This site always has interesting discussions.

Today I'll point out the latest paper in a debate over whether there are limits to human life span. As everyone in the audience here is no doubt aware, human life expectancy is gently trending upward. Life expectancy at birth is rising at about two years with every decade, while life expectancy at 60 is rising at about a year with every decade. The evidence in support of this trend is robust, thanks to the enormous demographic databases collected over the past few decades. Is this trend approaching any sort of limit to human life span, however? Can historical data even be used to answer that question? This is a much more challenging proposition, as the available data for the oldest humans, the population of supercentenarians older than 110, is sparse. Very, very few people survive to these ages, to the point at which statistical methods operating on this data become ever more dubious with each additional year of age.

Still, people crunch the numbers and try to extract meaning. You might recall that last year, Jan Vijg's group put forward their argument for the data to show there to be a limit to human life span over the years in which that data was collected. It was coupled to some unexpectedly pessimistic commentary on the future development of longevity science, given that Vijg has for some time been counted among those researchers openly in favor of extending healthy life spans by treating aging as a medical condition. The paper sparked some occasionally heated discussion. I don't think the researchers expressed their argument all that well in their publicity materials, and the popular science press then generated more than the usual degree of mess and confusion when they pitched in.

So to the casual observer, it was a little difficult to see whether Vijg and company were making the obvious point, which is that human life span is effectively limited by the present level of medical technology, or whether some more subtle argument was being made. I think it is hard to disagree with the statement that medical technology determines limits to human life span. Where we can debate, given the sparse nature of the evidence to hand, is whether or not there exists one or more mechanisms of aging that have not been impacted in any meaningful way by improvements in medical technology over the past century, and which, on their own, can produce a very high rate of mortality in late life. That circumstance would look a lot like a limit when examining the consequent demographic data.

One mechanism that springs to mind here is the accumulation of transthyretin amyloid, found in one small study to be the majority cause of death in supercentenarians, but which appears to have only a smaller impact on mortality in younger old age - it is implicated in something like 10% of heart failure cases, for example. Can we argue that advances in medicine and public health over the past century have had little to no impact on the accumulation of misfolded transthyretin deposits in tissues, and thus this mechanism acts as a limit on life span? Or do some of these improvements in fact produce an small, incidental reduction in amyloid burden in later life? I think that the evidence to support any of the possible positions on these questions is presently lacking.

Whatever the state of effective limits on life span today, however, the limits on life span tomorrow are determined by progress towards rejuvenation therapies. There are treatments under development that can clear transthyretin amyloid from tissues, for example. The same is true for many of the other forms of molecular damage and waste accumulation that cause aging. Thus any debate over what the present demographics do or do not show is more academic than it might otherwise be. The natural state of human aging, already largely paved over by medicine, will be buried completely, made irrelevant in the decades ahead by the advent of means to repair the damage, restore youthful function, and eventually to indefinitely postpone all of the symptoms of aging.

No detectable limit to how long people can live

Supercentenarians, such as Jeanne Calment who famously lived to be 122 years old, continue to fascinate scientists and have led them to wonder just how long humans can live. A study published last October concluded that the upper limit of human age is peaking at around 115 years. Now, however, a new study comes to a starkly different conclusion. By analyzing the lifespan of the longest-living individuals from the USA, the UK, France and Japan for each year since 1968, researchers found no evidence for such a limit, and if such a maximum exists, it has yet to be reached or identified.

"We just don't know what the age limit might be. In fact, by extending trend lines, we can show that maximum and average lifespans, could continue to increase far into the foreseeable future." Many people are aware of what has happened with average lifespans. In 1920, for example, the average newborn Canadian could expect to live 60 years; a Canadian born in 1980 could expect 76 years, and today, life expectancy has jumped to 82 years. Maximum lifespan seems to follow the same trend. Some scientists argue that technology, medical interventions, and improvements in living conditions could all push back the upper limit. "It's hard to guess. Three hundred years ago, many people lived only short lives. If we would have told them that one day most humans might live up to 100, they would have said we were crazy."

Many possible maximum lifespan trajectories

A recent analysis of demographic trends led to the claim that there is a biological limit to maximum human lifespan (approximately 115 years). Although this claim is not novel - others have also identified a biological 'barrier' at 115 years - the methodology that the authors used is. Here we show that the analysis does not allow the distinction between the hypothesis that maximum human lifespan is approximately 115 years and the null hypothesis that maximum lifespan will continue to increase. The central difficulty with this exercise is accurately extrapolating onwards from a limited, noisy set of data.

Beyond a plateauing of maximum life span, there are other different trajectories that maximum lifespan could follow over time if the null hypothesis (that maximum lifespan will continue to increase) were true, with maximum lifespans continuing to increase to an eventual future plateau or continuing to increase indefinitely. All three models appear equally consistent with the known maximum lifespan data used. How the authors differentiated between these possibilities is important. Their claim rests on their identification of a plateau in the ages of maximum lifespan beginning around 1995. They separated the data into two groups, 1968-1994 and 1995-2006, and modelled each group using linear regression. While the first partition shows a trend for increasing maximum lifespan, the second partition does not. It is this latter partition upon which their conclusions are largely based. This is problematic, because, even within a dataset showing an overall trend for an increase with ti me, normal variability can generate apparent plateaus and even temporary decreases over small intervals.

Furthermore, the authors do not describe how they identified the lifespan plateau, nor the partition site, indicating that these were products of casual visual inspection. This is a critical point for the validity of their argument because even slight changes to the assumptions that they made can notably alter the results of their analysis, with markedly different outcomes. In conclusion, the analyses do not permit us to predict the trajectory that maximum lifespans will follow in the future, and hence provide no support for their central claim that the maximum lifespan of humans is "fixed and subject to natural constraints". This is largely a product of the limited data available for analysis, owing to the challenges inherent in collecting and verifying the lifespans of extremely long-lived individuals.

A reply from Jan Vijg's research group

The authors of the accompanying comment disagree with our finding of a limit to human lifespan. Although we thank them for alerting us to other work reporting a limit of around 115 years, we disagree with the arguments presented and remain confident in our results. We feel that the scenarios presented, although imaginative, are not informative. They argue that their three different models (which they extrapolate until the year 2300) are not statistically differentiable based on the data available. We used a data-driven approach to identify the trend in the maximum reported age at death (MRAD) by analysing actual data rather than arbitrary simulations; although the authors criticize us for visually inspecting our data, graphing data in order to evaluate the choice of model has long been acknowledged as a useful and important technique by statisticians. Taken together, and in the absence of solid statistical underpinning of various possible future scenarios, we feel that our interpretation of the data as pointing towards a limit to human lifespan of about 115 years remains valid.

This is, it has to be said, exactly the sort of exchange one might expect to see between researchers who are working with a very sparse set of data. It is always interesting to watch the ongoing efforts to better refine, mine, and interpret this data, but it is of limited relevance to the near future of therapies to treat aging. All of the present well-known demographics of later life will be changed greatly for the better as therapies capable of addressing the causes of aging emerge.