Subclinical Hypothyroidism

ჯანდაცვა · §660

Your thyroid lab comes back with one number flagged: TSH a little above the reference range, free T4 normal. That's subclinical hypothyroidism — it shows up in about one in ten adults and roughly one in five women past 60. The instinct it triggers, a daily thyroid pill, is right in pregnancy and right when the number is above 10. Most of the time in between, the best modern trials say the pill doesn't actually change how you feel.

Decide · As-needed Evidence Moderate თავი ჯანდაცვა

The number on the lab is real; the decision it points to is stratified. A TSH above 10 carries a real bump in heart-disease risk and earns a pill. Pregnancy and pregnancy-planning earn one too, with a lower cutoff. Between 4.5 and 10 in everyone else — the band most people land in — the largest trials of the last decade show the pill doesn't move energy, mood, weight, or how the day feels. The honest move there is repeat the test, check antibodies, and mostly wait.

The thyroid runs on a thermostat. The pituitary gland releases TSH to tell the thyroid to make more hormone; the thyroid releases hormone (mostly T₄, converted to active T₃ in your tissues); that hormone signals back to the pituitary to dial TSH down. The loop runs tight. A small drop in thyroid hormone produces an outsized rise in TSH, which is why one TSH reading tells you more about thyroid function than any other single test Cooper 2001.

“Subclinical” describes the gap between what the thermostat sees and what your body feels. The pituitary detects the early slowdown and cranks up TSH; the thyroid catches up enough to keep free T₄ in range. The lab flags it, but at this stage the rest of you may not. The open question for the last twenty years has been whether the gap matters — whether tissue-level hormone signaling has already slipped before the free-T₄ number says so. The honest answer from the contemporary trials is: usually not very much.

What the trials actually showed

The story of subclinical hypothyroidism over the last decade is the story of mainstream endocrinology slowly accepting that the older instinct — abnormal lab, daily pill — doesn’t hold up when you actually test it. Three trials carry the weight.

The follow-up trial in adults 80 and over found the same nothing Mooijaart et al. 2019. And when researchers pooled twenty-one randomized trials of levothyroxine in mild subclinical hypothyroidism, covering more than two thousand patients, the answer didn’t move: no change in quality of life, depression scores, fatigue, body weight, or cognition Feller et al. 2018. A panel that reviewed this evidence for the BMJ in 2019 took the unusual step of issuing a strong recommendation against routine treatment for non-pregnant adults with TSH under 20 Bekkering et al. 2019.

Heart disease is the place where the evidence shifts. Pooling data from eleven prospective cohorts — 55,287 people watched for years — the risk of coronary heart disease events climbs at TSH 10 and above: about 1.89 times the rate of heart-disease events, 1.58 times the rate of heart-disease deaths. Between 4.5 and 10, the signal disappears Rodondi et al. 2010. A separate analysis of six cohorts found heart failure risk roughly doubled at TSH above 10, again with nothing meaningful below it Gencer et al. 2012. A UK general-practice analysis of nearly five thousand patients suggested that treating mild subclinical hypothyroidism in adults aged 40 to 70 cut heart-disease events by about 39% — with no benefit, and possible harm, in those over 70 Razvi et al. 2012. The catch is that this last result is observational, not a trial. No randomized study has been large enough or long enough to confirm a heart-disease benefit from treatment.

The summary the evidence supports: a lab number alone is not a reason to take a pill. A lab number above 10, or a lab number in a pregnant woman, is.

What most guides get wrong

“The reference range is one number that fits everyone.” It isn’t. The median TSH in healthy adults drifts up roughly 0.3 mIU/L per decade after 40, and by age 80 the top of the normal range climbs to around 7.5 in the US population data Surks & Hollowell 2007. A lot of what gets labeled subclinical hypothyroidism in older adults is, biologically, normal aging. The major US endocrinology bodies have started to argue for age-adjusted reporting precisely for this reason Pearce et al. 2023.

“If I’m tired and my TSH is high, the thyroid is why.” Tempting, common, usually wrong. Sleep debt, depression, anemia, deconditioning, and ordinary stress all produce the same morning. The cleanest test of the thyroid-is-why theory is the trial that picked exactly those patients — lab abnormal, symptoms present — and gave half of them the pill. Tiredness didn’t move Stott et al. 2017 Feller et al. 2018. Treating the thyroid is the diagnostic shortcut a lot of people reach for first; the trials say it usually doesn’t close the loop.

“One high lab is a diagnosis.” Half to two-thirds of mildly elevated TSH readings normalize on their own within a few years, especially when thyroid antibodies are negative and the original number was under 7 Biondi et al. 2019. The first elevated reading is a signal to repeat the test in six to twelve weeks with antibodies added, not a verdict.

“Once on the pill, on the pill for life.” Not necessarily. A fair share of people started on levothyroxine for mild subclinical hypothyroidism would test normal off it. A monitored taper, with a TSH check six to eight weeks after stopping, is a reasonable conversation to have with a clinician once you’ve been stable Biondi et al. 2019.

“The lab number is always real.” Biotin supplements at 5 mg or more a day — common in hair and nail products — can scramble the TSH assay and give a falsely low reading; stop them three days before the test. A recent flu can transiently bump TSH. So can starting amiodarone, lithium, or lithium-class medications Biondi et al. 2019.

