What a blood biomarkers test measures, and how to read your own results

A blood biomarkers test is a set of quantitative measurements made on a serum, plasma, or whole-blood sample: concentrations of metabolites, proteins, lipoproteins, hormones, and cell counts. Each result is a number with units, a method, and an uncertainty. The useful ones tell you something about a physiological process you cannot observe directly. The main skill in reading them is separating a real change in your body from measurement noise and normal day-to-day biological fluctuation.

That skill matters more than panel size. A 100-analyte panel run once gives you a hundred numbers of unknown personal baseline. A 30-analyte panel run four times a year gives you trajectories, and trajectories are what you can act on with a clinician.

What the common analytes measure

Blood analytes fall into a few functional groups, and it helps to know which process each one reports on.

Lipids and lipoproteins report on atherogenic particle burden. Total cholesterol, HDL-C, and triglycerides are measured enzymatically. LDL-C is usually calculated, not measured, which makes it unreliable when triglycerides are high. Apolipoprotein B (ApoB) counts atherogenic particles directly by immunoturbidimetric assay, and lipoprotein(a) is measured separately because it is largely genetically set and does not track with the rest of the lipid panel.

Glycemic markers report on glucose exposure over different windows. Fasting glucose covers minutes, fructosamine roughly two to three weeks, and HbA1c roughly the preceding two to three months weighted toward the recent weeks. HbA1c is a proxy that assumes normal red cell lifespan. Anemia, hemoglobin variants, and recent blood loss distort it.

Organ function markers are mostly enzymes and clearance products. ALT and AST report hepatocyte turnover. Creatinine and cystatin C are used to estimate glomerular filtration rate, and cystatin C is less affected by muscle mass. Inflammation is usually captured by high-sensitivity C-reactive protein, an acute-phase protein made in the liver.

Cardiac injury markers are a separate class with a different purpose. High-sensitivity cardiac troponin I or T detects myocyte necrosis at nanogram-per-liter concentrations, and the assay generation matters enormously for interpretation ¹. These are acute-care tests interpreted against serial draws hours apart. A single outpatient troponin is not a screening test, and any abnormal result belongs with a clinician immediately.

Analytical noise versus biological variation

Every result carries two independent sources of scatter. Analytical imprecision (CVa) is the assay’s own reproducibility, typically 2 to 4 percent for a well-run chemistry analyte and worse for immunoassays. Within-subject biological variation (CVi) is how much your true value oscillates around your own set point. The EFLM Biological Variation Database catalogs published estimates for hundreds of analytes.

The two combine into the reference change value, which is the threshold a difference between two of your own results must exceed before it means anything:

RCV = 2^0.5 * Z * sqrt(CVa^2 + CVi^2)

Use Z = 1.96 for a two-sided 95 percent criterion. For an analyte with CVa of 3 percent and CVi of 6 percent, the RCV is about 19 percent. A move from 100 to 110 is noise. For high-sensitivity CRP, where CVi is very large, the RCV can exceed 100 percent. That is why a single CRP of 3.1 mg/L followed by 1.4 mg/L tells you almost nothing.

Compute the RCV for each analyte you track and store it alongside the result. It converts a wall of numbers into a short list of changes worth discussing.

Prediction is weaker than most panels imply

There is a systematic gap between a biomarker being associated with disease and a biomarker improving prediction for an individual. A large review of the most popular cardiovascular blood biomarkers found that added predictive value over standard risk factors was minimal or null for most of them, even when the association with outcomes was statistically solid ². Association in a cohort and discrimination for one person are different quantities.

The same pattern appears in cancer detection. Blood-based gene expression panels for colorectal cancer have been derived with reported sensitivity and specificity that look strong in the discovery set, then shrink under independent validation ³. Careful work on colorectal adenoma and cancer markers has made the point explicitly, showing that candidate markers must be validated in samples collected the same way as the intended clinical use ⁴.

Read this as a discipline rather than a reason to skip testing. Track analytes with clear mechanistic interpretation and established decision thresholds. Treat novel multi-analyte scores as research-grade until validated in a population like you.

Preanalytics dominate your error budget

Most surprising results are collection artifacts. Fix the collection protocol before you argue with the number.

Fast 10 to 12 hours if you are measuring triglycerides, and hold the fast constant across draws even for analytes that do not strictly require one. Draw at the same time of day. Cortisol, testosterone, and iron all follow diurnal rhythms with swings large enough to swamp any real trend. Avoid vigorous exercise for 48 hours before a draw if you are measuring creatine kinase, ALT, AST or creatinine.

