Age, sex and TSH-FT4-FT3 relationships: Advanced lessons

No one wants to be ageist or sexist in thyroid screening or diagnosis.

But if we are blind to the way that age and sex can distort the hypothalamus-pituitary-thyroid (HPT) axis, our interpretations of thyroid hormone levels and TSH will inevitably become ageist and sexist by refusing to respect and accommodate human diversity.

In a previous post, “Age bias may hide hypothyroidism under a normal TSH,” I discussed the transformations in TSH-FT4 relationships with age, focusing on adults. Then, I followed up with a post focusing on “Pediatric and teenage TSH, FT4, and FT3 levels.”

Those previous posts discuss mainly focus on different reference intervals and medians for TSH, FT4 and FT3 across thyroid-healthy age and sex cohorts.

For advanced readers, this posts moves to an advanced level. It examines the changing relationships between two or three of these hormones by displaying graphs and talking about negative and positive correlations between TSH and thyroid hormones.

This advanced knowledge can strengthen clinical discernment when diagnosing and deciding to treat, or not to treat, people who are currently not on treatment for a thyroid disease.

This is the kind of knowledge a thyroid expert and/or educated thyroid patient needs if a case is complex, or if a person’s hormone levels are inconclusive when assessed in isolation from each other.

Hormone relationships are at the core of the healthy HPT axis function, which is often misleadingly referred to as “thyroid function.” Hormone relationships reflect three dimensions of the HPT axis:

  • healthy thyroid function (response to TSH),
  • healthy hypothalamus and pituitary function (response to FT3 and FT4), and
  • healthy thyroid hormone metabolism (intracellular enzymes’ responses to FT3 and FT4 carried into cells all over the body).

What is a physiologically appropriate set of TSH-FT4-FT3 hormone relationships, in contrast to relationships that reveal an HPT axis dysfunction within or beyond the thyroid? The answer partly depends on one’s age and sex.

In this article, I provide a review of scientific research by Gilad Karavani, David Strich and their team, who examined a large, well-screened population from young children over 1 year old, to people over 80 years old, with TSH levels no higher than 7.5 mIU/L.

They presented research findings regarding four hormone relationships:

  • TSH-FT4 relationships
  • TSH-FT3 relationships
  • FT3:FT4 ratios
  • TSH relationships to FT3:FT4 ratios

They revealed many insights into the youthful and aging HPT axis that most physicians and scientists overlook. Ignorance of these dynamics may lead to mistakes in diagnosis and incorrect assessments of thyroid hormone health risk.


First, cautions are in order about misapplications of Karavani and Strich’s descriptive science, since even scientists find it tempting to turn a description of healthy HPT axis into a prescription for people who are not so healthy.

In 2014, Karavani and colleagues found that in children, FT4, FT3 and the FT3:FT4 ratio differ at rising quartiles of TSH up to 7.5 mIU/L in a well-screened pediatric population up to age 20. Childhood and youth enhanced the FT3:FT4 ratio per TSH quartile, and obesity (high BMI) enhanced it further.

Then, in 2016, Karavani’s research team published a follow-up article with Strich as first author. They expanded their research to include adults across the lifespan and found very different patterns in age than in youth.

They double-checked the association mathematically, and revealed that correlation coefficients between TSH and the thyroid hormones wobble and shift significantly among age and sex cohorts.

They discovered that the “inverse (negative) log-linear relationship between TSH and FT4” was not as strong or even statistically significant across age and sex cohorts. Since other research shows that the negative correlation strengthens in hypothyroidism, but not subclinical hyperthyroidism, did some of their cohorts have more hypothyroid individuals in them? Or was this the effect of biological age and sex?

If the latter is true, as suggested by the data, it means that a given TSH level between 0.2 mIU/L or 7.5 mIU/L, does not correlate with a predictable range of FT4 levels in 15 year old males, 45 year old females, and 85 year old females.

But their most important finding was that aging inverts the TSH-FT3 relationship and TSH relationship to the FT3:FT4 ratio. These hormone relationships transform from strongly positive correlations in youth to a mildly negative correlation in old age.

In fact, the loss of the TSH-stimulated FT3:FT4 ratio is a distinctive feature of the aging HPT axis when combining data from both sexes.

All these findings break down the idea that there is a static mechanism called “thyroid function” or “the” HPT axis, that is constant throughout life.

In contrast to this linear relationship between TSH and FT3:FT4, TSH in isolation has a U-shaped relationship with age: TSH is high in youth, it is reduced in middle adulthood, and then it rises again with age. When TSH is high for the second time in the human lifespan, it leaves behind the FT3 and may either maintain or raise the FT4 in thyroid-healthy people.

But one must be very careful not to justify the loss of the TSH-stimulated FT3:FT4 ratio as “adaptive” in all cases in the elderly. A reduced FT3 and FT3:FT4 ratio is also a distinctive feature of nonthyroidal illness syndrome (NTIS) that increases mortality and morbidity at all ages, especially in the elderly (Langouche et al, 2019).

To discuss the findings, I briefly review the biological necessity of sufficient circulating FT3, the role of metabolism in maintaining the FT3:FT4 ratio, and the risks of high-normal FT4 as articulated by leading scientists.

To conclude, I reflect on free thyroid hormone measurement and assessment of hormone relationships and the FT3:FT4 ratio as part of a complete and accurate diagnosis of HPT axis function.

COPYRIGHT NOTICE: Reproduction of a copyrighted article’s data, graphics, and quotations falls within US copyright law as “fair use” and within Canadian copyright law as “fair dealing.” — “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” (Stanford University).

A caution against misapplication

While Karavani and Strich gave important insights into the healthy HPT axis in youth and age, males and females, one should not jump to conclusions about which hormone levels, ratios and relationships are “adaptive” or “maladaptive” to all humans of a given age or sex.

