Which health biomarkers are worth measuring, and how often

If you want a short answer: measure a small set of analytes that have low analytical noise, meaningful within-person variation you can act on, and evidence linking them to hard outcomes. A short list of analytes covers most of the ground for a healthy adult.

• apolipoprotein B
• lipoprotein(a) once in your life
• HbA1c
• fasting insulin and glucose
• hs-CRP
• ALT and GGT
• creatinine and cystatin C
• albumin
• ferritin with transferrin saturation
• TSH
• a complete blood count

Then measure them more than once. A single draw tells you almost nothing about where your personal setpoint sits, because for most analytes the within-person variation between draws is large relative to the difference you care about.

The interesting engineering problem is deciding whether a change between two numbers is signal. Choosing which test to order is not the interesting problem.

The two variances that decide everything

Every result you get is your homeostatic setpoint plus two error terms: analytical imprecision (CV_A, the assay) and within-person biological variation (CV_I, your own day-to-day fluctuation around your setpoint). The reference change value tells you how big a difference between two serial results must be before it exceeds noise:

RCV = 2^0.5 * Z * sqrt(CV_A^2 + CV_I^2)

With Z = 1.96 (two-sided, 95%), the leading constant is 2.77. Work an example. HbA1c has CV_I near 1.5–2% and a good lab CV_A near 2%, so RCV is roughly 7–8%. A move from 5.4% to 5.6% is inside noise. A move from 5.4% to 5.9% is not. The opposite extreme is hs-CRP, with within-person variation often 30–60% between draws in healthy people. A doubling there can be nothing at all. Studies that sample the same healthy donors repeatedly over weeks are where these estimates come from, and they also set the quality goals labs are supposed to hit (imprecision under half of CV_I).¹

The second number worth computing is the index of individuality, CV_I divided by between-person variation CV_G. When it is below about 0.6, the population reference interval is nearly useless for you: you can double your personal value and still sit comfortably inside “normal.” TSH, ferritin, creatinine, and most thyroid and iron markers behave this way. This is the actual argument for longitudinal self-measurement. Your own prior result is a far tighter comparator than the lab’s 2.5th-to-97.5th-percentile band.

The third number is reproducibility over long intervals: the intraclass correlation between samples drawn a year or more apart. In the Nurses’ Health Studies, ICCs across plasma, red cell, and urine analytes varied widely by marker, which means some single measurements represent long-run exposure well and others do not.² If you only ever plan one draw, pick analytes with high ICC. If you plan quarterly draws, the low-ICC markers become usable because you can average them.

What to measure, and why those

Start from outcomes. A systematic review of cohort studies of blood-borne mortality biomarkers found consistent associations for markers of inflammation, renal function, glycemia, and nutritional status, with CRP and albumin recurring across cohorts.³ That is a decent prior for what belongs on a baseline panel.

Our default set, grouped by what it tells you:

• Lipids and atherogenic particle count: apoB, standard lipid panel, Lp(a) once. ApoB counts particles rather than the cholesterol carried in them, and diverges from LDL-C when triglycerides are high. Lp(a) is largely genetic and does not need repeating.
• Glycemic regulation: fasting glucose, fasting insulin, HbA1c. Compute HOMA-IR yourself from glucose and insulin. HbA1c is biased by red cell lifespan, so interpret it next to reticulocyte count and haptoglobin if it disagrees with your glucose data.
• Inflammation: hs-CRP, plus ESR and one cytokine (IL-6) if you have a reason. Draw hs-CRP at least twice, separated by weeks, and never within two weeks of an infection or hard training block.
• Liver and kidney: ALT, AST, GGT, ALP and bilirubin. Also creatinine, cystatin C and urine albumin-to-creatinine ratio. Cystatin C is less muscle-dependent than creatinine, which matters if you lift.
• Iron and nutrition: ferritin, transferrin saturation and B12. Also folate, 25-OH vitamin D and albumin.
• Endocrine and blood: TSH with free T4, CBC with differential.

A good biomarker needs analytical validity, a plausible link to the biology, and demonstrated association with an outcome in a population resembling yours.⁴ Most consumer panels fail on the third criterion while marketing hard on the first two.

Pre-analytics ruin more results than assays do

Fix the protocol before you fix the panel. Draw at the same time of day (cortisol, testosterone, and iron markers have strong diurnal swings). Standardize fasting state, 12 hours for triglycerides and insulin. Avoid strenuous exercise for 48 hours before drawing CK, AST, ALT, or hs-CRP. Sit for five minutes before venipuncture, since posture shifts plasma volume and moves protein-bound analytes by several percent. Use the same lab and the same assay method, and record the method: switching from an immunoassay to LC-MS/MS for testosterone will produce a step change that looks like biology.

Dried blood spots are a reasonable option for high-frequency sampling when venipuncture logistics dominate, and they raise participation in population studies substantially, though the analyte menu and precision are narrower than serum.⁵ Wearable and continuous platforms extend the sampling rate further, with their own calibration and drift problems.⁶

Where blood chemistry stops

A blood panel gives you 40 scalars sampled a few times a year. The molecular layers underneath answer different questions. Whole-genome sequencing gives you fixed risk alleles, pharmacogenes and carrier status, and never needs repeating. The pharmacogenes are CYP2C19, CYP2D6, DPYD, TPMT and SLCO1B1. RNA sequencing of whole blood gives cell-type composition and pathway activity that shift on the timescale of weeks. Plasma proteomics gives thousands of analytes at once, which is how blood-based detection of cancer and neurodegeneration is being built.⁷ Continuous glucose data gives you 288 readings a day, enough to estimate post-meal excursion, overnight variability, and coefficient of variation, none of which HbA1c can resolve.