Posture changes plasma volume. Standing for 15 minutes before a draw can raise protein-bound analytes by several percent relative to a supine draw. Prolonged tourniquet time does the same through hemoconcentration. Hemolysis during draw or transport raises potassium and LDH and interferes with several immunoassays, and the lab should report a hemolysis index.

Remote and mail-in collection adds another layer. Dried blood spot and self-collected capillary samples are workable for some analytes, and population research has used remote collection to track biomarkers longitudinally at scale ⁵. Capillary and venous values are not interchangeable for every analyte, so do not mix collection modes inside one time series. Point-of-care biosensor devices are improving quickly, but their imprecision is generally wider than a central laboratory analyzer ⁶.

Storing and analyzing results

Get results as structured data rather than PDFs. The cleanest path is a lab portal that exposes HL7 FHIR Observation resources. Each result then carries a LOINC code, a value, a unit in UCUM notation and a reference range. If you only have PDFs, transcribe once into a tidy table and never transcribe again.

A minimal schema that has served us well:

date, loinc, analyte, value, unit, method, lab_id, ref_low, ref_high, fasting, time_of_day

Keep method and lab_id. Changing labs or assay platforms creates step changes that look like biology. Keep LOINC codes because analyte names are ambiguous across labs. Verify each code at loinc.org rather than trusting a mapping table. HbA1c mass fraction is 4548-4, serum glucose is 2345-7, and serum creatinine is 2160-0.

For analysis, resist fitting anything elaborate to eight points. Plot the series, overlay your personal RCV band around the previous value, and flag crossings. A Theil-Sen slope estimator handles the small-n, outlier-prone case better than ordinary least squares. When the same direction of change appears across two mechanistically linked analytes, take it more seriously than a single outlier.

Reference intervals are the central 95 percent of a reference population, which means roughly one in twenty healthy results falls outside by construction. Order 40 analytes and expect two flags with no underlying pathology. Decision thresholds, where they exist, are different objects derived from outcome data. Interpretation of any out-of-range result, and any decision that follows, belongs with a clinician who knows your history.

Questions people also ask

How often should I repeat a panel?

Match the interval to the analyte’s biology. HbA1c integrates about three months, so quarterly is the fastest meaningful cadence. Lipids respond to change within four to six weeks. Lipoprotein(a) is largely genetic and needs measuring once, with a confirmation.

Are at-home finger-stick tests as good as a venous draw?

For some analytes they are close, for others they are not, and the difference is analyte-specific. Capillary samples can differ systematically from venous ones, and imprecision is usually wider. Keep one collection mode per time series so your trend is not confounded by method ⁶.

Why did two labs give me different numbers for the same analyte?

Different assay platforms and calibrators produce different values. This is especially true for immunoassays such as thyroid hormones, ferritin, and testosterone. Bias between platforms can exceed the reference change value. Stay with one lab if you want an interpretable trend.

Does a large panel find things a small one misses?

Sometimes, at the cost of more false flags from the 95 percent reference interval convention. We would rather run a focused panel more often than a wide panel once. Repeat measurement is what separates signal from noise.

Footnotes

1. Ioan Tilea, Andreea Varga, Razvan Constantin Serban. Past, Present, and Future of Blood Biomarkers for the Diagnosis of Acute Myocardial Infarction—Promises and Challenges. Diagnostics, 2021. https://doi.org/10.3390/diagnostics11050881 ↩

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 ↩

3. Mark Han, Choong Tsek Liew, Hong Wei Zhang, et al. Novel Blood-Based, Five-Gene Biomarker Set for the Detection of Colorectal Cancer. Clinical Cancer Research, 2008. https://doi.org/10.1158/1078-0432.ccr-07-1801 ↩

4. Lawrence C. LaPointe, Susanne K. Pedersen, Robert Dunne, et al. Discovery and Validation of Molecular Biomarkers for Colorectal Adenomas and Cancer with Application to Blood Testing. PLoS ONE, 2012. https://doi.org/10.1371/journal.pone.0029059 ↩

5. Kumar B. Rajan, Neelum T. Aggarwal, Elizabeth A. McAninch, et al. Remote Blood Biomarkers of Longitudinal Cognitive Outcomes in a Population Study. Annals of Neurology, 2020. https://doi.org/10.1002/ana.25874 ↩

6. Mengxing Ouyang, Dandan Tu, Lin Tong, et al. A review of biosensor technologies for blood biomarkers toward monitoring cardiovascular diseases at the point-of-care. Biosensors and Bioelectronics, 2021. https://doi.org/10.1016/j.bios.2020.112621 ↩ ↩²