These are not health risk studies.

These are not thyroid therapy clinical trials.

Instead, these HPT axis norms are derived from a well-screened population with unique exclusion criteria.

Their purpose is to understand how the HPT axis functions in populations believed to be free of thyroid disease and free of certain known conditions that alter the HPT axis.

Strich and colleagues did not determine the thresholds of health risk or treatment targets for any of the following groups:

  1. Pregnant individuals,
  2. People on thyroid medications, whether anti-thyroid treatment or T4 and/or T3 hormone treatment, which distort TSH-FT4-FT3 interrelationships to varying degrees in different individuals (Midgley et al, 2015),
  3. People with thyroid peroxidase antibodies (Brown et al, 2016) or TSH-receptor antibodies (Paragliola) that are known to distort TSH-FT4 relationships as well as thyroid function,
  4. People with autonomously functioning thyroid nodules, which in many cases do not lower TSH but can nevertheless distort FT4 and FT3 levels (Giovanella et al, 2016),
  5. People who have severe chronic illnesses such as heart failure, liver failure, kidney failure, cancer, and diabetes, which can distort intracellular T4 and T3 metabolism (Langouche et al, 2019; van den Berghe et al, 2014), and which may involve drugs that distort pituitary response (Haugen, 2009; Benvega et al, 2018),
  6. People with a kidney disorder like nephrotic syndrome, which can alter TSH, T4 and T3 urinary clearance rates (Li et al, 2019; Yoshida et al, 1988),
  7. Individuals with mild or severe pituitary dysfunctions, including hypothalamic central hypothyroidism which can be misdiagnosed as subclinical hypothyroidism (Beck-Peccoz et al, 2017), or underrecognized autoimmune pituitary diseases (autoimmune hypophysitis, Frasca et al, 2021).

An advanced reader will interpret these research results by Karavani and Strich and colleagues with discernment as an aid to diagnosis, but they will not transform a description of a healthy population into a prescription for a vulnerable, handicapped individual.

Another common type of misapplication is to use research like this to “predict” unknown hormone levels based on insufficient data. This could occur if people expect men or women of a certain age to fit a certain FT4 and FT3 profile if they have a given TSH level, but they do not check to verify their FT4 or FT3. That would be the fallacy of stereotyping, of generalizing from the populations featured in these studies to the individuals not featured in these studies.

TSH-enhanced FT3 levels and FT3:FT4 ratios in youth

Hormone relationships in youth provide the context for interpreting the hormone relationships in aging. Without looking at childhood, a person may think that it’s always unhealthy to have mildly high or high-normal FT3 or high FT3:FT4 ratios. It’s an important lesson in the flexibility of the HPT axis.

Karavani and colleagues’ publication in 2014 highlighted TSH-enhanced FT3 levels and FT3:FT4 ratios in children and youth aged 1–20 years. They also studied the effect of body mass index (BMI) on thyroid hormones.

They analyzed lab results from 3,276 subjects in Jerusalem, Israel (1,317 from age 1–10, and 1,959 aged 11–20 years), using a Cobas Roche assay.

Their exclusion criteria were:

  • No positive titers of TPOAb or TGAb on their medical record,
  • No past or current thyroid treatment with levothyroxine, methimazole, propylthiouracil, recombinant TSH,
  • No treatment with drugs known to affect thyroid function, such as antiepileptic drugs, lithium, or glucocorticoids.
  • No samples in which TSH levels were above 7.5 mIU/L “because such patients would have been under medical follow-up or evaluation and were more likely to be ill.”

They divided this population aged 1-20 into TSH quartiles (819 samples per quartile) to examine the correlations. The relationships remained when analyzed by age groups and males vs. females, so the data is presented for the entire cohort.

These are high levels compared to adults. Adult reference ranges for FT3 on a Roche assay are around 3.1 to 6.8 pmol/L. As you will see below, the adult average FT3:FT4 ratio is 0.33 mol/mol.

Across their entire cohort aged 1-20, there was a positive linear correlation of r = 0.12 (p <0.0001) between TSH and FT3. In other words, young people with higher levels of TSH had higher levels of FT3.

The same strength of correlation and significance, r = 0.12 (p <0.0001), was found between TSH and the FT3:FT4 ratio.

In contrast, no correlation was found between TSH and FT4 (r = -0.02; N.S.) in this age cohort.

Interestingly, they discovered that a subgroup of youth with obesity was responsible for amplifying the FT3 levels in the total population graph shown above:

“Thyroid hormone levels differed by weight groups; TSH and FT3 increased significantly in the obese group, whereas FT4 declined significantly with increasing BMI. There was a minor positive correlation (linear coefficient = 0.07) between BMI percentile and TSH (P = 0.002).”

(Karavani et al, 2014)

When the population with normal weight was studied, the rise in FT3 and FT3:FT4 ratio from the third to fourth TSH quartile was eliminated, as shown.

In the normal weight subset,

  • FT4 levels were slightly higher in all TSH quartiles (mean 16.32, versus 16.06 in total population), and
  • TSH levels were slightly lower overall (mean 5.93, versus 6.15 in total population).

Obesity enhances the FT3 elevation at every level of TSH, but the influence of age is more powerful than that of obesity.

Overall, in the pediatric group, pituitary TSH still rises in response to FT4 decreases, but as TSH increases, FT3 rises in parallel.

If there is a strongly positive TSH-FT3 relationship in childhood, does it fade as we age? These dynamics raised questions about the entire lifespan, which were answered two years later by the same research team.

