I queried Claude Opus 4.8 on this to see what the optimal levels are for lowest all cause mortality, and this is what it came back with:
Here’s the evidence-based synthesis. The crucial framing up front: for most of these markers the “lowest-mortality” level is where healthy long-lived people sit naturally, not a number you should drug a healthy person toward. Much of the “low FT3 / high rT3 = death” signal is reverse causation — sick people generate those patterns — and the levothyroxine RCTs (TRUST, IEMO) in subclinical hypothyroidism failed to improve hard outcomes. So read these as a physiological signature of slow aging, not a treatment target list.
Optimal-longevity targets
| Marker |
Longevity-optimal zone |
Direction of the mortality signal |
| TSH |
~1.0–2.5 mIU/L (high-normal tolerated, esp. with age) |
U-shaped: both suppressed and ≥10 raise mortality |
| Free T4 |
Lower third of range (~0.8–1.1 ng/dL / ~13–15 pmol/L) |
Linear — higher FT4 is consistently worse
|
| Free T3 |
Mid-normal (~2.8–3.4 pg/mL / ~4.3–5.2 pmol/L) |
Low FT3 tracks mortality, but mostly as illness marker |
| Reverse T3 |
Low-normal |
Higher = worse, but it’s a sickness flag, not a lever |
|
FT3/FT4 ratio(derived) |
Higher |
Efficient peripheral conversion = lower mortality |
| TPO antibodies |
Negative / undetectable |
Positivity → inflammation, AITD progression, stroke |
| Tg antibodies |
Negative / undetectable |
Same; weaker independent signal |
The reasoning per marker
TSH — aim mid-to-high-normal, not low. The single most striking longevity finding: families bred for exceptional survival run higher TSH. Leiden Longevity Study offspring averaged ~0.8 mIU/L higher TSH around the clock than spousal controls, and lower family mortality among nonagenarian siblings’ parents was associated with higher serum TSH and lower free T4. This replicates in Ashkenazi centenarians. The mechanism is a lower thyroidal response to TSH — the gland is set to idle. Meanwhile the danger is at the edges: subclinical hyperthyroidism (suppressed TSH) and TSH ≥10 both raise total/CVD mortality, heart failure, and AF in the large IPD meta-analyses. In a 6,054-person prospective cohort, higher TSH was not associated with all-cause mortality overall, but cardiovascular-mortality estimates diverged — increased risk in those under 72 and decreased risk in the elderly, which is why the high-normal tolerance widens with age. PubMednih
Free T4 — the cleanest, most actionable target: keep it low. This is the most reproducible signal in the whole field. Higher FT4, even within the normal range, predicts more death. The same cohort found higher free thyroxine associated with all-cause mortality (HR 1.18 per unit) and with cardiovascular mortality in the elderly (HR 1.61). It holds in elderly men (MrOS-Sweden), the Rotterdam Study, and frailty/cognition endpoints. Low-normal FT4 is the cell-level correlate of low metabolic “burn rate.” If any single number matters for your stack, it’s this one. nih
Free T3 — wants nuance, not maximization. Two opposing truths. In hospitalized/elderly populations, low FT3 robustly predicts mortality — but that’s largely non-thyroidal illness syndrome (NTIS), where inflammation downregulates D1/D2 deiodinase. Low-T3 syndrome carried 15.1% vs 4.1% 30-day mortality in non-critically-ill internal-medicine patients, which is prognosis, not a deficiency to correct. Yet in healthy people the picture inverts toward efficiency: the NHANES sensitivity analysis found the highest FT3/FT4 ratio quartile had lower all-cause mortality than the lowest (HR 0.70). Leiden offspring had slightly lower absolute FT3 but a higher FT3/FT4 ratio — they convert efficiently from a low T4 base. So the target is mid-normal FT3 achieved through good conversion, not a high FT3 propped up by high FT4. PubMed CentralPubMed Central
Reverse T3 — a sickness barometer, not a target organ. A 2021 review reports serum rT3 positively correlates with all-cause mortality (HR 1.23), likely due entirely to the adverse effects of non-thyroidal illness, of which rT3 is a sensitive marker. Translation: elevated rT3 tells you the body is shunting T4 down the inactivating D3 pathway because of illness, stress, calorie restriction, or drugs (amiodarone, glucocorticoids). Worth noting for your context — rT3 can rise with carbohydrate restriction, short-term calorie deprivation, poor glucose control, and high-dose soy isoflavone, so aggressive fasting/keto can transiently push it up without pathology. Keep it low-normal; don’t chase it pharmacologically. Sage JournalsOptimalDX
TPO & Tg antibodies — you want them gone. Negative is the target. Beyond predicting progression to overt hypothyroidism and Hashimoto’s, even sub-positive detectable TPOAb has been proposed as a sign of low-grade inflammation with negative health effects in euthyroid people, with associations to arterial wall thickening and stroke. One genuine wrinkle: TgAb behaves oddly metabolically — a large Chinese national survey linked TgAb presence to lower glucose/lipid disorder risk, and the NHANES autoimmunity cohort found TgAb inversely associated with hypertension — so its independent mortality contribution is muddier than TPOAb’s. Net: undetectable is ideal; isolated low-titer TgAb in a euthyroid person with normal FT4 is a softer flag than positive TPOAb. Oxford Academic
The integrated longevity phenotype
The slow-aging thyroid signature is internally coherent: high-normal TSH + low-normal FT4 + adequate FT3 via a high FT3/FT4 ratio + low rT3 + no antibodies. It’s a gland set to low output with efficient peripheral conversion — the endocrine equivalent of the mTOR/IGF-1 downregulation you already track. The two findings with the strongest causal plausibility (versus reverse-causation confounding) are low FT4 and high TSH; the FT3/rT3/antibody panel is more diagnostic-of-context than target-able.
One caveat for the report you’re likely building from this: nearly all of it is observational, and no trial has shown that nudging a euthyroid person’s FT4 to the low-normal end extends lifespan. The honest claim is association with a longevity phenotype, not demonstrated benefit from intervention.