A prime example of bad misleading research

[diogenes]

A little bit of scientific comment and analysis to discriminate plausible papers from rubbish.

In the latest journal Thyroid I've picked out a paper that could mislead readers in its misuse of statistical analyses These are the kind of sloppy papers that all too easily get picked on as support for a particular belief. The almost laughable part is to find probabilities of 0.049 and 0.025 in their estimation of heart attack and stroke as being significant. It's generally accepted that probabilities greater than 0.01 are of doubtful significance and greater than 0.05 as being insignificant. Now the above numbers are either of marginal and uncertain significance or of no meaningful significance at all. Note the confidence limits almost reach 1.000 (ie insignificant). Also note that T3 users were mixed in as combined therapy and T3 only. I have just posted this "paper" as a fine example of publishing bad work and worse, getting the esteemed journal Thyroid to accept it.

ThyroidVol. 32, No. 7

Heart Failure and Stroke Risks in Users of Liothyronine With or Without Levothyroxine Compared with Levothyroxine Alone: A Propensity Score-Matched Analysis

Wook Yi, Bo Hyun Kim, Mijin Kim, Jinmi Kim, Myungsoo Im, Soree Ryang, Eun Heui Kim, Yun Kyung Jeon, Sang Soo Kim, and In Joo Kim

Published Online:8 Jul 2022Methods: We conducted a retrospective multicenter study across four hospital databases encoded in the Observational Medical Outcomes Partnership (OMOP) CDM. LT3 users were defined as those who received an LT3 prescription for at least 90 days (with or without LT4), and their safety outcomes were compared with those in LT4-only users after 1:4 propensity score matching. Safety outcomes included the incidences of osteoporosis, cardiovascular disease, cancer, anxiety disorder, and mood disorder.

Results: We identified 1434 LT3 users and 3908 LT4-only users. There was a statistically significant difference in the incidence rate of safety outcomes between LT3 users and LT4-only users. The risks of heart failure (incidence rate ratio [IRR] = 1.664, 95% confidence interval [95% CI] 1.002–2.764, p = 0.049) and stroke (IRR = 1.757, CI 1.073–2.877, p = 0.025) were higher in LT3 users than in LT4-only users. When subgroup analysis was performed according to the presence/absence of thyroid cancer history and duration of thyroid hormone replacement, the risk of heart failure was higher in LT3 users with a history of thyroid cancer and those who underwent ≥52 weeks of LT3 therapy. In addition, the risk of stroke was higher in LT3 users without thyroid cancer history and those who underwent ≥52 weeks of LT3 therapy.

Conclusions: The use of LT3 was associated with increased incidence of heart failure and stroke in patients with a longer duration of LT3 use and history of thyroid cancer. Therefore, clinicians should consider the risk of heart failure and stroke in thyroid cancer patients with long-term use of LT3. These findings require confirmation in other populations.

Dear me!

[DippyDame]

That nonsense should please the endos!!

[Lilian15 in reply to DippyDame]

That is the problem rather than decide what is good research and what is bad, they just cherry pick the ones that come to the conclusion they want.

[Rapunzel]

As this is such a bad paper, it's to be hoped that there was no sponsorship from pharmaceuticals who manufacture Levo...whether it's declared or not. Don't prescribe that! Prescribe this! Excellent reinforcement for the many medics who liken Lio to Columbian marching powder

🙄

[tattybogle]

on the bright side ... does this mean they couldn't show even an 'insignificant' association between T3 use and incidences of osteoporosis, cancer, anxiety disorder, and mood disorder ?

[Musicmonkey]

How does something like this get published if it is so poor? I thought these studies were checked for appropriateness.

[LindaC in reply to Musicmonkey]

By whom? Their own system - 'The System They Put Together'!? Sickening!

Our only hope is via the good guys! xox

[Musicmonkey in reply to LindaC]

Thank goodness for the good guys!

[LindaC in reply to Musicmonkey]

💚❤️💛🎶🔍 Exactly!

Endo Ducking Stool remaining in this 21st C

[TSH110 in reply to LindaC]

Nice one!

[TSH110 in reply to Musicmonkey]

I get the impression very few scientists really have a thorough grasp of statistics or how to employ them correctly including the gatekeepers for these journals

[LindaC in reply to TSH110]

Absolutely and even on the basic level that anyone with half a cell would laugh at these grossly constructed 'ranges'. But for doctors, with some mere smattering of 'science', saying 'Normal' to really ill people sat before them, is anathema. Quite embarrassing really, is that level of incompetence!