TSH-FT4-FT3 relationships across the lifespan

The remainder of this article reviews the intriguing research by Strich and colleagues (coauthored by Karavani) in 2016, which built on the methods and findings from their 2014 publication.

Their purpose was not to establish age-specific reference intervals, but to analyze how age and sex affected hormone levels.

They demonstrated how TSH, FT3 and FT4 levels differ by age and by sex in 10,227 males and 17,713 females (27,940 people) in Israel.

They used the same exclusion criteria mentioned above in Karavani’s study, but also added a low TSH cutoff. They included no samples in which TSH levels were above 7.5mIU/L or below 0.2mIU/L.

Therefore, the population still included a percentage of ambulatory people who may have had severe chronic illnesses that can compromise the HPT axis, because nonthyroidal illness syndrome (NTIS) and drugs can keep TSH normal or mildly low while thyroid hormone losses occur, and mild TSH elevations may occur during recovery (Braithwaite, 2015).

Strich and team unfortunately didn’t give their laboratory’s FT4 or FT3 reference ranges. However, they used a Cobas Roche platform for all their assays, and these usually have medians and reference intervals close to the following:

  • Median TSH: approximately 1.5 to 1.6 mIU/L; Median FT4, approximately 15-16 pmol/L; Mean FT3, approximately 5.0 pmol/L.
  • Ranges from Hoermann et al, in various studies: TSH, 0.4 to 4.0 mIU/L; FT3 3.1 to 6.8 pmol/L, FT4, 11-23 pmol/L.
  • Ranges from Ganslmeier, et al, 2014, who selected a very healthy population, which narrowed all the ranges while keeping the medians approximately the same: TSH 0.58–3.49 mIU/L; FT4 11.58-20.46 pmol/L; FT3, 3.56 – 5.88 pmol/L.

1a. TSH-FT4 relationships by age

In the bar graphs below, each age group has four vertical bars expressing their four “TSH quartiles” with increasing TSH levels.

(If you wish to see Strich’s data tables visualized another way, as heat maps, see a post that complies data from Strich as well as other studies — “Normal FT3:FT4 thyroid hormone ratios in large populations.”)

Strich and colleagues only included data from patients with TSH levels between 0.2 and 7.5 mIU/L. Therefore,

  • The darkest blue “TSH Q1” vertical bar refers to the hormone level in the study population with the lowest 25% of TSH levels, but no lower than 0.2 mIU/L.
  • The lightest blue “TSH Q4” vertical bar refers to the hormone level in the study population with the highest 25% of TSH levels, but no higher than 7.5 mIU/L.
NOTE: The X axis has irregular age groupings because Strich and colleagues categorized their data in this way. The groups 40-60 and 60-80 span 20 years each, while others cover only 10 years each.

The horizontal dotted lines provide population-wide, and adult-wide, means for comparison. They reveal that:

  • Overall, FT4 levels per unit of TSH drop in the second decade of childhood.
  • FT4 patterns across TSH quartiles are similar in age cohorts for 10-20 and 20-30 years of age.
  • The negative relationship between TSH and FT4 is most extreme in ages 30-80 (at higher TSH levels, FT4 is lower).
  • After age 80, the population’s slightly higher TSH maintains higher FT4 levels than they did in the 40-60 and 60-80 age group.

These patterns are relevant because they are not linear, but slightly U-shaped across age cohorts.

Why does the TSH escalate over age 80 while FT4 does not fall? There are many possible reasons. TSH escalation may be permitted by reduced pituitary sensitivity to thyroid hormone, or may be necessary due to reduced TSH bioactivity, or it may be caused by slower TSH metabolic clearance rates, as seen in aging mice (Silva and Larsen, 1978).

1b. TSH-FT4 correlations by age and sex

In ages 30-80, are the women or the men in the study population mainly responsible for the lower FT4 per increasing TSH quartile?

To answer this question and others, Strich and colleagues also did something interesting. In addition to analyzing each hormone separately by age and TSH quartile, they provided tables with positive or negative Pearson’s correlation coefficients between the hormones so that they could examine shifts in the HPT axis based on sex as well as age.

Many scientists create correlations between TSH or FT4 and other biomarkers or health outcomes, but they often leave the TSH-FT4 correlation unexamined, treating it as a given.

Pearson’s correlation is mainly used to discover the strengths of linear relationships, not log-linear ones. Strich and colleagues did not specify whether TSH was log-transformed into linear units before calculation (see Log Transformations in a Linear Model, Ford, 2018). However, their results are quantitatively similar to those of researchers who log-transformed TSH (Lamichhane et al, 2020).

As above, I have turned Strich’s tabulated data into graphs to visualize the transformation.

They confirm the shifting TSH-FT4 correlations with age and sex. Overall, females’ pituitaries have a stronger sensitivity to negative feedback from FT4, and males lose the negative correlation above age 60:

NOTE: In these correlation graphs, I expand the data from cohorts 40-60 and 60-80 into two equal data points so that the X axis represents a timeline with 10-year increments. Their correlation data did not include an age 80+ category.

All the correlations were relatively weak and fell mainly on the negative side.

The strongest negative linear correlations were seen in three age groups:

  1. Females aged 40-59, r = -0.19
  2. Males aged 30-39, r = -0.14
  3. Females aged 60-80, r = -0.13

Surprisingly, NO statistically significant correlation was found in:

  • Males aged 60-80, r = 0.00
  • Males and females aged 20-29, r = 0.03, r = -0.01
  • Males aged 1-9 r = -0.04

Nevertheless, the statistical significance (p < 0.05) lends weight to a weak negative TSH-FT4 relationship in Strich’s research results, since it raises confidence that it is a pattern likely to be found in a larger population of a similar type.

