What a Longevity Blood Test Should Measure
A “longevity blood test” is a standard clinical chemistry panel plus a handful of additions, scored against age norms and tracked over time. There is no assay that measures how long you will live. What exists is a set of analytes with strong associations to mortality and multimorbidity in large cohorts, and a set of composite scores built on top of them. The value comes from repeated measurement on yourself under controlled conditions and not from a single 70-marker snapshot. This page covers which markers carry signal and how the composites are constructed. It also covers how to keep pre-analytical noise from swamping your trend and what schema to store the results in.
What belongs on the panel
Start with the analytes that have been standardized across labs and have decades of outcome data behind them. Apolipoprotein B is the one I would insist on. It counts atherogenic particles directly, is measured by immunoturbidimetry against an international reference preparation, and does not require fasting. A standard lipid panel gives you LDL-C by calculation (Friedewald or Martin-Hopkins), which diverges from particle count in people with high triglycerides.
Lipoprotein(a) is largely determined by the LPA gene and is close to constant across adult life. Measure it once, in nmol/L rather than mg/dL, since mass units depend on isoform size. HbA1c integrates glycemia over the preceding two to three months and is constrained by red cell lifespan, so anemia or hemoglobin variants distort it. Fasting insulin and glucose together let you compute HOMA-IR, which is noisy in a single draw and useful only as a repeated measure.
For renal function, ask for cystatin C alongside creatinine and use the 2021 CKD-EPI combined equation. Creatinine alone tracks muscle mass as much as filtration, which makes it a poor longitudinal marker in anyone changing their training load. High-sensitivity CRP is the standard inflammation marker, and it is the most volatile item on the list. A recent infection or hard training session can move it by an order of magnitude, so interpret it only from three or more draws.
Round it out with a few standard panels: a complete blood count with differential, a comprehensive metabolic panel and thyroid function. Add ferritin with transferrin saturation, homocysteine and 25-hydroxyvitamin D. The CBC matters more than it looks: red cell distribution width and lymphocyte percentage are among the strongest mortality-associated indices in routine hematology, and they cost nothing extra. Reviews of candidate aging biomarkers keep converging on this set of inflammatory, metabolic, and hematologic measures rather than on exotic assays.12
Composite scores and what they are worth
The interesting layer sits above the raw analytes. PhenoAge is the most useful clinical-chemistry composite. Levine and colleagues fit a mortality model and then expressed the result in years. The model takes the following inputs.3
- albumin
- creatinine
- glucose
- log CRP
- lymphocyte percent
- mean cell volume
- red cell distribution width
- alkaline phosphatase
- white blood cell count
- chronological age
Every input is on a standard panel, so you can compute it yourself from a CSV.
The same paper transferred that phenotype onto DNA methylation and produced DNAm PhenoAge, which is the reason epigenetic clocks became interesting to people who care about outcomes rather than about age prediction. Clocks trained purely to predict chronological age saturate at high accuracy and then carry little residual information. Clocks trained on a mortality-weighted phenotype keep a signal after adjusting for age.45
Machine learning on blood chemistry alone can also predict chronological age reasonably well, which tells you that a modest panel of common analytes carries a real age signal.6 Treat the resulting “biological age” number as a compressed summary of your panel. It is not an independent measurement. If your score moves, open the components and find out which one moved.
Two newer directions are worth watching. Large cohort work has identified blood biomarkers shared across multiple chronic disease clusters, which supports scoring a panel as a whole rather than marker by marker.7 And intervention studies have shown that composite scores respond to short-term dietary interventions, which means the number is modifiable and therefore sensitive to what you did the week before the draw.8
Pre-analytical control decides whether your data is usable
Most of the variance in a personal time series is procedural. Fix these before you fix anything else.
Draw at the same time of day, ideally between 07:00 and 09:00. Cortisol, testosterone, and iron all follow strong diurnal patterns. Fast 12 hours for insulin and triglycerides, and keep the fasting duration constant rather than maximal. Stay seated for 10 minutes before venipuncture, because standing shifts plasma volume and inflates every concentration by several percent. Avoid vigorous exercise for 48 hours before a CRP or creatine kinase draw.
Homocysteine rises in the tube if plasma is not separated from cells promptly, so it needs an EDTA tube on ice or rapid centrifugation. Potassium is falsely elevated by hemolysis, which is the most common reason for an absurd result. Use the same laboratory and ask which method each assay uses. A change in vitamin D platform or ferritin immunoassay produces a step change in your series that looks like biology.
Deciding whether a change between two draws is real
This is where technical readers get the most out of their own data. Every analyte has analytical variation (CVa, from the instrument) and within-person biological variation (CVi, from you). The reference change value gives the smallest difference that is unlikely to be noise:
RCV = 1.96 * sqrt(2) * sqrt(CVa^2 + CVi^2) # 95%, two-sided
Published CVi estimates live in the EFLM Biological Variation Database. Plug them in and you find that HbA1c moves meaningfully on small changes, while hs-CRP and ferritin need very large swings before a two-point comparison means anything. For the volatile markers, the right approach is a rolling median over three or more draws. A difference of two is not enough.
