The Ninth Supercentenarian Workshop was held May 11 and 12 at the Palazzo Brancaccio in Rome, Italy.
Sponsored by the Max Planck Institute for Demographic Research, the meeting was led by MPIDR director James Vaupel, who is also co-director of DuPRI's Center for Population Health and Aging, and locally organized by Graziella Caselli of the Department of Social, Economic, Actuarial and Demographic Studies of the University of Rome La Sapienza.
Over two full and lively days, researchers from Europe, Scandinavia, the U.S., Canada, Japan and Brazil presented their latest data and analyses of the phenomenon of "supercentenarians" — people who reach the age of 110 or more — and "semi-supercentenarians," those between 105 and 110 years old. Many of the presentations updated research described in the group's most recent monograph, "Supercentenarians," published in 2010. They included trends in mortality at extreme old ages, comparisons to historical rates of centenarian survival, evidence on biological traits associated with reaching supercentenarian status and statistical issues raised by the nature of available data on the oldest old.
The workshop participants also reported progress in their efforts to establish an International Database on Longevity (www.supercentenarians.org) for the use of population researchers. Ultimately, the web portal will also offer information about supercentenarians for the general public. DuPRI will be serving as the repository for U.S. data destined for the IDL, and will be helping the U.S. Social Security Administration's Bert Kestenbaum to gather and validate information on U.S. semi- and supercentenarians. The nature of data required to maximize the usefulness of the database as a research tool was an important topic of the meeting, as were privacy issues the resource may raise in various contributing countries.
Glimpses of the Plateau
Caselli and Vaupel opened the workshop by noting that the circa-1880 palazzo where the researchers were gathered was built five years after the birth of Jeanne-Louise Calment, the Frenchwoman who died in 1997 at age 122 and a half. Hers is the longest human lifespan ever documented and Calment would be invoked several times during the sessions as participants discussed whether she was an "outlier," and natural human longevity, as indicated by a mortality rate plateau, might be stabilizing closer to the age of 110.
Jutte Gampe of the MPIDR presented a preliminary analysis of mortality rate trajectories among centenarians in a range of countries, incorporating the latest data contributed to the IDL. Although she said she had difficulty "taming" some of the data "without making extra assumptions" for the highest ages, overall she said it looks as though a plateau might be starting around ages 106, 107 and 108. Vaupel noted that in the Swiss data, it appeared that a plateau began around age 104, but in general the plateau seems to be higher for individual countries than for the whole dataset.
Whether these plateaus in recent semi-supercentenarian cohorts are significant, or influenced by early-20th-century events like World War I and the "Spanish flu," is a question that would arise repeatedly as participants presented new data. For true supercentenarians, another challenge is the fact that too few "cases" make it hard to establish a trend amid the fluctuations in mortality rates, explained Jean-Marie Robine of INSERM, the French national science and medicine research agency, in an interview after the meeting. Robine, based in Montpellier, created the original database of the oldest old, upon which the new IDL builds. To date, the collection contains data on some 800 supercentarians and more than 1,000 semisupercentarians, but many of the meeting participants reported their plans to contribute hundreds of new cases this year.
After Gampe's talk, Vaupel commented that some 20 presentations on the subject of mortality at older ages were given at the Population Association of America meeting held a week before the Rome gathering, indicating a strong interest in the topic and the potential for a lot of researchers to use the IDL when it becomes available. One of the most pressing questions validated data can answer, he noted, is whether mortality rates are indeed plateauing (some aging researchers think they are not) and at what age.
In this context, Vaupel recapped the principles of the Gamma-Gompertz model for calculating mortality trajectories: "The observed increase in the force of mortality for the population is equal to the actual average increase for the individuals who make up the population minus the variance in the force of mortality among the people in the population...if you look at the rates of change...then the rate (at which) mortality goes up for the population is equal to the average rate it goes up for the individuals in the population minus the coefficient of variation of mortality among the individuals in the population, a measure of the heterogeneity of the population, multiplied by the level of mortality in the population.
