Best Biomarker Testing Companies, and How to Choose One

There is no single best biomarker testing company, and the ranking changes depending on what you are optimizing for. If you want breadth of analytes at a fixed annual price, Function Health and Superpower are the obvious candidates. If you want control over which assays are run and cash pricing per test, order directly from Quest or Labcorp through a consumer ordering service. Ulta Lab Tests, Marek Health, and Empirical Health all work for this. If you want data you can compute on, the deciding factor is not the vendor’s dashboard but whether they return structured results with LOINC codes and whether they hold assay methods constant across draws. Our recommendation for most technically fluent readers: order a defined panel through a direct-to-consumer front end on a single reference lab, keep the same lab and the same assay for every repeat, and export results as structured data from day one.

What a 100-biomarker panel contains

The marketing number is a count of analytes. It does not count independent signals. A typical 100-to-160 marker panel is assembled from a small number of standard clinical orders. A comprehensive metabolic panel contributes 14 analytes. A complete blood count with differential contributes another 15 to 25, depending on whether you count percentages and absolutes separately. A lipid panel contributes five, and a thyroid panel three or four. Add the following and you are already near 100.

• iron studies
• a handful of vitamins
• sex hormones
• hs-CRP
• HbA1c
• insulin
• urinalysis

That matters for interpretation. Neutrophil count and neutrophil percentage are not two pieces of evidence. Calculated LDL-C, non-HDL-C, and total cholesterol are algebraically linked. The number of genuinely independent axes in a 100-marker panel is closer to 20 or 30. Any product that promises insight proportional to analyte count is overstating what the panel can resolve.

The second consequence is statistical. Reference intervals are conventionally the central 95% of a healthy reference population, so each test flags roughly one healthy person in twenty. Run 100 of them and the expected number of out-of-range values in a perfectly healthy person is about five. Panels of this size produce false alarms by construction. Interpreting which flags matter is a clinical judgment, and persistent or large deviations belong in front of a physician rather than an LLM.

The criteria we use to judge a testing company

Analytical validity comes first. A biomarker becomes a usable medical test only when the assay, the intended population, and the decision it informs are specified together. Most consumer panels leave the last two implicit ¹. Ask which platform runs each analyte and what the coefficient of variation is at your concentration range. Vendors that cannot answer are reselling a black box.

Longitudinal consistency comes second, and it is the criterion most consumer products fail. A vendor that silently switches reference labs between your March and September draws has introduced a method shift you cannot distinguish from a biological change. The same problem has dominated the blood-based Alzheimer’s biomarker field, where pre-analytical handling and platform differences drove much of the early irreproducibility ². If you plan to track anything over years, pin the lab.

Data access comes third. You want per-analyte values with units, reference intervals, LOINC codes, collection timestamps, and the method name. A PDF is a dead end. An API returning HL7 FHIR Observation resources, or even a clean CSV with one row per analyte per draw, is what you need to build a time series.

Price is fourth. It does not come first. Annual memberships run roughly $400 to $500 for the broad panels. Ordering an equivalent set of cash-price tests through a lab aggregator often lands between $200 and $600, depending on how many specialty assays you include. The spread is small enough that assay quality should decide.

How the main options compare

Function Health and Superpower are the packaging layer. They order from Quest or Labcorp, wrap the results in a clinician review, and present a dashboard with trends. They are a reasonable starting point if you have never had a broad panel run and want one decision instead of thirty. The limitation is control: you take the panel they define, and substituting a better assay for a given analyte is generally not an option.

Ulta Lab Tests, Marek Health, and Empirical Health sit closer to the lab. You choose individual test codes, pay per test, and draw at a Quest or Labcorp patient service center. This is our default. You can specify, for example, an isoform-independent Lp(a) assay reported in nmol/L rather than a mass assay in mg/dL. You can also skip analytes you will not act on.

Quest and Labcorp direct consumer portals are the same labs without the ordering layer. Coverage is good, pricing is transparent, and the result PDFs are clean. Both offer patient result APIs of varying quality, and Labcorp’s NMR LipoProfile remains the practical route to lipoprotein particle counts.

