Fight Aging!

Members of the life insurance industry have typically been far ahead of the rest of the general public outside the life sciences when it comes to an appreciation of progress towards therapies to treat aging as a medical condition. It is a very large industry, and thus has significant funding to direct towards analysis and prediction of trends in medicine. The prospect of increasing healthy human longevity, of a change in the way in which aging is addressed by medical research and development, is both an existential threat and opportunity for the life insurance industry. Those who predict correctly will thrive, and those who do not will suffer.

Thus it is always interesting to see how insurance industry researchers and analysts react to developments in the medical life science space. Here, the focus is on aging clocks, ways to combine biological data that predict mortality risk across populations. If biological age could be measured accurately for an individual via any specific variety of aging clock, and thus a good estimate of intrinsic mortality risk derived that individual, one would imagine that the life insurers would adopt this technology very rapidly. They have every motivation to do so. The reasons why they have not so far done are the same reasons as to why clocks are not yet the gold standard for assessing the quality of potential rejuvenation therapies: the clocks are not accurate for individuals, and their underlying connections to biological age are not fully understood.

Biological Clocks: Ready for Prime Time?

John Smith is a 50-year-old male applying for a life insurance policy. His medical history is unremarkable, and his recent medical visits record good health. He exercises regularly and is an enthusiastic member of several wellness programs. A recent test from a longevity company reports a biological age of 46, and records that he is aging at 0.7 years per year. Both the company and Mr. Smith are excited that his life expectancy is well above normal. Are they right? And if they are, should life insurers be excited too?

Chronological age is the widely accepted starting point of mortality assessment. It is also time-honored, having first appeared in insurance life tables in the 17th century. Yet it is a blunt metric: two 50-year-olds may have quite different health statuses and life expectancies. Biological age is a term that is widely used in the aging literature. But curiously, it has no accepted definition. This reflects both the complexity of aging and the lack of any gold-standard metric. It reflects how old the body has become, functionally and biologically. It incorporates dimensions of health, such as current physiological state, and the cumulative molecular and cellular damage that has accrued over time. Thus, at face value, biological age would seem to provide more useful underwriting information than chronological age. If one of our 50-year-olds had a biological age of 46 and the other 54, mortality projections would be quite different, and premiums could reflect these.

So, where did John Smith's biological age determination come from? It was likely provided by an "epigenetic clock." which is an algorithm that estimates chronological age from patterns of cellular DNA methylation. Epigenetic clocks are statistical models, trained on methylation status of selected subsets of CpG sites - typically ranging from a few hundred to a few thousand - chosen to optimize predictive performance.

What are the rubs against epigenetic clocks? There are quite a few. Epigenetic clocks are not trained to provide reliable predictions at the individual level. Rather, they are statistical models designed to minimize error across thousands of samples. Consequently, when applied to a single person, their estimates are biologically noisy. Epigenetic age is not a traditional biomarker, such as BMI or serum glucose, which can generate reliable individual-level information. Thus, to equate a younger predicted epigenetic age with a younger biological age, even though this is common practice, is an overextension. Epigenetic clocks are highly dependent on the training data and the populations from which they are derived. If a clock is applied to different populations - such as the very fit (Mr. Smith) or self-selected individuals (those likely to buy a commercial test) - the predictions may be inaccurate.

Are epigenetic clocks of value to life insurers? Not at present. While biological age, to the extent it is equated with epigenetic age, does predict mortality, it does not outperform traditional mortality risk predictors such as age, sex, smoking, blood pressure, BMI, and medical history. Similarly, pace-of-aging, although an outwardly attractive metric, does not outperform traditional measures of current health status. One exception might be the young or apparently healthy, where traditional risk factors are absent, and early deterioration might be captured. Another might be the small number of older individuals where all traditional risk markers are negative; epigenetic clocks may provide better insight into current health. But both scenarios would require longitudinal analyses to prove clock utility.