"If there's no plateau, we might as well go home!""So we have a correction factor," Vaupel continued, "so the rate of increase of mortality in the population is lower than the rate of increase for the individuals in the population. That correction factor is a product of two terms, one is how heterogenous the population is and the other is the level of mortality, so as the level of mortality goes up that reduction gets bigger and bigger. And at the point where we have the plateau, then the rate of increase for the population is zero, so the rate of increase for the individuals in the population has to be exactly equal to this measure of the heterogeneity of the population multiplied by the level of mortality in the population. And the fact that these two things have to be equal, and if you have a plateau, they have to be equal at all ages after the plateau, puts an enormous constraint on the mortality model. This equation has to hold at all ages. And that implies that there's one and only one valid mortality model, and that is that death rates have to be going up exponentially for individuals (as in Gompertz)...and there has to be a proportional hazards model so individuals differ from each other in their relative risks, and there has to be a distribution of relative risk that's gamma distributed. So the only model that works is the Gamma-Gompertz model. And one of the really amazing results of the work we've done is to show that this one and only mortality model describes mortality at the oldest ages where we have this plateau!
"How quickly do you get to the plateau?" he asked. "Well it turns out if you have the apparent increase in death rates in the population and you compare that to the actual increase in death rates in individuals in the population, and the ratio between the two is given by a very simple term, namely how far, relatively, are you away from the plateau. Suppose the plateau is .7 and the observed mortality is .35, then you're halfway to the plateau. When you're halfway to the plateau, the observed rate of mortality increase is half the rate of the individuals in the population. And as you get closer and closer to the plateau, the observed rate of increase goes to zero."
Referring to a recently published paper claiming there is no plateau, Vaupel joked, "If there's no plateau, we might as well go home!" Practically speaking, he said debates about maximum natural lifespans are likely to linger given that no one alive today is even close to Mme. Calment's record-setting age at death. And 1997, her year of death, is a long time ago, Vaupel said. Her record is "not going to be broken next year because nobody's 121. The oldest person now is 115, and there's only one, so it's going to be seven years at least, maybe more." Since fewer than 10 living people are 114 years old, he continued, if half die each year, "we're going to have to put up with these questions for 15 years or more."
To make matters worse, the number of people reaching the age of 110 appears to be stagnating in some countries. In the United States the rate of supercentenarian increase has slowed considerably, according to Kestenbaum, and is now expected to double every 20 years, instead of every 10 as was true in the latter half of the 20th century. In Japan the number of people reaching 110+ does not appear to be stagnating, but in the Nordic countries it is. Vaupel noted that the Japanese "take care of their old people — there's no euthanasia or deciding to let people die a natural death," whereas in many other countries different attitudes toward life-extending interventions might account for some of the differences. Others remarked that in Switzerland the rate is not slowing either, and perhaps an effect from World War I — which began nearly a century ago — could partly explain the supercentenarians who appear to be missing in some countries but not others.
Robine gave a summary of efforts toward "legalizing" the IDL since the group's last workshop in Madrid in early 2011. Their desire to centralize the collection and validation of data on semi- and supercentenarians, and to open it to other researchers, led to the decision to move it from INSERM to the Max Planck Institute in Rostock, Germany. MPI has extensive experience in database management, but can only handle "nominative" data on behalf a third party. Therefore, Robine reported, he and MPIDR's Vladimir Jdanov have spent considerable time finalizing a contract between MPI and INED, the French national statistics agency, which has agreed to act as the third-party "client."
What has also become apparent during that time, however, is that going forward, national laws regarding data privacy will add complexity to the handling of personal information about centenarians. Researchers contributing data may have to anonymize it, depending on their nation's policies, but doing so may interfere with the validation of individual entries and checks to avoid duplicates, as well as with users' ultimate applications for the data. For now, the supercentenarian group can continue to do research with the material they've compiled and add new data, but MPIDR will also work with contributors to ensure their respective national laws are observed.
Hints of Rising Lifespans in North America and the UK
After a break, Kestenbaum gave participants a look at new U.S. data he'll be contributing to the IDL. First, though, he described how his team finds candidate supercentenarians and validates their dates of birth and death. Starting with electronic SSA records, including applications for a Social Security Number, welfare rolls, the SSA Master Beneficiaries Record and the Master Death File, he also cross-references Medicare part B files. That paid insurance program's list is more likely to be up to date than the free Medicare part A program, he explained. Census records are also useful to verify that an individual's recorded age at various points in their lifetime is consistent with birthdate information. With three independent sources confirming a candidate's age, Kestenbaum adds that person to the U.S. list of validated semi- or supercentenarians. Unfortunately, older SSA records are kept on microfiche, "in a cave in Pennsylvania 200 miles from headquarters," he said, making validation time-consuming.
So he has focused on thoroughly validating a five-percent sample of U.S. supercentenarians. The first cohort he completed and added to the IDL was born between 1870 and 1889. Of 341 supercentenarian cases in the sample, 90 percent were female, 15 percent were African American and the greatest age at death in the group was 119. The next cohort he's working on is made up of those born between 1890 and 1894. Including deaths through 2009, there are 148 cases in the sample — also 90 percent female and 15 percent black. Their greatest attained age at death so far is 116. But, Kestenbaum said, he and his colleague Renee Ferguson hope to verify some additional cases, so the total of supercentenarians in the sample should go up.