What happens if you do nothing

For most people who land in the 4.5–10 TSH band, the honest answer is: probably not much. The lab number drifts; in a fair fraction of cases it comes back into range on its own; the rest of life feels the same it did the year before. The trials that asked “does treating this make people feel better” came up empty Feller et al. 2018. The bigger risk for someone in this band is that the lab finding gets treated as a diagnosis for symptoms that have nothing to do with the thyroid — the fatigue stays, but now the patient and clinician have an explanation that lets them stop looking.

The exceptions matter. If you’re positive for the thyroid peroxidase (TPO) antibody — the immune-system marker for the autoimmune process that drives most cases — your odds of slipping into actual hypothyroidism creep up at about 4 percent per year rather than the background 2 percent; over a decade that becomes a real probability rather than a remote one Vanderpump et al. 1995. If your TSH is above 10, the heart-disease numbers turn from background noise to a real signal: in the large pooled cohorts, somewhere around 1.9 times the heart-disease event rate and 1.6 times the heart-disease death rate compared with people whose TSH sits normal Rodondi et al. 2010. Heart failure risk roughly doubles in the same band Gencer et al. 2012. None of this lands in a week; it’s the kind of risk that compounds quietly across a decade, the kind that doesn’t announce itself until something stops working.

If you’re pregnant or trying to be, the time horizon collapses. Untreated subclinical hypothyroidism in early pregnancy is linked with higher rates of miscarriage and preterm delivery, and the fetus depends on the mother’s thyroid supply before its own gland comes online around the twelfth week Alexander et al. 2017. That window is the strongest reason this lab is worth knowing your number for at all.

What to actually do with the number

Step one is to repeat it. A single elevated TSH is a question, not an answer.

If the second number is normal, you’re done; the first reading was noise. If the second is still elevated, the diagnosis is biochemically real and the next call is whether to treat it. The threshold has converged across most current guidelines into a few clear cases.

TSH above 10, persistent: treatment is recommended across all major guidelines, with the strongest case in adults under 70 Garber et al. 2012 Pearce et al. 2013 NICE NG145 2019.
Pregnancy: treat above 10 always; treat between 2.5 and 10 if TPO antibodies are positive; consider treatment in the 2.5–10 band even without antibodies for women planning conception or undergoing IVF Alexander et al. 2017.
TSH 4.5–10, non-pregnant, under 65, symptomatic: a three- to six-month trial of levothyroxine is reasonable. Set a clear stop date and a clear measure of success up front (the symptom that bothers you most), and stop the pill if the symptom hasn’t budged when the lab does NICE NG145 2019.
TSH 4.5–10, otherwise: recheck every six to twelve months. The current BMJ Rapid Recommendations panel takes the strongest stance against routine treatment for this group Bekkering et al. 2019.

If you do start the pill, levothyroxine is the only first-line treatment supported by the evidence. Starting dose is usually 25–50 mcg for older adults and people with heart disease, 50–75 mcg for healthy younger adults. The dose is set against a target TSH in the lower half of the reference range — not below it. Recheck TSH at six to eight weeks after any dose change.

When the pill is the wrong move

Over 65 or 70 with TSH 4.5–10. Two large randomized trials in this group (TRUST in adults 65+, IEMO80+ in adults 80+) found no benefit on symptoms or quality of life from treatment. Over-replacement in older adults raises atrial fibrillation and fracture risk in the same population. Watch and wait is the default Stott et al. 2017 Mooijaart et al. 2019.
TSH suppressed below the reference range on therapy. Driving TSH below normal — over-replacement — raises atrial fibrillation risk by roughly 40–70% and accelerates bone loss in postmenopausal women. If your TSH on treatment is under the reference range, the dose needs to come down Selmer et al. 2014.
Untreated adrenal insufficiency. Rare but dangerous: if you have undiagnosed adrenal insufficiency — common in autoimmune polyglandular syndromes that also produce thyroid autoimmunity — starting thyroid hormone before steroid replacement can trigger an adrenal crisis. Worth asking about before starting the pill if you have other autoimmune conditions, persistent low blood pressure, or unexplained weight loss Garber et al. 2012.
Compounded T3/T4 mixtures and desiccated thyroid. Marketed aggressively in functional-medicine practices, popular on patient forums. The evidence base for using them instead of levothyroxine in subclinical hypothyroidism is weak; dose consistency batch-to-batch is worse; T3 content can swing TSH out of range and raise the same arrhythmia risk above. If you go down this road, do it with an endocrinologist who’ll monitor you closely.

Who this looks different for

Pregnant or trying to conceive. The threshold drops and the urgency rises. Trimester-specific TSH ranges replace the standard ones; the first-trimester upper bound sits around 4.0 mIU/L in current ATA guidance, lower if your lab has local pregnancy ranges. Above 10 means treat now. Between 2.5 and 10 with positive TPO antibodies means treat. Between 2.5 and 10 without antibodies is less settled, but most reproductive endocrinologists treat to a TSH under 2.5 in women trying to conceive or going through IVF, and the cost-risk math for a generic pill that may improve miscarriage rates is generous Alexander et al. 2017. One caveat from the largest trial in the area: starting levothyroxine after the eighth to twentieth week of pregnancy didn’t improve the children’s IQ at five years Casey et al. 2017. The window where treatment helps is preconception and the first weeks of pregnancy. Check the number before you start trying, not after the second missed period.

