This supports my speculation from another study, found here.

Brilliant study. Astonishingly, ~500 mcg thyroxine only increased FT3 about 20% in depressed patients. In controls, it nearly doubled FT3 (about 80# increase), Thyroxine doubled total T4 and total T3 in both depressed and controls. That lines up with the FT3 rise in controls, but is still much lower than I'd expect from this dose.

Note that the depressed patients had equal thyroid symptoms before taking thyroxine, so experienced no increase. The pre-treatment depressed patients, treated depressed patients, and treated controls all had an equal number of thyroid symptoms. Thus, only the pre-treatment healthy patients lacked hyperthyroid symptoms. This is interesting because the treatment was 500 mcg thyroxine (average 484 mcg), which you'd expect to cause more symptoms.

This could mean either that the depressed patients had as many symptoms alleviated as they had caused by thyroxine, or that depressed patients simply had lower thyroidergic activity. The latter is supported by the fact that free T3 and free T4 rose 2 or 3 times less in treated depressed patients compared to treated controls (from full text).

It's also possible that the pre-existence of symptoms masked their appearance (in which case increased severity might be expected). I doubt that this is the case. A few symptoms trended toward improvement.

The individual change in REE to a given change in dose varied radically. The dose change was only 50 mcg, which seemed to change TSH about half what they've calculated for 15% REE change. That is to say, it looks like a 50 mcg dose reduction will only raise TSH from 1.0 to 5.0, not to 10.0. This is my personal guesstimate, and will need to be properly calculated from other studies. My point is merely that 50 mcg will not cause the full 15% REE increase. Additionally, the response to 50 mcg may depend on initial TSH.

It looks like TSH changes that remained hyperthyroid (i.e. bellow 1.0) had little effect on REE. This might be because the body is maintaining thyroid status, or that REE is more like an on/off switch. However, this only covers relatively small changes in thyroxine dose.

It is unclear how supraphysiological doses effect REE. It seems likely that the ten times greater dose (500 mcg) used for depression would significantly increase REE.

This is the same dosage that another study found to be indiscernible from placebo. I suspect a couple things are at play.

The first is that these are treatment-resistant patients, meaning that we already know they are not susceptible to placebo; given that we are effectively comparing combination therapy to SSRIs alone, this means that the SSRI only stage is not experiencing placebo, thus making the therapeutic effect visible by comparison (retrospective comparison, given that there is no control group).

The second is that this is combination therapy. It seems likely that thyroxine is more effective when combined with SSRIs.

Note that this study uses much lower doses than similar studies I've seen, yet had similar results in terms of efficacy rate (roughly 50% responders). I'm not currently certain how the magnitude of effects compares with those studies. Namely, comparing with Pfeiffer et al (350 mcg), Rudas et al (235 mcg) and Bauer et al (482 mcg), and also Bauer et al from 2016 (300 mcg)

This is another study consistent with the 50% remission figure. The dose is also similar: between 235 mcg (Rudas et al) and 482 mcg (Bauer et al).

Fascinating. At 100 mcg, I'd expect one to be able to tell the difference. This is especially surprising given the two groups. Namely, the symptomatic group and the healthy control group. I'd expect at least one group to be able to tell the difference. However, it's worth noting that the TSH between the two groups were virtually identical. The groups were selected based on hypothyroid symptoms rather than actual thyroid status.

The fact that healthy controls could not tell the difference is odd. It is both not what I expected and cuts against what the related lesswrong article says about discernibility to healthy subjects (odd given that lesswrong mentions this specific study). What may be happening is that the question was not specific enough. If subjects interpreted reduced vitality as a sign that they were not receiving thyroxine, then many might get is wrong. I wish they has a more thorough questionnaire on perception of drug effects. Even an informal "describe the experience" question would be nice.

All in all, I find it unlikely that this is accurate. There are at least two possibilities. It could indicate that the dose was not high enough, in which case the study is not testing what it thinks it's testing. That is supported the the fact that the TSH of the thyroxine group is barely out of the normal range. The other option, as mentioned above, is that they are not asking the right questions. If that is the case, that also seems to invalidate the findings.

In conclusion, this result means that we can't trust the study. I'm certain higher doses would be discernable from placebo to subjects. It is likely that it was discernible at this dose (100 mcg) if the right questions were asked. I'm thankful they included this outcome.

P.S. I found this study completely by accident, then found out it was related to that lesswrong article that I've always appreciated.

This is just as I expected, though I'd also be unsupervised by opposite findings. I'm yet to find why symptoms such as insomnia are so often cited for hyperthyroidism. I see a few possibilities. One is that it's very slow onset effect that takes more than 2 months to develop (personally, this seems unlikely). Another is that it only effects people with preexisting anxiety or hyper-arousal (strikes me as a likely partial explanation). A third is that it only effects people with the most extreme hyperthyroidism (also strikes me as partial explanation). Finally, it's possible that the entire thing is a myth. Authors of another study from 2011 noted that "[sleep] is being characterized as poor without further elaboration." I think it may be the case that people are just assuming patient's sleep issues are caused thyrotoxicosis because it seems like it would, when in fact only a small fraction actually are. Perhaps thyrotoxicosis even turns depressive insomnia into anxious insomnia, thereby confusing physicians.

The next question is whether treating light to moderate hyperthyroidism would resolve insomnia. This would answer some of the above possibilities. However, I'm uninterested in severe hyperthyroidism because it is above the maximum treatment dose I commonly see, namely 500 mcg thyroxine.

25.2 out of 33 is quite high (i.e. very positive results). In fact, those results seem too good to be true, and are likely due to the limitations of the study. In particular, by 54 months of treatment the non-responders or those with side effects would have already stopped taking thyroxine. Nonetheless, the data is consistent with other studies indicating that supraphysiological doses of levothyroxine has few and mild side effects (regardless of the sample population).

Thyroxine working in about 50% of patients was also found in Bauer et al. This dose was not quite as high as Bauer, but 235 mcg is still above a full replacement dose.

I'd like to see results on minor depression. However, such studies are unlikely.

This is the highest dose of levothyroxine that I've seen administered. Even treatment for thyroid cancer rarely goes beyond 300 mcg (or even 200 mcg, which is more common).

This is the only group with a TSH significantly below the reference range. The reason, of course, is that the patients with the most symptoms ended up getting the highest doses of thyroxine (T4) by the end of the study.

This is all very interesting. It is also true that, in the treatment of depression, there is a relatively low dropout rate due to side effects from treatment with T3 (liothyronine).

liothyronine (T3) augmentation in the treatment of depression