Every month, a new longevity biomarker gets surfaced in popular health media. Some are legitimately important. Many are expensive to test, difficult to interpret, and not meaningfully actionable at the individual level. The problem is that when everything is framed as critical, nothing is prioritised correctly.
What follows is an attempt to separate the markers with genuine predictive value from those that are either not well-validated in non-clinical populations or not actionable without medical intervention.
The tier-one markers: most predictive, most actionable
VO₂ max is the single strongest predictor of all-cause mortality in the published epidemiological literature. A 2018 study in JAMA Network Open (n=122,000) found that low cardiorespiratory fitness was associated with higher mortality risk than smoking, hypertension, or diabetes. The fifth percentile of VO₂ max carried 5x the mortality risk of the 97.7th percentile — a larger effect size than any blood biomarker currently in routine use. It's also directly modifiable: VO₂ max responds significantly to structured aerobic training in 8–12 weeks.
Grip strength is a reliable proxy for overall musculoskeletal function and is inversely associated with cardiovascular mortality, cancer mortality, and cognitive decline. It's not causal — grip strength measures overall muscle integrity — but it's one of the simplest tests with the most longitudinal validation across populations. Declining grip strength over time is a meaningful signal. Absolute values matter less than trajectory.
Fasting insulin and HOMA-IR are vastly more informative than fasting glucose alone for metabolic health assessment. Many individuals have normal fasting glucose but elevated fasting insulin, indicating the pancreas is compensating for reduced insulin sensitivity. HOMA-IR (homeostatic model assessment for insulin resistance) = (fasting glucose × fasting insulin) / 405. A score above 2.0 indicates insulin resistance; below 1.0 is optimal. This is not a standard test in routine panels — you usually have to request it.
The most underutilised test: Fasting insulin. Cheap, widely available, rarely ordered. If your fasting glucose is normal but your fasting insulin is elevated (above ~8 uIU/mL), you have meaningful metabolic information that fasting glucose alone would have missed — and that information is both actionable and time-sensitive.
Tier two: valuable with caveats
ApoB (apolipoprotein B) is superior to LDL-C as a predictor of cardiovascular risk. Each LDL particle carries one ApoB molecule — so ApoB counts atherogenic particles directly, while LDL-C estimates the cholesterol content (which can vary between particles). Mendelian randomisation studies consistently show that ApoB is causally associated with ASCVD risk, independent of other lipid markers. The issue: optimal ApoB ranges are still debated in non-clinical contexts, and treatment thresholds require individual risk stratification with a clinician.
Lp(a) (lipoprotein (a)) is genetically determined (~90% heritable) and causally associated with cardiovascular disease and aortic stenosis. Testing is valuable because it identifies a fixed, inherited risk that standard lipid panels don't capture. The limitation: there is currently no lifestyle intervention that meaningfully lowers Lp(a). If elevated, it informs risk awareness and may warrant more aggressive management of modifiable risk factors — but the marker itself isn't actionable in isolation.
HbA1c provides a 90-day average of blood glucose. It's useful for trend tracking but less sensitive than fasting insulin for detecting early metabolic dysfunction. Values can be influenced by haemoglobin variants, iron status, and erythrocyte turnover rate — which is why it's a population screening tool rather than a precise individual metric.
The overhyped tier
Biological age clocks (epigenetic methylation clocks like GrimAge, PhenoAge) are scientifically legitimate in population research. As individual clinical tools, they're premature. Measurement variability between samples from the same individual can exceed reported "biological age" differences. The within-person reliability is not yet sufficient to base decisions on a single test. They're worth watching as the technology matures, not worth paying $400 for a snapshot in 2026.
Telomere length is inversely associated with ageing-related disease in population studies. In individuals, the test-retest variability is high, and telomere length is influenced by factors (stress, smoking, sleep) that are worth addressing regardless of telomere status. The test adds limited information beyond what's already captured by lifestyle markers.
Micronutrient panels — comprehensive panels testing 30+ nutrients — are typically low-value for people eating a varied, adequate diet. The exceptions: vitamin D, iron/ferritin, B12 (particularly for those on plant-forward diets or over 60), and zinc in vegetarians. Testing everything without a specific deficiency hypothesis produces noise and spurious interventions.
The framework for a useful panel
A practical annual panel for a healthy adult focused on longevity:
- Full blood count + metabolic panel
- Fasting glucose and fasting insulin (calculate HOMA-IR)
- HbA1c
- ApoB and Lp(a) (test Lp(a) once — it's stable; retest ApoB annually)
- hs-CRP (high-sensitivity C-reactive protein — inflammatory marker)
- Vitamin D (25-OH)
- Ferritin
- TSH (thyroid)
Complemented by functional assessments: annual VO₂ max test (or reliable proxy via structured running/cycling protocol), grip strength measurement, and a timed walking speed test (gait speed below 0.8 m/s in adults over 60 is a validated mortality predictor).
The goal of biomarker tracking is not the data itself. It's to identify leverage points — places where a specific intervention will move a specific metric and produce a health outcome that matters. Without that link, data is just noise with a price tag.