How to Test Your Biomarkers and Read the Results Yourself

Oak
biomarkers blood-testing lab-data measurement longitudinal
Two studio-lit mounts of the same iridescent winged reptile on black plinths, their translucent throats and wings glowing with amber vessels, one brighter.

A biomarker test is a quantitative measurement of a specific molecule in a specific matrix (serum, plasma, whole blood, urine, tissue) by a specific method, reported with units and a reference interval. To check yours: order a panel from a lab that will give you the numeric results, not a grade. In the US, cash-price direct-to-consumer ordering through a CLIA lab runs roughly $10–40 per routine chemistry or lipid analyte, $40–150 for specialty immunoassays like Lp(a), insulin, or IGF-1, and $500–$3,000 for bundled consumer platforms that wrap 50–100 analytes plus an interface. Insurance typically covers biomarker testing when there is a diagnosis code and a guideline behind it (lipids in a cardiovascular workup, tumor markers in oncology) and typically does not cover screening panels you order for yourself out of curiosity. Interpretation of any result that could change your care belongs with a clinician. What follows is about measurement.

What a biomarker test measures, precisely

A result like LDL-C 118 mg/dL is the output of a pipeline with four layers, and each one has an error term:

  1. Pre-analytical. Fasting state, posture at draw, tourniquet time, tube type (EDTA vs serum separator vs citrate), time to centrifugation, freeze-thaw cycles. Potassium rises with hemolysis. Glucose falls ~5–7% per hour in an uncentrifuged tube without fluoride. Standing up for 30 minutes can drop plasma volume enough to shift albumin-bound analytes several percent.
  2. Analytical. The method matters as much as the number. LDL-C can be calculated (Friedewald, invalid above ~400 mg/dL triglycerides), estimated (Martin-Hopkins), or measured directly. Testosterone by immunoassay and by LC-MS/MS disagree materially at low concentrations. Vitamin D, thyroglobulin, and insulin all have method-specific biases large enough to matter across labs.
  3. Biological. Within-subject variation over days to weeks, independent of any change in your physiology’s set point.
  4. Interpretive. The reference interval is the central 95% of a reference population measured on that platform, not a target and not a decision limit. Population-based intervals and clinical decision limits are different objects, and the literature on introducing new markers into routine practice is blunt that traceability, method comparison, and interval derivation are the hard parts, not the discovery.1

Any downstream analysis you do that ignores layers 1–3 will produce noise you mistake for signal.

Get the data out in a computable form

Ask for three things from whichever lab you use, and treat the third as non-negotiable:

  • The numeric result with units and the lab’s reference interval.
  • The method and platform (e.g. “immunoturbidimetric, Roche cobas c502” vs “nephelometric”).
  • A structured export: HL7 v2 ORU, FHIR Observation JSON, or at minimum a CSV with LOINC codes.

LOINC is what makes multi-year, multi-lab data joinable. 2093-3 is total cholesterol in serum/plasma, mass/volume. 4548-4 is HbA1c by the IFCC-standardized measure. Without codes you end up string-matching “Cholesterol, Total” against “CHOL TOTAL” against “Cholesterol (total), serum” and silently dropping half your history.

A minimal working schema we use:

draw_id, collected_at (ISO 8601 with timezone), fasting_hours,
loinc, analyte, value, unit, method, instrument,
lab_id, ref_low, ref_high, flag

Store the raw value and unit as reported. Do unit conversion at query time, never at ingest. mg/dL vs mmol/L for glucose (÷18.016), mg/dL vs mmol/L for cholesterol (÷38.67), ng/mL vs nmol/L for 25-OH-D (×2.496): baking a conversion into stored data is the single most common way people corrupt their own lab history. If your only source is PDFs, OCR them and hand-verify every extracted number against the PDF for the first year of data. Tesseract will read 1.8 as 18 often enough to matter.

The math that makes a single result interpretable

Two draws differ. Is that a change or is it noise? The reference change value answers it:

RCV = Z × √2 × √(CVa² + CVi²)

where CVa is analytical imprecision (ask the lab for the CV at your concentration, or use their published method CV) and CVi is within-subject biological variation. For a two-sided 95% statement, Z√2 ≈ 2.77.

