Multi-clock consensus scoring in practice.

A signal, not a verdict.

Biological-age clocks have multiplied faster than clinical practice has learned to use them. Each clock measures a genuinely different aspect of ageing — DNA methylation patterns, circulating plasma proteins, clinical variables collected at a routine visit — and each was validated against a different outcome, in a different population, using a different mathematical method. Managing any single one of them as a definitive verdict on a patient's biological age is a category error most longevity clinics make anyway, because a single number is easier to sell than an honest range.

We do not use a single clock at Clinic82. We run four, and we combine them through a consensus methodology developed with Dr Oliver Zolman, because the disagreement between clocks is itself clinically informative — it tells us where the signal is strong and where it is not.

Four methods, four signals.

Each of the four clocks in our panel was built to answer a slightly different question, using different underlying data:

• Horvath clock (2013) — the original epigenetic clock, built on DNA-methylation patterns read across more than 50 tissue types. It was the first clock to demonstrate that biological age could be read directly from methylation data, and it remains a reference point for the field, with a reported accuracy of roughly ±3.6 years.
• GrimAge (2019) — developed by Lu and colleagues, GrimAge integrates epigenetic markers with twelve circulating plasma proteins associated with mortality risk. Of the widely used clocks, GrimAge is the strongest predictor of mortality and years-to-death, which is why it carries particular weight in our consensus score for older patients or those with elevated cardiovascular or metabolic risk.
• LinAge2 (2025) — built by NUS Medicine on clinical variables rather than methylation or proteomic data, LinAge2 requires no sequencing and is designed to scale to population-level screening. Its clinical-variable basis makes it faster and cheaper to run, at some cost in the granularity that epigenetic clocks provide.
• Zolman Clocks — the consensus-scoring methodology itself, developed with Dr Oliver Zolman for clinical decision-making inside the 82-organ-type framework. Rather than an independent fifth measurement, this is the integration layer that reconciles the other three into a single clinically usable score.

Weighting, not averaging.

The Zolman consensus methodology is not a simple average of the three underlying clocks. Each clock is weighted according to what is clinically most relevant for a given patient — GrimAge's mortality association carries more weight where cardiovascular or metabolic risk is already elevated; the Horvath clock's tissue-level resolution carries more weight where organ-specific ageing is the clinical question; LinAge2's clinical-variable basis provides a lower-cost cross-check that flags when the epigenetic and proteomic clocks disagree with what the patient's clinical presentation would predict.

When the clocks disagree substantially — for example, when GrimAge indicates elevated mortality-associated ageing while the Horvath clock reads close to chronological age — the discrepancy itself is flagged in the patient's report rather than resolved silently into a single averaged number. A discrepancy of that kind is often the most clinically useful part of the result, because it points toward a specific driver rather than a generic ageing signal.

The consensus score is also tracked longitudinally rather than read as a single snapshot. A patient's repeat panel at six or twelve months matters more than any single reading, because clock measurements carry biological and assay noise that a trend line filters out far better than one number can.

What patients actually see.

A patient's report presents the consensus biological age alongside the individual clock readings that produced it, rather than the consensus number alone. This is a deliberate transparency choice: a patient whose consensus score reads four years younger than chronological age, but whose GrimAge reading alone is elevated, needs to know that nuance rather than see a single reassuring headline figure.

Clinically, the consensus score functions as a triage signal into the 82-organ-type framework rather than as an endpoint in itself. A consensus reading meaningfully older than chronological age directs attention toward which specific clock is driving that result, and from there toward which organ domain is the most plausible driver — cardiovascular, metabolic, or otherwise. The number opens an investigation; it does not close one.

We do not present the consensus score as a guarantee of years of life gained or lost. It is a measurement with a defined confidence range, tracked over time, used to direct clinical attention — not a verdict on a patient's future.

Still calibrating.

Two areas we are actively refining in how we run multi-clock scoring:

• Evaluating newer clocks as they publish. The field is moving quickly — LinAge2 itself is a 2025 addition — and we are currently evaluating several newer proteomic and organ-specific clocks for inclusion in the consensus methodology. We have been deliberately conservative about adding a clock before its validation data is mature enough to weight reliably.
• Standardising discrepancy communication. When clocks disagree significantly, how we explain that disagreement to a patient without either overstating the uncertainty or glossing over it is something we are still improving. Early reports erred toward technical language that did not always land clearly; we are revising the reporting format with Dr Oliver Zolman to make discrepancy flags more immediately actionable for the patient reading them.

The four-clock consensus approach is, we think, the right architecture. Which clocks sit inside it, and how clearly we communicate what they disagree about, is the part still under active revision.