In Winter, everyone gains T3 except thyroidless patients on T4

Winter-Summer-FT3A 2017 article reported research on 11,806 healthy controls and 3,934 thyroidless people treated with L-T4 monotherapy in Sicily, and then did a more in-depth “retrospective longitudinal” study on 119 of the thyroid patients compared to 156 matched controls.

They discovered that the thyroid-disabled people lost a “significant” amount of Free T4 and Free T3 hormone in winter, on average, in addition to a loss of T4 that caused an increase in TSH.

In contrast, people with healthy thyroids experience the very opposite effect — a statistically significant increase in average Free T3 during the colder months of the year. Unlike the thyroidless, healthy controls kept their average TSH and FT4 levels almost exactly the same all year long.

This seasonal effect in winter exaggerates the average Free T3 deficit in people treated with T4 hormone alone.

Even in summer, the thyroidless on LT4 already have significantly lower Free T3 than healthy controls. Thyroid scientists all agree, and we have known for many decades, that their year-long significantly lower T3:T4 ratio is induced by T4 monotherapy itself, not by illness.

But when winter comes, the difference is even more extreme between the lower FT3 levels of thyroidless T4-monotreated people and the higher FT3 of healthy people.

Specifically, in the longitudinal study, the healthy controls increased FT3 from 4.31 pmol/L in the warmest four months of Summer to 4.40 pmol/L in the coldest four months of Winter (+2.9% of the FT3 reference range), while the thyroid-disabled decreased FT3 from 4.07 in Summer to 3.80 pmol/L in Winter (-8.7% of the reference).

Because the thyroidless patients already had a lower FT3 in summer, when the Winter came, thyroid-disabled patients had a net deficit of -0.60  pmol/L less FT3 than healthy controls, an amount that represents 19.4% of the FT3 reference range. The thyroidless patients’ relative loss of FT4 in winter still kept their FT4 higher than healthy controls at a surplus of +21% of the FT4 reference range.

In this post, I’ll provide their basic data.  Then I’ll calculate their data as percent of their reference ranges so that they can be analyzed and compared, both summer to winter, and between “healthy” people and the thyroidless.


The study did not have a methodology that could examine the “clinical significance” of the data, that is, whether the FT3 loss worsened any of the patients’ symptoms or changed their body temperature or affected any concurrent medical conditions. It also had some shortcomings in terms of not revealing the full spread of the data and the reference ranges that could be used to interpret them in context with the larger population.

What the study is good for is critically examining the theoretical assumptions about the HPT axis (the relationship between hypothalamus, pituitary and thyroid hormone secretions) and the degree to which it is distorted and modified by not having a thyroid and being on T4 monotherapy.

This research data and its discussion is important because the HPT axis model of health is currently being used to judge the “euthyroid” status of patients on this therapy and its boundaries with underdose and overdose.

It has strong clinical implications because patients at the boundaries of the current diagnostic criteria can be misdiagnosed, underdosed, or overdosed based on the theory’s assumptions about what their TSH concentration means, and based on a false idea that they may only need testing or thyroid dose adjustments once a year.

If the data on such a large population show that this supposedly unchanging axis undergoes significant and unexplained distortions with potential for clinical effect, the HPT axis theory ought to be updated and made more accurate to account for the shifts and distortions.


[ETHICS NOTE: My reproduction of the copyrighted article’s data and quotations falls within US copyright law as “fair use” and within Canadian copyright law as “fair dealing.” According to Stanford university’s explanation of the US law, “a fair use is any copying of copyrighted material done for a limited and “transformative” purpose, such as to comment upon, criticize, or parody a copyrighted work. Such uses can be done without permission from the copyright owner.” I engage in commentary, criticism, and mathematical transformation of the data, below.]

W = Winter, their coldest 4-month period, December to March
S = Summer, their warmest 4-month period, June to September.

Cross-sectional study: Euthyroid subjects (or “HC,” healthy controls)
W: 3,819 people.  S: 3,703  people.
Hormone (units) W/S average (IQR, interquartile range, central 50% of data)
TSH (mU/L) W = 1.41 (0.90-2.18); S = 1.40 (0.90-2.14)
FT4 (pmol/L) W = 14.0 (12.4-15.4); S = 13.9 (12.2-15.4)
FT3 (pmol/L) W= 4.47 (3.96-5.00); S = 4.34* (3.85-4.93)
FT3/FT4 ratio W = 0.32 (0.28-0.36); S = 0.31 (0.27-0.36)