[TSH110 in reply to LindaC]

It’s scandalous and I’ve been there when you are too ill and ill informed to fight your corner, you have to rely on them to know what they’re doing. The most frightening thing is they believe they know what they’re doing and change is not possible with a closed, unenquiring mind that ignores the evidence before their very eyes and simply does not listen - they doctor for status and money patients are just money making units to get out of the surgery as fast as possible job done.

[LindaC in reply to TSH110]

Absolutely - I've just recently posted the latest horror - unexpected after all of these years... this is Counterfeit Medicine, for the reasons you state above.

[helvella Administrator]

From what we see, few get LT3 at all (that is, with LT4 or on its own) until they have tried LT4.

Thus, I suggest, at least the majority of those getting LT3 will have gone through a period of not doing well on LT4. Maybe that is an important factor in ending up with heart failure and/or stroke?

Furthermore, those who had thyroid cancer were likely to have been on higher LT4 doses (to achieve TSH suppression). Which skews the treatment histories even more.

This at least needs to be considered and excluded before they make any claims.

[tattybogle in reply to helvella]

would it be too childish to suggest that in order to get any T3 they would have had several stressful and upsetting encounters with endocinologists ?

pretty sure that ought to be included as a statistical factor ... i mean , just reading about other people's 'encounters' is enough to give me a raised pulse.. god knows what effect actually seeing one for myself would have on my heart ......

[TSH110 in reply to tattybogle]

Not to mention the anger side of it that no CBT could quell!

[jimh111]

It's very difficult to make any judgement without studying the full paper. This one is behind a paywall, are you able to obtain a copy for us?

I'm not too worried about using p=0.05 as the cut-off for significance as this is used for nearly all medical studies. It would be good to see 0.01 used but that would mean losing many studies because of cost or unavailability of sufficent data. It is important to bear in mind how relevant the results are in every study. e.g. a study of 20 patients that showed drug x cured most patients p=0.05 is much more convincing than a study of 10,000 patients that found 'significant' benefit of drug y p=0.01. Drug companies invariably use large studies to demon strate 'significant' benefit of drugs that have marginal benefit.

This paper is a surprise to me as I recently posted on the subject and the evidence I found suggested T4 was more harmful than T3, see healthunlocked.com/thyroidu... . It seems that T4 has specific harmful actions in the heart. I assume this study was controlled for factors such as age, sex, BMI, smoking etc. I would also expect it to compare dose, either by looking at the dose taken or TSH. Dose is important because we need to know whether the fT4 levels of the T3 group. If they had higher fT4 it would be silly to blame the T3.

[diogenes Remembering in reply to jimh111]

Studies using patient panels with many confounding variables cannot in all honesty be acceptable at p = 0.05. There is no strict classification of subgroups. If studies cannot be done to get a much better outcome than this, it's better that they are not done at all.

[jimh111 in reply to diogenes]

Studies usually make allowance for confounding variables (is this called multivariate analysis? I can't remember). I assume this is done in this study, if not, it is worthless and so should have been rejected.

[Charlie-Farley]

Guffaw

We were made to do stats at Plymouth for the environmental science degree compulsory for two years. By the end I would have gladly done a third - how can anyone conduct valid studies or experiments without a basic understanding of stats? Even if you have the basics pays to call in the stats gods to give some input. Interestingly, at Lancaster I continued to consider ‘the stats’ and booked myself chat with the stats department so someone with a far higher knowledge than my still relatively rudimentary grasp could cast their eye over my proposed experimental design. Why collect data if the design is flawed and it cannot then be analysed? I was (apparently) the first person to do this! Bad science can be proliferated by people rubber stamping buddies through peer review- or you rubber stamp mine and I’ll rubber stamp yours. I’ve see some utter b0!!k$ published in the field.

Could go on but I’ve put that disillusionment to bed years ago shan’t rake up dead embers 😂👍

[TSH110]

Is there any way to challenge this rubbish?

[helvella Administrator in reply to TSH110]

Anyone could try doing so by writing to the journal.

I hold out little hope on the basis I suspect they will not take seriously criticism from people not involved in the area. But that is (in my view) no reason to give in and NOT write. The worst they could do is ignore your letter/email (unless you have some sort of professional reputation which could be damaged).

Just keep it accurate, simple, to the point. Such that they cannot claim that it is wrongly based, inaccurate, too complicated, or nonsensical.

[TSH110 in reply to helvella]

I think we are involved in the area being the ones with it! Although I know you mean researchers.

[Poniesrfun]

A group I chat with was discussing this paper - someone mentioned “incorrect statistical methodology” which I read as “statistical mythology”. It does seem that much thyroid "research" seems to be based on mythology. And they keep re-searching the same thing over and over. Time for new search instead?