy, -irritable, +well, +connected). This is good news, but look at the effect sizes (supp table 9)!

All outcome was measured on a scale of 100.
The largest difference was observed on the 'happy' scale where the difference is ... a grand total of 3 points! The avg effect size is merely 2 points (supp table 9)! Std effect sizes are not reported, but these are very small effects!

Effect sizes matter when evaluating efficacy, we should always put that next to p-values (tinyurl.com/mrpbtjkh). Effect size particularly matters here, because effects are also transient

Importantly, as the paper notes "the LSD group was partially unblinded". To overcome blind breaking, a subsection of data was re-examined, where participants guessed their drug as 'I do not know', corresponding to a 'blinding worked' subsample.

In this 'blinding worked' subsample 5/7 acute effects are gone (all except 'well' & 'energy') and effect sizes remain tiny (for this 'blinding worked' subsample there is no frequentist analysis, by 'no effect' I mean that the Bayesian credible interval crosses 0; supp fig 7).

🆒side note: in our blinding focused reanalysis of the self-blinding trial (psyarxiv.com/cjfb6/), we concluded "microdosing increases self-perceived energy beyond what is explainable by weak blinding", in sync with these results! (we did not have an item analogous to 'well')

The paper also has some analysis on the difference between what participants expected and then what they experienced, but I found this analysis secondary and not directly relevant to efficacy

As for adverse events, "participants who reported at least one AE in the LSD group was 85.0% and in the placebo group was 80.0%" or 337 vs 216 total adverse events (supp table 7). These numbers are obviously fairly close, but 10% in the LSD group stopped due to adverse events

Most classic antidepressant trials only include dropout rate without the reason for dropping out (could be also due to the adherence inconvenience, lack of efficacy etc.). Thus, I cant compare this 'dropout due to AE' rate to classic ADs - let me know if you know more here

In summary:
(1) v small, but significant acute effects if blinding is ignored
(2) effects from 1 are mostly gone and remain v small in the truly blind subsample
(3) no persistent long term benefits even if blinding is ignored
(4) 10% dropout due to adverse events
... exciting?

cant resist my ego's urge to highlight that qualitatively we found the EXACT same results with the self-blinding microdose study (tinyurl.com/2as4n566) for ~1% of the costs of this study: very small transient effects that are mostly gone when blinding integrity is considered.

Overall, I feel these results are similarly lukewarm as other trials in healthy samples (e.g. tinyurl.com/yv2636rv & tinyurl.com/ym9rd29u) - best case scenario is a small effect.
#microdose research needs to turn towards patient populations to see if there is more there.