Cross-sectional study: LT4-treated athyreotic patients (“Rx”)
W: 1,315 people. S: 1,241 people
TSH*** (mU/L) W = 0.70 (0.20-1.68); S = 0.38* (0.10-1.20)*
FT4 (pmol/L) W = 16.4 (14.2-18.4); S = 16.7* (14.3-19.3)
FT3 (pmol/L) W = 3.86 (3.39-4.44); S = 4.00** (3.47-4.47)
FT3/FT4 ratio W = 0.23 (0.20-0.25); S = 0.24 (0.20-0.27)
LT4 dose (μg/kg/d) W = 1.64 (1.39-1.86) S = 1.65 (1.42-1.91)

Statistical significance of change from W to S: *P = <.001, **P <.008 (from text, p. 211)

“In the athyreotic patients, the slope of the FT4/TSH inverse correlation was much steeper in summer (−1.52) than in winter (−0.85) (P<.001, ANCOVA test).”

*** TSH was suppressed by therapeutic policy in many of the Rx patients because all had been diagnosed with thyroid cancer. Their total thyroidectomy, with or without subsequent radioiodine treatment, had occurred at least 1 year before their data was collected, and their dose of LT4 was stable and constant throughout the study.

Longitudinal study data:

As reported in this same article, researchers also conducted a more rigorous and in-depth “retrospective longitudinal study” of data within their larger, “cross-sectional” data set.

They focused on 119 of the athyreotic patients for whom they had two full thyroid hormone tests of TSH, FT3 and FT4 during both the coldest and hottest months of the same year. Their data was matched for comparison with 159 age-matched controls who also had similarly rich data per patient.

In the longitudinal study, “the seasonal difference was more substantial,” and even the values the FT3 change from winter to summer (which had achieved a somewhat lesser significance calculation of P= <0.008 in the cross sectional study), achieved the statistical significance of P=<0.001 for the degree of change from winter to summer.

Longitudinal study: Euthyroid subjects / HC
W: 159 people.  S: 159  people.
Hormone (units) W/S average (IQR, interquartile range, central 50% of data)
TSH (mU/L) W = 1.41 (0.90-2.40); S = 1.43 (1.00-2.33)
FT4 (pmol/L) W = 14.2 (12.5-15.4); S = 14.0 (12.1-14.9)
FT3 (pmol/L) W = 4.40 (3.88-4.84); S = 4.31 (3.85-4.77)
FT3/FT4 ratio W = 0.32 (0.27-0.36); S = 0.31 (0.27-0.36)

Longitudinal study: LT4-treated athyreotic patients / Rx
W: 119 people. S: 119 people.
TSH (mU/L) W = 0.80 (0.22-1.44); S = 0.20* (0.06-0.70)
FT4 (pmol/L) W = 16.3 (14.2-17.7); S = 17.8* (15.4-19.9)
FT3 (pmol/L) W = 3.80 (3.36-4.19); S = 4.07* (3.80-4.51)
FT3/FT4 ratio W = 0.23 (0.20-0.24); S = 0.24 (0.21-0.26)
LT4 dose (μg/kg/d) 1.59 (1.30-1.80) 1.59 (1.30-1.80)

Statistical significance of change from W to S: *P = <.001


The researchers pointed out that a change in healthy controls from S = 4.34 to W = 4.47 pmol/L is a 2.9% relative increase.

They appear to have calculated this relative increase by considering the lower value of 4.34 as 100%, so that an increase to 4.47 in winter would be 102.99% of the summer value.

However, this type of calculation is unhelpful in many ways.

First of all, it is puzzling that they selected only the FT3 increase from the euthyroid controls to highlight as “% change” when the Rx patients’ TSH, FT4, FT3 and ratio change summer to winter was more pronounced than theirs. The HCs’ FT3 seasonal change failed to remain statistically significant in the longitudinal study, while the Rx patients’ seasonal change became more distinct, and their FT3 change’s statistical significance, in particular, increased.

It is important to highlight the degree of seasonal change, since that is the main point of the article. However, doing so by this means of this calculation is not very helpful.

“Percent of relative change” is not a physiological index but rather an arbitrary, mathematical calculation that can introduce non-biological parameters and numerical distortions into the comparison.

The first problem with such a calculation is that the Free T3 range is not 0 to 4.34 pmol/L. There is no such thing in reality as a lab result of 0 pmol/L FT3.  If you truly had a FT3 of zero, you would likely be dead already or would die very soon. Therefore, it does not make biological sense to consider 0 pmol/L as 0% and 4.34 pmol/L as a value of 100%. 

If one were to apply a similar calculation to all their data, further distortions would appear. Hormones like FT3 and TSH can have very small numbers, compared to FT4.  The percent of change from 0.5 to 1 mU/L of TSH is a 100% increase (it doubled the number) which can be psychologically misleading because the TSH range is usually only 4.5 mU/L wide. This exaggerates the change when it is only a small part of the total reference range’s width. This is another distortion that fails to put the significant change in a biological frame.

