I would interpret your findings about the links between lithium, weight gain, hypothyroidism, and DI differently.
Subclinical hypothyroidism affects an estimated 5% of the population, and causes both weight gain and fatigue. So we can point out that we do see an appreciable amount of both subclinical hypothyroidism and obesity in the general population, and they were found to be associated in 1104 subjects ages 10-19 years old.
Lithium may also cause weight gain by other mechanisms, as yet unknown.
The numbers you found show that going on a clinical dose of lithium causes an average of 0-6 kg weight gain above whatever weight gain they may have experienced from a dietary dose. We can imagine that going from a very low dietary dose of lithium (say < 10 ug/day) to a small but higher daily dietary dose (say > 40 ug/day) is responsible for more marginal weight gain than going from 20 ug/day to 200 mg/day. So it may be that an increase over time in dietary lithium doses is responsible for most of whatever weight gain lithium is causing in the population, with a much smaller amount of additional weight is caused by putting people on a clinical dose.
Lithium may cause DI only starting at clinical doses. It may then show a dose-dependent association. We would therefore expect to see a dose-dependent relationship between DI and lithium, but no DI epidemic.
The fact that lithium shows a dose-dependent response at therapeutic doses allows for dietary doses in the neighborhood of 20 ug/day to cause weight gain. What we’d need is evidence that the weight gain caused by lithium disappears at dietary doses.
We don’t have that here. The closest is the Rinker study, which found 13⁄28 MS patients reported weight gain and 10⁄28 reported weight loss. More patients gained than lost weight, so if anything it supports the conclusion that low-dose lithium can lead to weight gain. MS can cause both weight loss and gain generally on its own, so we probably shouldn’t update too much on this one small study.
After learning that 5% of the population has subclinical hypothyroidism, I’m actually more inclined to believe that normal doses of dietary lithium may be causing the obesity epidemic than I was before reading this post. However, my credence is still well below 50%, so I also still agree with your post’s title.
Note: This was originally part of a different comment, but I split it off for clarity.
After learning that 5% of the population has subclinical hypothyroidism, I’m actually more inclined to believe that normal doses of dietary lithium may be causing the obesity epidemic than I was before reading this post. However, my credence is still well below 50%, so I also still agree with your post’s title.
I’m a bit confused about this update. A lot of chronic diseases are very common. And as I showed, hypothyroidism has not been increasing over time. It might be decreasing. And, again, the geographical pattern of hypothyroidism does not match the geographical pattern of obesity.
And as I showed, hypothyroidism has not been increasing over time. It might be decreasing.
I’m specifically considering subclinical hypothyroidism. From the paper you link:
A significant limitation to this report is that the incidence rates of thyroid disorders were based on diagnoses recorded on standardized medical records. Because of this, the findings reflect the rates of thyroid functional abnormalities that were clinically detected and exclude subclinical dysfunction since not all service members are tested for thyroid disorders.
And, again, the geographical pattern of hypothyroidism does not match the geographical pattern of obesity.
If we’re working with the “lithium causes a desire to overeat and be sedentary (via hypothyroidism and possibly other mechanisms) in young people, which leads to lifelong obesity-promoting eating and activity habits” version of the hypothesis, then it may just be that Brazil and China are only now gaining the wealth and food supply to permit this to occur in a large enough swathe of the population. We see obesity growing exponentially in China and India (I couldn’t find a graph for Brazil), while it’s looking like it’s on the concave part of the sigmoid curve in the USA.
I’m inclined to think that if lithium is causing obesity in this manner, that subclinical hypothyroidism is one of its important mechanisms of action. If treating subclinical hypothyroidism in young people did not lead to long-term lower rates of obesity, I would lower my credence in an important dietary lithium-obesity link significantly.
The other paper I linked, which uses nationally representative data from the 2007-2012 NHANES, estimated the prevalence of subclinical hypothyroidism to be 3.5% (lower than the prevalence in the 1988-1994 NHANES, which was 4.3%), and noted that the prevalence of at-risk TSH levels seems to have decreased or remained stable with time. (Although mean TSH levels have increased.)
(I apologize for not having replied earlier — I was curious and wanted to check the raw NHANES data on TSH levels myself before replying to you, so I tried to, but found that parsing one of the datasets was too hard and ended up not doing it.)
[ETA: actually, the 4.3% number is based on a different definition of hypothyroidism. With regards to the definition used in the newer study (TSH > 4.5 mIU/L in the absence of clinical hypothyroidism), it says the following:
Percent reference population with TSH > 4.5 mIU/L for this study was found to be 1.88% which is similar to what was found by Hollowell et al. [Hollowell et al. is the group that analyzed ’88-’94 data]. It would indicate that at risk TSH levels in the reference U.S. population may have decreased a bit or remained at the same level for reference US population.