Postpartum. Roughly 1 in 20 women develop a transient thyroid disturbance in the year after delivery, typically running through a brief hyperthyroid phase before settling into a hypothyroid stretch. A persistent post-pregnancy fatigue you can’t shake is a reason to check the number rather than to assume sleep deprivation Pearce et al. 2003.

Older adults. The strongest case against routine treatment. A mildly elevated TSH at 75 is more likely to be normal-for-your-age physiology than disease — the top of the healthy range climbs with the decades. The two large trials that tested treatment in adults 65 and over and 80 and over both came up empty Stott et al. 2017 Mooijaart et al. 2019. The default in this group is to recheck, watch for symptoms or TSH climbing past 10, and otherwise leave the pill bottle on the shelf.

Younger adults with TSH above 10. The clearest case for treatment outside pregnancy. This is the band where the observational cardiovascular signal lives and where the modeled benefit of treatment is highest Rodondi et al. 2010 Razvi et al. 2012. If you’re under 65 and your TSH is persistently above 10, the conversation with the clinician should be when to start the pill, not whether.

What treatment actually delivers when it does

The pill works quickly on the lab number and slowly, if at all, on everything else. Your TSH will be back in range within six to twelve weeks of finding the right dose. Your LDL cholesterol will probably drop somewhere between 5 and 10 mg/dL over the same period, more if your starting TSH was high Biondi et al. 2019 Cappola & Ladenson 2003. If you’re a younger adult with a TSH that had climbed past 10, observational data suggests your heart-disease risk over the following decade is closer to where it would have been if the lab had been normal all along — though no randomized trial has been big enough to clinch that Razvi et al. 2012.

If you’re pregnant, the payoff is in what doesn’t happen: a miscarriage that doesn’t occur, a preterm delivery that doesn’t. The size of that benefit is hardest to pin down because the cleanest trial timed treatment too late to test it Casey et al. 2017 Negro et al. 2010. The biology, the observational data, and the low cost of being wrong about it line up on the side of treating early.

On energy, focus, weight, and mood: don’t expect the morning to suddenly look different. The trials that measured it found, on average, no change. Some people do feel better — the variation across individuals is real — but there’s no test or lab number that predicts who. The honest version of the conversation with the prescribing clinician is to pick the one symptom you most want addressed, set a date three to six months out, and stop the pill if that symptom hasn’t moved.

The real-world friction

Testing is cheap and routine. A TSH-with-reflex-to-free-T₄ panel runs about $30 to $80 at most US labs without insurance and is usually fully covered when ordered for a symptom; TPO antibodies add another $30 to $60. Any primary care clinician can order it.

The treatment, when indicated, is among the cheapest prescription drugs in the pharmacy. Generic levothyroxine runs about $50 to $100 a year through standard pharmacies and considerably less through GoodRx or mail-order. The branded versions (Synthroid, Tirosint) cost meaningfully more without much benefit for most people; the case for the brand is real only for the small minority with absorption issues or sensitivities to fillers.

The ongoing friction is monitoring. Plan on a TSH recheck six to eight weeks after starting or changing dose, then every six to twelve months once stable. That’s the part of the protocol that quietly drifts — people skip the follow-up, the dose drifts off-target, and either the lab over-suppresses or the original problem creeps back. A single recurring reminder on the phone solves it.

Related, but not this

Overt hypothyroidism — the same axis pushed further: TSH high and free T₄ low. Different threshold for treatment (always) and different conversation. Worth its own entry.
Hashimoto’s thyroiditis — the autoimmune process behind most subclinical hypothyroidism in the developed world. The TPO antibody test mentioned above is what catches it.
Subclinical hyperthyroidism — the opposite lab pattern: TSH low, free T₄ and T₃ normal. Different stakes — atrial fibrillation and bone loss top the list rather than heart disease and miscarriage.
Postpartum thyroiditis — a transient thyroid disturbance in the year after delivery. Often resolves on its own; sometimes settles into permanent hypothyroidism.
Iodine intake — the upstream nutritional input to thyroid function. Iodine deficiency in pregnancy is the leading cause of preventable cognitive impairment worldwide; intake is adequate in most US adults but trending down in some subgroups.
Cardiovascular risk numbers more directly — ApoB and Lp(a) testing if the LDL bump on your lab is the part you care about.

Related in the handbook

Worsens / aggravates

— One overlooked cause of a sluggish thyroid is low iodine, often from quietly switching off iodized salt.

Diagnosis

— Sluggish bowels are a classic low-thyroid sign; stubborn constipation is a reason to check your TSH.

— Subclinical hypothyroidism is frequently the opening act of autoimmune Hashimoto's.
— Iodine intake — both deficiency and excess — is one of the inputs behind a sluggish thyroid.
— The whole decision hinges on reading where in the range your TSH sits — above 10 is a different call than 4.5.
— Subclinical hypothyroidism is exactly the gray-zone result thyroid screening turns up — high TSH, normal T4, treat-or-watch.
— If you do end up on levothyroxine, it's a narrow-therapeutic-index drug where consistency matters.
— An underactive thyroid can show up as carpal tunnel; new wrist symptoms are a reason to check thyroid.
— Low thyroid is a known risk factor for frozen shoulder — one of the quiet ways it shows up in the body.
— Selenium is one of the nutrients that interacts with the autoimmune thyroid disease behind subclinical hypothyroidism.

Substance and claimed effects

Subclinical hypothyroidism (SCH) is a biochemical diagnosis: serum TSH above the population reference range with free thyroxine (free T₄) within range. Population prevalence in iodine-sufficient countries is approximately 4–10% of adults, rising to 10–20% in women over 60 Hollowell et al. 2002 Cooper 2001. The most common etiology in iodine-replete regions is chronic autoimmune (Hashimoto’s) thyroiditis, identifiable by elevated thyroid peroxidase (TPO) antibodies Pearce et al. 2003. Claimed downstream consequences across the literature: fatigue / low energy, weight gain and unfavorable body composition, low mood and clinical depression, dyslipidemia (elevated LDL-C and total cholesterol), reduced fertility and adverse obstetric outcomes, and increased risk of coronary heart disease, heart failure, and cardiovascular mortality — though the magnitude and even existence of each effect is contested in the modern RCT literature Biondi et al. 2019. The entry covers each of these consequences holistically, plus the treatment-threshold question, and the diagnostic pitfalls (age effects, assay variation, transient elevation, non-thyroidal illness).

Evidence by addressing question

mechanism

The hypothalamic-pituitary-thyroid axis: hypothalamic TRH stimulates pituitary TSH, which stimulates thyroidal production of T₄ (and a small amount of T₃). Peripheral deiodination converts T₄ to active T₃. TSH is regulated by negative feedback from circulating thyroid hormone via a log-linear relationship: a small drop in free T₄ produces a much larger relative rise in TSH, making TSH the most sensitive single index of thyroid status Cooper 2001. “Subclinical” reflects this sensitivity: the pituitary detects mild hypothyroidism before peripheral hormone levels fall out of range. The mechanistic implication is that mild thyroid hormone insufficiency may already produce tissue-level effects (cardiac, vascular, neuropsychiatric, lipid-metabolism) before free T₄ falls below the assay reference. However, the strength of this argument depends on (a) whether the elevated TSH reflects true hormone insufficiency vs. set-point variation, and (b) whether end-organ thyroid hormone signaling correlates well with TSH at the mild end. Both assumptions are weaker than typically presented in pre-2017 literature Bekkering et al. 2019.

evidence

Effect on quality of life and symptoms. The most rigorous current evidence is the TRUST trial: 737 adults aged 65+ with persistent SCH (mean TSH 6.4 mIU/L) randomized to levothyroxine vs. placebo, followed for a median of one year. No difference in hypothyroid symptom score or tiredness score, no difference in any pre-specified secondary outcome Stott et al. NEJM 2017. The IEMO80+ trial extended this to adults aged 80+: again, no benefit on thyroid-related symptoms Mooijaart et al. JAMA 2019. A 2018 systematic review and meta-analysis of 21 randomized trials (n=2,192) found no clinically meaningful effect of levothyroxine on quality of life, depressive symptoms, fatigue, body mass index, blood pressure, or cognitive function in adults with SCH Feller et al. JAMA 2018. The BMJ Rapid Recommendations panel synthesized this evidence into a strong recommendation against routine levothyroxine treatment for most non-pregnant adults with SCH (TSH 4–19.9 mIU/L) Bekkering et al. BMJ 2019.

Cardiovascular outcomes. Observational pooled data from 11 prospective cohorts (n=55,287) showed an increased risk of coronary heart disease events (HR 1.89) and CHD mortality (HR 1.58) at TSH ≥10 mIU/L, with no increased risk at TSH 4.5–6.9 mIU/L Rodondi et al. JAMA 2010. A similar individual-participant analysis of six cohorts (n=25,390) found heart failure risk elevated at TSH ≥10 (HR 1.86) but not at lower TSH bands Gencer et al. Circulation 2012. The Whickham 20-year follow-up (n=2,779) found increased ischemic heart disease incidence among the SCH subgroup; reanalysis suggested levothyroxine treatment was associated with reduced events in younger participants but not older ones Vanderpump et al. 1995 Razvi et al. 2008. A UK General Practice Research Database analysis of 4,735 SCH patients (40–70 years) found levothyroxine treatment associated with a 39% reduction in fatal and nonfatal IHD events in those aged 40–70, but no benefit (or possible harm) in those over 70 Razvi et al. Arch Intern Med 2012. A Danish nationwide cohort (n=628,953) confirmed graded mortality risk with increasing TSH severity Selmer et al. 2014. No RCT has shown CV-event reduction from levothyroxine in SCH; the TRUST trial was underpowered for hard CV endpoints.

Lipid effects. The HUNT study (n=30,656 euthyroid Norwegians) showed a positive linear association between TSH (within and above reference range) and total cholesterol, LDL-C, non-HDL-C, and triglycerides — the effect was present even at TSH values within the conventional reference range Asvold et al. 2007. Trials of levothyroxine in SCH show modest LDL-C reductions, on the order of 5–10 mg/dL, with larger effects at higher baseline TSH; meta-analyses suggest the effect is real but small in absolute terms Biondi et al. 2019 Cappola & Ladenson 2003.

Mood and depression. A 2019 meta-analysis of 25 studies found SCH associated with depression with an OR of ~2.3 cross-sectionally, but the effect was heavily attenuated in higher-quality studies and not consistent in prospective designs Loh et al. 2019. Levothyroxine RCTs (including TRUST and the Feller meta-analysis) show no effect on depressive symptoms in unselected SCH Feller et al. 2018 Stott et al. 2017. There is some evidence that women with postpartum or autoimmune SCH have elevated depression risk independent of TSH Pop et al. 2006.

Fertility and pregnancy. SCH (definitions vary; trimester-specific TSH ranges apply) is associated with miscarriage, preterm delivery, gestational hypertension, and possibly impaired offspring neurodevelopment in observational data Alexander et al. ATA 2017. The Casey NICHD RCT (n=677 with SCH, n=526 with isolated hypothyroxinemia) tested levothyroxine vs. placebo starting at ≤20 weeks gestation: no difference in offspring IQ at 5 years, no difference in obstetric outcomes Casey et al. NEJM 2017. Earlier observational work suggested levothyroxine reduced miscarriage in TPO-positive SCH women, but the Cochrane review found insufficient evidence to support routine treatment outside high-risk subgroups Reid et al. Cochrane 2013. ATA 2017 guidelines recommend levothyroxine for pregnant women with TSH >10, and to consider treatment for TSH 2.5–10 if TPO-positive Alexander et al. 2017.

protocol

Diagnostic confirmation. A single elevated TSH is not a diagnosis: 50–60% of mildly elevated TSH values (4.5–10 mIU/L) normalize on repeat testing within 2–5 years, particularly when TPO-negative Biondi et al. 2019. Standard workup repeats TSH plus free T₄ in 6–12 weeks (sooner if symptomatic or planning pregnancy); persistence over two measurements separated in time defines the diagnosis. Add TPO antibodies to stratify progression risk — antibody-positive SCH progresses to overt hypothyroidism at ~4% per year vs. ~2% per year if antibody-negative Vanderpump et al. 1995.

Treatment thresholds. Guideline convergence:

ATA/AACE 2012: treat all SCH with TSH >10; consider treatment for TSH 4.5–10 in symptomatic patients, those with TPO antibodies, those with cardiovascular risk factors, or those under 70 Garber et al. 2012.
ETA 2013: treat TSH >10; in TSH 4.5–10, treat if symptomatic or under 65 with goiter, antibodies, or cardiovascular risk; reserve in elderly Pearce et al. ETA 2013.
BMJ Rapid Recs 2019: strong recommendation against routine treatment for TSH <20 in non-pregnant adults regardless of age — based on TRUST/IEMO80+/Feller meta-analysis Bekkering et al. 2019.
NICE NG145 (2019): consider trial of levothyroxine in adults under 65 with TSH >10 on two occasions 3 months apart; consider 6-month trial in those with TSH 4–10 if symptomatic and under 65 NICE NG145 2019.
ATA pregnancy 2017: treat all pregnant women with TSH >10 or TSH 2.5–10 with TPO antibodies; consider treatment for TSH 2.5–10 if TPO-negative Alexander et al. 2017.

Dosing. Initial levothyroxine 25–50 mcg daily for older adults or those with cardiac disease, 50–75 mcg for healthy younger adults; recheck TSH at 6–8 weeks; titrate to TSH within the lower-normal half of the reference range. Take on empty stomach 30–60 min before food; separate from calcium, iron, PPIs by 4 hours. Pregnant women typically need a 25–30% dose increase from preconception baseline Alexander et al. 2017.

contraindications

Levothyroxine itself has few absolute contraindications. Practical caveats:

Older adults / cardiac disease. Over-replacement increases atrial fibrillation and fracture risk; the TRUST/IEMO data suggest no benefit and possible harm in over-70s — treatment is rarely indicated unless TSH >10 with clear symptoms Stott et al. 2017 Mooijaart et al. 2019.
Suppressed TSH. Levothyroxine dosing that drives TSH below the reference range elevates atrial fibrillation risk by ~40–70% and contributes to bone loss in postmenopausal women Selmer et al. 2014.
Adrenal insufficiency. If undiagnosed adrenal insufficiency coexists (e.g., autoimmune polyglandular syndrome), starting thyroid hormone before glucocorticoid replacement can precipitate adrenal crisis Garber et al. 2012.

misconceptions

“Normal TSH range” is one number. Reference ranges shift meaningfully with age: median TSH in healthy iodine-replete adults rises ~0.3 mIU/L per decade after age 40, and the 97.5th percentile in adults over 80 reaches 7.5 mIU/L in NHANES III — people called “subclinically hypothyroid” by a static cutoff may simply be exhibiting normal age-related set-point shift Surks & Hollowell 2007. The 2023 AACE/ATA recommendation now favors age-adjusted reporting Pearce et al. 2023.

