AT SOME POINT IT DOES NOT GET WORSE: LATE LIFE AND MORTALITY PLATEAUS
Review of: L. D. Mueller, C. L. Rauser, and M. R. Rose. 2011. Does Aging Stop? Oxford University Press, Oxford and New York. xx, 204 Pages. ISBN: 978019975422
“Evolution coddles you when you are young and forsakes you when you are old”–Wachter (1997, p. 3)
In humans, as in many other vertebrates, mortality rates are lowest in late childhood, right before the start of adolescence. After that, they increase approximately exponentially. Specifically, mortality rates double approximately every eight years or so in contemporary developed countries such that the probability that a 70-year-old woman will die in a given period is about eightfold higher than that of her 46-year-old daughter and about 64-fold higher than that of her 22-year-old granddaughter.
Such exponential changes in mortality rate have a long history in demography. In fact, the British mathematician Benjamin (Gompertz 1825) formalized these patterns in a series of exponential mortality rate schedules nearly two centuries ago. Gompertz mortality curves fit equation (1), where Mt is the mortality rate at time t, Ml is the mortality rate at time i (i < t), and k is the exponential rate of mortality increase,
For any given Gompertz curve, k is a constant that can be thought of as a rate of aging. Across different Gompertz curves, k can and does vary. The eight-year doubling time in contemporary human societies translates into a k of approximately 0.09 per year. In a complete Gompertz curve, i is the time when mortality begins to increase exponentially; it usually occurs in late childhood, just before adolescence. To generate partial Gompertz curves, one can set i to later times. Gompertz curves not only fit human mortality schedules, but also those of other primates (e.g., Bronikowski et al. 2011). They even fit mortality patterns of a large number of invertebrates, at least for a portion of their adult lifespan.
Why do mortality rates increase with age? Given the multitude of other subjects in which Darwin made seminal early contributions, it is a little surprising that he did not devote much time on an evolutionary explanation for aging. This task was left to such luminaries as August Weissman, Sir Ronald Fisher, J. B. S. Haldane, Peter Medawar, and George Williams. They observed that the efficiency of natural selection decreases with age, especially after the age of first reproduction. Natural selection thus influences genetic variations that have effects early in life more than those with effects later in life. One consequence is that the late acting mutations should accumulate in the population via random genetic drift even though deleterious in their effects. This is the mutation–accumulation model (Medawar 1952). Another consequence is that alleles that have advantageous effects early in life will be favored by natural selection despite having deleterious effects later on. A hypothetical example would be an allele affecting glucose metabolism that allows for more rapid growth in the young but increases susceptibility to diabetes in later life. This is the antagonistic pleiotropy model (Williams 1957). William Hamilton (1966) and Brian Charlesworth (1980) formalized the verbal arguments by Medawar and Williams.
During the 1980s, the subfield of evolutionary biology of aging emerged, as various labs used experimental evolution (mainly with invertebrates such as nematodes and flies) and other tools to investigate these models of aging. By the early 1990s, the consensus had been reached that the Hamiltonian paradigm was correct: increasing mortality rates with age were a consequence of the declining power of natural selection with age. Then, the big controversy was about the relative importance of antagonistic pleiotropy versus mutation accumulation (Rose 1991), a debate that continues till date (Hughes 2010). Regardless of which side was correct, the prevailing view was that aging occurs because natural selection coddles the young and forsakes the old.
Two experimental studies published in 1992, one in Drosophila (Curtsinger et al. 1992) and another in medflies (Carey et al. 1992), seemed to challenge this paradigm. These studies showed that late in life, mortality rates do not continue to increase as expected by the Gompertz equation. Instead, mortality increases first decelerates and then stops. Late in life, mortality rates plateau. Both these two papers used large cohorts of individuals (needed to detect small changes in demographic patterns late in life when a small proportion of the initial individuals were still alive), and were well-controlled lab studies.