The next cohort, born between 1895 and 1899, contains just 58 cases right now, but those born in 1899 would only have reached 110 in 2009, so Kestenbaum expects many more to be added. Whether an upward trend in the number of Americans reaching supercentenarian status will continue, however, remains to be seen. There were 37 supercentenarians in the 1870-1874 cohort, 67 in the next five-year group, 99 in the next and 133 in the next. Looking at the equivalent moment in the 1890-1894 cohort's history, he said, "they had 115 of their 148 deaths by that time, so they were three-quarters of the way there." The 1895-1899 cohort may end up being "late bloomers" and catch up to their predecessors, he noted. But at present, "It looks low, and we don't know why." Casselli commented that a similar dip seen in Italian data has been attributed to WWI and the 1918 flu pandemic — both events that killed disproportionate numbers of young adults at the time when this cohort would have been in their late teens and early twenties.
Bernard Desjardins and Robert Bourbeau from the University of Montreal took a different approach to sampling the Canadian population, in part, they explained, to ease the validation process in a nation of 10 provinces, where all civil registration is done on the provincial level. Civil registration in Canada is fairly recent as well, Bourbeau said, so they knew they would need to rely on baptismal records. They decided to focus on Quebec, which has 25 percent of Canada's population, and is largely Catholic. Beginning with the provincial death registry, they found that between 1985 and 2009 there were 6,503 deaths of centenarians, of whom 5,008 were born in Quebec. (They eliminated immigrants because their birth records would be difficult to retrieve). Further narrowing the sample to deaths over the age of 105 left 372 candidates.
"Politicians are taking note of the implications"
Limiting themselves to Quebec-born French-Canadian Catholics gave the team a 90 percent success rate in validating individuals through parish and census records. In a cohort born between 1877 and 1904, they found 259 candidates, of whom only 11 were supercentenarians, all female. Together with the 332 cases the team has already sent to the IDL from Quebec, the data show an age-specific mortality rate of around .5, they noted, in keeping with a plateau beginning around age 105. Citing data on mortality among French Canadians born in the 17th and 18th centuries, Borbeau and Desjardins remarked that a mortality rate plateau started a decade earlier — around age 93 — during that period.
In the United Kingdom, a similar gain of nearly a decade is detectable in more recent mortality statistics, said Adrian Gallop of the Government Actuary's Department. In the 1960s, an 85-year-old man had a 20 percent chance of dying, but by the early 1990s that same risk level was not reached until age 92. To identify British supercentenarians, Gallop explained, his team also begins with a list of deaths, from the Office of National Statistics, then seeks out birth and death certificates to glean more information. Since 1992, he said, there have been 97 deaths at age 110 or above in the UK, of which 71 cases have been age-validated. Among 13 of those individuals born overseas, half were from Pakistan. For the great majority, the listed cause of death was merely "old age" (55 cases). Of the remaining individuals, pneumonia is the cause given for 23 cases, Alzheimer's disease for six and cardiovascular/cerebrovascular disease for another six.
Noting some other interesting trends in the data he'd analyzed, Gallop saw a likely effect of WWI in the form of missing male centenarians. In addition, he observed, people born in the 1930s had higher mortality rates than those born before or afterwards. Still, the rate at which people are reaching ages of 100 and more is rising — the UK had a projected 11,600 centenarians in 2010 and 100,000 are projected for 2035. Politicians are taking note of the implications, as well Gallop said: In May a proposal was made in parliament to raise retirement benefits based on longevity projections for the nation.
Politics and Population in Italy
The workshop's host country also has an eye toward its growing elderly population, according to Marco Marsilli of Istat, the Italian National Institute for Statistics. The government wants to improve the accuracy of population registries in general, he explained, and Istat's semisupercentenarian survey is a formal part of the National Statistical Plan. Finding and validating Italy's semi- and supercentenarians is especially difficult, though, because civil registries have traditionally been kept by Italy's 8,000-plus municipalities, and population data generated locally tend to be considerably inflated to bolster funding. As an example, Marsilli offered, when he first looked at population registries for the city of Rome, he found 900 male and 1,665 female living Romans born before 1904. Demographic models had projected there would be just 54 men and 69 women over age 105. Once he completed his validation process, the final count was one man and 18 women.