Specialty vendors occupy a different tier. Boston Heart and Cleveland HeartLab run extended cardiometabolic panels. Research-grade proteomics from Olink or SomaScan is now reachable through several services. Understand what those return: Olink reports NPX values on a log2 relative scale, and SomaScan reports aptamer RFU. Neither is an absolute concentration, and agreement between the two platforms for the same named protein is often poor. Treat proteomics as a discovery layer. Do not treat it as a clinical result.

Assay method matters more than the vendor

For several analytes, the method changes the number more than your physiology does. Testosterone by immunoassay is unreliable at low concentrations, where liquid chromatography tandem mass spectrometry is the accepted approach. 25-hydroxyvitamin D by LC-MS/MS resolves D2 and D3 separately, while immunoassays report a combined value with variable cross-reactivity. LDL-C is usually calculated, not measured, and the Friedewald equation degrades when triglycerides are elevated.

Within-person biological variation sets the floor on what a change means. Fasting glucose and total cholesterol vary by roughly 5% to 6% within a person across days. Triglycerides vary by around 20%. A single 15% move in triglycerides is noise. Three consecutive draws trending the same direction is signal. Collect at a fixed time of day and fasted. Note recent illness and exercise, because hs-CRP responds to both.

Getting your data out

Export at every draw, and keep the raw records rather than screenshots. A workable minimum schema is one row per result with the following fields.

• analyte
• loinc
• value
• unit
• ref_low
• ref_high
• method
• lab
• collected_at

Store it as Parquet or SQLite and version it. Use LOINC as the join key, because vendor display names drift. The same analyte appears as “Vitamin B12”, “B12”, and “Cobalamin” across three reports.

Once you have three or more draws, the analysis worth doing is simple. Plot each analyte against time with the reference interval shaded. Compute the within-person standard deviation and compare each new value against it, rather than against the population interval. Multi-analyte models built on standard blood panels can predict chronological age with useful accuracy, which is a reminder that ordinary panels carry more structure than single-marker reading extracts ³.

Is biomarker testing worth it

For general wellness panels, the answer is that the evidence for annual broad screening in asymptomatic adults is thin, and the value comes mostly from establishing your own baseline for later comparison. For specific clinical questions the picture is different. Plasma p-tau217 assays now match or exceed cerebrospinal fluid testing for Alzheimer’s pathology in reported accuracy ⁴. Circulating tumor DNA has moved from research assay to a tool for detecting residual disease and tracking treatment response ⁵. Blood-based colorectal cancer screening has been modeled as cost-effective relative to no screening, though it underperforms colonoscopy ⁶.

The distinction to carry is that a measured molecule becomes a useful test only after it has been qualified against a defined context of use ⁷. Most of what appears on a 100-marker panel has been qualified for diagnosis in symptomatic patients. It has not been qualified for optimization in healthy ones. Buy the panel for the baseline and the trend. Bring anything persistent or extreme to a clinician.

Footnotes

1. 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 ↩

2. Harald Hampel, Sid E. O’Bryant, José L. Molinuevo, et al. Blood-based biomarkers for Alzheimer disease: mapping the road to the clinic. Nature Reviews Neurology, 2018. https://doi.org/10.1038/s41582-018-0079-7 ↩

3. 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 ↩

4. Nicolas R. Barthélemy, Gemma Salvadó, Suzanne E. Schindler, et al. Highly accurate blood test for Alzheimer’s disease is similar or superior to clinical cerebrospinal fluid tests. Nature Medicine, 2024. https://doi.org/10.1038/s41591-024-02869-z ↩

5. Elizabeth L. Christie, Sarah-Jane Dawson, David D.L. Bowtell. Blood Worth Bottling: Circulating Tumor DNA as a Cancer Biomarker. Cancer Research, 2016. https://doi.org/10.1158/0008-5472.can-16-2281 ↩

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 ↩

7. Virginia B. Kraus. Biomarkers as drug development tools: discovery, validation, qualification and use. Nature Reviews Rheumatology, 2018. https://doi.org/10.1038/s41584-018-0005-9 ↩