The statistical significance of Strich’s correlations are based on more than 1,000 data points in each of Strich’s age/sex cohort — except ages 60-80, which had only 300 to 400 data points per sex.

How to interpret correlation coefficients

Click to view this section

As correlations come closer to the bold horizontal line at r = 0.00, they lack a negative or positive linear pattern. As they depart from the 0.00 line in either direction, they become more strongly positive or negative.

According to classical theory, the HPT axis is characterized by a negative (or “inverse”) log-linear correlation between TSH and FT4.

This correlation only expresses the way the two hormone levels usually move in opposite directions during primary thyroid dysfunction:

  • When FT4 falls slightly, one expects TSH to rise in an attempt to compensate for thyroid hypofunction by stimulating the thyroid more strongly.
  • When FT4 rises slightly, one expects TSH to fall in an attempt to compensate for thyroid hyperfunction by reducing stimulation of the thyroid.

A positive or absent correlation will occur when the two hormone levels move in parallel, either rising or falling together, or when TSH secretion is no longer regulated by FT4 (and/or FT3). This implies abnormal pituitary TSH secretion:

  • When TSH rises or remains inappropriately normal while FT4 (and/or FT3) rises in syndromes of “resistance to thyroid hormone,” (RTH) the hypothalamus and pituitary seem insensitive to negative feedback from thyroid hormone. It is also seen when hyperthyroidism is driven by a TSH-secreting pituitary adenoma.
  • When TSH falls or remains inappropriately normal while FT4 (and/or FT3) falls in central hypothyroidism or disorders of inhibited TSH secretion, the hypothalamus and pituitary do not respond appropriately to the loss of thyroid hormone, and even a healthy thyroid gland would be insufficiently stimulated to produce thyroid hormone.

The common belief that pituitary TSH disorders are “rare” helps to justify TSH-only screening. Once implemented at a population level, TSH-only screening ensures that such diagnoses are rare. Diagnoses remain rare because informed physicians are rare. In this way, TSH-only screening systems remain apathetic and blind to the true prevalence rates of pituitary TSH secretion disorders.

A statistically significant correlation (dots highlighted in color) expresses strong confidence that a linear correlation exists between TSH and FT4 levels within that specific age and sex cohort in the general population that the sample represents.

The test of statistical significance revealed that some relatively weak (low) correlations had a significant trends in a population subgroup.

Why is the TSH-FT4 linear correlation no stronger than -0.19?

One might expect, given the strong messaging about the “log-linear relationship between TSH and FT4,” that it would be stronger than r = -0.19 no matter the age or sex or TSH level of the population.

For more than a decade, scientists have known that the TSH-FT4 logarithmic relationship is not log-linear across the full range of TSH values (See Hoermann et al, 2010; Hadlow et al, 2013).

As shown below, the curved “double negative sigmoidal” trendline flattens to become nearly horizontal while TSH and FT4 are within and close to their reference intervals (Brown et al, 2016).

In this graph, the TSH units on the Y axis are in logarithmic scale, while FT4 changes on the X axis are expressed in linear scale.

  • False: The straight line on the left expresses the incorrect concept of the log-linear relationship that maintains the same degree of linearity throughout. This straight line would represent a log-transformed Pearson’s correlation coefficient such as -0.25, and it would enable prediction of a range of expected FT4 levels for every TSH level, and vice versa.
  • True: The double sigmoidal curves on the right express the more accurate TSH-FT4 relationship. The negative correlation between TSH and FT4 becomes weaker or stronger depending on where the FT4 is located, and the where the concurrent FT3 is (not shown), and one’s age, and whether or not TSH-inflating thyroid peroxidase antibodies are present, and so on.

The trendline on the right becomes almost horizontal at the point where the two population medians for TSH and FT4 intersect. A horizontal trendline represents no correlation, r = 0.

Therefore, the log-transformed TSH (lnTSH) correlation coefficient is expected to be weaker within a well-screened, healthy population. Strich’s data set focused on a well-screened population with TSH levels between 0.2 and 7.5 mIU/L.

However, the correlation does not entirely disappear when one includes TSH values from 0.2 to 7.5 mIU/L because this range still includes FT4-TSH data points that depart from the population median in opposite directions (TSH high-normal, FT4 low-normal).

How weak were the TSH-FT4 correlation coefficients in other studies?

Click to expand section

Ren and colleagues (2021) compared TSH-FT4 Pearson’s correlations in populations with different iodine status. While excluding all signs of thyroid disease, they also found the following r values:

  • r = -0.046 (no correlation) mean urinary concentration 100-199 mcg/L, “adequate” iodine
  • r = -0.16 (low negative correlation) mean urinary concentration 200-299 mcg/L, “above requirements”
  • r = -0.177 (low negative correlation) mean urinary concentration >300 mcg/L, “excess”

Lamichhane and team (2020) found TSH-FT4 Pearson’s correlations to be weakly negative while FT4 was within reference range:

  • r = -0.04 in “subclinical hypothyroidism” (p < 0.26)
  • r = -0.11 in “euthyroidism” (p < 0.001)
  • r = -0.13 in “subclinical hyperthyroidism” (p < 0.07)

In subclinical thyroid disease, TSH and FT3 are more strongly correlated than TSH and FT4, as shown by the clear vs. solid bars in the graph below, derived from Lamichhane and colleagues:

NOTE: The FT3 vs. FT4 correlation is not the same as the FT3:FT4 ratio. It would have been helpful for Lamichhane to have calculated the lnTSH vs. FT3:FT4 ratio.

This graph of Lamichhane’s data gives us some clues to interpret Strich’s correlation data as relatively stronger or weaker than it is in certain thyroid diagnoses.

Which types of hidden thyroid hormone imbalance might have increased the negative correlation between TSH and FT4 in Strich’s population of women aged 40-60, but made the negative correlation disappear in men aged 60-80?