“Fatigue + high TSH = thyroid is the cause.” Symptom overlap with depression, sleep disturbance, anemia, and deconditioning is enormous. The strongest test of this assumption — the TRUST trial — randomized patients precisely on symptoms + biochemistry and found no symptom benefit from normalization Stott et al. 2017. The implication is that in most adults with mild SCH, fatigue is not caused by the thyroid and won’t resolve with treatment.

“Once on levothyroxine, always on levothyroxine.” Roughly 30–60% of mild SCH cases revert to euthyroidism spontaneously, especially with negative antibodies and TSH <7 Biondi et al. 2019. A monitored discontinuation trial is reasonable in selected patients after sustained euthyroidism on therapy.

“The TSH lab can lie.” Transient elevation occurs in: recovery from acute illness (non-thyroidal illness syndrome rebound), within ~6 weeks of starting amiodarone, lithium, or interferon; biotin supplementation (≥5 mg/day) interferes with the immunoassay and can produce spuriously low TSH; assay heterophile antibodies can produce spuriously high TSH Biondi et al. 2019.

audience

Pregnancy and fertility. The lowest treatment threshold and the strongest case for treatment. Trimester-specific TSH ranges apply (first trimester upper bound ~4.0 mIU/L per ATA 2017, lower if local data available); TSH 2.5–10 with TPO antibodies merits treatment; isolated SCH without antibodies is more controversial but commonly treated to TSH <2.5 in women planning conception or undergoing IVF Alexander et al. 2017. The Casey trial caveat: treating SCH detected after 8–20 weeks gestation did not improve offspring neurodevelopment, suggesting any benefit requires preconception or very early initiation Casey et al. 2017.

Women in general. Female-to-male prevalence ratio ~3:1, sharper after menopause; postpartum thyroiditis affects ~5% of women in the year after delivery and presents as transient SCH or hypothyroidism Pearce et al. 2003.

Older adults (65+). The strongest case against routine treatment. TRUST, IEMO80+, and the Feller meta-analysis all support a watch-and-wait approach with treatment reserved for TSH >10 with clear hypothyroid symptoms or progression Stott et al. 2017 Mooijaart et al. 2019.

Adults under 65 with TSH >10. Strongest case for treatment outside pregnancy — the cardiovascular signal from Rodondi 2010 and Razvi 2012 is concentrated in this group, and the BMJ guideline still allows treatment at this severity Rodondi et al. 2010 Razvi et al. 2012.

stakes

Untreated SCH carries graded but mostly modest absolute risks:

Progression to overt hypothyroidism at 2–4% per year, accelerated by elevated TPO antibodies and higher baseline TSH Vanderpump et al. 1995.
Cardiovascular events at TSH ≥10: ~1.9-fold risk of CHD events and ~1.6-fold risk of CHD mortality vs. euthyroid — baseline absolute risk dependent on age and other CV risk factors Rodondi et al. 2010.
Heart failure at TSH ≥10: ~1.9-fold risk in pooled cohorts Gencer et al. 2012.
Pregnancy outcomes: miscarriage risk and preterm-delivery risk modestly elevated; effect on offspring IQ contested Alexander et al. 2017 Casey et al. 2017.
Mood/QoL: No replicated longitudinal causal signal in mild SCH; depression association attenuates in higher-quality studies Loh et al. 2019.

For TSH 4.5–10, the absolute consequence in non-pregnant adults is small enough that the strongest contemporary guidelines (BMJ Rapid Recommendations) actively recommend against treatment Bekkering et al. 2019.

payoff

Where treatment is indicated (pregnancy, TSH >10, symptomatic younger adults):

Pregnancy: reduction in miscarriage in TPO-positive women treated preconception (observational); attenuation of preterm delivery in some trials Negro et al. 2010 Maraka et al. 2017.
Lipids: LDL-C reduction of 5–10 mg/dL on average, larger when baseline TSH is higher Biondi et al. 2019.
Cardiovascular events: Observational evidence of CHD-event reduction in adults 40–70 (Razvi 2012); no RCT confirmation Razvi et al. 2012.
Symptoms / energy: Trials in unselected SCH show no average benefit; some clinicians report individual responders but these are not predictable from TSH or antibody status Feller et al. 2018.

Time-to-effect: TSH falls into target range within 6–12 weeks of dose stabilization; symptom or lipid effects, where present, are usually visible by 12 weeks. The honest framing is that for the majority of mildly SCH adults, the felt-experience payoff is minimal.

The credibility range

Optimist case. Thyroid hormone is a global metabolic regulator; mechanistic plausibility for downstream effects on lipids, cardiovascular tissue, mood, and pregnancy outcomes is strong. Observational data are consistent: graded risk relationships across TSH bands for IHD, heart failure, mortality, and miscarriage Rodondi et al. 2010 Gencer et al. 2012 Selmer et al. 2014. The Razvi 2012 UK GPRD analysis found a 39% reduction in CHD events with levothyroxine in adults 40–70 with SCH — if causal, this is a meaningful intervention Razvi et al. 2012. The Whickham reanalysis adds supporting longitudinal signal in younger adults Razvi et al. 2008. Pregnancy is the clearest case: observational and biological-plausibility evidence converge on real harm from untreated SCH in conception and early gestation, and treatment is cheap and low-risk. Treatment is also inexpensive (generic levothyroxine, ~$50–100/year) with a long safety record at physiological doses; the asymmetry of low cost vs. potentially meaningful effect (especially at TSH >7–10, in symptomatic patients, or with positive antibodies) supports a low treatment threshold in selected groups.

Skeptic case. The best contemporary RCT data — TRUST, IEMO80+, Casey NICHD, and the Feller meta-analysis — consistently show no clinically meaningful benefit of levothyroxine on the outcomes patients actually care about: symptoms, quality of life, depression, body composition, cognition, or (in pregnancy after 8–20 weeks) offspring neurodevelopment Stott et al. 2017 Mooijaart et al. 2019 Casey et al. 2017 Feller et al. 2018. The observational cardiovascular signal is concentrated at TSH ≥10 — a relatively small subgroup, and one where treatment is uncontroversial anyway. The lower TSH band (4.5–10) is where most SCH lives and where benefits are thinnest. Age-adjusted reference ranges suggest a substantial fraction of currently labeled SCH cases — particularly in older adults — are simply normal physiology mis-labeled Surks & Hollowell 2007 Pearce et al. 2023. Over-replacement (driving TSH below normal) is iatrogenic harm with documented cardiovascular and skeletal consequences Selmer et al. 2014. The BMJ Rapid Recommendations panel reviewing the contemporary trial evidence issued a strong recommendation against routine treatment for almost all non-pregnant adults with TSH <20 Bekkering et al. 2019. USPSTF (2015) concluded insufficient evidence to recommend screening — not because screening can’t detect cases, but because detection doesn’t reliably improve outcomes USPSTF 2015.

The author’s call. The credibility range collapses around a stratified position: treat liberally in pregnancy and pregnancy-planning, treat in non-pregnant adults with TSH >10, and adopt watchful waiting for non-pregnant adults with TSH 4.5–10 — with a time-limited trial in symptomatic adults under 65 reasonable but unlikely to produce meaningful benefit on average. The contemporary evidence has shifted decisively against the older “treat-most” instinct: TRUST and the Feller meta-analysis are well-conducted, well-powered, and consistent. The pregnancy exception is anchored in mechanism (fetal dependence on maternal thyroid hormone before fetal thyroid function begins ~12 weeks), real observational risk for miscarriage, and the low cost/risk of treatment. Evidence rating sits at 4 (multiple high-quality RCTs, strong guideline alignment in recent years). Controversy rating sits at 3–4 because guideline bodies still disagree on the 4.5–10 band and clinical practice lags behind the BMJ Rapid Recommendations.

Stakeholder and incentive map

Commercial: Levothyroxine is generic and cheap; no large branded incentive remains except for the few branded preparations (Synthroid, Tirosint). Compounded T3/T4 combinations and desiccated thyroid (Armour) have a small but vocal practitioner-marketed segment. Diagnostic-testing volume is a much larger revenue driver than the drug itself.
Professional bodies: ATA, AACE, ETA historically issued treat-more guidelines; the BMJ Rapid Recommendations (Bekkering 2019) and the underlying evidence have produced friction with that history. Newer ATA position statements (e.g., Pearce 2023) acknowledge the issue with static reference ranges Pearce et al. 2023.
Patient communities: A substantial online community (thyroid-focused subreddits, “Stop the Thyroid Madness”-style advocacy) lobbies for lower TSH treatment thresholds, T3-containing regimens, and desiccated thyroid. Their core observation — that some patients feel better on treatment despite RCT averages showing no benefit — is real but selection-biased.
Functional / integrative medicine: Pushes treatment of “subclinical” and even “suboptimal” thyroid (TSH >2.5) much more aggressively than mainstream endocrinology. Often pairs with T3 supplementation and desiccated thyroid. Evidence base is weak; commercial incentive is significant.
Skeptic / counter-incentive: USPSTF, BMJ Rapid Recommendations panel, and an increasing minority of endocrinologists arguing against over-diagnosis, particularly in the elderly USPSTF 2015 Bekkering et al. 2019.

Population variability

Age. TSH set-point rises with age — 97.5th percentile reaches ~7.5 mIU/L by age 80 in iodine-replete populations Surks & Hollowell 2007. Older adults with mildly elevated TSH likely include many false positives. Older adults also show no benefit and possible harm from treatment Stott et al. 2017.
Sex. Female prevalence ~3x male; postpartum and perimenopausal windows show transient spikes in incidence.
Antibody status. TPO-positive SCH progresses ~2x faster to overt disease and is the most informative single risk-stratifier outside of TSH severity Vanderpump et al. 1995.
Iodine status. Iodine-deficient regions show different prevalence and etiology distributions (more nodular disease, fewer autoimmune cases); the US has been iodine-replete since the 1950s but recent data suggest declining intake in subgroups.
Pregnancy. hCG cross-reacts with TSH receptor and physiologically suppresses TSH in the first trimester; estrogen elevates TBG and increases thyroid hormone demand. Trimester-specific ranges are mandatory; using non-pregnant ranges over-diagnoses SCH in normal early pregnancy and under-diagnoses it in mid-to-late Alexander et al. 2017.
Drugs. Amiodarone, lithium, interferon, tyrosine kinase inhibitors, immune checkpoint inhibitors all raise TSH; biotin supplements distort the assay; metformin lowers TSH Biondi et al. 2019.
Cardiovascular phenotype. The CHD-event signal in observational pooled data is concentrated at TSH ≥10 and absent in the 4.5–6.9 mIU/L range Rodondi et al. 2010.