In addition, starting with a smaller number will exaggerate a 1.0 change, and starting with a larger number will minimize a 1.0 change. 1 to 2 mU/L of TSH is a 100% increase but 4 to 5 mU/ L of TSH is only a 20% increase. Such distortions would make comparisons across hormones more difficult as well, because the low and high boundaries of FT3, FT4 and TSH have very different low and high numeric values, and the FT3 range is often 3 pmol/L wide, while the FT4 range is often 10 pmol/L wide. This mode of calculation introduces a distortion that can make comparisons unfair within one hormone or between two hormones.


Therefore, a better way to compare the hormone changes is in relationship to the width of the reference range, by “% of reference range.”

Despite the many controversies that surround the methods and uses of reference ranges, at least they aim to be based on a human population’s range of variation. They are more biologically useful as a way of weighing the relative magnitude of a change.

Percent of reference also provides a less distorted comparison among the three hormones at all their numeric concentrations.

Below, I obtain reference ranges for the hormones and then provide their data “transformed” into calculations of “percent of reference.”


In this study, the TSH reference range was provided: 0.4 to 4.0 mU/L, a range of 3.6 units wide.

A frustrating feature of this study is that Gullo et al, 2017 did not provide local laboratory reference ranges for FT3 or FT4, but we do have access to the manufacturer’s recommended reference ranges in a fairly recent article (Fillee et al, 2012), which is a reasonable estimate. For the Abbott Architect assay,

The FT3 range was 3.1 pmol/L wide (2.6 – 5.7 pmol/L).

The FT4 range was 10 pmol/L wide (9.0 to 19.0 pmol/L).

As confirmation, these ranges roughly match the reference ranges represented only visually as shaded areas in graphs from Gullo et al’s previous study in 2011, Figure 1.


As I make these “percent of reference” calculations below, I arrange the data differently so that the Rx and HC data is more easily compared hormone by hormone across the “cross-sectional” and “longitudinal” data sets.


CS = cross sectional study; LS longitudinal study

HC = Healthy controls, Rx = Thyroidless on L-T4

W = Winter, their coldest 4-month period, December to March
S = Summer, their warmest 4-month period, June to September.

“% of ref” = 0% bottom of reference range, 100% top of reference range.

CS/LS; Hormone; HC/Rx, W/S average % of range  (IQR, interquartile range, central 50% of data)

Free T3

CS FT3 (% of ref) HC, W = 60.3 (43.9 – 77.4); S = 56.1 (40.3 – 75.2)
CS FT3 (% of ref) Rx, W = 40.6 (25.5 – 59.4); S = 45.2 (28.1 – 60.3)

LS FT3 (% of ref) HC, W = 58.1 (41.3 – 72.3); S = 55.2 (40.3 – 70.0)
LS FT3 (% of ref) Rx, W = 38.7 (24.5 – 51.3); S = 47.4 (38.7 – 61.6)

Rx / HC Comparisons:

CS FT3 HC S to W +4.2% of ref
CS FT3 Rx S to W -4.6% of ref

LS FT3 HC S to W +2.9% of ref
LS FT3 Rx S to W -8.7% of ref

CS FT3 Rx lower than HC in Winter = -19.7%; and in S = -10.9%
LS FT3 Rx lower than HC in Winter = -19.4%; and in S = -7.8%

Free T4

CS FT4 (% of ref) HC, W = 50.0 (34.0 – 64.0); S = 49.0 (32.0 – 64.0)
CS FT4 (% of ref) Rx, W = 74.0 (52.0 – 94.0); S = 77.0 (53.0 – 103.0).

LS FT4 (% of ref) HC, W = 52.0 (35.0 – 64.0); S = 50.0 (31.0 – 59.0)
LS FT4 (% of ref) Rx, W = 73.0 (52.0 – 87.0); S = 88.0 (64.0 – 109.0)

Rx / HC Winter Comparisons:

CS FT4 HC S to W +1.0% of ref
CS FT4 Rx S to W -3.0% of ref

LS FT4 HC S to W +2% of ref
LS FT4 Rx S to W -15% of ref

CS FT4 Rx higher than HC in Winter = +24%; and in S = +28%
LS FT4 Rx higher than HC in Winter = +21%; and in S = +38%

T3:T4 ratio (not calculable as % of reference; Gullo et al’s data is simply rearranged below)

CS FT3/FT4 ratio HC, W = 0.32 (0.28-0.36); S = 0.31 (0.27-0.36)
CS FT3/FT4 ratio Rx, W = 0.23 (0.20-0.25); S = 0.24 (0.20-0.27)