Worked examples, using commonly cited biological variation estimates:

  • HbA1c: CVi ≈ 2%, CVa ≈ 2% → RCV ≈ 8%. A move from 5.4% to 5.6% is 3.7%, inside the noise. From 5.4% to 6.0% is 11%, outside it.
  • LDL-C: CVi ≈ 8%, CVa ≈ 3% → RCV ≈ 24%. Most “my LDL dropped 15 points” stories are a single draw against a single draw with no replication.
  • hs-CRP: CVi is large, commonly estimated above 40% → RCV well over 100%. A single CRP tells you almost nothing about your baseline. Three draws, median, and you have something.
  • TSH: CVi ≈ 19% → RCV ≈ 55%.
  • Serum creatinine: CVi ≈ 4.5%, CVa ≈ 3% → RCV ≈ 15%, which is why small eGFR wobbles are usually nothing.

The companion number is the index of individuality, CVi / CVg, where CVg is between-subject variation. When it is below about 0.6, the population reference interval is nearly useless for you and your own prior results are the correct comparator. Creatinine, calcium, and most of the thyroid panel are in that regime. This is the entire argument for longitudinal self-measurement over annual one-shot panels: for low-II analytes you can sit inside the reference interval while having moved 40% from your own baseline.

Practical consequence: for any marker you intend to act on, draw it twice, 1–2 weeks apart, same lab, same fasting state, same time of day, and use the mean as your baseline. Then compute RCV against that baseline going forward.

What we would order

Opinionated, and assuming you want maximum information per dollar rather than a long list:

  • Lipids with apoB and Lp(a). ApoB counts atherogenic particles directly and has lower CVi than calculated LDL-C. Lp(a) is largely genetic, so one good measurement (ideally in nmol/L on a non-size-sensitive assay) is close to a lifetime number.
  • CMP, CBC with differential, liver enzymes. Cheap, standardized, and the backbone of any trend model.
  • HbA1c plus fasting insulin and glucose. HbA1c has low CVi and integrates ~90 days, which makes it the most trend-friendly metabolic marker you own. Note it is distorted by hemoglobin variants and altered red cell lifespan.
  • Ferritin plus transferrin saturation, TSH with free T4, 25-OH-D, uric acid, hs-CRP (×3).
  • If you can pair blood with continuous glucose data, do. A CGM turns one fasting glucose into ~1,400 readings per day and lets you estimate postprandial excursions that HbA1c averages away.

Beyond routine chemistry, the useful frontier is multi-analyte panels rather than single markers. Metabolomic panels have been evaluated for lung adenocarcinoma detection in serum,2 multi-system endocrine, immune, inflammatory, and metabolic panels have been studied in obstructive sleep apnea,3 and blood-based molecular panels have been tested for identifying risk before clinical onset in psychiatric disease.4 These are research-grade signals, not screening tools you should act on alone.

Standardization, and why switching labs breaks your series

Two labs measuring the same tube can return different numbers, and the difference is not random. Before you merge histories, look for a method comparison: Deming or Passing-Bablok regression between the platforms, with bias reported at clinically relevant concentrations. If you cannot find one, keep the series separate and label the discontinuity. IFCC-standardized HbA1c and isotope-dilution-MS-traceable creatinine travel between labs reasonably well. Immunoassay-based insulin, ferritin, free T4, and most tumor markers do not. Adopting a new marker into routine use requires exactly this kind of traceability and comparability work.1

Where single blood markers are doing real diagnostic work

The cases where a blood biomarker genuinely substitutes for a harder test are worth knowing, because they set the bar. Blood markers have been evaluated for distinguishing ischemic from hemorrhagic stroke in the acute setting, where the decision window is minutes.5 Plasma biomarkers in Alzheimer’s diagnosis have been compared against conventional CSF and imaging workups on cost-effectiveness grounds.6 Blood-based colorectal cancer screening has been modeled against colonoscopy and stool testing, where the tradeoff is adherence against per-test sensitivity.7 In oncology, blood-based biopsy has been compared with tissue biopsy on cost and outcomes for molecular profiling,8 and testing approach correlates with treatment selection and total cost of care in advanced disease.9 In post-transplant medicine, panels are used to predict outcomes rather than to make a diagnosis.10

None of those is a wellness panel. They are validated in defined populations against defined endpoints, and the analytical requirements are stricter than anything a consumer platform is holding itself to.

Failure modes to watch for

  • A single draw treated as a baseline. Covered above: replicate.
  • Reference interval treated as a target. The interval describes a population’s central 95%, including people with undiagnosed disease.
  • Multiple comparisons. Order 80 analytes with 95% intervals and you should expect roughly four flagged values in a person with nothing wrong. Flag counts are not information.
  • Unit and method drift across years, merged silently.
  • Non-fasting draws mixed into a fasting series, especially triglycerides and insulin.
  • Acting on a marker that is moving because you have a cold. CRP, ferritin, and white count are acute-phase reactants. Wait two to three weeks after any illness.
  • Skipping the clinician on anything abnormal and persistent. A confirmed out-of-range result with a plausible mechanism needs a physician, not a dashboard.