Knowledge gaps

RCT data on hard CV endpoints. No RCT has had the size and follow-up to test whether levothyroxine reduces MI, stroke, or CV mortality in SCH. TRUST was symptom-focused and underpowered. A definitive trial would require many thousands of patients followed 5–10 years; unlikely to be commercially funded given drug-generic status.
Identifying responders. Roughly some fraction of SCH patients do feel better on levothyroxine despite the null-on-average trials. No biomarker, antibody, or clinical phenotype currently predicts response.
Optimal TSH target on therapy. Whether targeting mid-normal vs. low-normal TSH improves outcomes is unsettled; some patients report symptom improvement only at lower TSH targets but the evidence is mostly anecdotal.
Preconception TSH cutoff. The benefit of treating women with TSH 2.5–4.0 trying to conceive is plausible but not established by RCT.
Combination T4/T3 therapy. Whether a subset of patients with normal TSH on T4 but persistent symptoms benefit from T3 addition is the subject of ongoing trials; current evidence does not support routine use.
The contested 4.5–10 band. What changes the call here is a well-powered RCT in younger adults (40–65, TSH 4.5–10, symptomatic) of levothyroxine vs. placebo on quality of life and CV biomarkers. The closest existing data — Razvi 2012 — is observational.

The brief named six consequences (energy, weight, mood, cholesterol, fertility, cardiovascular risk) plus the treatment-threshold question. The article covers all six, but the scoring lands lower on most of them than the brief might imply, because the contemporary RCT evidence (TRUST, IEMO80+, Feller meta-analysis, Casey NICHD) is squarely negative on QoL/symptom endpoints. The honest framing was to score the consequences holistically against the actual evidence rather than the popular framing — energy and mood get 1s rather than 2-3s for this reason.

Weight specifically did not get its own dimension or paragraph as a primary consequence: trials show no average BMI change with treatment in mild SCH, so it’s mentioned only in passing under payoff. If a future reviewer expected a dedicated weight discussion, the call was that the evidence didn’t support one.

Hard scoping calls:

Overt hypothyroidism deliberately excluded. Different threshold (always treat), different conversation, different decision tree. Belongs in its own entry — flagged in out-of-scope.
Hashimoto’s thyroiditis kept as background mechanism only; the autoimmune story warrants its own entry once that exists.
Subclinical hyperthyroidism excluded — opposite axis, different stakes (atrial fibrillation, bone loss); deserves its own entry.
Postpartum thyroiditis mentioned briefly in audience but not given full coverage; transient course makes it a different management problem.
Combination T4/T3 therapy and desiccated thyroid covered briefly in contraindications as a warning rather than a fair alternative. Evidence base is weak; full discussion would require a separate entry on the levothyroxine-vs-combination-therapy question.

Rating difficulties:

Evidence: 4 not 5. Multiple high-quality RCTs exist, but they don’t cover the full question — no RCT has tested hard CV endpoints, guidelines still disagree on the 4.5–10 TSH band, and contemporary clinical practice continues to treat more liberally than the BMJ Rapid Recommendations panel advises. Withholding the 5 reflects the unresolved 4.5–10 question rather than weakness in the trial evidence itself.
Controversy: 3. Genuine disagreement among bodies (ATA/AACE/ETA vs. BMJ Rapid Recs vs. USPSTF non-recommendation on screening) on the threshold question. Not 4 because the consensus on TSH >10 and pregnancy is solid; the disagreement is bounded.
Longevity: 2. The CV signal is real (Rodondi 2010, Gencer 2012, Selmer 2014) but concentrated at TSH ≥10 and not RCT-confirmed for treatment benefit. A 3 would be defensible if Razvi 2012’s 39% IHD-event reduction were RCT-confirmed; it isn’t.
Mood: 1. Cross-sectional association exists (Loh 2019 meta), treatment effect doesn’t (Feller 2018). Score reflects the genuine epidemiologic signal without overstating the actionable one.

Future-link candidates (once entries exist): overt-hypothyroidism, hashimotos-thyroiditis, subclinical-hyperthyroidism, postpartum-thyroiditis, iodine-intake, apob-testing, lp-a-testing, preconception-labs.

Separate-entry candidates surfaced during writing: combination T4/T3 therapy (its own evidence base, contested), desiccated thyroid (commercial / advocacy story large enough to warrant standalone coverage), age-adjusted lab reference ranges (broader applicability than thyroid alone).

One pragmatic note: the action verb is decide rather than test because the entry’s utility is the threshold call, not the test itself. Most readers landing here will already have a lab result; the question they actually need answered is “does this need treatment.”