LS FT3/FT4 ratio HC, W = 0.32 (0.27-0.36); S = 0.31 (0.27-0.36)
LS FT3/FT4 ratio Rx, W = 0.23 (0.20-0.24); S = 0.24 (0.21-0.26)

Rx / HC Winter Comparisons:

CS Ratio Rx lower than HC in W = -0.09; and in S = -0.07
LS Ratio Rx lower than HC in W = -0.09; and in S = -0.07 (the same in CS and LS studies)


CS TSH (% of ref) HC, W = 28.1 (13.9 – 49.4); S = 27.8 (13.9 – 48.3)
CS TSH (% of ref*) Rx, W = 08.3 (-05.6 – 35.6); S = -00.6 (-08.3 – 22.2)

LS TSH (% of ref) HC, W = 28.1 (13.9 – 55.6); S 28.6 (16.7 – 53.6)
LS TSH (% of ref) Rx, W = 11.1 (-5.0 – 28.9); S -5.6 (-9.4 – 8.3)

Rx / HC Comparisons:

CS TSH HC S to W +0.3% of ref
CS TSH Rx S to W +7.7% of ref

LS TSH HC S to W -0.5% of ref
LS TSH Rx S to W +16.1% of ref

*CS TSH Rx lower than HC in Winter = -19.8%; and in S = -28.4%
*LS TSH Rx lower than HC in Winter = -17.0%; and in S = -34.2%

*NOTE:  TSH was often medically suppressed below reference by policy in the Rx cohort, which skews the Rx data lower in comparison to HC all year long.


The Longitudinal Study data is the basis of these comparisons because its data is more equally distributed patient by patient.


The major difference between Rx and HC occurred in winter:

The average Rx patient lost -8.7% of reference range in their FT3 levels.
The average HC patient gained +2.9% of reference. 

The total spread between Rx and HC patients in Winter placed Rx at an average deficit of -19.4% of reference in their FT3 levels (HC, W = 58.1%, Rx, W 38.7% of reference)

At the low end of Rx patient FT3, the 25th percentile had FT3 levels 24.5% of reference, placing them at a -33.3% deficit in comparison to the HC average at 58.1% of reference.

At the high end of Rx patient FT3, the 75th percentile had FT3 levels 51.3% of reference, which was still a -6.8% deficit in comparison to the HC average at 58.1% of reference.


The bottom of the FT4 Rx W interquartile range (52.0) was almost the average FT4 for HC W. This means ~75% of Rx patients’ data was higher than the healthy average Free T4.


The top of ratio’s IQR Rx W interquartile range (0.25) was significantly lower than the bottom of the average ratio for HC W.

The ratios were entirely different. They do not overlap even in summer.


The graphs in the article’s Figure 2 revealed that some of the Rx patients had a fully suppressed TSH (it was not reported how many), while the healthy controls’ TSH data fell within reference.

This alone would have biased the Rx TSH averages and IQRs to be lower than the HCs.

The comparison of averages and “% of reference” between the two cohorts was certainly not on equal grounds because Rx data was not kept within the reference range.

On the other hand, the “slope” of the FT4 correlation plotted across TSH values shown in their Figure 2 is a fair comparison between the two cohorts and is both statistically and biologically significant in its implications for the HPT axis theory.


The main thing you should notice in all this data is the degree of difference and inequity that exists between thyroidless on T4 and the “healthy” population.

This data alone should explode any false impression that there is one single HPT axis that applies to all people across the hormone-shifting conditions of thyroid disease and therapy. Future clinical research going forward ought to be based on a theory of the HPT axis that has evolved to account for research evidence.

My goal in this post has mainly been to share a treasure-trove of data that can ought to be analyzed and pondered by others, and which I can draw on later, too.

I’ll follow up later with further posts that may draw on this data set and the researchers’ own discussion of it.


Gullo, D., Latina, A., Frasca, F., Le Moli, R., Pellegriti, G., & Vigneri, R. (2011). Levothyroxine Monotherapy Cannot Guarantee Euthyroidism in All Athyreotic Patients. PLoS ONE, 6(8).

Gullo, D., Latina, A., Frasca, F., Squatrito, S., Belfiore, A., & Vigneri, R. (2017). Seasonal variations in TSH serum levels in athyreotic patients under L-thyroxine replacement monotherapy. Clinical Endocrinology, 87(2), 207–215.



Categories: Healthy thyroid axis, Research Reviews, T4-monotherapy

3 replies


  1. Principles & Practical tips for Reverse T3, FT3, FT4 – Thyroid Patients Canada
  2. 7 ways to raise TSH without reducing thyroid dose – Thyroid Patients Canada
  3. Japanese thyroid scientists examine symptoms in relation to FT3 and TSH – Thyroid Patients Canada

Leave a public reply here, on our website.

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: