Midlife labs
The untargeted annual physical does not extend life, but a short list of cheap, specific blood markers — read more carefully than the "normal" flag on the report — does change what you do. This is the practical version: what to draw every year, what to draw once in a lifetime, what to leave for a symptom, and what to refuse outright.
Decades of randomized trials of general health checks converge on a sobering conclusion: in a Cochrane review, they made little or no difference to deaths from any cause (11 trials, about 233,000 adults) or from cancer (8 trials, about 139,000 adults), and probably none to cardiovascular deaths [1]. A 2021 review reached the same verdict on deaths and cardiovascular events, while noting that checks do pick up more high blood pressure and depression and improve uptake of proven screens [2]. The cost — financial, psychological, and physical — of indiscriminate testing of asymptomatic adults is real: cascades of follow-up imaging, biopsies of incidentalomas that would never have caused harm, lifelong medications started on the basis of a single borderline value. The proactive answer is not "test more"; it is to test a targeted, biologically meaningful panel and to interpret it against guideline cut-offs and your own trend over time, not just the population reference range printed on the result — while remembering that for most markers no tighter "optimal" value has been tested against outcomes.
What this page is and isn't
This is a midlife screening cadence for an otherwise healthy adult who is paying attention. It is not a workup for a specific symptom and it is not a substitute for a clinician's judgement on a particular patient. The thresholds below come from named guidelines and the studies linked inline, so you can see what each number rests on; where no outcome-tested target exists, the page says so.
A useful rule of thumb running through every section: the "normal" range on a lab report describes the population tested, not the best outcome. For some markers (apoB, HbA1c, hs-CRP, ALT/AST) guidelines or outcome data point to a tighter line than the lab flag; for others (fasting insulin, HOMA-IR, ferritin, uric acid, eGFR, vitamin D) no "optimal" value has been tested against outcomes, and pushing toward one can mislead.
The annual core panel
Six markers, drawn fasting, ideally on the same morning. Together they cover the four largest modifiable drivers of biological ageing in midlife: atherogenic lipid burden, glycaemic control and insulin sensitivity, systemic inflammation, and iron handling.
Apolipoprotein B (apoB)
Evidence: Strong that apoB is at least as good a predictor as LDL-C and superior under discordance; Moderate for the incremental benefit of acting on apoB rather than LDL-C, which rests on observational cohorts.
For half a century cardiovascular risk has been screened on LDL cholesterol (LDL-C) — the mass of cholesterol inside low-density-lipoprotein particles. The mechanistic driver of atherosclerotic cardiovascular disease (ASCVD) is not cholesterol mass but the number of apoB-containing particles crossing the endothelium and lodging in the artery wall.[3] Each atherogenic particle — LDL, very-low-density lipoprotein (VLDL), intermediate-density lipoprotein (IDL), and Lp(a) — carries exactly one apoB molecule, so a single apoB measurement gives a direct, standardized count.
The clinical edge of apoB over LDL-C appears in discordance: measured LDL-C and apoB disagree in around 1 in 5 patients [4], and the risky version — a "normal" LDL-C masking a high count of small, dense LDL particles — is concentrated in people with insulin resistance, high triglycerides, or central adiposity. In those phenotypes, an LDL-driven decision underestimates real risk. How LDL-C is calculated matters too: the newer Martin/Hopkins method markedly reduces the mismatch [5]. The 2026 ACC/AHA dyslipidaemia guideline keeps LDL-C and non-HDL-C as the primary targets. It calls measuring apoB reasonable for people already on treatment who have reached those goals — particularly with established cardiovascular disease, cardiovascular-kidney-metabolic syndrome, type 2 diabetes or high triglycerides — and says it may be reasonable before treatment, to refine risk [6]. For a longevity-oriented baseline workup, a one-time apoB is probably worth more than the LDL-C on the same panel.
No guideline sets one apoB target for everyone; each scales it to risk, and the numbers differ between societies (apoB in mg/dL; 100 mg/dL = 1.0 g/L):
| Guideline | apoB goal, or level at which to step up treatment |
|---|---|
| European (ESC/EAS 2019) | Below 100 at moderate risk, below 80 at high risk, below 65 at very high risk [7] |
| National Lipid Association (2024) | Consider stepping up treatment above 90 at borderline-to-intermediate risk, above 70 at high risk, above 60 at very high risk [8] |
| US (ACC/AHA 2026) | Below 100, alongside LDL-C below 100, when the coronary calcium score is 1–99; consider below 55 in very-high-risk cardiovascular disease once LDL-C and non-HDL-C goals are met [9] |
→ Full apoB-first treatment framework in Lipid management.
Fasting insulin + HbA1c (and HOMA-IR)
Evidence: Moderate that fasting insulin and HOMA-IR pick up insulin resistance before glucose or HbA1c become abnormal; Weak for any specific "optimal" value — none has been tested against outcomes.
Standard primary care screens for type 2 diabetes use fasting glucose and HbA1c (glycated haemoglobin, a three-month average of blood-glucose exposure). Both lag: in a cohort of 6,538 British civil servants, insulin sensitivity fell steeply over the five years before a diabetes diagnosis, the pancreas briefly pushed out more insulin to compensate, and the sharp rise in fasting glucose began only about three years out [10]. Fasting insulin is the early-warning marker that fills the gap: as tissues become insulin-resistant, the pancreas compensates by pushing more insulin, so insulin can rise while glucose still looks normal. Insulin immunoassays are not standardised between laboratories — between-method differences are worst at low concentrations, exactly where healthy values sit — so track the trend in one lab rather than treating a single value as a universal line.[11]
A common pattern in apparently healthy 40-somethings: fasting glucose 5.3 mmol/L (95 mg/dL) — "normal"; fasting insulin about 110 pmol/L (18 μIU/mL) — well above the range seen in lean, healthy adults. The two combine into HOMA-IR (homeostasis model assessment of insulin resistance): fasting glucose in mmol/L × fasting insulin in μIU/mL, divided by 22.5. Higher means more resistant; the example works out at about 4.2. No "optimal" HOMA-IR or fasting insulin has been tested against outcomes. In 21,684 lean adults with normal glucose and lipids, used to set one laboratory's reference interval (mostly women, aged 20–60, one assay), the median fasting insulin was about 40 pmol/L (6.5 μIU/mL) and the median HOMA-IR about 1.3 in both sexes, with 95% of HOMA-IR values between 0.39 and 2.86 [12]. A target below 1.0 would put most of that healthy group above target. The outcome evidence points at the high end instead: among 5,511 US adults without diabetes, the quarter with HOMA-IR above 2.8 had about 64% higher all-cause mortality than the quarter at 1.4 or below (HR 1.64), a link seen only in people with a body mass index below about 25 [13]. In one Spanish population, the cut-off for insulin resistance sat near 2 (1.85–2.05) when tied to metabolic-syndrome risk and at 3.46 on a purely percentile basis [14]. So treat a HOMA-IR of about 2 or more as worth acting on, and a lower value as reassuring rather than a number to push down.
| Marker | Conventional "normal" | What to aim for |
|---|---|---|
| Fasting insulin | <150 pmol/L (<25 μIU/mL) | Within your lab's reference interval, tracked in the same lab; no outcome-tested target |
| HbA1c | <5.7% | 5.0–5.4% |
| HOMA-IR | <2.5 | Below about 2; no outcome-tested target |
Lower is not always better below the prediabetes range. In 11,092 middle-aged US adults without diabetes followed for about 15 years, an HbA1c below 5.0% went with about 48% higher all-cause mortality than 5.0–5.4% (HR 1.48), and the pattern held after people with anaemia were excluded [15]. Pooled data from six cohorts of adults aged 50 and over (28,681 people) found no clear excess mortality below 5.0%, and in the cohort with the most pronounced J-shape (US NHANES), the excess faded once anaemia, liver markers and other confounders were accounted for [16]. So a very low HbA1c is a prompt to look for anaemia or liver disease rather than a hypoglycaemia warning, and the band to aim for is 5.0–5.4%, not the floor.
A fasting draw covers the metabolic panel; a continuous glucose monitor (CGM) does not belong on it. In non-diabetic adults the evidence that wearing one improves any hard outcome is thin, brand-to-brand readings disagree, and the most common result is spike-anxiety over post-meal excursions that are entirely normal. See Continuous glucose monitors.
High-sensitivity C-reactive protein (hs-CRP)
Evidence: Strong as a long-horizon risk marker; Moderate that acting on it (beyond the metabolic and lipid levers that drive it) changes outcomes.
The most-validated circulating marker of the chronic, low-grade inflammation ("inflammaging") that underlies atherosclerosis, neurodegeneration, and several cancers. The long-horizon evidence is unusually clean: in 27,939 initially healthy women in the Women's Health Study followed for 30 years, baseline hs-CRP in the top quintile carried about a 70% higher risk of a major cardiovascular event than the bottom quintile — a stronger and more durable signal than LDL-C gave over the same interval (about 36% higher), and independent of it (hs-CRP hazard ratio 1.70, 95% CI 1.52–1.90; LDL-C 1.36, 1.23–1.52).[17] A confidence interval (CI) is the range in which the true effect most plausibly lies; when it sits entirely above 1.0, as both do here, the association is unlikely to be chance. hs-CRP remains a risk-enhancing factor in the 2026 ACC/AHA guideline: for adults without cardiovascular disease whose estimated 10-year risk is borderline (3% to under 5%), an hs-CRP of 2 mg/L or more on two successive occasions, with no other explanation, means high-intensity statin therapy can be useful [18]. The JUPITER trial selected apparently healthy adults on an hs-CRP of 2.0 mg/L or more and still cut events sharply on a statin. See Lipid management.
| hs-CRP | Interpretation |
|---|---|
| <1.0 mg/L | Lowest-risk range; no target below the guideline line (2 mg/L twice) has been tested against outcomes |
| 1.0–3.0 mg/L | Elevated background inflammation; investigate metabolic and lifestyle drivers |
| ≥3.0 mg/L (twice, weeks apart) | High-risk; consider intensified lipid and metabolic management |
hs-CRP is an acute-phase reactant: a fresh viral infection, a hard workout in the last 24–48 hours, or a recent injury will all push it up transiently. Any elevated value should be repeated several weeks later before acting on it.
A useful systemic check: in a cross-sectional study of 2,588 untreated Korean adults, those with fasting insulin above 90 pmol/L (15 μIU/mL) had nearly five times the odds of an hs-CRP above 2 mg/L compared with those below 30 pmol/L (5 μIU/mL), although the comparison did not account for body weight [19]. Insulin resistance and inflammation travel together, so improving the metabolic picture (weight, activity, diet) is a sensible first step; recheck hs-CRP afterwards before adding anything aimed at inflammation itself.
Ferritin
Evidence: Moderate for the diagnostic cut-offs below; Weak for any narrower "optimal" window — none has been tested in a trial. In about 162,000 Copenhagen adults whose doctors had ordered blood tests, the lowest mortality sat near 60 µg/L in women and 125 µg/L in men without marked inflammation, an association rather than a tested target [20].
Standard reference ranges for ferritin span from about 10 to 300+ µg/L. That range is useful for detecting severe iron-deficiency anaemia or full-blown haemochromatosis; it is not useful for midlife optimisation. Both ends of the standard "normal" carry signal:
- Low ferritin (under about 30 µg/L) should not be treated as low-normal. Below 15 µg/L iron stores are essentially empty, and below 30 µg/L they are generally low, which is why most labs set their lower limit somewhere between 15 and 30 µg/L [21]. It is common in premenopausal women, distance runners and people on long-term plant-based diets. A value in this range usually means iron deficiency — something to discuss with a clinician, who may look for a cause, especially in men and postmenopausal women — not a fine-tuning target.
- High ferritin carries signal at the top end: over 23 years in a Danish general-population cohort, a ferritin of 200 µg/L or more went with about 10% higher mortality (HR 1.1) and 600 µg/L or more with about 50% higher (HR 1.5), although the same paper's meta-analysis found no excess risk for the top quarter or third of values across population studies [22]. Ferritin is also an acute-phase reactant, so a high value needs to be read alongside hs-CRP and transferrin saturation before being called "iron overload."
| Ferritin | What it usually means in a healthy adult |
|---|---|
| Under 15 µg/L | Iron stores empty: iron deficiency [23] |
| 15–30 µg/L | Low iron stores [24] |
| Under 70 µg/L with infection or inflammation | Can still mean iron deficiency [25] |
| Over 150 µg/L (women who menstruate) or 200 µg/L (men and other women) | Possible iron overload: check transferrin saturation and hs-CRP [26] |
If ferritin is high with a normal hs-CRP and a transferrin saturation of 45% or more, that pattern points toward genuine iron loading, and US liver-society guidance advises testing for the hereditary haemochromatosis gene (HFE) [27]. Why ferritin matters for ageing, and the dietary levers that move it, are covered under Iron and aging.
Uric acid
Evidence: Weak. Both very low and very high uric acid track higher mortality in large cohorts, but no study has defined an "optimal" window, and lowering uric acid in people without gout has not been shown to prevent cardiovascular events.
Uric acid sits on a U-shaped mortality curve: in 375,163 Korean adults, both the lowest levels (under about 210 µmol/L, 3.5 mg/dL, in men and 150 µmol/L, 2.5 mg/dL, in women) and the highest (about 565 µmol/L, 9.5 mg/dL, or more in men and 505 µmol/L, 8.5 mg/dL, in women) went with higher death rates [28]. A 2025 analysis found the same U-shape in US adults but not in an older Chinese cohort followed for only two years [29]. Very low uric acid in older adults frequently reflects malnutrition or sarcopenia. Very high uric acid causes gout and travels with hypertension and metabolic syndrome, but lowering it has not prevented cardiovascular events: among 5,721 adults aged 60 or over with ischaemic heart disease but no gout, a heart attack, stroke or cardiovascular death occurred in 11.0% of those on allopurinol and 11.3% on usual care over about five years [30].
There is no outcome-tested "optimal" range, so the practical use is to flag values at either extreme and read them in the context of diet (fructose load, alcohol, purines) rather than as a stand-alone diagnosis. Rheumatology guidance conditionally advises against urate-lowering drugs for high uric acid without gout; for people who do have gout, it aims for below 360 µmol/L (6 mg/dL) [31].
Foundational organ function — CBC and CMP
Evidence: Strong. These are the cheapest, best-validated markers of organ integrity in medicine; the readings below follow current kidney and liver guidance rather than tighter "optimal" numbers.
A complete blood count (CBC) and a comprehensive metabolic panel (CMP) remain useful annual baselines despite being unsexy:
- eGFR (estimated glomerular filtration rate, calculated from creatinine on the CMP) tracks kidney filtration, but creatinine also reflects muscle and diet: high-protein diets, creatine supplements, heavy strength training, vegetarian or keto diets and changes in muscle mass all shift it without any change in kidney function. Above about 60 mL/min/1.73 m², eGFR on its own says little. In general-population cohorts mortality was flat between 75 and 105 and rose only lower down — about 18% higher at 60 than at 95 (HR 1.18) [32]. A drift below 90 with age is common and is not kidney disease unless there are other signs of damage, and a very high creatinine-based eGFR is not automatically good news, because it can reflect low muscle mass [33]. The 2024 KDIGO guideline therefore recommends combining creatinine with cystatin C when creatinine may mislead and the result matters, and testing people at risk — including anyone with diabetes, high blood pressure or cardiovascular disease — with a urine albumin-to-creatinine ratio (uACR) as well as eGFR, because albumin in the urine can flag kidney damage while eGFR is still normal [34].
- ALT (alanine aminotransferase) and AST (aspartate aminotransferase) can hint at fatty liver disease — now called metabolic dysfunction-associated steatotic liver disease (MASLD) [35], and closely tied to the insulin-resistance story above. But a normal result does not rule it out. The enzymes are often normal even with advanced scarring, and US liver-society guidance treats ALT or AST above about 30 U/L, whether intermittently or persistently, as a possible sign of ongoing liver injury — a level below most labs' printed upper limit. For adults with prediabetes, type 2 diabetes or two or more metabolic risk factors, the same guidance recommends a Fibrosis-4 index (FIB-4, calculated from age, AST, ALT and platelet count) every 1–2 years; a score of 1.3 or more (above 2.0 after age 65) should lead to elastography, an enhanced liver fibrosis (ELF) blood test or referral [36].
- CBC indices like mean corpuscular volume (MCV) and red cell distribution width (RDW) can hint at B12, folate or iron problems, but they are late and insensitive signals: in screened populations a raised MCV caught only about 1 in 6 cases of low B12 [37], and in pernicious anaemia macrocytosis was absent in a third of patients, sometimes masked by coexisting iron deficiency [38]. A normal CBC does not rule out a deficiency; test B12 or ferritin directly when there is a reason to suspect one.
Targeted micronutrients — B12 and vitamin D
Evidence: Moderate for B12 checks in long-term metformin users (a secondary analysis of a large randomised trial, now reflected in diabetes guidelines); Weak for routine micronutrient screening in healthy adults without symptoms.
Two micronutrient labs that earn their place only in defined subgroups:
- Vitamin B12 for adults on long-term metformin: diabetes guidelines advise periodic checks, especially after four to five years of use, at doses of 1,500 mg a day or more, with another risk factor for deficiency, or with anaemia, peripheral neuropathy or chronic kidney disease [39]. In long-term follow-up of a large randomised diabetes-prevention trial, each year of metformin use raised the odds of low B12 by about 13% [40]. Strict vegans without a supplement or fortified foods, and long-term users of proton-pump inhibitors (PPIs) or H2 blockers, are also at higher risk, but UK guidance tests them only when there is also a suggestive symptom or sign [41]. Serum B12 alone is insensitive; confirm with methylmalonic acid (MMA) or holotranscobalamin (the fraction of B12 actually available to cells) in anything ambiguous. A total B12 above about 258 pmol/L (350 pg/mL) makes deficiency unlikely, and below 133 pmol/L (180 pg/mL) confirms it [42]. See B vitamins for the form and dosing detail.
- 25-hydroxyvitamin D is not a routine screen. The 2024 Endocrine Society guideline found no trial evidence that routine testing helps the general population, including people with obesity or darker skin, and no clear evidence defining an optimal blood level for preventing disease, so it suggests against routine testing without an established medical reason [43]. Test when a clinician has a specific reason, such as a bone-density or malabsorption work-up, and treat any target range as a convention rather than a level proven by outcomes. See Vitamin D for the threshold-by-threshold breakdown.
The annual panel at a glance
| Marker | Conventional reference | What to aim for | Cadence |
|---|---|---|---|
| ApoB | <130 mg/dL | Risk-based: below about 90–100 mg/dL at moderate (borderline-to-intermediate) risk, 70–80 at high risk, 55–65 at very high risk (see the apoB table) | Annual |
| Fasting insulin | <150 pmol/L (<25 μIU/mL) | Within lab reference interval; trend in one lab | Annual |
| HbA1c | <5.7% | 5.0–5.4% | Annual |
| HOMA-IR | <2.5 | Below about 2 | Annual (calculated) |
| hs-CRP | <3.0 mg/L | <1.0 mg/L (lowest risk; not an outcome-tested target) | Annual |
| Ferritin | 10–300+ µg/L | Above 30; below 150 (W who menstruate) / 200 (M and other W) µg/L | Annual, with transferrin saturation |
| Uric acid | broad | Avoid either extreme (no outcome-tested window) | Annual |
| eGFR (CMP), plus uACR if at risk | >60 mL/min/1.73m² | Stable over time; uACR normal | Annual |
| CBC, ALT/AST | by lab | CBC within reference; ALT/AST under about 30 U/L; FIB-4 below 1.3 if metabolic risk factors | Annual |
| Vitamin B12 | lab-specific | >258 pmol/L (deficiency unlikely) | Periodically on long-term metformin (especially after 4–5 years); otherwise if symptoms plus a risk factor |
| 25(OH)D | varies | No outcome-proven target | Only with a specific indication |
One-time tests
A small set of markers is determined by genetics or fixed structural traits. They settle in early adulthood and don't move with diet, exercise, or standard pharmacotherapy. Drawing them more than once in a lifetime is waste.
Lp(a) — one lifetime measurement
Evidence: Strong. Lp(a) is an independent, genetically determined causal risk factor, and universal one-time screening is now guideline-backed; what a high result should change in practice is still evolving.
Lipoprotein(a) is an LDL-like particle with an extra protein attached (apolipoprotein(a)), long suspected of promoting clots as well as plaque, although that clotting role is unproven in people. About 90% of a person's Lp(a) level is inherited, mainly through variants in the LPA gene; it is fairly stable from about age five, barely moved by diet or exercise, and not lowered by statins, which in some studies have raised it [44]. Roughly 1 in 5 people has an Lp(a) of 125 nmol/L (50 mg/dL) or more [45] — high enough to materially elevate cardiovascular and aortic-valve risk.
The 2026 ACC/AHA dyslipidaemia guideline and the National Lipid Association both recommend universal one-time Lp(a) screening in every adult.[46] Action thresholds start at about 125 nmol/L (50 mg/dL), which carries about 1.4 times the cardiovascular risk of someone at the population median; at 250 nmol/L (100 mg/dL) the risk roughly doubles [47]. A high result doesn't change the available drugs much: statins do not lower Lp(a), PCSK9 inhibitors lower it by about 15–30%, and lipoprotein apheresis is approved in the US only for people with an Lp(a) of 130 nmol/L (60 mg/dL) or more who also have familial hypercholesterolaemia and coronary or peripheral artery disease [48]. The first outcome trial of a drug designed to lower Lp(a) — pelacarsen, in 8,323 people with established cardiovascular disease — lowered Lp(a) but missed its primary endpoint of reducing major cardiovascular events in topline results released in September 2026; the full data have not yet been presented [49]. Trials of lepodisiran, olpasiran and muvalaplin are still ongoing [50]. What a high result does change is how aggressively the other levers are pulled — driving apoB to the lowest tolerated level, controlling blood pressure tightly, and intensifying surveillance. See Lipid management for the Lp(a)-specific drug landscape.
APOE — only as pharmacogenetics, never as predictive screening
The Apolipoprotein E ε4 allele is the largest common genetic risk factor for late-onset Alzheimer's disease and modestly elevates cardiovascular risk. The temptation to "find out your ε4 status" as a longevity test is strong; the clinical case against it is stronger.
Three reasons to refuse predictive APOE genotyping in healthy asymptomatic adults:
- Poor discrimination for dementia. About half of people clinically diagnosed with Alzheimer's carry no ε4 copy [51], although among those with amyloid-confirmed disease about two-thirds are carriers [52]. Two copies (ε4/ε4, about 2% of people) are a different story: in brain-donor and clinical cohorts, nearly all homozygotes had Alzheimer's brain changes, and by age 65 nearly all had abnormal amyloid in spinal fluid and 75% had a positive amyloid scan. Earlier studies, summarised in the same paper, put their lifetime risk of Alzheimer's dementia at up to about 60% by age 85 [53], so many never develop dementia. Either way the result rarely changes individual management, because the universal mitigations (aerobic exercise, sleep, blood pressure control, hearing, social engagement) apply regardless.
- No disease-modifying prophylaxis exists for asymptomatic carriers. The result delivers anxiety without an action.
- Insurance and disclosure risk in some jurisdictions, particularly for long-term care and disability cover.
APOE testing does belong in one specific context: before starting an anti-amyloid antibody (lecanemab or donanemab) for mild cognitive impairment due to Alzheimer's or early Alzheimer's. People with two ε4 copies have far more amyloid-related imaging abnormalities (ARIA) — brain swelling and microbleeds — on these drugs: on lecanemab, symptomatic brain swelling occurred in 9% of homozygotes, against 2% of single-copy carriers and 1% of non-carriers [54]. The EU licences for both drugs, and the UK licence for lecanemab, exclude homozygotes, so the genotype is checked first [55], [56], [57]. US labels say testing should be performed but allow treatment without it, and homozygotes remain eligible after counselling about their higher risk [58], [59]. That is pharmacogenetic testing, not screening. The right time to draw it is at the start of an amyloid-clearing therapy decision, not at age 45 with no symptoms.
Hormones — symptom-driven, not screen-driven
Hormonal screening of asymptomatic adults is the area where direct-to-consumer "longevity" panels overreach the hardest, so the discipline here is short and firm.
- Testosterone — measure on the basis of clinical symptoms (low libido, fatigue, lost morning erections, body-composition change, persistent low mood) plus the supporting workup, not as a general "optimisation" lab. Two morning fasting draws are required because diurnal variation is large. Pursuing high-normal levels in asymptomatic men is not supported by outcome data and carries real risks (polycythaemia, exacerbation of occult prostate disease, blood-pressure elevation). See Testosterone therapy.
- Estradiol, follicle-stimulating hormone (FSH), and progesterone in women rarely give actionable information outside of symptomatic perimenopause workup, fertility evaluation, or titration of indicated menopausal hormone therapy. The decision to start MHT is symptom- and timing-driven, not lab-driven.
- Thyroid-stimulating hormone (TSH) — the case against routine asymptomatic screening is laid out in detail in Thyroid management. The short version: a mildly elevated TSH in a well midlife adult is overwhelmingly likely to be transient or physiological, and the TRUST trial in 737 adults aged 65 or older found that treating subclinical hypothyroidism did not improve symptoms, tiredness or thinking scores; it was too small and short to test effects on heart attacks or strokes, and it did not include midlife adults.[60] Order TSH when there's a reason — fatigue, weight change, mood, cold intolerance, dyslipidaemia, fertility evaluation, or pregnancy.
- DHEA-S (dehydroepiandrosterone sulfate) declines linearly with age, and the cortisol/DHEA-S ratio is an interesting research biomarker for adrenal resilience. As a clinical screen in healthy adults it doesn't change management; DHEA supplementation is controversial and the lab is most useful longitudinally rather than diagnostically.
- Random serum cortisol is useless as a general "stress" lab — see the tests-to-skip section below for the salivary version of the same mistake.
Biological age clocks
Evidence: Moderate that second- and third-generation clocks predict mortality and disease; Weak that a clock reading should drive any specific individual decision — they are surrogate biomarkers, not hard outcomes.
Where the annual panel measures the substrates of ageing, DNA-methylation-based epigenetic clocks attempt to measure ageing itself — the integrated rate at which a person's biology is drifting from their chronological age. Three generations of clocks now exist, with sharply different uses:
| Generation | Examples | Trained on | What it is good for |
|---|---|---|---|
| First | Horvath, Hannum | Chronological age | Little beyond estimating your age |
| Second | GrimAge, PhenoAge | Mortality and clinical biomarkers | Predicting mortality and disease in groups |
| Third | DunedinPACE | Longitudinal rate of organ-system decline within a single birth cohort | Tracking the pace of change in research trials |
The clock built to measure the pace of ageing is DunedinPACE.[61] It was trained on the Dunedin longitudinal study — a single birth cohort of New Zealanders born in 1972–1973 — by tracking 19 organ-system biomarkers across four timepoints over two decades. Because the algorithm was built within one birth year, it sidesteps the cohort confounders that plague cross-sectional clocks (leaded petrol, vaccination history, dietary environment). The output is a single number:
- 1.0 — ageing at one biological year per chronological year
- <1.0 — decelerated ageing (0.85 = 15% slower than calendar)
- >1.0 — accelerated ageing
DunedinPACE gives nearly the same number when the same blood sample is run twice: an intraclass correlation (a reliability score where 1.0 is perfect agreement) of 0.96 across 36 technical replicates [62]. That is lab precision, not stability in the person. When the same people were re-sampled within hours or days — around meals, acute stress, air-pollution exposure or a change of altitude — most clocks were only moderately stable, and DunedinPACE was among the less stable ones [63]. In 899 generally healthy older adults (mean age 70) tested yearly, DunedinPACE did not change on average over two years, and across the clocks studied, one-year rises tended to reverse the next year [64]. So a change between two personal results a year apart can easily be noise: clocks can detect average effects across a trial group, but none is validated for tracking one person. In its development paper, DunedinPACE's links with disease and mortality were similar in size to GrimAge's [65].
Honest caveats:
- These are surrogate biomarkers, not hard outcomes. A slowed clock under a particular intervention predicts, but doesn't prove, a longer or healthier life.
- Only a few randomized trials show a clock slowing, all as secondary analyses with small effects. In DO-HEALTH, omega-3 alone (1 g a day) slowed three clocks in adults over 70 — by about 1% on DunedinPACE — while vitamin D and exercise on their own moved none, and the three combined added up on only one clock (PhenoAge) [66]. Two years of calorie restriction in CALERIE, where people achieved about half the prescribed 25% cut, slowed DunedinPACE by 2–3% but did not move PhenoAge or GrimAge [67]. Semaglutide slowed DunedinPACE by about 9% over 32 weeks in 84 people with HIV-associated fat gain [68]. See Omega-3 and GLP-1 receptor agonists.
- The commercial clock-testing market is uneven on analytical quality. Ask any provider three questions before paying: which published clock algorithm they report, whether the assay is run in an accredited clinical laboratory, and what their test-retest reliability is on repeat samples from the same person. Then run every serial measurement through the same provider and the same sample type — absolute scores are not comparable across platforms, and in some clocks the same person's estimate differed by almost 30 years between a cheek-swab or saliva sample and blood [69] — so a change of vendor or sample destroys the only thing the test was for.
- Testing more often does not help: one person's result already varies over hours to days, so frequent repeats mostly sample noise.
Tests to skip
The flip side of a disciplined panel is the discipline to refuse the marketed extras. Five widely advertised tests routinely fail their own claims and reliably trigger downstream harm.
NMR LipoProfile (and other LDL-particle counts) when apoB is available
NMR spectroscopy returns an estimated LDL-P (particle number). Because every atherogenic particle carries exactly one apoB, the two largely track each other: in 27,673 initially healthy women followed for 11 years, the top fifth for either measure carried about 2.5 times the cardiovascular risk of the bottom fifth [70], and a review of 25 studies judged them nearly equivalent, preferring apoB for its availability, standardisation and low cost [71]. If you already have an apoB result, ordering NMR is paid-for redundancy.