Questions people also ask

How much do biomarker tests cost? Individual routine analytes are $10–40 at cash-price CLIA labs, specialty immunoassays $40–150, and bundled consumer panels $500–$3,000 depending on breadth and whether a clinician review is included. Cost-effectiveness analyses in specific clinical settings, such as blood biomarkers versus conventional workup in Alzheimer’s diagnosis, evaluate a different question: total cost per correct diagnosis, not the sticker price of a draw.6

How can I check my biomarkers? Order through a physician, a direct-to-consumer lab portal, or a consumer platform, then insist on the structured export with LOINC codes and the method. Draw twice, a week or two apart, fasting, same time of day, same lab, and keep every result in one table with collection timestamps.

Does insurance cover biomarker testing? Usually yes with a diagnosis or guideline indication, usually no for self-ordered screening panels. In oncology the coverage question is entangled with which testing approach is used, and testing approach tracks with both treatment selection and total cost of care.9

What are biomarker tests? Measurements of molecules (proteins, metabolites, nucleic acids, cells) in a body fluid or tissue, used to estimate the state of some biological process. The word covers everything from a $12 potassium to a multi-analyte panel with a trained classifier behind it, and the analytical rigor varies by orders of magnitude across that range.

Oak builds longitudinal molecular profiles of individuals: whole-genome sequencing, RNA sequencing, proteomics, blood biomarkers, and continuous glucose data, integrated into one model of you. Build your profile.

Footnotes

  1. Catharine Sturgeon, Robert Hill, Glen L. Hortin, et al. Taking a new biomarker into routine use – A perspective from the routine clinical biochemistry laboratory. PROTEOMICS – Clinical Applications, 2010. https://doi.org/10.1002/prca.201000073 ↩ ↩2

  2. Johannes F. Fahrmann, Kyoungmi Kim, Brian C. DeFelice, et al. Investigation of Metabolomic Blood Biomarkers for Detection of Adenocarcinoma Lung Cancer. Cancer Epidemiology, Biomarkers & Prevention, 2015. https://doi.org/10.1158/1055-9965.epi-15-0427 ↩

  3. Wesley Elon Fleming, Aliya Ferouz-Colborn, Michael K. Samoszuk, et al. Blood biomarkers of endocrine, immune, inflammatory, and metabolic systems in obstructive sleep apnea. Clinical Biochemistry, 2016. https://doi.org/10.1016/j.clinbiochem.2016.05.005 ↩

  4. M K Chan, M-O Krebs, D Cox, et al. Development of a blood-based molecular biomarker test for identification of schizophrenia before disease onset. Translational Psychiatry, 2015. https://doi.org/10.1038/tp.2015.91 ↩

  5. Alejandro Bustamante, Anna Penalba, Cyrille Orset, et al. Blood Biomarkers to Differentiate Ischemic and Hemorrhagic Strokes. Neurology, 2021. https://doi.org/10.1212/wnl.0000000000011742 ↩

  6. Kenta Noda, Yeongjoo Lim, Rei Goto, et al. Cost-effectiveness comparison between blood biomarkers and conventional tests in Alzheimer’s disease diagnosis. Drug Discovery Today, 2024. https://doi.org/10.1016/j.drudis.2024.103911 ↩ ↩2

  7. John M. Inadomi. Cost-Effectiveness of Blood-Based Biomarkers for Colorectal Cancer Screening—An Ounce of Prevention Is Worth a Pound of Cure. JAMA Network Open, 2023. https://doi.org/10.1001/jamanetworkopen.2023.43346 ↩

  8. A Arnaud. Costs and Outcomes Comparison of Tissue and Blood Based Biopsies for the Purpose of Biomarker Testing. Value in Health, 2016. https://doi.org/10.1016/j.jval.2016.03.1556 ↩

  9. Stacey DaCosta Byfield, Bela Bapat, Laura Becker, et al. Biomarker Testing Approaches, Treatment Selection, and Cost of Care Among Adults With Advanced Cancer. JAMA Network Open, 2025. https://doi.org/10.1001/jamanetworkopen.2025.19963 ↩ ↩2

  10. Sophie Paczesny. Biomarkers for posttransplantation outcomes. Blood, 2018. https://doi.org/10.1182/blood-2018-02-791509 ↩