How to Analyze Your Own Blood Biomarkers

A single blood biomarker result is a noisy point estimate of a quantity that moves on its own. The useful unit of analysis is not one value against a population reference interval, it is a series of values from you against your own history, with the analytical and biological variation of that particular assay attached. Most of the value in blood biomarker analysis comes from three things. The first is measuring the same analyte on the same platform repeatedly. The second is knowing the reference change value for that analyte. The third is controlling the pre-analytical conditions (fasting state, time of day, posture and tube type) so that the differences you see are differences in you.

That framing changes what you do with the data. You stop asking “is 4.7 normal?” and start asking “is the move from 3.9 to 4.7 larger than this assay and this analyte produce on their own?”

Start with the variation numbers and not the reference range

Every analyte has three variation components. Analytical imprecision (CV_A) is the assay’s own noise. Within-subject biological variation (CV_I) is how much you fluctuate around your personal set point. Between-subject variation (CV_G) is how much set points differ across people. You need CV_A and CV_I before a delta means anything.

The reference change value is:

RCV = 2^(1/2) * Z * sqrt(CV_A^2 + CV_I^2)

with Z = 1.96 for a two-sided 95% threshold, or 1.65 for a one-sided 95% test of “did it go up.” For an analyte with CV_A = 3% and CV_I = 6%, RCV ≈ 18.6%. A move from 100 to 112 is inside the noise. A move from 100 to 130 is not.

The other number worth having is the index of individuality, CV_I / CV_G. When it is below about 0.6, the population reference interval is nearly useless for you: your personal range is narrow relative to the spread across people, so you can be well inside “normal” while sitting far off your own baseline. Many immunological and protein markers fall here. A study of healthy individuals sampled repeatedly reported within-subject variation and derived analytical quality goals for a panel of immunological blood markers, which is the kind of source you want when you are deciding what delta to take seriously.¹

Practically: for each analyte you track, store cv_a (ask the lab, or estimate it from split samples) and cv_i (literature), and compute RCV programmatically rather than eyeballing the trend line.

Control pre-analytics or throw the series away

Most “surprising” biomarker moves are pre-analytical. The ones that bite hardest:

• Posture. Standing shifts water out of plasma and raises protein-bound and cell-associated analytes by roughly 5–10%. Sit for 10 minutes before the draw, every time.
• Tourniquet time. Over a minute of stasis inflates potassium, lactate, and protein measures.
• Fasting state and time of day. Cortisol, testosterone, and iron have large diurnal swings. Triglycerides and glucose are dominated by the last meal.
• Tube type and processing delay. Serum versus EDTA plasma versus citrate gives different values for the same analyte. Delayed centrifugation causes hemolysis and potassium leak from red cells.
• Hemolysis, icterus, lipemia indices. Ask for them. They are the first thing to check on an outlier.

Fix the protocol: same lab, same time of day, same fasting window, same posture and same tube. Write it down and repeat it. If you change labs or platforms mid-series, treat it as a new series and re-baseline. Immunoassays from different manufacturers are frequently not interchangeable, and the difference can exceed anything biological you would be looking for.

Which markers earn their place

Be skeptical of any marker sold on the strength of a hazard ratio. A large review of the most popular blood biomarkers for cardiovascular disease found that, added on top of conventional risk factors, most contribute minimal or null improvement in discrimination.² A marker can be strongly associated with an outcome across a population and still move almost no one’s individual risk estimate. The quantities to ask about are incremental C-statistic, net reclassification, and calibration. Relative risk is not among them.

Alzheimer’s blood markers are the current test case. Plasma assays for amyloid and phosphorylated tau have moved quickly from discovery into clinical use, and the field has been explicit about the work needed to get from candidate to validated test: standardized pre-analytics, multi-cohort replication, and defined intended-use populations.³ Earlier plasma protein panels reported strong discrimination that did not hold up under independent replication.⁴ A 2025 analysis argues that substantial evidence gaps remain, especially around how these tests behave in unselected, non-specialist populations where prior probability is low.⁵ That is the general lesson: performance measured in a memory clinic does not transfer to a healthy 40-year-old.

Cancer detection markers deserve the same scrutiny. Multi-cancer and colorectal blood tests are progressing, and cost-effectiveness modeling suggests blood-based colorectal screening can be worthwhile mainly when it reaches people who would otherwise not be screened at all, and when adherence to follow-up colonoscopy is high.⁶ A positive result on any screening blood test is a referral to a clinician for confirmatory imaging or endoscopy, never a conclusion.