These fly studies also sparked interest about the human mortality plateaus. Gompertz himself recognized that mortality curves might level off late in life. Periodically, especially since the 1930s, demographers saw evidence of deceleration and plateaus of human mortality rates very late in life. By the early 1990s, this evidence had become more solid. Although the initial fly results were met with some skepticism, there was the sense that if these plateaus were real and general, then the evolutionary theory of aging was in trouble.
Lawrence Mueller, Casandra Rauser, and Michael Rose (henceforth, MRR), three biologists at the University of California at Irvine, have worked separately and together on the evolutionary biology of aging and other aspects of life-history evolution, mainly performing experimental evolution studies with Drosophila. Initially skeptical of the plateaus, MRR were later convinced of their reality and importance. In their provocative, slim book, Does Aging Stop?, they lay out the evidence for the plateaus showing that they are not just artifacts of environmental conditions. They also demonstrate that the patterns, being observed in humans and several orders of insects, are general. To these researchers, however, the existence of plateaus is not reason to suspect the validity of the evolutionary theory of aging. Rather, they argue that the evolutionary theory of aging predicts such plateaus. MRR view these plateaus as a revolution for the study of gerontology, and suggest that their existence may allow us to greatly extend the human lifespan.
Why Does Aging Stop?
During the 1990s, studies in insects, including some from MRR's group, ruled out possible artifacts that could cause the mortality plateaus. For instance, the plateaus are not due to the older animals experiencing reduced population density, as they persist even when care is taken to ensure constant density. Nor are they owing to changes in activity patterns as individual's age.
Instead, MRR argue that the mortality plateaus are a consequence of the theory of evolution, as articulated by Hamilton and others. Through mathematical models, they (e.g., Mueller and Rose 1996) and others show that under a variety of conditions, antagonistic pleiotropy and mutation accumulation can generate mortality plateaus.
Experimental evolution studies by MRR and others show that mortality plateaus can evolve to start at different ages. In particular, selecting for later reproduction results in the plateaus starting later. These responses are rapid and reversible, which suggest that they are due to the antagonistic pleiotropy of standing genetic variation, and not new mutational inputs.
The papers of Carey et al. (1992) and Curtsinger et al. (1992) clearly showed the mortality patterns share a common author, James Vaupel, who (e.g., Vaupel 1997) has championed the idea that the plateaus may result from some individuals being more robust throughout life than others, owing to genetic and/or environmental factors. MRR do not deny that such lifelong heterogeneity may exist, but argue that it is insufficient to account for the plateaus.
Mortality Plateaus in Humans
In contemporary, developed societies, human mortality clearly plateaus sometime in the 1990s, a period that MRR call “late life.” Hence, demographic aging slows, and even stops. What is the significance of this finding for gerontology?
As MRR note, human late life begins much later than the comparable stage in flies or in nematodes. In human late life, mortality rates are extraordinarily high, asymptotically approaching a figure of between a one-third and a half dying each year. Even though mortality rates do not continue to rise, this elevated plateau is still little comfort.
Why does late life occur so late in humans? MRR point to three possible explanations: adaptation to agriculture, adaptation to changed demography (including age at first reproduction), and increases in effective population size. All of these are plausible. I would add two other explanations: extended parental care and what I’ll call “80 is the new 60.” Due to the extended period of childhood dependency and parental care in humans, reproductive value does not go to zero when reproduction ceases. In a hunter-gatherer society, a 55-year-old woman with a 12-year-old son would likely enhance her fitness by staying alive a few more years. In addition, it is likely that effects of late-acting, deleterious mutations occur at more advanced ages in contemporary populations than they did in our evolutionary history due to our changed environments. One such change is reduced early exposure to pathogens, which likely led to lower mortality rates late in life (Finch and Crimmins 2004).
All of the above about late plateaus in humans is rank speculation. The problem is that we have essentially zero knowledge about late life and mortality plateaus for our species, except for a very small sliver of our evolutionary history. We do not know whether, and if so, when mortality plateaued late in life in 17th century London, much less in populations transitioning to agriculture 8000 years ago. Hence, we do not know whether these plateaus are ancestral or derived.