The neurodegeneration field is the clearest case study in what happens when a biomarker moves from research to clinic. Plasma p-tau217 and related assays discriminate well in specialist cohorts, and the roadmap for clinical adoption has been written out explicitly by task forces.⁸⁹ The counter-argument is equally explicit: performance in unselected primary-care populations, cutoff transferability across platforms, and the absence of outcome data in asymptomatic people remain unresolved.¹⁰ Treat any proteomic risk score you get back as a research-grade number, and take interpretation of it to a clinician rather than acting on it.

Storing it so you can compute on it

Keep one long-format table, one row per measurement: date, analyte, value, unit, loinc, method, lab, fasting_hours, notes. Wide formats break the moment a lab renames a column or changes units. Store lab PDFs alongside, but parse to the table. From there, computing RCV between consecutive draws, fitting a per-analyte linear trend with statsmodels, and flagging points outside your personal ±2 SD band is twenty lines of pandas. Keep raw molecular data in its native formats: CRAM plus gVCF for the genome, quant.sf or a gene-level counts matrix for RNA-seq, NPX or concentration tables for proteomics, and raw 5-minute CGM exports rather than the vendor’s summary.

The point of this structure is that an analysis agent can read it. A single PDF of last week’s panel is nearly uninterpretable without your history. A five-year table with units and methods attached is a different object entirely.

Questions people also ask

What is a full biomarker test? There is no standard definition, which is why the phrase shows up in marketing. In practice it means a broad blood panel of 60 to 100 analytes. The panel covers the areas below.

• lipids
• glycemia
• liver
• kidney
• thyroid
• iron
• inflammation
• CBC

A full molecular characterization would add genome, transcriptome, proteome, and continuous sensor data.

How do I check my biomarkers? Order through a clinician or a direct-access lab, standardize the draw conditions above, and repeat at a fixed interval (quarterly is a reasonable default for an adult tracking a specific question, annually for baseline surveillance). Export results to a structured table rather than accumulating PDFs. Interpretation of any abnormal result belongs with a physician.

What are the three inflammatory markers? The three usually meant are C-reactive protein (hs-CRP for cardiovascular-range measurement), erythrocyte sedimentation rate, and interleukin-6. Procalcitonin and ferritin get used as acute-phase markers too. All of them rise nonspecifically with infection, injury, and recent hard exercise, so timing matters more than assay choice.

Which biomarker company is the best? Ask instead which one returns raw, machine-readable data with assay methods and LOINC codes attached, uses the same platform run-over-run, and lets you export history. Panels that only return a PDF and a proprietary “score” cannot be compared against your own prior results, which is the only comparison that carries much information.

Footnotes

1. Najib Aziz, Roger Detels, Joshua J. Quint, et al. Biological variation of immunological blood biomarkers in healthy individuals and quality goals for biomarker tests. BMC Immunology, 2019. https://doi.org/10.1186/s12865-019-0313-0 ↩

2. Joanne Kotsopoulos, Shelley S. Tworoger, Hannia Campos, et al. Reproducibility of Plasma, Red Blood Cell, and Urine Biomarkers among Premenopausal and Postmenopausal Women from the Nurses’ Health Studies. Cancer Epidemiology, Biomarkers & Prevention, 2010. https://doi.org/10.1158/1055-9965.epi-09-1318 ↩

3. Evelyn Barron, Jose Lara, Martin White, et al. Blood-Borne Biomarkers of Mortality Risk: Systematic Review of Cohort Studies. PLOS ONE, 2015. https://doi.org/10.1371/journal.pone.0127550 ↩

4. Robert L Holland. What makes a good biomarker?. Advances in Precision Medicine, 2016. https://doi.org/10.18063/apm.v1i1.17 ↩

5. Michael W. Ostler, James H. Porter, Orfeu M. Buxton. Dried Blood Spot Collection of Health Biomarkers to Maximize Participation in Population Studies. Journal of Visualized Experiments, 2014. https://doi.org/10.3791/50973-v ↩

6. Daniel Sim, Michael C. Brothers, Joseph M. Slocik, et al. Biomarkers and Detection Platforms for Human Health and Performance Monitoring: A Review. Advanced Science, 2022. https://doi.org/10.1002/advs.202104426 ↩

7. Samir M. Hanash, Christina S. Baik, Olli Kallioniemi. Emerging molecular biomarkers—blood-based strategies to detect and monitor cancer. Nature Reviews Clinical Oncology, 2011. https://doi.org/10.1038/nrclinonc.2010.220 ↩

8. Davide Angioni, J. Delrieu, O. Hansson, et al. Blood Biomarkers from Research Use to Clinical Practice: What Must Be Done? A Report from the EU/US CTAD Task Force. The Journal of Prevention of Alzheimer’s Disease, 2022. https://doi.org/10.14283/jpad.2022.85 ↩

9. Oskar Hansson, Kaj Blennow, Henrik Zetterberg, et al. Blood biomarkers for Alzheimer’s disease in clinical practice and trials. Nature Aging, 2023. https://doi.org/10.1038/s43587-023-00403-3 ↩

10. Willem A Van Gool, J A Floria Siebrand, Carol Brayne, et al. Evidence gap in blood biomarkers for Alzheimer’s disease. BMJ, 2025. https://doi.org/10.1136/bmj-2025-084781 ↩