The graph above shows that the relative strength of the negative TSH-FT4 log-linear correlation only distinguishes between two diagnoses: euthyroidism and overt hypothyroidism. The only subgroup with a strong negative correlation between logarithmically-transformed TSH and FT4 is “overt hypothyroidism,” with r = -0.75.

Subclinical hypo had a weaker correlation and subclinical hyper had almost an unchanged TSH-FT4 correlation compared to euthyroidism. This is because endogenous subclinical hyper and hypo are both more strongly characterized by a rising FT3:FT4 ratio:

  • In subclinical hypo, the thyroid’s successful maintenance of near-median FT3 levels under a mildly elevated TSH may prevent the TSH from rising higher.
  • In subclinical hyper, the FT3 may oppress TSH long before FT4 exceeds the reference limit.

Both of these are compensatory mechanisms that would weaken the negative TSH-FT4 correlation.

Therefore, Strich and colleagues would only see a strongly negative TSH-FT4 correlation if more people in a given age cohort had a significantly reduced FT4 toward the lower end of reference, poorly compensated by FT3, while TSH rose toward its upper reference limit and beyond.

2a. TSH-FT3 relationships by age

Next, Strich and colleagues revealed that the relationship between TSH and FT3 shifted over the lifespan. Finally, we get to examine the most potent thyroid hormone concentration entering cells all over the body, the concentration so many researchers continue to ignore.

During early youth, as TSH rose and FT4 fell through the quartiles, FT3 rose in parallel with the TSH. This positive TSH-FT3 relationship flattened in early to middle adulthood, and then inverted so that the relationship became negative in older people.

In children and youth, notice that the rising pattern of FT3 is the opposite of the TSH-FT4 relationship. Before age 30, a rise in TSH provides additional FT3, or maintains FT3. This compensates for the FT4 losses per quartile shown in section 1a, above.

Notice that the mean FT3 level in this well-screened population does not fall below the lower limit of its reference range, which is placed at 3.1 to 3.5 in most Roche assays.

How low does the FT3 fall in males and females over 80? Their mean remains above 3.7 pmol/L, but what was their median and full range? Strich and colleagues did not provide a +/- SD or range to give us an idea of the distribution of FT3 levels, but they did provide a correlation coefficient shown below.

It is unclear whether anyone in the >80 age group had nonthyroidal illness syndrome lowering their FT3, because age-associated illnesses were not excluded from this population. What is the age-appropriate FT3 reference interval in an 80+ population whose normal RT3:FT4 relationship proves they are not in a state of “nonthyroidal illness syndrome” in which RT3 builds up relative to its only source, FT4?

A physician should not presume that a FT3 of 3.7 is sufficient for health in every elderly individual, but should consider the concurrent FT4 level and clinical presentation.

2b. TSH-FT3 correlations by age and sex

The relationships outlined in the bar graph of section 2a are further analyzed by sex:

This line graph reveals a significant positive correlation between TSH and FT3 transforms into a negative correlation with age.

“Until 30 years of age, there was a significant positive linear correlation of TSH with FT3 (r = 0.14; P < 0.001), while in the above 30 groups, no positive correlation was noted.”

(Strich et al, 2016)

Noteworthy features:

  • The high TSH and FT3 in the first months and years of life are confirmed in our post “Pediatric and teenage TSH, FT4, and FT3 levels.”
  • Teenage boys as well as women aged 30-39 and 60-80 had no significant TSH-FT3 correlation, either negative or positive. It is possible that these cohorts’ data points were randomly distributed or more densely clustered around the population means without revealing a trend.
  • Surprisingly, females over 60 tend to get slightly more FT3 per unit of TSH stimulation than males do, on average, although that does not mean they have higher FT3 levels than males.

At this point, I must dispel a common myth.

TSH-stimulated FT3 does not come from peripheral metabolism, but from the thyroid.

This was proven by studies in people with no thyroid glands who are injected with recombinant human TSH (rhTSH, see Beukhof et al, 2018). While the patients were maintained on a steady dose of LT4 hormone, TSH injection escalated their circulating TSH levels far above reference range. On average, all their thyroid hormone levels stayed the same, even RT3 levels, while only their FT3 and TT3 experienced a mild reduction.

In contrast, when people with healthy thyroids undergo rhTSH injection, their FT3 and FT3:FT4 ratios rise (Fast et al, 2010).


  • An increase in TSH-receptor signaling enhances T4-T3 conversion and T3 synthesis within the healthy thyroid gland.
  • An increase in TSH-receptor signaling does not directly enhance net T4-T3 conversion outside the thyroid gland to maintain or elevate circulating FT3.

The thyroid’s T3:T4 secretion ratio may average approximately 1:16 in healthy people, but the full range is very wide. Thyroidal T3 secretion is very flexible and varies widely from person to person. There is no throttle on human thyroidal T3 secretion. One healthy subject obtained more than 40% of their daily T3 production from their thyroid gland (Pilo et al, 1990; see “Meet a person with the perfect T3:T4 thyroid secretion ratio“).

Scientists have revealed that as TSH levels rise, it stimulates the rate of thyroidal T4 and T3 synthesis, but the rate of T3 synthesis is enhanced to a greater degree (Citterio et al, 2017). This elevates the T3:T4 secretion ratio on the T3 side.

In addition, thyroids function as metabolic engines. Unlike the thyroids of mice and rats, which mainly express D1 enzyme, the human thyroid expresses both D1 and D2 enzymes that convert T4 to T3 hormone (Salvatore et al, 1996). Compared to all other human organs and tissues, the thyroid has the highest density of mRNA for DIO1 and DIO2, the genes that regulate these enzymes. Circulating FT4 re-enters the thyroid through the bloodstream, which is the same way that circulating TSH reaches thyroid tissue.