(I apologize for not having replied earlier — I was curious and wanted to check the raw NHANES data on TSH levels myself before replying to you, so I tried to, but found that parsing one of the datasets was too hard and ended up not doing it.)
No worries, thank you for all the great research you’re doing.
The analysis of ’88-’94 data says:
Hypothyroidism was found in 4.6% of the U.S. population (0.3% clinical and 4.3% subclinical)
The analysis of ’07-’12 data says:
The prevalence rate of clinical hypothyroidism in general U.S. population was 2.4%… The percent population with subclinical hypothyroidism was 3.5% (Table 3).
So we see a 20% decrease in subclinical hypothyroidism (4.3% → 3.5%), but an 800% increase in clinical hypothyroidism (0.3% → 2.4%).
My original argument was based on prevalence in the population, not rate of change across time. If anything, given that (as these papers state), clinical hypothyroidism most definitely is associated with BMI, I think they lend support to the hypothyroidism/obesity explanation. Perhaps what we are seeing is a more rapid move in individual patients from subclinical to clinical hypothyroidism, as a result of the hypothesized lithium contamination. In affected patients, by the time the medical system catches it, it’s usually already clinical, whereas in the (postulated) less lithium-contaminated 80s, there was a longer period of time in the subclinical phase per patient when contact with the medical system could catch and diagnose subclinical hypothyroidism.
I will however make a couple meta notes:
a) I’m not putting as much time into this as you, so I’m worried I’m losing track of the details of the argument.
b) I’m trying to salvage the lithium theory because I don’t think it’s utterly destroyed by this data, not because I think it’s extremely likely.
So I generally just have to apologize if, on reflection, my overall argument here is full of incoherencies and inconsistencies. I’m forming my thoughts as I write these comments, and I expect to change my mind in the future—I’m just not sure in which direction.
So we see a 20% decrease in subclinical hypothyroidism (4.3% → 3.5%), but an 800% increase in clinical hypothyroidism (0.3% → 2.4%).
The abstract of the paper analyzing ’88-’94 data says that they used a different definition of “subclinical hypothyroidism” than the definition that is commonly used today (I had edited my comment to reflect that a few seconds before you replied. I am so sorry for the error!!). Quoting from the paper:
(Subclinical hypothyroidism is used in this paper to mean mild hypothyroidism, the term now preferred by the American Thyroid Association for the laboratory findings described.)
So it seems that the prevalence of hypothyroidism was 4.6% in this survey, not 0.3%. So the prevalence of clinical hypothyroidism has decreased.
With regards to what we nowadays call subclinical hypothyroidism (TSH > 4.5 mIU/L in the absence of clinical hypothyroidism), the paper that analyses ’07-’12 data does say:
Percent reference population with TSH > 4.5 mIU/L for this study was found to be 1.88% which is similar to what was found by Hollowell et al. [Hollowell et al. is the group that analyzed ’88-’94 data]. It would indicate that at risk TSH levels in the reference U.S. population may have decreased a bit or remained at the same level for reference US population.
Note: I’m a little bit sick today, and it’s possible I made a mistake in my stoichiometry or in converting from math to reasoning. If so, I will happily stand corrected if anybody points out my error.
The change in terminology is just verbiage. In fact, it appears they have narrowed the definition of both subclinical and clinical hypothyroidism in the newer paper. In light of how they changed the definitions, we should think that a definition-neutral rate of both subclinical and clinical hypothyroidism has gone up even more than I’d described in my previous comment.
Hypothyroidism is defined in part by lower-than-normal thyroxine (T4). In the earlier paper, T4 levels are defined as “clinical” that would be defined as “normal” or “subclinical” in the later paper. According to the definitions of the later paper, all “subclinical” patients in the earlier paper would have been considered “normal.”
They switched from measuring bound + unbound thyroxine (T4) to free thyroxine (FT4) in the second paper. So the numbers aren’t directly comparable because they’re measuring the molecule in two different states the body. I don’t know whether we can do more than rely on the researchers’ implied claim that the definitions of normal vs. subclinical vs. clinical hypothyroidism remain comparable under the new definition.
Extracts and calculations for legibility:
’88-’94 paper:
… high T4 is a concentration 169.9 nmol/liter and low T4, a concentration 57.9 nmol/liter...
Hypothyroidism was defined as clinically significant if TSH > 4.5 mIU/liter and T4 < 57.9 nmol/liter and as subclinical or mild when TSH > 4.5 mIU/liter and T4 >= 57.9 nmol/liter...
Subclinical hypothyroidism was defined as having TSH levels ≥ 4.5 mIU/L and FT4 within the normal reference range [of 0.6-1.6 ng/dL]. Those who had TSH levels ≥ 4.5 mIU/L and FT4 below 0.6 ng/dL were defined as having clinical hypothyroidism.
Thyroxine (T4) has molecular weight 776.87 g/mol.
In adults, normal levels of total T4 range from 5–12 micrograms per deciliter (mcg/dl) of blood. Normal levels of free T4 range from 0.8–1.8 nanograms per deciliter (ng/dl) of blood.
I haven’t converted these densities to molarities, so I haven’t compared these ranges with those provided by the ’88-’94 paper, but this distinction seems relevant.
Good catch, I will edit the previous comment tomorrow when I’m on my computer. Given that the sub vs clinical distinction turns on T4/FT4 and these papers test for different values, I’d need to give more thought about how comparable they are.
Update: I have now looked into the raw TSH data from NHANES III (1988-1994) and compared it with data from the 2011-2012 NHANES. It seems that, although median TSH levels have increased a bit, the distribution of serum TSH levels in the general population aged 18-80 (including people with thyroid disorders) has gotten more concentrated around the middle; both very high levels (characteristic of clinical or subclinical hypothyroidism) and very low levels (characteristic of clinical or subclinical hyperthyroidism) are less common in the 2011-2012 NHANES compared to NHANES III. You can see the relevant table here. There might be bugs in my code affecting the conclusion of the analysis.
This paper, which pretty much used the same NHANES surveys, looked at a somewhat different thing (thyroid levels in a reference population without thyroid disorders or other exclusion criteria) but it seems to report the same finding w.r.t. high TSH levels: a lower proportion of the population in the latest survey meets the TSH criteria for clinical or subclinical hypothyroidism.
The fact that lithium shows a dose-dependent response at therapeutic doses allows for dietary doses in the neighborhood of 20 ug/day to cause weight gain. What we’d need is evidence that the weight gain caused by lithium disappears at dietary doses.
We don’t have that here. The closest is the Rinker study, which found 13⁄28 MS patients reported weight gain and 10⁄28 reported weight loss. More patients gained than lost weight, so if anything it supports the conclusion that low-dose lithium can lead to weight gain.
Let us not conflate different meanings of “low-dose lithium” here. In the Rinker study, the dose that patients took was still more than 1000x greater than what people in e.g. New Zealand get from their food. That is not a small difference.
Relevantly, this study found no association between serum lithium concentration in the general population in Germany and BMI. Also, as I mentioned in the post, the correlation between log(water lithium levels) and log(obesity %) across Texas counties is negative.
Relevantly, this study found no association between serum lithium concentration in the general population in Germany and BMI. Also, as I mentioned in the post, the correlation between log(water lithium levels) and log(obesity %) across Texas counties is negative.
I was primarily trying to address your local argument about interpreting obesity, DI, and hypothyroidism in this comment. These other pieces of evidence seem important, but just aren’t what I’m focused on.
I’m aware of that. By “low-dose lithium,” I specifically meant in the 30mg/day range, as opposed to the 1000x lower levels in a typical diet. My point is that none of the studies you list can help very well to understand what lithium’s effects are at a normal dietary level. We should not extrapolate our certainty too far out of the range they’ve studied, and as you point out, dietary lithium is very far out of that range. We can and should make that same argument about the link between lithium and weight gain outside of the clinical dose. Just because it causes weight gain at a low or normal clinical dose does not mean it causes weight gain at a dietary dose.
Overall, we should be uncertain about this hypothesis, not certain that it is right or wrong. To me, uncertainty means “worthy of more study.” Of course, it also means “accurately reporting information,” and SMTM has really managed to set themselves up for a takedown by omitting and distorting so much of the information they use to build their case. That said, I retain let’s say 40% credence for a fairly strong version of their “environmental contamination” hypothesis, with perhaps 40% credence that that low ug levels of lithium in a typical diet are perhaps a 30% contributor to contamination-linked obesity.
I’d very much like to understand how your credences can be so high with nothing else to back them up than “it’s possible and we lack some data”. Like, sure, but to have credences so high you need to have at least some data or reason to back that up.
When I wrote the comment, it was mainly because of the prevalence of subacute hypothyroidism in the general population. However, one of Natalia’s studies that I hadn’t been focusing on persuaded me that variations in dietary lithium intake is unlikely to be responsible for so much weight gain. Serum lithium isn’t associated with BMI.
none of the studies you list can help very well to understand what lithium’s effects are at a normal dietary level.
This study found no association between serum lithium concentration in the general population in Germany and BMI. Also, as I mentioned in the post, the correlation between log(water lithium levels) and log(obesity %) across Texas counties is negative.
Sorry, I specifically meant the studies you list under the “Lithium weight gain seems to (perhaps) be dose-dependent even at therapeutic doses” section, which is what I was focused on when I wrote my comment. My bad for not being clear.
I’m skeptical of how much we can learn from the Texas study. If harder tap water causes people to drink more bottled water, that could explain the negative correlation.
However, I do think the study showing a lack of correlation between lithium and BMI is important, and cuts strongly against the argument I was making. I would expect to see a striking correlation there if large relative variations at low absolute levels of dietary lithium were an important mechanism contributing to obesity. That’s just not the case.
My only reservation is that we’re trying to answer SMTM’s hypothesis using studies that weren’t specifically designed to test it. I don’t feel confident enough to completely dismiss the idea based on this alone. However, I do think it’s substantially more fragile than I thought before I took this into account. Epistemic status: blowin’ in the wind.
Update: I found a few papers from villages in Argentina with very high lithium exposure, one of which (whose subjects had urinary lithium concentrations ranging from 0.1-14 mg/L) found a positive association between lithium excretion and BMI (r=0.11), and did find that at such levels lithium increased plasma TSH levels. But only 17% of participants were obese,[1] even though the average urinary concentration was > 4 mg/L.
In the other one (which I think I’d seen before, actually) serum lithium concentration (which strongly correlated (0.84) with urinary excretion) was not found to be associated with BMI. The range of urinary concentration was 0.105–4.600 mg/L. Interestingly, higher exposure seemed to have been associated with smaller body size (both height and weight) for the adults in the study as well as the newborn children. The obesity rate among adults in this sample is only 7%.
Those kind of mixed results don’t seem to paint a compelling picture of dietary lithium-induced obesity. So I’m roughly where I was last we exchanged, when you highlighted the lack of serum lithium and BMI correlation in the other study.
In this case, the thing that jumps out at me is that these are studies in Argentinian villages, with people of Indigenous backgrounds eating relatively traditional and likely constrained diets, perhaps with a higher level of compulsion to work than in wealthier areas.
all women almost exclusively drank tap water, their diets (mainly corn, beans, chicken, and pork) were very similar, and only 3 of 202 women reported ongoing use of any medication
The mechanism that seemed most likely to me was that lithium increases both thirst and fatigue (perhaps mediated via subclinical hypothyroidism), leading to increased consumption of sugary beverages in areas with a lot of soda pop around, and decreased activity levels when a sedentary lifestyle is tractable. Since in these areas, the soda pop and sedentary lifestyles are perhaps less available, this might prevent lithium from showing up more strongly as a driver of obesity. If lithium is driving obesity through thirst and sedentary behavior, the women of SAC may just be satisfying their thirst with water and being stuck working even when they’re very tired.
I’d need more data on how obesity rates vary from village to village in this region to know if 17% is low or high. This study in the same village (San Antonio de los Cobres) found an 11% obesity rate among schoolchildren, as compared to 35% in the USA. That’s just to say that we may be looking at a different overall diet/lifestyle/genetic/contaminant reference class than we see in the USA as a whole.
All that said, I don’t see lithium and obesity leaping out of the data. It’s possible that an abundance of diverse low-level environmental contaminants is responsible, with lithium just one of many. Hence, harvesting data from studies looking at specific individual contaminants won’t ever solve the whole puzzle. “It’s probably not the lithium,” but maybe it’s the lithium + 999 other chemicals? Perhaps having a bunch of weird industrial chemicals floating around in the air, water, soil, and food affects people’s metabolism and energy by hammering at all sorts of different bodily systems—inflammation, hunger, mood, who knows what else?
If the explanation is just “a thousand weird chemicals, each with a small and possibly interacting effect across diverse bodily systems making you tired, hungry and thirst” we’d have to ask why, on average, exposure to random chemicals makes you fat and hungry rather than thin and satiated. Perhaps the brain sends hunger and thirst signals when it’s out of whack, on the off chance that the reason it’s having problems is lack of energy? Certainly it makes sense to me that overall, bodily contamination causes fatigue rather than an increase in energy and drive. And I would also expect that the body’s ability to continue extracting nutrients from whatever food you put into it is pretty robust, since that’s mission-critical. So “body gets a chemical shock, guesses hunger/thirst is the cause, can’t maintain focused energy due to chemical interference, still absorbs nutrients and turns them into fat” as explanation?
And what sort of study would we run or look for to investigate this extremely messy hypothesis?
But I have to admit that “people have more junk food and time on their hands, so people sit around, pig out, and get fat” is also a pretty compelling explanation for the obesity epidemic. The only issue is that McDonald’s has been around since 1940, the country’s been wealthy and working desk jobs for a long time, but obese adults were < 15% of the population in most US states in 1990, but 36 states have obesity rates at 25%+. It could be generational decay in how people spend their time and how they eat. But that is getting handwavey in the same way that the contaminant hypothesis is handwavey.
As a followup, the portrait I painted of “chemically-induced tired brain misfiring syndrome”-induced obesity (CITBMS-obesity) seems to suggest a link between obesity and depression would be associated. There does appear to be a link or a U-shaped association between BMI and depression (I’ve only scanned the abstracts).