MRR see the existence of mortality plateaus, and their late appearance, as a potential way to monumentally extend the human lifespan. For instance, they state: “It is a simple demographic point that greatly extending the human functional lifespan, or healthspan, would be much more easily achieved by shifting the age at which human aging stops to much earlier ages. If the process of aging were stopped at the age of 40 years, for example, then the capacity of modern medicine to sustain the survival and function of people over that age would be greatly increased.” (p. 140).
The key word in this utopian passage is “if.” If we could adjust the cessation point of human aging even to age 65—much less 40—that would be a dramatic and revolutionary advance in medicine. But can we do it? My sense is that we cannot in the foreseeable future.
One demographic puzzle, neglected by MRR, is that populations of humans and other primates (although the data are not as robust) exhibit a negative correlation between the baseline mortality rate (Mi) and the rate of increase in mortality (k) (e.g., Hawkes 2010). Such a negative correlation is counter to the Williams–Hamilton explanation for aging, under which we expect a positive correlation (lower baseline mortality being associated with a lower rate of mortality increase). Indeed, positive correlations have been observed in populations of other animals, most notably guppies (e.g., Reznick 1997).
Heterogeneity might explain the negative correlation between baseline mortality rates and the rates of increase in mortality. In a harsh environment with high baseline rates of mortality, the frail die young, leaving the population that comprised more hearty individuals, who then age more slowly. In a less harsh environment with a low baseline mortality rate, the frail persist longer. As a consequence, the population as a whole has a more rapid increase in mortality rates in populations (Hawkes 2010). Lifelong heterogeneity could also explain why human females that give birth well into midlife live longer than average. Females remaining fertile are more likely to be hearty than those who do not, and, thus, should have a greater life expectancy despite the increased resources devoted to reproduction.
Could it be that mortality plateaus in flies (and presumably other invertebrates) arise from Hamiltonian theory and the action of selection acting on antagonistic pleiotropic genetic variants, but heterogeneity is a major factor in the evolution of mortality plateaus in humans?
Here and elsewhere (e.g., Rose et al. 2006), MRR present their work providing an explanation for mortality plateaus based on Hamiltonian theory as a scientific revolution, sensu Kuhn (1962), and on par with the Einstein's theory of relativity. This, to me, is a gross overstatement of the importance of the claims. Even if we had conclusive evidence that mortality plateaus in humans as well as those of invertebrates are owing to the Hamiltonian model (and we are not there yet), it is not clear how much the way we view aging would change. Whether this would radically change the way medicine is practiced is also dubious, especially given the very late start and high mortality of the plateau. Moreover, plateaus have been hinted at for almost two centuries, and especially since the 1930s. The accomplishment is the combination of the detailed experimental evolution studies and the extension of established evolutionary theory of aging. It is not a revolution in the strict Kuhnian sense either. No paradigms have been overturned; instead, they have incorporated the observations of the plateaus into a preexisting, well-established paradigm. If the Hamiltonian explanation for the plateaus holds up to be general, this is a laudable achievement, but no special or general theory of relativity.
If not a scientific revolution such as Einstein's theories of relativity, how then should we consider these mortality plateaus? I am reminded of the protein electrophoresis studies of the 1960s. As told by Lewontin (1974), the findings of abundant genetic variation in natural populations did not arrive in a vacuum, but came into preexisting debates (classic vs. balance school). Such was the case for the mortality plateaus as well; Vaupel et al. thought that lifelong heterogeneity was rampart, and MRR and their colleagues operated in the framework of the Hamiltonian theory. The classical and balance school participants, that would become the neutralists (neoclassicists) and the selectionists, each incorporated the electrophoresis results into their models. The same has occurred with the Vaupel and Hamiltonian schools with the mortality plateaus.
The self-aggrandizement does detract from the value of the book. The book is also hampered by its narrow focus. Little, if any mention, is devoted to nonhuman primate studies, or even nonhuman vertebrate studies. In addition, other relevant areas of study, such as reliability theory (e.g., Gavrilov and Gavrilova 2001), are given short shrift. This is unfortunate because the mortality plateaus are an interesting and not fully explained feature of demography that could influence how we view both evolutionary biology and medicine.
Associate Editor: M. Wade