TSH-receptor signaling upregulates both DIO1 and DIO2 in the thyroid gland, enhancing the rate of T4-T3 conversion. Studies of Graves’ hyperthyroidism exemplify the incredible power of D1 and D2 enzymes within the TSH-receptor-stimulated thyroid gland to convert circulating T4 to T3 (Salvatore et al, 1996).

Why do higher levels of TSH enhance thyroidal T3 secretion to maintain or raise FT3?

The thyroid gland’s flexible T3:T4 secretion ratio is necessary to maintain or recover homeostasis (balance; euthyroidism) during or after a challenge to circulating T3 supply:

  • To compensate for a thyroid’s reduced ability to synthesize T4, either in early mild thyroid hypofunction and/or iodine deficiency (Hoermann et al, 2020)
  • To compensate for shortfalls in peripheral T4-T3 conversion outside the thyroid gland (See our review “Thyroid T3 secretion compensates for peripheral T4-T3 conversion“)
  • To support recovery from a transient T3 deficiency caused by “nonthyroidal illness syndrome” during a severe illness (Braithwaite, 2015; van den Berghe, 2014).

Why does TSH-stimulated FT3 elevation weaken and fail during aging?

Apparently, the “TSH-T3 shunt” within the thyroid gland (Berberich et al, 2018) is not working at the same rate as it does in earlier adulthood and childhood. In aging, one or more mechanisms may become weaker:

  • TSH molecules may become less bioactive
  • The thyroid’s TSH receptors may become less sensitive, or other factors may interfere with thyroid cell function in response to TSH
  • Metabolic T3 losses are too significant to be fully compensated by an aging thyroid.

3a. TSH relationships to the FT3:FT4 ratio during aging

When the FT3 is divided by the FT4, for instance, 5.0 pmol/L FT3 divided by 15 pmol/L FT4, the quotient yields their relative ratio in blood, such as 0.33. This ratio of 0.33 was the average FT3:FT4 ratio in the healthy adults in Strich’s study. The clinical and metabolic significance of the FT3:FT4 ratio is explained further below.

The graph below sows how the ratio shifts with increasing levels of TSH stimulation.

The FT3 and FT4 patterns shown in the graphs above combine to create a more uniform linear trend in FT3:FT4 ratios across age groups. The quartiles of increasing TSH generally enhance this hormone ratio — until after age 80:

NOTE: An error in Table 1 reported the ratio as “0.41” for 60-80 TSH Q1. This is highly unlikely given the trends across age groups, and given the quartile’s mean FT3 of 4.45 and mean FT4 of 15.26 (the ratio of these averages is 0.29). It is also inconsistent with this group’s data in Tables 2 and 3. Therefore, I have corrected the value, presuming it was likely a typographical error for 0.31.

By looking at the graphs for FT4 and FT3, one can see that the lower ratio in populations over 80 is not only due to a decrease in FT3 but partly due to a mild increase in the FT4 per unit of TSH.

In seniors, instead of boosting FT3, a rising TSH may maintain or even augment FT4 in old age. If this were not the case, the FT3 may fall too low for health.

3b. Correlations between TSH and the FT3:FT4 ratio, by age and sex

This line graph has the strongest positive correlation in female children aged 1-9.

Among the three graphs presented in this section, this graph has the most linear trajectory.

The main exception is for their cohort of males aged 30-39, who appear quite youthful in their ratio’s positive correlation with TSH.

3c. TSH-stimulated FT3/FT4 ratios across both sexes

Finally, a graph provided by Strich combines correlation data from both sexes across the age groups. The combination of both sexes reduces the wobbles in the line graph.

Strich explain what they found in the correlations as they combined data from both sexes:

“In the pediatric and young adults, until age 40, there was a positive and significant correlation between TSH and FT3/FT4 ratio (r = 0.08; P < 0.001), but in the older
groups, this correlation decreased to nil as age increased (from 0.04 to −0.08).

This trend, i.e. the decreasing correlation with age was linear and significant (r = −0.94, P = 0.02).

(Strich et al, 2016)

It’s not clear whether a linear trendline is the best fit to the data, but they drew a trendline through the wobble from early childhood to the senior years. This trendline line expressed an extremely high correlation of r = -0.94, which is very close to a perfect correlation of -1.0.

But keep in mind this r = -0.94 is not a correlation between hormones. It is the correlation between two different variables:

  1. the variable of age (shown on the X axis) and
  2. the variable of the correlation coefficient between TSH and the FT3/FT4 ratio (shown on the Y axis).

The very different U-shaped trend of mean TSH levels across age groups (from high to low to high) distracts many people from noticing this more fundamental downward shift in TSH-stimulated FT3 per unit of FT4:

NOTE: The X axis is missing ages 20-30 in the original article. Nevertheless, the trend in mean TSH levels is well visualized.

People who generalize about the benefits of the (re)elevated TSH in aging often have no idea what is actually going on with FT3, FT4 and the ratio underneath that superficial TSH measurement.

Two senior citizens may have the same TSH level of 7.5:

  • One may have compensated hypothyroidism (euthyroid hyperthyrotropinemia) providing a sufficient supply of both FT4 and FT3 for healthy aging.
  • Another may have a FT3 and FT4 both at the basement of their age-specific intervals while a chronic nonthyroidal illness, and a lack of TPO antibodies, prevents their TSH from rising up to 15. It’s not their fault if their inhibited secretion of pituitary TSH fails to signal to a physician the true severity of their hypothyroidism.