Cell-based assays inherit their own variability. When aging markers are measured in peripheral blood mononuclear cells, replicate handling and isolation protocol contribute a substantial share of the observed spread.9 If you are paying for anything measured in isolated cells, ask about replicate agreement before you ask about the result.
Schema and pipeline
Request results as structured data and not as PDFs. Most large labs can export HL7 FHIR Observation resources, and every result carries a LOINC code. Store one row per measurement:
draw_datetime, lab_id, loinc, analyte, value, unit_ucum,
method, ref_low, ref_high, fasting_hours, notes
Normalize units with UCUM so that nmol/L and mg/dL never sit in the same column. Keep method populated, because it is the only way to detect an assay change later. In Python, pandas plus a small YAML file of CVi values is enough to compute RCVs and flag significant changes. Compute PhenoAge from the nine analytes directly rather than trusting a vendor’s implementation, and log the version of the coefficients you used.
Cadence: quarterly for the metabolic and inflammatory markers, annually for thyroid and vitamin status, once for Lp(a) and any genotype. More frequent testing mostly buys you noise.
What blood cannot tell you
A panel describes circulating chemistry at one moment. It does not describe tissue-level processes, and it does not diagnose anything. Several findings are reasons to involve a physician rather than a spreadsheet. These include abnormal ferritin, an unexplained CBC shift, a rising calcium and a persistently elevated hs-CRP. The same applies to any composite score that moves sharply without an explanation in its components.
Questions people also ask
How often should I get a longevity blood panel? Quarterly is enough for the metabolic and inflammatory markers, and annually is enough for vitamins and thyroid. The limiting factor is within-person biological variation. Access to testing is not the constraint.
Is the “biological age” number meaningful? It is a compressed summary of the inputs, with real mortality associations behind the weighting.3 Read it as a flag that points you back to specific analytes, and do not treat a one-year change as a measurement of your aging rate.
Should an epigenetic clock be part of the panel? Only if it is a phenotype-trained clock and you understand its reproducibility limits. Clocks trained to predict chronological age add little on top of a good chemistry panel.45
Do I need to fast? Fast for insulin, glucose and triglycerides. Keep the duration constant across draws. ApoB, HbA1c, and hs-CRP do not require fasting.
Can diet change these markers quickly? Yes, and that is a reason to standardize the week before a draw. Short dietary interventions have been shown to shift hepatic and blood markers used in composite aging scores.8
Footnotes
-
P. M. Engelfriet, E. H. J. M. Jansen, H. S. J. Picavet, et al. Biochemical Markers of Aging for Longitudinal Studies in Humans. Epidemiologic Reviews, 2013. https://doi.org/10.1093/epirev/mxs011 ↩
-
Karl-Heinz Wagner, David Cameron-Smith, Barbara Wessner, et al. Biomarkers of Aging: From Function to Molecular Biology. Nutrients, 2016. https://doi.org/10.3390/nu8060338 ↩
-
Morgan E. Levine, Ake T. Lu, Austin Quach, et al. An epigenetic biomarker of aging for lifespan and healthspan. Aging, 2018. https://doi.org/10.18632/aging.101414 ↩ ↩2
-
Steve Horvath, Kenneth Raj. DNA methylation-based biomarkers and the epigenetic clock theory of ageing. Nature Reviews Genetics, 2018. https://doi.org/10.1038/s41576-018-0004-3 ↩ ↩2
-
Ran Duan, Qiaoyu Fu, Yu Sun, et al. Epigenetic clock: A promising biomarker and practical tool in aging. Ageing Research Reviews, 2022. https://doi.org/10.1016/j.arr.2022.101743 ↩ ↩2
-
Evgeny Putin, Polina Mamoshina, Alexander Aliper, et al. Deep biomarkers of human aging: Application of deep neural networks to biomarker development. Aging, 2016. https://doi.org/10.18632/aging.100968 ↩
-
Alice Margherita Ornago, Caterina Gregorio, Federico Triolo, et al. Shared and specific blood biomarkers for multimorbidity. Nature Medicine, 2026. https://doi.org/10.1038/s41591-025-04038-2 ↩
-
Sebastian Brandhorst, Morgan E. Levine, Min Wei, et al. Fasting-mimicking diet causes hepatic and blood markers changes indicating reduced biological age and disease risk. Nature Communications, 2024. https://doi.org/10.1038/s41467-024-45260-9 ↩ ↩2
-
Brady M. Owen, James Phie, Jennifer Huynh, et al. Evaluation of quantitative biomarkers of aging in human PBMCs. Frontiers in Aging, 2023. https://doi.org/10.3389/fragi.2023.1260502 ↩
Pothos builds longitudinal molecular profiles of individuals: whole-genome sequencing, single-cell RNA sequencing, and plasma proteomics, delivered as a detailed report, every data file, and stored cells. Build your profile.