Biological age, explained.

“Biological age” is shorthand for a real question: compared to a population-level reference, how worn is your body? Three established methods answer that question — and they’re not interchangeable.

The three clocks

1. Methylation clocks (epigenetic age)

The most studied class. They measure DNA methylation at a few hundred CpG sites — chemical tags that change predictably with age — and regress them against chronological age. The original Horvath clock (2013) and the more recent GrimAge and PhenoAge are this family.

What it’s good for: long-horizon mortality and disease-risk prediction. GrimAge in particular outperforms chronological age at predicting all-cause mortality.

The catch:methylation drifts slowly. A clock built on lifetime drift moves slowly back. You won’t see a 90-day readout shift much, even if the underlying biology has improved dramatically.

2. Phenotypic age

Built from a panel of nine clinical biomarkers — albumin, creatinine, glucose, CRP, lymphocyte percent, mean corpuscular volume, red cell distribution width, alkaline phosphatase, and white blood cell count — plus chronological age. Levine’s PhenoAge (2018) is the canonical implementation.

What it’s good for: capturing the systems that actually move on a treatment timeline. Inflammation drops, glucose normalizes, kidney markers improve — phenotypic age tracks all of it within months.

The catch:it’s only as good as your most recent labs. It’s a snapshot, not a trajectory.

3. Pace of aging (DunedinPACE)

A second-generation methylation clock that measures the rateof biological aging — not where you are, but how fast you’re moving. A pace of 1.0 is average. 0.85 means you’re aging 15% slower than the population. 1.2 means 20% faster.

What it’s good for:evaluating whether an intervention is changing the trajectory, even when the absolute biological age hasn’t had time to catch up.

The catch:needs a methylation array, and it’s less established than the older Horvath/GrimAge family.

Why Halo uses phenotypic age as the primary readout

Phenotypic age has three properties methylation doesn’t:

• It moves on a treatment timeline.The biomarkers PhenoAge tracks shift within the first 3–6 months of a real protocol. That’s the window where members need feedback.
• It runs on standard labs. No specialty array, no $300 swab — the panel is part of the comprehensive workup everyone gets at intake.
• It’s explainable. If your phenotypic age drops, we can show you which biomarkers moved. Methylation clocks, even when accurate, are largely a black box.

For members who want the deeper signal, we layer DunedinPACE on top — quarterly methylation reads to confirm the trajectory the biomarkers suggest. But the primary readout, the number on your dashboard, is phenotypic.

The right clock is the one that responds to what you’re actually doing. Otherwise you’re measuring inertia.Halo editorial team

What “cellular age” means in our framework

Halo separates three concepts deliberately:

• Chronological age— your DOB. Doesn’t move.
• Biological age — the system-level snapshot. Phenotypic age, with DunedinPACE as a secondary read.
• Cellular age — the deeper layer: NAD⁺ status, mitochondrial markers, inflammatory tone. The substrate that biological age is built on.

The configurator on the homepage projects all three. You enter your chronological age and a few inputs about your starting state, and the model returns where biological and cellular age likely sit today — with confidence intervals, because point estimates of anything in this space are dishonest.

The honest limitation

No biological-age estimate is exact. The best published clocks have median absolute errors around 2–4 years against chronological age in validation cohorts, and the variance grows in subgroups underrepresented in training data. We treat the number as a moving signal, not a verdict — and we report the change over time more prominently than the absolute value, because the change is the thing that’s actually meaningful.