How Advanced Biological Age is Calculated

There is no single lab test that directly measures biological age. Any biological age result is a model, and models differ enormously in how they are built. This guide explains how the Advanced BioAge Report is calculated, and what that method can and cannot tell you.

The Advanced BioAge Report is powered by VOLO Health’s modeling platform. We collect and process your blood work, and VOLO’s models turn those results into system-by-system age estimates.

We chose this platform for one reason above all others: the models are anchored to real health outcomes rather than to expert consensus about what “healthy” should look like. The model was never told which values are good. It learned that from data on what actually happened to hundreds of thousands of people over many years.

Step 1: Your blood work

Everything starts with measurement. Your sample is processed in a CLIA-certified, CAP-accredited laboratory, and the resulting biomarker values are the only inputs to your report.

The models draw on markers across several categories, including metabolic and glycemic markers, a full lipid picture, liver and kidney function, inflammatory markers, and blood count markers. Report completeness scales with panel breadth: the more of these markers your panel includes, the more confident each system result becomes.

Step 2: Removing the effect of age and sex

Age is by far the strongest predictor of most chronic disease, and sex matters substantially too. If a model does not handle this carefully, it mostly ends up rediscovering that older people get sick more often, which tells you nothing about you.

So the first modeling step builds a baseline expectation of disease prevalence as a function of age and sex, separately for each condition. Every individual in the training data is then measured against their own demographic baseline rather than against everyone else. What comes out is a relative signal: how does this person’s risk compare to what would be expected for someone of their age and sex?

This is why your report is explicitly peer-calibrated. Two people with identical lab values but different ages will receive meaningfully different reports, because the question being asked is always relative to your own demographic group.

Step 3: Learning from outcomes, not from age

The models were trained on two large bodies of data: a major long-term population health study following roughly half a million participants with linked medical records, plus a much larger set of US clinical records used for external validation and calibration.

For each health system, the outcomes being modeled are defined using standardized diagnostic code sets applied to those linked medical records, over a multi-year observation window following the original blood draw. In other words, the model looked at people’s biomarkers at one point in time, then looked at what conditions they went on to develop years later, and learned which biomarker patterns preceded which outcomes.

This is the single most important thing to understand about the method. The model was not trained to guess your age. It was trained to estimate long-term disease risk. The age you see in your report is the final translation step, not the modeling target.

Step 4: Capturing how markers interact

The models used are gradient-boosting machines, chosen specifically because they handle two things that simpler models struggle with:

Nonlinear relationships. Risk does not always rise in a straight line as a marker rises. Some markers carry risk at both ends of their range. Some have a threshold effect. The model can represent those shapes rather than forcing a straight line through them.

Interactions between markers. This is the part that a marker-by-marker reading of your labs cannot give you. Certain combinations carry more risk together than the sum of their individual contributions. Two people can have similar-looking individual results and meaningfully different combined risk pictures. Because the models were trained on the whole panel at once, those combinations are captured.

Step 5: Modeling each system separately

Every system in your report is modeled independently, using the same underlying data and the same set of biomarker inputs. No single condition gets privileged treatment.

This matters more than it sounds. Because the systems are modeled simultaneously on the same population, the models can capture trade-offs where a biomarker value that looks favorable for one outcome is less favorable for another. A single-condition model would miss that entirely.

It also means your ten system results are directly comparable to each other. A given ranking in the kidney system means the same thing, relative to your peers, as the same ranking in the brain system.

Step 6: Translating into an age

Once your relative risk estimate exists for a system, it is converted into an age by working backwards through the reference curve for people of your sex. The model asks: at what age would the average person in the reference population carry this same level of predicted risk?

If the answer is four years below your chronological age, your delta reads as four years younger. That is the whole of what the number means.

Why this differs from your standard lab reference range

The reference range printed next to a lab value is designed to flag clinical outliers in a general population. It is typically the central range of a healthy reference group, sometimes adjusted for perceived healthy limits.

These models are looking for something different: the biomarker values associated with the most favorable long-term outcomes across multiple systems at once, for someone of your specific age and sex. That target range is usually narrower than the standard lab range and can sit in a different place within it.

The consequence is that a value can be entirely normal on your lab report and still appear as a contributing factor pulling a system’s age upward. Both things are true at once. The lab range is telling you there is no immediate clinical concern. The model is telling you the value is not at the modeled optimum for your profile. Only the second one is a target for proactive attention, and neither is a diagnosis.

Honest limitations

• This is a model, not a measurement. Every result is an estimate built on population-level associations, and population associations are not evidence of causation for any individual.
• Blood biomarkers cannot see everything. Genetics, fitness, body composition, sleep quality, imaging findings, and medical history all shape aging and none of them are inputs here.
• Some systems are inferred, not measured. The respiratory, bone, and brain results are derived from blood patterns, not from lung function testing, bone density scanning, or cognitive assessment.
• Single results are snapshots. Lab values move with hydration, fasting state, time of day, sleep, illness, medication, and training load. The trend across multiple reports is more informative than any single one.

References and further reading

1. Sudlow C, Gallacher J, Allen N, et al. UK Biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age. PLoS Med. 2015;12(3):e1001779. doi:10.1371/journal.pmed.1001779
2. Levine ME, Lu AT, Quach A, et al. An epigenetic biomarker of aging for lifespan and healthspan. Aging (Albany NY). 2018;10(4):573-591. doi:10.18632/aging.101414
3. Rutledge J, Oh H, Wyss-Coray T. Measuring biological age using omics data. Nat Rev Genet. 2022;23:715-727. doi:10.1038/s41576-022-00511-7
4. Oh HSH, Rutledge J, Nachun D, et al. Organ aging signatures in the plasma proteome track health and disease. Nature. 2023;624(7990):164-172. doi:10.1038/s41586-023-06802-1
5. Oh HSH, Le Guen Y, Rappoport N, et al. Plasma proteomics links brain and immune system aging with healthspan and longevity. Nat Med. 2025;31:2703-2711. doi:10.1038/s41591-025-03798-1

Your blood work is processed in a CLIA-certified, CAP-accredited laboratory. The Advanced BioAge Report is generated using VOLO Health’s modeling platform and is provided for educational and informational purposes only. It is not intended to diagnose, cure, mitigate, treat, or prevent any disease or medical condition and does not constitute medical advice. Statistical associations described here are population-level findings and are not evidence of causation at the individual level. Always consult a qualified healthcare professional with questions regarding your health.