Routine IGF-1 screening in healthy adults
Insulin-like growth factor 1 (IGF-1) is medically essential for diagnosing acromegaly and growth-hormone deficiency. As a "longevity biomarker" to be optimised in healthy adults, it fails on two counts:
- Bidirectional risk. Cohort data show a U-shape: in a pooled analysis of 12 studies, people with low IGF-1 had about 27% higher all-cause mortality (HR 1.27) and those with high IGF-1 about 18% higher (HR 1.18) than people in the middle range [72], a pattern repeated in about 381,000 UK Biobank adults, where the highest levels also carried higher cardiovascular mortality [73].
- No safe modulation pathway. There is no randomized evidence that nudging IGF-1 in either direction in an asymptomatic adult improves outcomes, and genetic studies suggest that lifelong higher IGF-1 modestly raises cancer risk at a few sites — about 8% higher colorectal cancer risk per standard deviation of IGF-1 [74].
Test IGF-1 to rule in or out a pituitary disorder, not to "track ageing."
"Adrenal stress" salivary cortisol panels
"Adrenal fatigue" — the idea that chronic stress depletes cortisol output and produces a clinical syndrome — is not a recognized medical diagnosis. A systematic review of 58 studies found no evidence that it exists, and noted that no endocrinology society recognises it [75]; the Endocrine Society states that saliva and blood tests sold to diagnose it "are not based on scientific facts" [76]. In that review, salivary cortisol rhythms did not reliably separate fatigued from healthy people: 16 of 26 studies found no difference.
Real disorders of the adrenal–pituitary axis (Cushing's syndrome, Addison's disease, mild autonomous cortisol secretion in adrenal incidentalomas) require specific, validated workups, not a four-tube saliva kit.
IgG (and IgG4) food-sensitivity panels
The premise — that food-specific IgG antibodies indicate intolerance — gets the immunology backwards. Food-specific IgG and IgG4 mostly reflect normal exposure to, and tolerance of, foods people eat often [77], and a rise in IgG4 is one of the typical immune changes during oral immunotherapy for genuine, IgE-mediated food allergy [78]. The Canadian Society of Allergy and Clinical Immunology strongly discourages IgG testing to identify or predict adverse reactions to food [79], and the American Academy of Allergy, Asthma & Immunology and the European Academy of Allergy and Clinical Immunology advise against it for diagnosing food allergy or intolerance [80].
The downstream harm is concrete: in people without symptoms, a positive result mostly reflects foods eaten often, and the elimination diets it prompts narrow nutrition and strain social eating for no proven gain. The one place the idea has shown a signal is irritable bowel syndrome (IBS): in a sham-controlled trial in 223 people with IBS who had tested IgG-positive, about 60% of those on an IgG-guided diet had their abdominal pain fall by at least 30%, against 42% of those on a sham diet, using a purpose-built 18-food assay rather than a consumer panel [81]. That is a question for a gastroenterologist, not a reason for a healthy adult to buy a panel.
Provoked (post-chelation) heavy-metal urine testing
Routine screening of asymptomatic adults for heavy metals (lead, mercury, cadmium, arsenic) is already low-value unless there's an occupational or known-exposure history. The variant marketed in alternative-medicine clinics — administering a chelating agent (DMSA, DMPS) and then collecting urine — has not been scientifically validated, has no demonstrated benefit and may be harmful, according to the American College of Medical Toxicology [82].
The mechanism of the deception is straightforward: chelating agents raise urinary metal output in healthy people too, so provoked urine is expected to show higher levels. No reference range exists for provoked urine, yet some labs compare it with the range for normal, unprovoked urine, so results look falsely high. The chelation that often follows is costly and can harm: risks include fatal low blood calcium with edetate disodium, loss of essential minerals such as zinc, copper and iron, and kidney damage with some agents.
The harms of overtesting
The reason a disciplined panel matters isn't just cost. Every avoidable test in an asymptomatic adult carries a measurable downside:
- Overdiagnosis — finding a real biological anomaly that would never have caused harm in the patient's natural lifetime. The reference example is small, indolent papillary thyroid cancer found by high-resolution neck ultrasound; a modelling analysis of US registry data estimated that 72–94% of papillary thyroid cancers diagnosed between 1991 and 2019 were overdiagnoses — cancers that would never have shortened life — and for most of that period the usual response was surgery, often followed by lifelong thyroid hormone and ongoing surveillance.[83] The 2025 American Thyroid Association guidelines had already made active surveillance a formal, if conditional, option for small (1 cm or less), low-risk papillary cancers [84]. See Thyroid management and Cancer screening cadence.
- Cascade testing. A single ambiguous result invites confirmatory imaging, biopsies, and procedure-related complications. The base-rate problem is unforgiving: even a 95%-specific test on a 1%-prevalence finding produces more false positives than true positives.
- Iatrogenic medication starts. Borderline labs (a TSH of 5.2, an "elevated" ferritin in the upper quartile, a single high LDL on a stress week) frequently trigger lifelong prescriptions that confer no benefit and impose real side-effect burdens.
- Anxiety and its knock-on effects. A false alarm is not free even when the follow-up is clear: after false-positive screening mammograms, women report worry specific to the condition for up to 3 years, more so after more invasive follow-up, and they are slightly less likely to come back for their next routine screen [85]. That evidence comes from breast screening, but every borderline flag on a broad panel is another chance for the same cycle to start.