Why a sufficient FT3:FT4 ratio and circulating FT3 are still vital in old age

The FT3:FT4 ratio matters to health because both FT3 and FT4 are carried into cells side by side. The rate at which they enter cells is partly determined by their relative availability in blood.

Every organ and tissue requires classical T3 hormone signaling to send commands to build new proteins. A healthy level of intracellular T3 signaling depends on both the local rate of intracellular T4-T3 metabolism plus a “top-up” of direct FT3 from circulation.

A variable percentage of intracellular signaling will always derive from Free T3 being carried into cells. No tissue can remain healthy for very long on “normal FT4” levels alone. (See “How do we get enough T3 into thyroid hormone receptors?“)

To meet the needs of all tissues and cells regardless of their diverse enzyme expression, sufficient FT3 is needed. This is why Abdalla and Bianco titled their 2014 scientific article “Defending plasma T3 is a biological priority.”

But how much is “sufficient”? It always depends on the concurrent FT4 supply and the intracellular T4-T3 conversion rate, as well as one’s age, sex and thyroid gland status.

  • If there is less FT4 in circulation (below the population median), more FT3 is needed to top up T4-T3 conversion.
  • If there is more FT4 in circulation (above the population median), less FT3 may be needed … as long as T4 is converting efficiently to T3 within all tissues.

As explained by Hoermann’s 2016 scientific review article on “Relational Stability” among TSH and thyroid hormones (which I’ve paraphrased in several posts), the HPT axis adjusts TSH and FT4 to compensate for metabolic stressors. This occurs in both mild hypo- and hyperthyroidism:

  • When FT4 falls in early or mild hypothyroidism, the TSH rises not only to maintain FT4 but to defend FT3 by stimulating the thyroid more vigorously. (Hoermann et al, 2020).
    • In thyroid hormone metabolism, as FT4 falls, T4-T3 conversion via Deiodinase type 2 becomes more efficient.
  • When FT3 rises in early or mild hyperthyroidism, the TSH falls in an effort to minimize the further escalation of both FT3 and FT4, which is usually driven by TSH-receptor stimulating antibodies (Graves’ disease) or autonomously functioning thyroid nodules.
    • In thyroid hormone metabolism, as FT4 rises, T4-T3 conversion via Deiodinase type 2 becomes less efficient. As both hormones rise, deiodinase type 3 is upregulated, enhancing the metabolic clearance of T4 and T3.

These HPT axis responses attempt to adjust the FT3:FT4 ratio in an effort to maintain appropriate levels of circulating hormone to enable euthyroid levels of intracellular T3 signaling.

But will the TSH adjustment and metabolic adjustment succeed or fail to maintain euthyroidism in the face of a mild thyroid hormone imbalance?

It depends on the health of the thyroid gland and the health of the body-wide system of thyroid hormone metabolism.

The FT3:FT4 ratio as an index of thyroid hormone metabolism

“Metabolism” involves not only T4-T3 conversion but also T3 and T4 metabolic losses to various non-T3 metabolites such as Tetrac, Reverse T3, and T3-sulfate. Clearance also occurs as T4 and T3 are lost in urine. (See “A complete pathway map of T4 and T3 metabolism and clearance“).

Active transporters are the gatekeepers of thyroid hormone metabolism and signaling. A large family of transmembrane transporters exists to carry T3 and unconverted T4 out of cells.

Each tissue and cell type expresses its unique set of transporters. Many prefer to carry T3, and some prefer to carry T4, but most carry both hormones across the membrane, hand in hand.

Therefore, the FT3:FT4 ratio expresses the balance of continuous cellular influx and efflux of T3 and T4, and it is a major factor co-regulating the rate of intracellular T3 signaling.

Some cells are 100% dependent on Free T3 being carried into cells from circulation. These T3-dependent cells are the ones with nuclear thyroid hormone receptors that express neither deiodinase type 2 (D2), the most efficient enzyme for T4-T3 conversion, nor deiodinase type 1 (D1) the dominant enzyme for T4-T3 conversion in the thyroid, liver, and kidney.

For example, neuronal cells are 100% dependent on FT3. They express neither D2 nor D1, so they cannot produce their own intracellular T3 from T4. Instead, they express D3, the enzyme that converts T4 to RT3 and inactivates T3 into an inactive form of T2. This protects neurons from mildly excess T3, a situation they can handle well. But neurons have thyroid hormone receptors. They still need enough T3 to float unconverted past these D3 enzymes and bind to receptors in the nucleus (Bianco et al, 2019).

Metabolism and clearance pathways can easily get rid of a mild isolated excess of FT3, when there is not a concurrent excess of FT4 placing high demands on the time and energy of enzymes. Both D1 and D3 are upregulated by T3 hormone signaling, and they both clear T3 to T2. Therefore, mild FT3 excess is not harmful if FT4 is low enough to prevent hyperthyroidism from occurring due to excess intracellular T4-T3 conversion in addition to excess intracellular influx of FT3.

The major danger to health occurs when both FT3 and FT4 are above reference range on a chronic basis, as occurs in Graves’ hyperthyroidism. In that situation, the D1 and D3 enzymes cannot clear intracellular T3 and T4 at a rate faster than transporters carry them into cells, and some cells will not express D1 or D3 and will have no defense system.

In contrast, insufficient T3 is not healthy for the metabolic system, either. Low-normal and low FT3 downregulates D1 enzyme, which is most densely expressed in thyroid, liver, and kidney, and is responsible for a significant, variable percentage of T4-T3 conversion. In people with a genetically weaker D2 enzyme, there will be poor metabolic compensation for a weaker D1 enzyme, and the T3-poor may become T3-poorer, especially if the thyroid gland is failing (See “Meet deiodinase type 1 (D1): The philanthropist enzyme“)

The risks of high-normal FT4 in the low FT3:FT4 ratio

A low FT3:FT4 ratio can be created either by FT4 rising, or by FT3 falling, or by both alterations occurring simultaneously.