And as I understand it, we can totally have a CITMBS-obesity epidemic and also a CITMBS-underweight problem at the same time. “Chemically induced weight dysregulation” might be what we’d be looking at.
In such a case, actually, that might make it hard to use the studies we’ve looked at so far to gain information. If the curve is U-shaped, the two ends of the curve may cancel out when averaged together and disguise the effect.
I hadn’t thought of that before. Now my credences for lithium and contamination are back up. It doesn’t completely jive with the story I was telling earlier—now we have a piece where, perhaps, the chemically stressed brain gets hungry/thirsty in some people and loses appetite in others. Both seem plausible. I guess we’d want to look for a relationship between lithium (or other contaminants) and extremes of weight, rather than a correlation between BMI and lithium concentration.
In such a case, actually, that might make it hard to use the studies we’ve looked at so far to gain information. If the curve is U-shaped, the two ends of the curve may cancel out when averaged together and disguise the effect.
Note that this does not seem to be what has happened at a population level in the US. BMI seems to have increased pretty much at all levels — even the 0.5th percentile has increased from NHANES I (in the early 70′s) to the 2017-March 2020 NHANES, as has the minimum adult BMI. And the difference is not subtle.
For instance, here are the thinnest people in the last NHANES versus the first one:
Psychological health impact of chronic environmental contamination is understudied… The meta-analyses observed small-to-medium effects of experiencing CEC on anxiety, general stress, depression, and PTSD. However, there was also evident risk of bias in the data.
So not impossible? And perhaps there are methodologies here that could be useful for looking more broadly at a CEC theory of obesity?
I would interpret your findings about the links between lithium, weight gain, hypothyroidism, and DI differently.
Subclinical hypothyroidism affects an estimated 5% of the population, and causes both weight gain and fatigue. So we can point out that we do see an appreciable amount of both subclinical hypothyroidism and obesity in the general population, and they were found to be associated in 1104 subjects ages 10-19 years old.
Lithium may also cause weight gain by other mechanisms, as yet unknown.
The numbers you found show that going on a clinical dose of lithium causes an average of 0-6 kg weight gain above whatever weight gain they may have experienced from a dietary dose. We can imagine that going from a very low dietary dose of lithium (say < 10 ug/day) to a small but higher daily dietary dose (say > 40 ug/day) is responsible for more marginal weight gain than going from 20 ug/day to 200 mg/day. So it may be that an increase over time in dietary lithium doses is responsible for most of whatever weight gain lithium is causing in the population, with a much smaller amount of additional weight is caused by putting people on a clinical dose.
Lithium may cause DI only starting at clinical doses. It may then show a dose-dependent association. We would therefore expect to see a dose-dependent relationship between DI and lithium, but no DI epidemic.
The fact that lithium shows a dose-dependent response at therapeutic doses allows for dietary doses in the neighborhood of 20 ug/day to cause weight gain. What we’d need is evidence that the weight gain caused by lithium disappears at dietary doses.
We don’t have that here. The closest is the Rinker study, which found 13⁄28 MS patients reported weight gain and 10⁄28 reported weight loss. More patients gained than lost weight, so if anything it supports the conclusion that low-dose lithium can lead to weight gain. MS can cause both weight loss and gain generally on its own, so we probably shouldn’t update too much on this one small study.
After learning that 5% of the population has subclinical hypothyroidism, I’m actually more inclined to believe that normal doses of dietary lithium may be causing the obesity epidemic than I was before reading this post. However, my credence is still well below 50%, so I also still agree with your post’s title.
Note: This was originally part of a different comment, but I split it off for clarity.
I’m a bit confused about this update. A lot of chronic diseases are very common. And as I showed, hypothyroidism has not been increasing over time. It might be decreasing. And, again, the geographical pattern of hypothyroidism does not match the geographical pattern of obesity.
I’m specifically considering subclinical hypothyroidism. From the paper you link:
If we’re working with the “lithium causes a desire to overeat and be sedentary (via hypothyroidism and possibly other mechanisms) in young people, which leads to lifelong obesity-promoting eating and activity habits” version of the hypothesis, then it may just be that Brazil and China are only now gaining the wealth and food supply to permit this to occur in a large enough swathe of the population. We see obesity growing exponentially in China and India (I couldn’t find a graph for Brazil), while it’s looking like it’s on the concave part of the sigmoid curve in the USA.
I’m inclined to think that if lithium is causing obesity in this manner, that subclinical hypothyroidism is one of its important mechanisms of action. If treating subclinical hypothyroidism in young people did not lead to long-term lower rates of obesity, I would lower my credence in an important dietary lithium-obesity link significantly.
The other paper I linked, which uses nationally representative data from the 2007-2012 NHANES, estimated the prevalence of subclinical hypothyroidism to be 3.5% (lower than the prevalence in the 1988-1994 NHANES, which was 4.3%), and noted that the prevalence of at-risk TSH levels seems to have decreased or remained stable with time. (Although mean TSH levels have increased.)
(I apologize for not having replied earlier — I was curious and wanted to check the raw NHANES data on TSH levels myself before replying to you, so I tried to, but found that parsing one of the datasets was too hard and ended up not doing it.)
[ETA: actually, the 4.3% number is based on a different definition of hypothyroidism. With regards to the definition used in the newer study (TSH > 4.5 mIU/L in the absence of clinical hypothyroidism), it says the following:
]
No worries, thank you for all the great research you’re doing.
The analysis of ’88-’94 data says:
The analysis of ’07-’12 data says:
So we see a 20% decrease in subclinical hypothyroidism (4.3% → 3.5%), but an 800% increase in clinical hypothyroidism (0.3% → 2.4%).
My original argument was based on prevalence in the population, not rate of change across time. If anything, given that (as these papers state), clinical hypothyroidism most definitely is associated with BMI, I think they lend support to the hypothyroidism/obesity explanation. Perhaps what we are seeing is a more rapid move in individual patients from subclinical to clinical hypothyroidism, as a result of the hypothesized lithium contamination. In affected patients, by the time the medical system catches it, it’s usually already clinical, whereas in the (postulated) less lithium-contaminated 80s, there was a longer period of time in the subclinical phase per patient when contact with the medical system could catch and diagnose subclinical hypothyroidism.
I will however make a couple meta notes:
a) I’m not putting as much time into this as you, so I’m worried I’m losing track of the details of the argument.
b) I’m trying to salvage the lithium theory because I don’t think it’s utterly destroyed by this data, not because I think it’s extremely likely.
So I generally just have to apologize if, on reflection, my overall argument here is full of incoherencies and inconsistencies. I’m forming my thoughts as I write these comments, and I expect to change my mind in the future—I’m just not sure in which direction.
The abstract of the paper analyzing ’88-’94 data says that they used a different definition of “subclinical hypothyroidism” than the definition that is commonly used today (I had edited my comment to reflect that a few seconds before you replied. I am so sorry for the error!!). Quoting from the paper:
So it seems that the prevalence of hypothyroidism was 4.6% in this survey, not 0.3%. So the prevalence of clinical hypothyroidism has decreased.
With regards to what we nowadays call subclinical hypothyroidism (TSH > 4.5 mIU/L in the absence of clinical hypothyroidism), the paper that analyses ’07-’12 data does say:
Note: I’m a little bit sick today, and it’s possible I made a mistake in my stoichiometry or in converting from math to reasoning. If so, I will happily stand corrected if anybody points out my error.
The change in terminology is just verbiage.
In fact, it appears they havenarrowedthe definition of both subclinical and clinical hypothyroidism in the newer paper. In light of how they changed the definitions, we should think that a definition-neutral rate of both subclinical and clinical hypothyroidism has gone up evenmorethan I’d described in my previous comment.Hypothyroidism is defined in part by lower-than-normal thyroxine (T4). In the earlier paper, T4 levels are defined as “clinical” that would be defined as “normal” or “subclinical” in the later paper. According to the definitions of the later paper, all “subclinical” patients in the earlier paper would have been considered “normal.”They switched from measuring bound + unbound thyroxine (T4) to free thyroxine (FT4) in the second paper. So the numbers aren’t directly comparable because they’re measuring the molecule in two different states the body. I don’t know whether we can do more than rely on the researchers’ implied claim that the definitions of normal vs. subclinical vs. clinical hypothyroidism remain comparable under the new definition.
Extracts and calculations for legibility:
’88-’94 paper:
Subclinical hypothyroidism: TSH > 4.5 mlU/L and T4 >= 57.9 nM
Clinical hypothyroidism: TSH > 4.5 mlU/L and T4 < 57.9 nM
’07-’12 paper:
Thyroxine (T4) has molecular weight 776.87 g/mol.
(.6 ng/dL) * (10 dL/L) * (1E-9g/ng) * (1 mol/776.87 g) *(1E9 nmol / mol) = 7.7 nmol/L = 7.7 nM.
Subclinical hypothyroidism: TSH >= 4.5 mlU/L and 7.7 nM ⇐ T4 ⇐ 20.5 nM
Clinical hypothyroidism: TSH >= 4.5 mlU/L and T4 < 7.7 nM
FT4 is not the same thing as T4. From Medical News Today:
I haven’t converted these densities to molarities, so I haven’t compared these ranges with those provided by the ’88-’94 paper, but this distinction seems relevant.
Good catch, I will edit the previous comment tomorrow when I’m on my computer. Given that the sub vs clinical distinction turns on T4/FT4 and these papers test for different values, I’d need to give more thought about how comparable they are.
Update: I have now looked into the raw TSH data from NHANES III (1988-1994) and compared it with data from the 2011-2012 NHANES. It seems that, although median TSH levels have increased a bit, the distribution of serum TSH levels in the general population aged 18-80 (including people with thyroid disorders) has gotten more concentrated around the middle; both very high levels (characteristic of clinical or subclinical hypothyroidism) and very low levels (characteristic of clinical or subclinical hyperthyroidism) are less common in the 2011-2012 NHANES compared to NHANES III. You can see the relevant table here. There might be bugs in my code affecting the conclusion of the analysis.
This paper, which pretty much used the same NHANES surveys, looked at a somewhat different thing (thyroid levels in a reference population without thyroid disorders or other exclusion criteria) but it seems to report the same finding w.r.t. high TSH levels: a lower proportion of the population in the latest survey meets the TSH criteria for clinical or subclinical hypothyroidism.
Let us not conflate different meanings of “low-dose lithium” here. In the Rinker study, the dose that patients took was still more than 1000x greater than what people in e.g. New Zealand get from their food. That is not a small difference.
Relevantly, this study found no association between serum lithium concentration in the general population in Germany and BMI. Also, as I mentioned in the post, the correlation between log(water lithium levels) and log(obesity %) across Texas counties is negative.
I was primarily trying to address your local argument about interpreting obesity, DI, and hypothyroidism in this comment. These other pieces of evidence seem important, but just aren’t what I’m focused on.
In that paragraph, I was addressing your claim that we don’t have evidence that the weight gain caused by lithium disappears at dietary doses.
I’m aware of that. By “low-dose lithium,” I specifically meant in the 30mg/day range, as opposed to the 1000x lower levels in a typical diet. My point is that none of the studies you list can help very well to understand what lithium’s effects are at a normal dietary level. We should not extrapolate our certainty too far out of the range they’ve studied, and as you point out, dietary lithium is very far out of that range. We can and should make that same argument about the link between lithium and weight gain outside of the clinical dose. Just because it causes weight gain at a low or normal clinical dose does not mean it causes weight gain at a dietary dose.
Overall, we should be uncertain about this hypothesis, not certain that it is right or wrong. To me, uncertainty means “worthy of more study.” Of course, it also means “accurately reporting information,” and SMTM has really managed to set themselves up for a takedown by omitting and distorting so much of the information they use to build their case. That said, I retain let’s say 40% credence for a fairly strong version of their “environmental contamination” hypothesis, with perhaps 40% credence that that low ug levels of lithium in a typical diet are perhaps a 30% contributor to contamination-linked obesity.
I’d very much like to understand how your credences can be so high with nothing else to back them up than “it’s possible and we lack some data”. Like, sure, but to have credences so high you need to have at least some data or reason to back that up.
When I wrote the comment, it was mainly because of the prevalence of subacute hypothyroidism in the general population. However, one of Natalia’s studies that I hadn’t been focusing on persuaded me that variations in dietary lithium intake is unlikely to be responsible for so much weight gain. Serum lithium isn’t associated with BMI.
This study found no association between serum lithium concentration in the general population in Germany and BMI. Also, as I mentioned in the post, the correlation between log(water lithium levels) and log(obesity %) across Texas counties is negative.
Sorry, I specifically meant the studies you list under the “Lithium weight gain seems to (perhaps) be dose-dependent even at therapeutic doses” section, which is what I was focused on when I wrote my comment. My bad for not being clear.
I’m skeptical of how much we can learn from the Texas study. If harder tap water causes people to drink more bottled water, that could explain the negative correlation.
However, I do think the study showing a lack of correlation between lithium and BMI is important, and cuts strongly against the argument I was making. I would expect to see a striking correlation there if large relative variations at low absolute levels of dietary lithium were an important mechanism contributing to obesity. That’s just not the case.
My only reservation is that we’re trying to answer SMTM’s hypothesis using studies that weren’t specifically designed to test it. I don’t feel confident enough to completely dismiss the idea based on this alone. However, I do think it’s substantially more fragile than I thought before I took this into account. Epistemic status: blowin’ in the wind.
Update: I found a few papers from villages in Argentina with very high lithium exposure, one of which (whose subjects had urinary lithium concentrations ranging from 0.1-14 mg/L) found a positive association between lithium excretion and BMI (r=0.11), and did find that at such levels lithium increased plasma TSH levels. But only 17% of participants were obese,[1] even though the average urinary concentration was > 4 mg/L.
In the other one (which I think I’d seen before, actually) serum lithium concentration (which strongly correlated (0.84) with urinary excretion) was not found to be associated with BMI. The range of urinary concentration was 0.105–4.600 mg/L. Interestingly, higher exposure seemed to have been associated with smaller body size (both height and weight) for the adults in the study as well as the newborn children. The obesity rate among adults in this sample is only 7%.
their average age was 37, median 34, and at that age people are pretty close to the highest BMI they’ll ever have, according to NHANES data.
Very interesting, thanks for finding them.
Those kind of mixed results don’t seem to paint a compelling picture of dietary lithium-induced obesity. So I’m roughly where I was last we exchanged, when you highlighted the lack of serum lithium and BMI correlation in the other study.
In this case, the thing that jumps out at me is that these are studies in Argentinian villages, with people of Indigenous backgrounds eating relatively traditional and likely constrained diets, perhaps with a higher level of compulsion to work than in wealthier areas.
The mechanism that seemed most likely to me was that lithium increases both thirst and fatigue (perhaps mediated via subclinical hypothyroidism), leading to increased consumption of sugary beverages in areas with a lot of soda pop around, and decreased activity levels when a sedentary lifestyle is tractable. Since in these areas, the soda pop and sedentary lifestyles are perhaps less available, this might prevent lithium from showing up more strongly as a driver of obesity. If lithium is driving obesity through thirst and sedentary behavior, the women of SAC may just be satisfying their thirst with water and being stuck working even when they’re very tired.
I’d need more data on how obesity rates vary from village to village in this region to know if 17% is low or high. This study in the same village (San Antonio de los Cobres) found an 11% obesity rate among schoolchildren, as compared to 35% in the USA. That’s just to say that we may be looking at a different overall diet/lifestyle/genetic/contaminant reference class than we see in the USA as a whole.
All that said, I don’t see lithium and obesity leaping out of the data. It’s possible that an abundance of diverse low-level environmental contaminants is responsible, with lithium just one of many. Hence, harvesting data from studies looking at specific individual contaminants won’t ever solve the whole puzzle. “It’s probably not the lithium,” but maybe it’s the lithium + 999 other chemicals? Perhaps having a bunch of weird industrial chemicals floating around in the air, water, soil, and food affects people’s metabolism and energy by hammering at all sorts of different bodily systems—inflammation, hunger, mood, who knows what else?
If the explanation is just “a thousand weird chemicals, each with a small and possibly interacting effect across diverse bodily systems making you tired, hungry and thirst” we’d have to ask why, on average, exposure to random chemicals makes you fat and hungry rather than thin and satiated. Perhaps the brain sends hunger and thirst signals when it’s out of whack, on the off chance that the reason it’s having problems is lack of energy? Certainly it makes sense to me that overall, bodily contamination causes fatigue rather than an increase in energy and drive. And I would also expect that the body’s ability to continue extracting nutrients from whatever food you put into it is pretty robust, since that’s mission-critical. So “body gets a chemical shock, guesses hunger/thirst is the cause, can’t maintain focused energy due to chemical interference, still absorbs nutrients and turns them into fat” as explanation?
And what sort of study would we run or look for to investigate this extremely messy hypothesis?
But I have to admit that “people have more junk food and time on their hands, so people sit around, pig out, and get fat” is also a pretty compelling explanation for the obesity epidemic. The only issue is that McDonald’s has been around since 1940, the country’s been wealthy and working desk jobs for a long time, but obese adults were < 15% of the population in most US states in 1990, but 36 states have obesity rates at 25%+. It could be generational decay in how people spend their time and how they eat. But that is getting handwavey in the same way that the contaminant hypothesis is handwavey.
Thanks for finding those studies!
As a followup, the portrait I painted of “chemically-induced tired brain misfiring syndrome”-induced obesity (CITBMS-obesity) seems to suggest a link between obesity and depression would be associated. There does appear to be a link or a U-shaped association between BMI and depression (I’ve only scanned the abstracts).
And as I understand it, we can totally have a CITMBS-obesity epidemic and also a CITMBS-underweight problem at the same time. “Chemically induced weight dysregulation” might be what we’d be looking at.
In such a case, actually, that might make it hard to use the studies we’ve looked at so far to gain information. If the curve is U-shaped, the two ends of the curve may cancel out when averaged together and disguise the effect.
I hadn’t thought of that before. Now my credences for lithium and contamination are back up. It doesn’t completely jive with the story I was telling earlier—now we have a piece where, perhaps, the chemically stressed brain gets hungry/thirsty in some people and loses appetite in others. Both seem plausible. I guess we’d want to look for a relationship between lithium (or other contaminants) and extremes of weight, rather than a correlation between BMI and lithium concentration.
Note that this does not seem to be what has happened at a population level in the US. BMI seems to have increased pretty much at all levels — even the 0.5th percentile has increased from NHANES I (in the early 70′s) to the 2017-March 2020 NHANES, as has the minimum adult BMI. And the difference is not subtle.
For instance, here are the thinnest people in the last NHANES versus the first one:
The relevant Google Colab cells start here.
And then there’s this study: Chronic environmental contamination: A systematic review of psychological health consequences (again, I’m just skimming abstracts).
So not impossible? And perhaps there are methodologies here that could be useful for looking more broadly at a CEC theory of obesity?