When seeking and validating semi- and supercentenarians for the national registry, Marsilli employs a mixture of tactics. In addition to referencing official records, he tracks media mentions of centenarian birthday celebrations, gathers feedback from relatives about the candidates, interviews municipal officials and sometimes resorts to unofficial "head hunters" for additional detective work. Having completed three surveys, Marsilli is in the midst of his fourth. He has so far validated 633 individuals in Italy older than 105, 17 of whom reached at least 110. The oldest living person in his registry at present is Giuseppe Mirabella, of Sicily, who is about three months shy of his 111th birthday. A slightly more senior gentleman just died in March at age 111.
The incidence of semisupercentenarians in Italy is low but rising, according to Marsilli, who said the number grew by 120 percent between the 2001 and 2011 national censuses. The ratio of women to men is strikingly higher in the north than in the south, he added. But that is in line with trends at younger old ages in southern regions like Sicily and Sardinia, where mortality rates are about equal for men and women.
Data Gaps in South America and Asia
On day two of the workshop, discussion turned to countries beyond Europe and North America. Cassio Turra of CEDEPLAR at the Federal University of Minas Gerais in Belo Horizonte, Brazil, began the session by describing prospects for finding and validating Brazilian supercentenarians, based on his initial research.
Civil registry began with independence from Portugal in 1888, he explained, but registry didn't become systematic until after a 1930 military coup ended the Republic. More recent Brazilian governments have been focused on development concerns and therefore demographers have tended to be interested in tracking infant mortality and other measures of population health at younger ages, Turra said. But interest is now broadening to include compositional changes in the population and connections between early life conditions and mortality at all ages.
Turra thinks the population of semi- and supercentenarians is likely to be relatively small in Brazil, despite its large population, because mortality is high. Finding and validating cases will be a challenge. Census figures appear to be inflated and unreliable. For instance, the 2000 Brazilian census results show centenarians making up 5.1 percent of the population, while the figure for the U.S. the same year was 0.3 percent. The national social security administration has better records, but they are only accessible to researchers in anonymized form. Since the 1980s, all death certificates are sent to the national health ministry, making that agency a potential partner in identifying recent cases.
After describing several examples of validated Brazilian supercentenarians, Turra concluded that targeting a search among higher-SES populations — both because they tend to enjoy better health and to have good documentation — could be fruitful. Workshop participants discussed the pros and cons of that approach, including the major problem of age-ascertainment bias that could result. The same issue would come up again later with regard to non-academic efforts to track down supercentenarians through media mentions and similar sources that don't provide population-representative samples. The tactic was deemed more likely to find people at sensational ages like 115, leading to underestimation of 109- and 110-year-olds. (Reference)
Media attention creates a different kind of problem in Japan, where centenarians are extensively tracked and documented by the government, according to Yasuhiko Saito, of Nihon University Kaikan Daini Bekkan in Tokyo. Recent efforts to protect the privacy of Japan's extremely elderly have led to a closing-off of research access to some government data without explicit permission from each family. "Koseki" registries recording every birth, death and major life event in a family, which Saito described in detail in the group's last monograph, have been a valuable resource for centenarian research in Japan. In Taiwan, which was part of Japan until 1945, there is a similar family-register system, Saito noted. He is just beginning preliminary investigations of supercentenarians in Taiwan — which is not far from Okinawa with its famously long-lived residents — and in Singapore, a wealthy city-state in Southeast Asia with a high life-expectancy at birth that is similar to Japan's.
How to Make it to 105 and Beyond
Whether wealth, parenthood, lifestyle or any other factors stand out as potential contributors to extreme longevity was the topic of the next presentations. Axel Skytthe and Bernard Jeune, of the University of Southern Denmark, gave updated numbers for validated supercentenarians in Denmark. Based on deaths since 1968, they total 378 women and 58 men. At present, there are no living supercentenarians in the Nordic countries, they reported. But there's reason to hope there will be more, since the Danish civil registry has 925 living centenarians of January 1, 2012. And the percentage of centenarians reaching specific age milestones has been going up slightly, but steadily, according to Skytthe. The number reaching 105, for example, has risen by 1.4 percent annually.
"At present, there are no living super-centenarians in the Nordic countries."