However, in seniors who gain a mild increase in FT4 within the normal range, together with lowered FT3, this augmented FT4 level can become a health concern.

Recently, various researchers have cautioned that a high-normal FT4, especially in the presence of reduced FT3, poses health risk (Anderson et al, 2020; Ataoglu et al, 2018; Brozatiene et al, 2016). The risk is heightened in the elderly.

Anderson and colleagues’ 2020 appendix data revealed the association between levels of TSH, FT4 and FT3 and prevalence rates for many chronic diseases. In the following heat map, I have ranked the prevalence rates four cardiovascular diseases. Pink is high prevalence, and teal is low:

See “Prevalence rates for 10 chronic disorders at various FT4, TSH and FT3 levels.”

As one can see, FT3 is correlated with disease when it is low, and FT4 when it is high-normal, and TSH when it is high-normal. This is why the FT3:FT4 ratio is a rising biomarker in newer thyroid science. It is often more statistically significant, and clinically significant, than isolated thyroid hormone levels and TSH.

Many people misunderstand why high-normal FT4 may become a health risk even when the TSH is normal, even if the FT3 is not elevated, and even if the individual is clearly not hyperthyroid.

  1. Slower D2 enzyme function. High-normal FT4 in the face of low-normal FT3 is a concern because it is capable of slowing down intracellular T3 production. As FT4 rises above the population median, it reduces the activity of deiodinase type 2, the main enzyme that converts T4 to T3. It does this through a process called “ubiquitination” (See “Ubiquitination: The glass ceiling of T4 monotherapy“).
  2. Even slower D2 and D1 function during nonthyroidal illness. In severe illnesses, high-normal FT4 may become more of a metabolic burden than benefit. Both peripheral D2 and D1 enzymes are oppressed by tissue hypoxia (such as that caused by cardiac ischemia), by intracellular oxidative stress, and by inflammatory cytokines like IL-6.
  3. T4 activity at integrin αvβ3.  High-normal FT4 can also become a health concern because it signals at a T4-preferring integrin αvβ3 receptor on the membrane of certain cells located in rapidly-dividing blood vessels, cardiac tissue after trauma, and tumors. When T4 and its metabolite RT3 bind to this integrin receptor, they send different signals than T3, which only binds to the receptor at supra-physiological levels. This receptor’s signaling can increase blood clotting (platelet aggregation) and prevent cancer cell death. This is why researchers are developing a drug to block this non-essential receptor (Davis et al, 2018, 2019, 2020; see “Cancer scientists point finger at T4 & RT3 hormones“).

The shifts in the TSH-stimulated FT3:FT4 ratio with age are likely a physiological reason why many seniors with subclinical hypothyroidism do not benefit from TSH-normalizing, FT4-raising, FT3-lowering levothyroxine monotherapy (Midgley et al, 2015). It likely gives the 70 year old the hormone profile of an 80 year old, or a 90 year old, before their body has adapted compensatory mechanisms such as a slower T3 clearance rate.

Strich’s later suggestions (2018) that T3 hormone be included in thyroid hormone therapy is a wise and good one, but it will need to be a topic for a later post. The post will have to address the unreasonable fearmongering toward giving any T3 to people who clearly lack means of creating enough of their own T3.


An accurate, evidence-based assessment of thyroid function, more broadly, HPT axis function, involves answering questions about hormone relationships:

  • How appropriately is the thyroid responding to a given TSH level?
  • Is the metabolism supporting a healthy FT3:FT4 ratio?
  • Is the pituitary TSH responding appropriately to both FT3 and FT4 levels combined?

It seems that the declining TSH-stimulated FT3:FT4 ratio from youth to age is a major key to interpreting the function, and dysfunction, of the HPT axis.

With age, the TSH rises again to the levels seen in youth, but the FT3:FT4 ratio falls.

While this transformation occurs, thresholds of thyroid dysfunction and thyroid hormone health risk may also change.

Overt primary hypothyroidism is easy to see at any age — It’s clear that high escalations of TSH fail to compensate when the thyroid is too damaged, and the FT4 and FT3 fall too low for the individual body’s requirements in spite of that elevated ratio.

In subclinical hypothyroidism, if TSH rises in response to a low-normal FT4, and then the FT3 rises to compensate, and the person appears to be euthyroid and asymptomatic, then the system is working robustly as it does in youth.

But it’s not so easy when TSH is in the normal range or “subclinical hypothyroidism” zone in the elderly. A rising TSH may or may not be “subclinical” or compensated hypothyroidism. It depends on where the FT4 and FT3 are both located. That mildly elevated TSH may “fail” to compensate when FT4 falls low in old age.

In addition, is a spontaneously normalized TSH always a good thing in advanced age? The pituitary ages, too, and is not infallible. A normal TSH may also “fail” to rise high enough to maintain sufficient FT4 and FT3 for tissue euthyroid status in an elderly person.

This is why the measurement of FT3 and FT4 and the calculation of the FT3:FT4 ratio are so important to diagnosis in borderline or complex cases.

The measurement of TSH alone, and even the measurement of both TSH and FT4, can’t predict adequate T3 production or escalating T3 losses. Only direct measurement of FT3 and FT4 can determine whether any TSH level is succeeding or failing in its metabolic mission to “defend plasma T3” at a level appropriate for the individual’s health.


References for all articles cited in the “analyzing normal lab results” series are in a separate post.

One thought on “Age, sex and TSH-FT4-FT3 relationships: Advanced lessons

Leave a public reply here, on our website.

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