The asymmetry of these costs is the reason for the principle: order the smallest panel that answers a real question, and refuse the rest.
A practical annual cadence
For a generally healthy adult in midlife:
- Annual fasting morning draw: apoB, fasting insulin, fasting glucose, HbA1c (calculate HOMA-IR from the first two), hs-CRP, ferritin (with transferrin saturation if ferritin is high), uric acid, CBC, comprehensive metabolic panel (eGFR, ALT, AST). Add a uACR test if you have diabetes, high blood pressure or cardiovascular disease, and a FIB-4 liver score if you have prediabetes, type 2 diabetes or two or more metabolic risk factors.
- B12 periodically on long-term metformin, especially after four to five years; in vegans and long-term PPI/H2-blocker users, test if a suggestive symptom or sign appears.
- 25-hydroxyvitamin D only with a specific reason (for example a bone-density or malabsorption work-up), not as a routine screen.
- Once-in-a-lifetime: Lp(a). Add to the first draw a midlife adult ever has. The result is permanent.
- APOE genotype: only when an amyloid-clearing therapy decision is on the table in a patient with mild cognitive impairment.
- Hormonal labs: symptom-driven, not screen-driven. Two morning testosterones if low-T symptoms; TSH only with corroborating signs; estradiol/FSH only in the context of perimenopause workup or MHT titration.
- Biological-age clock: optional, and not a decision tool. Clocks such as DunedinPACE are research measures: no agreed standard yet says when one is accurate enough to guide an individual's care [86], and one person's results vary noticeably over hours to days, including around meals and acute stress [87]. If you test anyway, keep the same provider and sample type, draw at the same time of day under the same fasting conditions, and do not change treatment on the result alone.
- Coronary artery calcium (CAC) scoring — when a statin decision is genuinely on the fence. A non-contrast CT that converts a borderline-or-intermediate risk estimate into anatomy, used in men from 40 and women from 45. If the score is zero and you would rather avoid a statin, it is reasonable to defer one and repeat the scan in 3–7 years — but only when no higher-risk condition applies: familial or severe hypercholesterolaemia (LDL-C of 4.9 mmol/L (190 mg/dL) or more), diabetes after age 40, current smoking, or a strong family history of premature heart disease [88]. Any measurable calcium is subclinical atherosclerosis, and the guideline recommends starting lipid-lowering treatment, particularly at a score of 100 or more [89]. Not an annual test and not for everyone — the tool for the grey zone. See Lipid management for the score strata and how they change LDL-C targets.
- Imaging and procedural screens — biennial mammography 40–74 (US Preventive Services Task Force, USPSTF, 2024), colorectal screening 45–75 with the modality you'll actually complete, primary human papillomavirus (HPV) testing every 5 years 30–65, yearly low-dose CT at 50–80 for people with at least 20 pack-years who still smoke or quit within the past 15 years, one-time hepatitis C virus (HCV) antibody 18–79, one-time abdominal aortic aneurysm (AAA) ultrasound in men 65–75 who ever smoked. See Cancer screening cadence.
- Refuse: NMR LipoProfile when apoB is on the panel, routine IGF-1 in healthy adults, salivary "adrenal stress" mapping, IgG/IgG4 food-sensitivity panels, any provoked heavy-metal challenge, and whole-body magnetic resonance imaging (MRI) sold as a screen — in about 9,000 people without symptoms it found confirmed cancer in roughly 1.6% of people screened, with frequent incidental findings and no evidence that it improves outcomes [90].
Further reading
- Liss DT et al. General Health Checks in Adult Primary Care: A Review. JAMA 2021.[91]
- Krogsbøll LT et al. General health checks in adults for reducing morbidity and mortality from disease. Cochrane Database Syst Rev 2019.[92]
- De Oliveira-Gomes D et al. Apolipoprotein B: Bridging the Gap Between Evidence and Clinical Practice. Circulation 2024.[93]
- Ridker PM et al. Inflammation, Cholesterol, Lipoprotein(a), and 30-Year Cardiovascular Outcomes in Women. NEJM 2024.[94]
- 2026 ACC/AHA/Multisociety Guideline on the Management of Dyslipidemia. J Am Coll Cardiol 2026.[95]
- Mora S et al. Lipoprotein Particle Profiles by NMR Compared With Standard Lipids and Apolipoproteins in Predicting Incident CVD in Women. Circulation 2009.[96]
- Belsky DW et al. DunedinPACE, a DNA methylation biomarker of the pace of aging. eLife 2022.[97]
- Stott DJ et al. Thyroid Hormone Therapy for Older Adults with Subclinical Hypothyroidism (TRUST). NEJM 2017.[98]
- Aroda VR et al. Long-term Metformin Use and Vitamin B12 Deficiency in the DPP/DPPOS. J Clin Endocrinol Metab 2016.[99]
- Cadegiani FA, Kater CE. Adrenal fatigue does not exist: a systematic review. BMC Endocr Disord 2016.[100]
- American College of Medical Toxicology. ACMT Recommends Against Use of Post-Chelator Challenge Urinary Metal Testing. J Med Toxicol 2017.[101]
- Francis DO et al. Overdiagnosis of Papillary Thyroid Cancer. JAMA Netw Open 2026.[102]