Plasma as a discovery matrix is genuinely hard. The dynamic range of plasma protein concentrations spans more than ten orders of magnitude, and albumin and immunoglobulins dominate the mass, which is why untargeted plasma proteomics requires depletion or enrichment and why candidate markers so often fail to replicate.⁷ Circulating miRNA panels have a parallel problem: results depend heavily on whether you used serum or plasma, on hemolysis, on extraction chemistry, and on normalization strategy, which is a large part of why reported disease signatures disagree between labs.⁸ Blood transcriptomics carries its own caveat: expression in blood correlates imperfectly with expression in other tissues, so a blood RNA signal is evidence about blood unless a specific tissue correspondence has been demonstrated.⁹

A structure that works

Keep the data in long format, one row per measurement:

date, analyte, value, unit, loinc, lab_id, platform, method,
fasting_hours, draw_time, tube, hemolysis_index, cv_a, cv_i

LOINC codes matter more than they look. They are what let you join results from different labs without silently merging two assays that are not the same test. Most US lab PDFs and HL7/FHIR exports carry them. For ferritin alone you will find distinct codes by specimen and method.

Analysis steps we would run, in order:

1. Unit normalization and platform tagging. Convert to a canonical unit per LOINC. Never merge across platforms without a bridging study.
2. Flag pre-analytical violations. Non-fasting sample on a triglyceride series, hemolysis index above threshold on potassium or LDH: exclude or annotate, do not silently average.
3. Compute RCV per consecutive pair and per baseline-to-current. Report the delta as a multiple of RCV.
4. Estimate the personal set point once you have 4–5 clean draws. Mean plus or minus 1.96 × sqrt(CV_A² + CV_I²) gives your homeostatic range.
5. Trend test. With 6+ points, a Theil-Sen slope with a Mann-Kendall test is less sensitive to a single outlier draw than OLS, and you are not going to have enough points for anything fancier.
6. Multiplicity. A 60-analyte panel at a 95% interval produces about three flagged values on an unremarkable day. Expect them. Do not chase them individually.

Frame the whole thing as test evaluation rather than measurement: a marker is only useful within a stated intended use, in a defined population, with a decision that changes based on the result.¹⁰ If no decision changes, the measurement is data collection, which is a fine reason to do it as long as you are clear that is what you are doing.

Questions people also ask

Are blood biomarker tests worth it? Repeated measurement of a focused panel under controlled conditions is worth a great deal, because it gives you personal set points and detectable deltas. One-off broad panels in an asymptomatic person mostly generate false flags and follow-up costs, and most popular single markers add little predictive information beyond standard risk factors.²

How long do results take? Routine chemistry, CBC, and lipids are typically 1–3 days. Specialized immunoassays and send-out tests run 5–14 days. Research-grade proteomics and sequencing run weeks.

What does a blood biomarker test show? The concentration or activity of specific molecules at the moment of the draw. That covers metabolites, enzymes and hormones, along with proteins, cell counts and nucleic acids. It shows a state, not a trajectory, until you have several of them.

What blood biomarkers indicate cancer? No blood marker diagnoses cancer. Markers like PSA, CA-125, CEA, and AFP are used for monitoring known disease or as one input among several, and newer cell-free-DNA tests are screening tools whose value depends on adherence to confirmatory follow-up.⁶ Any abnormal result belongs with a clinician.

What are the risks? Physical risk is minimal (bruising, rarely vasovagal syncope). The real risks are false positives leading to invasive follow-up, incidental findings with no clear action, and over-interpreting a single value that is inside your own biological variation.

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. John P.A. Ioannidis, Ioanna Tzoulaki. Minimal and Null Predictive Effects for the Most Popular Blood Biomarkers of Cardiovascular Disease. Circulation Research, 2012. https://doi.org/10.1161/res.0b013e31824da8ad ↩ ↩²

3. Sid E. O’Bryant, Michelle M. Mielke, Robert A. Rissman, et al. Blood‐based biomarkers in Alzheimer disease: Current state of the science and a novel collaborative paradigm for advancing from discovery to clinic. Alzheimer’s & Dementia, 2016. https://doi.org/10.1016/j.jalz.2016.09.014 ↩

4. James D. Doecke. Blood-Based Protein Biomarkers for Diagnosis of Alzheimer Disease. Archives of Neurology, 2012. https://doi.org/10.1001/archneurol.2012.1282 ↩

5. 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 ↩

6. 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 ↩ ↩²

7. Jon M. Jacobs, Joshua N. Adkins, Wei-Jun Qian, et al. Utilizing Human Blood Plasma for Proteomic Biomarker Discovery. Journal of Proteome Research, 2005. https://doi.org/10.1021/pr0500657 ↩

8. Christina Backes, Eckart Meese, Andreas Keller. Specific miRNA Disease Biomarkers in Blood, Serum and Plasma: Challenges and Prospects. Molecular Diagnosis & Therapy, 2016. https://doi.org/10.1007/s40291-016-0221-4 ↩

9. Brandi Rollins, Maureen V. Martin, Ling Morgan, et al. Analysis of whole genome biomarker expression in blood and brain. American Journal of Medical Genetics Part B: Neuropsychiatric Genetics, 2010. https://doi.org/10.1002/ajmg.b.31062 ↩

10. Andrea R. Horvath, Sarah J. Lord, Andrew StJohn, et al. From biomarkers to medical tests: The changing landscape of test evaluation. Clinica Chimica Acta, 2014. https://doi.org/10.1016/j.cca.2013.09.018 ↩