Jeune described insights gained so far from four waves of interviews with centenarians since 1998. He found no large differences in survival by gender or socioeconomic status once individuals have passed age 100. Body mass index is not predictive, despite the importance of BMI in the 70s and up to about age 90. Social relations, such as contacts with children and other relatives as well as bereavement, all important for nonagenarians, also have little effect for centenarians. Similarly, a positive outlook marks a clear difference in probability of survival between ages 92 and 102, Jeune noted, but not between 100 and 105. The same goes for self-reported health status, he said. In short, all risk factors become less important with age, and lots of things that are predictive of survival for octogenarians and nonagenarians are no longer significant between 100 and 105. What is important, Jeune found, is grip strength — a significant survival difference goes along with each one-kilo increase in grip pressure. Housing situation, as a measure of independence, is also telling. Being independent, not being cognitively impaired, generally having few chronic diseases and good genes (i.e., APO-E status) are the characteristics that seem to be shared by those nearing semi-supercentenarian status, Jeune said.
Vaupel briefly mentioned an attempt to trace the roots of such robustness in a study of Sardinian centenarians on which he collaborated with Caselli and numerous Italian colleagues. The massive project involved four years of interviews and research, and more than 200 centenarians, about equally divided between men and women. Signs of general good health, even early in the centenarians' lives, were among the more striking findings. Most centenarians had had children, but fewer than the control subjects did, and often later in life. Women who had a child after age 40, for example, had a higher chance of living to 100. Longevity in a centenarian's mother also seemed to be predictive, and infant mortality among the children of both centenarian men and women was significantly lower than among controls. But such results could support a variety of conclusions, Vaupel remarked. Maybe they signal inherited biological sources of health and longevity, or maybe centenarians practiced better hygiene and thereby kept themselves and their children healthier, he speculated. Maybe the centenarians were better cared-for as they aged because their kids survived.
Another unknown that needs further study, Vaupel noted, is the reason behind a possible slowdown in the rate at which people are reaching semi- and supercentenarian status. Changes in modern environments and lifestyles are possibilities. It might also be because frailer people are making it to 100 thanks to medical interventions. That would lead to more deaths at low-centenarian ages and the statistical appearance of a slowdown, whereas in reality less-frail centenarians are living longer. How to get to age 100 and how to persist as a centenarian are, therefore, two different questions.
A Glance Backward
To further explore the question of whether human lifespans are indeed increasing, the meeting's final presentations looked at the mortality of centenarians and younger people in past centuries. Michel Poulain, of the Catholic University of Louvain in Belgium and Talinn University in Estonia, described how records improved in Belgium during the French regime in the late 18th century, when life tables and age-at-death statistics were first kept. A national census and precise statistical records were introduced in the 1840s. With the help of a retired physician, Dany Chambre, who investigated more than 1,000 candidates, Poulain used the Belgian data to validate 402 cases of centenarians between 1751 and 1915. Cases were sporadic until about 1815, according to Chambre, but after that they emerged almost yearly. Before 1840, the sex ratio of the group was 1.35 women for every man, Poulain noted, but between 1840 and 1915 the ratio rose to 2.33. And mortality rates among both men and women in this group ranged between 0.4 and 0.6.
Bourbeau and Desjardins presented their analysis of modal ages of adult deaths in Quebec and other countries. The earliest life tables they examined, from Breslau, Germany, covered the years 1687-1691. But, strangely, the numbers of deaths were about the same at every age between 40 and 70, so a peak age at which the most people died (the "modal age of death") was hard to pin down. Swedish data from the 1750s provided a modal age of death of about 72 (range 70-75). In France, Switzerland and The Netherlands the modal age also hovered in the early 70s (ranges 71-72, <70 and 73-74, respectively) between about 1850 and 1870. The team's best quality data was from Quebec, where they found a slight rise from the 18th century to the 19th. In a sample between the years 1750 and 1754, deaths after age 20 peaked in the early 70s, for example. Between 1740 and 1799, that age rises to the mid-70s. Workshop participants debated whether this change could represent part of a steady increase in lifespans. The beginnings of a well established "health transition" in the northern hemisphere and Europe between 1850 and 1950 might be reflected in the 19th century mortality patterns, some noted. Or, slightly lower modal ages of death in the mid-18th century could reflect a dip in lifespans due to wars and other hardships in the 17th and early 18th centuries, others suggested. Without more background on conditions at the cohorts' time of birth and of death, everyone agreed that little could be concluded but the data were intriguing.
With those last indications of how much more there is to discover about human lifespans and mortality, the workshop wrapped up. The question of when Mme. Calment's record longevity will be equaled or broken remains, but participants took their leave with renewed enthusiasm for adding new data to the IDL and using it to further explore what her example might say about human potential. They'll be reporting back at the next Supercentenarian Workshop, scheduled for the Spring of 2014.